# SOUND, MUSIC AND MOVEMENT IN PARKINSON'S DISEASE

EDITED BY: Marta M. N. Bien´kiewicz and Cathy Craig PUBLISHED IN: Frontiers in Neurology & Frontiers in Neuroscience & Frontiers in Psychology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2016 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

*All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-079-4 DOI 10.3389/978-2-88945-079-4

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **SOUND, MUSIC AND MOVEMENT IN PARKINSON'S DISEASE**

### Topic Editors:

**Marta M. N. Bien´kiewicz,** Aix Marseille Univ & CNRS, ISM **Cathy Craig,** School of Psychology, Queen's University Belfast, Northern Ireland

Author: Jeronimo Sanz, Brazil

(INSTAGRAM:@jeronimo\_dreams; WORDPRESS: https://sonhonosonho.wordpress.com/ FB: https://www.facebook.com/jeronimodreams/; FLICKR: https://www.flickr.com/photos/jeronimooo/)

Artwork based on self-portraits of Tim Andrews as a part of 'Over the Hill: A Photographic Journey' project (http://timandrewsoverthehill.blogspot. fr/2014/01/i-am-by-jeronimo-sanz.html)

Copyright - Creative Commons (CC BY-NC-ND 2.0) with a personal permission from the author to use as a cover for this e-book.

Recent years have brought new insights to the understanding of Parkinson's disease, impact of exercise and sound displays in rehabilitation and movement facilitation. There is growing evidence that auditory signals in the environment can provide a temporal template for movement and change the mode of motor control from intrinsic to extrinsic; habitual to goal-directed, enabling enhanced motor performance in patients. In addition, forced exercise rate studies show that exercising at the pace of healthy adults can have potential neuroprotective benefits for patients. Many research groups have explored the use of auditory cues (such as rhythmical auditory training) in improving gait and upper limb movement parameters. Cues are usually either intermittent (metronome) or continuous (dynamic sound displays). Similarly, dance based interventions suggest that patients benefit from additional sensory information (i.e. the temporal structure embedded in music and proprioceptive information from a dancing partner) that facilities movement. On the contrary, studies dedicated to auditory perception and motor timing report an impaired ability of patients to perceive and synchronise with complex rhythmical structures (i.e. causing an inability to play musical instruments).

With the growth of modern technology and the increasing portability of hi-specification devices (such as smart phones), new research questions on the design of interventions are beginning to emerge as we strive for more efficient therapeutic approaches. In this Research Topic we wanted to bring together top scientists from the movement disorder, motor control and sound related studies along with therapists. That way, we can engage in cross-disciplinary and challenging scientific debate about future rehabilitation avenues and frontiers for Parkinson's disease patients.

**Citation:** Bien´kiewicz, M. M. N., Craig, C., eds. (2016). Sound, Music and Movement in Parkinson's Disease. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-079-4

# Table of Contents


Angela L. Ridgel, Robert S. Phillips, Benjamin L. Walter, Fred M. Discenzo and Kenneth A. Loparo

*62 Cues and Attention in Parkinsonian Gait: Potential Mechanisms and Future Directions*

Daniel S. Peterson and Katrijn Smulders

*67 Beyond the Metronome: Auditory Events and Music May Afford More than Just Interval Durations as Gait Cues in Parkinson's Disease* Matthew W. M. Rodger and Cathy M. Craig

### **II. Motor Control Research**

*72 The Effect of Dopaminergic Medication on Beat-Based Auditory Timing in Parkinson's Disease*

Daniel J. Cameron, Kristen A. Pickett, Gammon M. Earhart and Jessica A. Grahn

*80 Parkinson's Is Time on Your Side? Evidence for Difficulties with Sensorimotor Synchronization*

Marta M. N. Bien´kiewicz and Cathy M. Craig

### *94 Regional Interplay for Temporal Processing in Parkinson's Disease: Possibilities and Challenges*

Michael Schwartze and Sonja A. Kotz

*100 Toward Precision Psychiatry: Statistical Platform for the Personalized Characterization of Natural Behaviors*

Elizabeth B. Torres, Robert W. Isenhower, Jillian Nguyen, Caroline Whyatt, John I. Nurnberger, Jorge V. Jose, Steven M. Silverstein, Thomas V. Papathomas, Jacob Sage and Jonathan Cole

### **III. Music Therapy**

*116 Exploring Music-Based Rehabilitation for Parkinsonism through Embodied Cognitive Science*

Andrea Schiavio and Eckart Altenmüller


### **IV. Dance Therapy**


Laura M. Blandy, Winifred A. Beevers, Kerry Fitzmaurice and Meg E. Morris

*150 The Embodied Self in Parkinson's Disease: Feasibility of a Single Tango Intervention for Assessing Changes in Psychological Health Outcomes and Aesthetic Experience*

Sabine C. Koch, Katja Mergheim, Judith Raeke, Clarissa B. Machado, Eliane Riegner, Joachim Nolden, Gudrun Diermayr, Dorothee von Moreau and Thomas K. Hillecke

*163 Verbal Auditory Cueing of Improvisational Dance: A Proposed Method for Training Agency in Parkinson's Disease*

Glenna Batson, Christina E. Hugenschmidt and Christina T. Soriano

*173 Dancing with Disease: A Dancer's Reflections on Moving with People with Parkinson's and Memory Loss*

David William Marchant

# Editorial: Sound, Music, and Movement in Parkinson's Disease

*Marta M. N. Bie´nkiewicz1 \* and Cathy Craig2*

*1CNRS, ISM, Aix Marseille University, Marseille, France, 2School of Psychology, Queen's University of Belfast, Belfast, UK*

Keywords: PD, exercise therapy, music therapy, dance therapy, cueing

**The Editorial on the Research Topic**

#### **Sound, Music, and Movement in Parkinson's Disease**

As editors of this special edition on sound, music, and movement in Parkinson's disease (PD), we are delighted with the final collection of papers that have been published in this research topic. We would like to express our sincere gratitude to all of the authors who accepted the invitation to participate: a list that includes not only full time researchers but also clinicians, movement therapists, and dance professionals. We would like to thank all the reviewers, who gave up their time to critique the submitted articles and also the substitute editors who stepped in whenever needed. We thank the participants for their significant efforts, without which, these research projects would not be possible. Finally, we thank the Frontiers editorial and technical staff for their guidance and patience.

Research published over the last few years has reinforced the idea that activity and vigorous exercise have an important role to play in ameliorating the disease progression and preventing secondary health problems in PD (1–3). However, for patients, every movement requires a lot of effort and can easily cause fatigue, a phenomenon that often discourages patients from actively participating in physical therapy. Over the last decades, various groups of researchers have looked at how cueing (i.e., providing an external sensory framework such as a beat) can help support and improve the initiation and timing of movement. Stemming from the seminal work of Martin (4), therapies using visually, acoustically, and somatosensorially enriched environments have been reported to improve motor function, posture, and well-being in patients with PD. This ability to pick up and use external sensory information to guide and time movement appears to remain intact in people with PD unlike the ability to initiate and control intrinsically driven actions that appear to be more adversely affected by the disease [see Ref. (5) for plausible physiological model].

As a final collection, this special issue allows us to disseminate state-of-the-art knowledge on the functional deterioration of motor control and present novel behavioral interventions that aim to alleviate symptoms in PD. In particular, we are interested in forms of movement therapy that are sustainable, focused on improving quality of life in the long term and feasible even where resources are scarce. Our parallel aim was to push the Frontiers of our understanding to see how sensory information can afford and shape movement facilitation in PD and how our knowledge can feed into the design of tailored rehabilitation programs. This is why experts from the fields of auditory stimulation, neuroimaging, motor control, and dance therapy were invited to engage in a dialog on the current and future management of PD, suggesting possible new routes for therapy while outlining the limitations of our current scientific understanding. The end result of this international effort is presented in this e-book.

In order to organize the articles, we have divided this collection into four sub-themes.

### *Edited and Reviewed by:*

*Alberto Albanese, Università Cattolica del Sacro Cuore, Italy*

#### *\*Correspondence:*

*Marta M. N. Bie ´nkiewicz mbienia@gmail.com*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 04 September 2016 Accepted: 16 November 2016 Published: 28 November 2016*

#### *Citation:*

*Bie ´nkiewicz MMN and Craig C (2016) Editorial: Sound, Music, and Movement in Parkinson's Disease. Front. Neurol. 7:216. doi: 10.3389/fneur.2016.00216*

### GAIT AND LOWER LIMB MOVEMENT THERAPY

This section is opened with a review article from Hackney et al., which provides a comprehensive introduction to the core theme of this research topic. The focus is on neural substrates used for internally and externally guided movement in healthy participants and PD patients. The compendium of articles presented in this review provides an overview of the possible functional basis for the efficacy of pace-based rehabilitation interventions and also identifies future directions that merit additional investigation. The next article by Ashoori et al. continues along this theme and takes an in-depth look at the subject of rhythmic auditory stimulation (RAS). The authors not only discuss the underlying mechanisms for its therapeutic power but also deliver a synopsis of the benefits stemming from RAS-based interventions and other technological innovations that enable the creation of online cues that are adapted to the needs of each patient. This article is supplemented by the work of Maculewicz et al. and presents a roundup of the technological solutions currently available that make use of instrumented footwear and that can also be used for RAS. We would like to redirect readers, who are interested in finding out more about the progress of health informatics in PD management, to the recently published article of Espay et al. (6).

In terms of original research, two studies offer a promising outcome. First, the work of Pau et al. looks in detail at the spatiotemporal and kinematic changes in gait patterns following a therapeutic program that encompasses RAS training. Second, Ridgel shares an update on the novel developments in the high-cadence cycling therapy and its prevalence over static cycling; another significant step forward following the previous seminal work of Ridgel et al. (7, 8). This study brings hope that effective therapeutic regimes, which exploit our knowledge of high-intensity training, will be available for PD patients in the near future. Finally, two opinion papers compliment this section – Peterson and Smulders provide insight into the attentional aspects of parkinsonian gait and its implications for the design of cueing interventions, while Rodger and Craig discuss how there is a need to go beyond the metronome and consider the wealth of additional information that music or action relevant sounds (9) can offer in terms of sensory cueing. More complex and rich auditory structures are postulated to grant a more flexible mapping between the timing of each step and the temporal structure afforded by the beats, melodies, and chord progressions (these points are further discussed by Schiavio and Altenmüller).

## MOTOR CONTROL RESEARCH

Difficulties with the spatiotemporal control of movement are the central theme of this section. Cameron et al. report on the effects of dopaminergic medication on two timing tasks that are based on rhythm discrimination and alignment. The authors conclude that medication supplementation and disease progression affect the ability to discriminate complex non-beat structures, but do not affect rhythm alignment ability compared to healthy adults. In addition, the authors find specific increases in the sensitivity to beat signals with dopaminergic medication. A study by Bieńkiewicz and Craig delivers some preliminary evidence for a correlation between severity of PD and difficulties in synchronizing movement to a simple beat, despite a preserved ability to apply as efficient movement strategies as healthy adults. Synchronization difficulty was found to be independent of movement amplitude and/or cognitive load. The interplay between the temporal processing and motor signs in PD is further discussed in the Perspective paper written by Schwartze and Kotz. These authors shift the focus from the functionality of the basal ganglia to a more widely distributed neural connectivity that includes the role of the cerebellum and the supplementary motor area in moderating the symptoms of PD. It is argued that impairment in temporal processing will have implications for the design of therapeutic interventions aimed at improving global motor function.

The closing contribution is a Method article presented by Torres et al. This paper demonstrates that by empirically estimating the family of probability distributions inherently present in the data, rather than *a priori* assuming a theoretical one, it is possible to extract the noise-to-signal ratio inherent in the data. This information, often called "noise" and traditionally smoothed out by averaging across epochs of data, actually contains a rich source of information about the integrity of the nervous system and progression of a neurological condition. The authors propose a novel platform for individual behavioral data analysis referred to as "precision phenotyping" and demonstrate its translatory power for the future development of personalized medicine as well as being a tool for distinguishing neurological conditions with often similar behavioral manifestations.

### MUSIC THERAPY

The music-based rehabilitation section is opened by a Hypothesis and Theory contribution by Schiavio and Altenmüller, which discusses the intricacies of the interactions between human cognition and the world from an embodied perspective. The authors point out the circular relationship between the body, brain, and the surrounding environment and the need to incorporate this into a rehabilitative context. From this perspective, motor rehabilitation interventions can be seen as reestablishing the lost relationship between the agent and the system, and not simple input/output dependencies. Music provides more than a simple timekeeping aid, by affording a variety of mind–body responses from self-regulation to sensorimotor coupling. Therefore, music motor therapy is not only effective from the mobility point of view but also from a psychological, socio-affective, and wellbeing standpoint. This message is repeated in two short Opinion papers that provide an additional overview of the power of music and RAS-based programs designed to alleviate parkinsonian symptoms. Mainka reinforces the idea that music offers a superior approach to cueing movement in PD as its stimulatory power exceeds simple pacing through the esthetic qualities of the music that induces affective changes in the listener, which impact on general well-being. The structured auditory signals offered through music are easy to memorize and allow for a carry over effect from training after the session has finished. Moreover, Raglio points out that strong methodological criterion should be employed for future studies investigating not only motor improvements but also psychological outcomes of these types of interventions to allow for a direct comparison with other exercise regimes and RAS programs.

### DANCE THERAPY

Apart from gait training, music is a canvas for dance-based rehabilitation in PD. The opening piece in this sub-theme is an Opinion paper by Dreu et al., which provides an essential introduction with a short summary of the body of research in this area. Evidence for the multi-faceted benefits of partnered dance is listed and includes augmented mobility, improved balance, and general improvements in well-being as the primary outcomes. In addition, the authors discuss the importance of an enriched environment for successful therapy along with the somatosensory cues available from bodily contact with another person during dance.

A further three original articles provide examples and guidelines for designing a feasible dance program for PD patients, with measures of psychological outcome being included as well as improvements in mobility. Blandy et al. disseminated their work on a partnered tango intervention with a proven safety and psychological health enhancement record. Similarly, Koch et al. describe an original non-partnered tango program and report increased well-being and self-efficacy measures for those

### REFERENCES


who participated, reinforcing the key outcomes mentioned by Schiavio and Altenmüller. Following an embodiment approach, Batson et al. present a methodological paper that focuses on training agency in PD and propose an active improvisation dance program. In this scenario, patients are encouraged to react freely to verbal cues, mirroring the unpredictability of daily interactions. Finally, the last paper is by Marchant, who writes from the perspective of a professional dancer and teacher and discusses the important points that need to be considered when designing therapeutic dance interventions. The author draws on his own experience and the many research projects he has worked on with vulnerable groups such as PD patients.

Our hope for this collection of papers is that we shine new light on PD rehabilitation and provide inspiration to anyone who may benefit, whether they are researchers, practitioners, therapists, or simply the wider public. Last but not least, we hope PD patients reading this will feel motivated to actively seek out programs that use sound, music, and movement so that they can lead more active and fulfilling lives.

### AUTHOR CONTRIBUTIONS

Both the authors (MB and CC) contributed equally to the ideas conveyed in the Editorial. MB wrote the main draft, which CC edited in terms of content and language. Both the authors approved the final manuscript and agreed to take responsibility for content contained.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Bieńkiewicz and Craig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# **Context-Dependent Neural Activation: Internally and Externally Guided Rhythmic Lower Limb Movement in Individuals With and Without Neurodegenerative Disease**

*Madeleine E. Hackney 1,2 \*, Ho Lim Lee<sup>3</sup> , Jessica Battisto<sup>3</sup> , Bruce Crosson1,4 and Keith M. McGregor 1,4*

*<sup>1</sup> Atlanta VA Center for Visual and Neurocognitive Rehabilitation, Decatur, GA, USA, <sup>2</sup> Division of General Medicine and Geriatrics, Department of Medicine, Emory School of Medicine, Atlanta, GA, USA, <sup>3</sup> Emory College of Arts and Sciences, Emory University, Atlanta, GA, USA, <sup>4</sup> Department of Neurology, Emory School of Medicine, Atlanta, GA, USA*

#### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*James Shine, Brain and Mind Research Institute, Australia Lawrence Mitchell Parsons, University of Sheffield, UK*

> *\*Correspondence: Madeleine E. Hackney mehack@emory.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology Received: 16 September 2015 Accepted: 16 November 2015 Published: 02 December 2015*

#### *Citation:*

*Hackney ME, Lee HL, Battisto J, Crosson B and McGregor KM (2015) Context-Dependent Neural Activation: Internally and Externally Guided Rhythmic Lower Limb Movement in Individuals With and Without Neurodegenerative Disease. Front. Neurol. 6:251. doi: 10.3389/fneur.2015.00251* Parkinson's disease is a neurodegenerative disorder that has received considerable attention in allopathic medicine over the past decades. However, it is clear that, to date, pharmacological and surgical interventions do not fully address symptoms of PD and patients' quality of life. As both an alternative therapy and as an adjuvant to conventional approaches, several types of rhythmic movement (e.g., movement strategies, dance, tandem biking, and Tai Chi) have shown improvements to motor symptoms, lower limb control, and postural stability in people with PD (1–6). However, while these programs are increasing in number, still little is known about the neural mechanisms underlying motor improvements attained with such interventions. Studying limb motor control under task-specific contexts can help determine the mechanisms of rehabilitation effectiveness. Both internally guided (IG) and externally guided (EG) movement strategies have evidence to support their use in rehabilitative programs. However, there appears to be a degree of differentiation in the neural substrates involved in IG vs. EG designs. Because of the potential task-specific benefits of rhythmic training within a rehabilitative context, this report will consider the use of IG and EG movement strategies, and observations produced by functional magnetic resonance imaging and other imaging techniques. This review will present findings from lower limb imaging studies, under IG and EG conditions for populations with and without movement disorders. We will discuss how these studies might inform movement disorders rehabilitation (in the form of rhythmic, music-based movement training) and highlight research gaps. We believe better understanding of lower limb neural activity with respect to PD impairment during rhythmic IG and EG movement will facilitate the development of novel and effective therapeutic approaches to mobility limitations and postural instability.

**Keywords: lower limb, motor control, neuroimaging, rhythm, externally cued, internally guided, Parkinson's disease**

### **REHABILITATION IN PARKINSON'S DISEASE**

Pharmacology and surgery do not fully address the motor, cognitive, and psychosocial needs of those with Parkinson's disease (PD), a neurodegenerative disorder that is related to dopamine depletion in the substantia nigra pars compacta, which in turn hinders processing in the basal ganglia (7). Several mobility programs are effective (e.g., mobility training, dance, tandem biking, and Tai Chi) for improving motor symptoms, lower limb control, and postural stability in people with PD (2–6). These programs use a mixture of internally guided (IG) and externally guided (EG) movement strategies, both of which have evidence to support their use in rehabilitative scenarios. However, little is known about rhythmic lower limb movement relating to locus of cue (EG vs. IG). Many rehabilitative programs preferentially select one locus over another, which may or may not be optimal for long-term improvement of mobility. Better mechanistic understanding of beneficial exercise effects on neural circuitry garnered under specific contexts could improve the design of motor rehabilitation interventions for particular symptoms (e.g., freezing and bradykinesia) and the various disease stages of PD.

The goal of this review is to provide rehabilitative specialists and researchers with the state of the rehabilitation science regarding the potential neural underpinnings of IG and EG movements. Respective of this, IG movement neural dynamics will be contrasted with those of EG movements. As implemented, the dichotomy provided in this review respective of movement locus of cue is didactically necessary but practically difficult to realize from a rehabilitative perspective. Ecologically speaking, human movement is rarely, if ever, purely IG or EG, whether in the case of daily activities or in rehabilitative settings. That said, determining the beneficial and most effective qualities and outcomes of IG and EG motor training could inform rehabilitation particularly for largely intractable conditions like PD.

### **INTERNALLY GUIDED MOVEMENT IN REHABILITATION**

Proper completion of IG movements relies on efficient function of subcortical loops involving the basal ganglia (8, 9). Due to dysfunction of the striato-thalamo-cortical (STC) circuit [also referred herein as the cortico-basal-ganglia-thalamic (CBGT)], people with PD have particular difficulty with IG tasks (10–12). However, this impairment can be remediated by motor rehabilitation that uses skills in which participants engage cognitively in planning and selecting movements (5). Specifically for individuals with PD, having complex movements broken down into simpler elements may facilitate motor performance. Employing a "movement strategy" that demands increased focus on movement plans and mentally rehearsing and/or preparing for selfinitiated movement may be helpful. For example, focusing on critical movement aspects (e.g., longer steps, quicker movements) helps individuals with PD to achieve nearly normal speed and amplitude (13). Thus, IG training may be helpful for individuals with PD.

### **EXTERNALLY GUIDED MOVEMENT IN REHABILITATION**

Abundant evidence also demonstrates benefits of rehabilitative exercise that exploits external cueing, which likely specifically targets neural systems that support balance (4, 14). EG strategies have improved movement initiation (15, 16). Other research has shown that people with PD have faster reaction times when externally cued compared to self-initiated (IG) movement (17). Synchronizing movement to rhythmic beats provided externally may enhance movement speed (18). There is also a well-known facilitating effect of cues for alleviating freezing of gait (FOG) (19). Furthermore, gait training with regular external rhythmic auditory cues has improved gait velocity, stride length, step cadence, timing of EMG patterns, and mobility in persons with PD (20–22). Evidence has begun to accumulate that suggests external cues access alternate neural pathways that remain intact in the individuals with PD, including the cerebellar-thalamo-cortical (CTC) network.

### **CHARACTERIZATION OF THE NEURAL CIRCUITRY INVOLVED WITH INTERNALLY AND EXTERNALLY GUIDED MOVEMENT**

This review covers available literature focused on imaging studies involving IG and EG upper and lower limb movement paradigms within the contexts of cortical and subcortical neural function. Our goal is to summarize findings from the available lower limb literature to inform future research with goals of characterizing neural areas involved in motor rehabilitation of PD. Numerous neural systems likely produce IG and EG movement and could be modulated by rehabilitative training. However, the current work is not intended to provide an encyclopedic reference to neural network function in motor systems, although we provide a reference in **Table 1** that catalogs a number of studies that involve IG and EG paradigms. Rather we will focus on the neural routes that likely assume multiple subcomponents to be explicated in future rehabilitation research. These routes are (a) cortically modulated (mainly in the frontal and parietal lobes), (b) subcortically modulated including the basal ganglia and thalamus, and (c) the cerebellum.

In the neurotypical model, the investigation of IG and EG movements of the upper extremity has received considerable attention in neuroimaging (41, 46, 49, 51, 52). These studies have suggested distinct cortico-cerebellar, cortico-cortico, and corticosubcortical neural pathways for IG vs. EG in a variety of contexts. The following sections will address differences between neural activity from both a region of interest (ROI) and neural network perspective. It needs to be noted that a vast majority of motor-related literature with such a focus has been done in upper extremity movements. Given the relatively young science of neuroimaging, this is somewhat understandable as there are considerable technical difficulties in controlling motion being translated from the legs to the head during movement. As such, studies investigating neural correlates of movements of the lower extremity in humans in the context of IG vs. EG control are rare. We will attempt to address the differences between upper extremity movements and lower extremity movements as available **TABLE 1| A summary of the relevant imaging studies in the context of IG vs. EG movement in healthy controls and individuals with Parkinson's disease**.


Frontiers in Neurology


www.frontiersin.org


literature can inform. However, the scarcity of data involving lower extremity movements does not afford us a complete understanding of the potential variability explained by contrasting upper vs. lower limb movements. While Keele et al. (53) demonstrated that finger tapping, forefoot tapping, and heel tapping are highly correlated in light of timing mechanisms (53), differences between hand and foot movements in functional brain anatomy have been reported in imaging studies (54, 55). An important example of these differences was recently reported by Volz et al. (30) who used functional magnetic resonance imaging (fMRI) to assess cortical network function in younger adults performing unimanual hand or foot tapping. The authors reported significant differences in premotor connectivity with hand movements having a much stronger cortical representation in the motor planning areas. Additionally, upper extremity performance appears to be highly lateralized as compared to lower limb movements. With respect to disease models, the examination of motor control in conditions like PD has largely focused on upper limb control. Respective of this, where applicable and available, we will attempt to differentiate reports involving lower extremity from upper extremity research (**Figure 1**).

**FIGURE 1 | A representative synopsis of the connectivity reported in the current review comparing neural connectivity in internally guided vs. externally guided movements**. The top panel represents connectivity within internally guided movements whose initiation from cortical regions (SMA, M1, and CMA) is mediated by the striatum (caudate for movement selection; putamen for execution) and thalamo-cerebellar bidirectional processing for movement execution. By contrast, the bottom panel represents connectivity during externally guided movements, which originate from sensorimotor integration (M1S1, SMA, and precuneus) due to the external cue for initiation. The progression of motor execution engages the lentiform nuclei (putamen, GP), which then influence cortico-thalamocerebellar processing during task execution. Abbreviations: SMA, supplementary motor area; M1, primary motor cortex; S1, primary somatosensory cortex; STN, subthalamic nucleus; CMA, cingulate motor area; GP, globus pallidus; Thal, thalamus.

**TABLE 1 |**

**Continued**

### **INTERNALLY GUIDED NEURAL CIRCUITRY – CORTICAL, SUBCORTICAL, AND PD**

## **Cortical Activity**

Cortical initiation of IG movements has received a fair amount of study with neuroimaging over the past 20 years. Prior to this, based on primate models, the role of the supplementary motor cortex (SMA) was considered crucial to the initiation of IG movements. Single unit recordings of neurons in macaque SMA showed greater spiking during IG movements rather than those prompted by external cues, which were thought to involve premotor areas (rostral to the anterior commissure in BA6) (56, 57). However, as our understanding of neural circuits improved with neural pathway tracings in primates, anatomical differentiation of these regions informed the functional distinction of premotor cortex, pre-SMA, and SMA (58, 59). When neuroimaging techniques were applied to probe the anatomical substrates involving IG movements, a much different picture with respect to cortical involvement emerged. Multiple labs used fMRI during upper extremity tasks to identify cortical activity during IG movement paradigms and found that the SMA (medial BA6, posterior to the anterior commissure) was not significantly driving the execution of movements in IG conditions (49, 60). Instead, there was a strong influence of a complex cortical and subcortical initiation and gating network that has since been largely confirmed by network-based modeling analysis (39, 50). Moreover, recent work has identified functionally distinct networks respective of the timing of the IG movements with alteration of cortico-basal ganglia involvement during initiation and then execution of the movement. During movement initialization (i.e., the process of identification/selection of motor responses), the anterior cingulate motor area (which may be considered an extension of pre-SMA) and pre-SMA appear to be critically involved as their damage results in improper suppression of selected motor responses (50, 61). However, the transition into direct engagement of skeletal muscle recruits additional cortical resources including lateral BA6, inferior parietal regions, and lateral inferior frontal areas (Broca's area – left BA44) (62). This lateral frontal recruitment could be interpreted as selective movement organization, as Broca's area is crucial for proper syntactic discourse (63, 64). For example, in the English language, syntax is heavily dependent on word order, and even more so in German. Broca's area is critical in identifying the sequential order of word placement (63). In an analogous motor role, activation in Broca's area may indicate order selection for organized movement (65).

## **Subcortical**

The cortical components of IG are strongly gated and in some cases are dependent on reciprocal input from subcortical structures, primarily the striatum. The caudolenticular gray bridges, topographically interposed between the caudate and the putamen, function as the primary efferent center from pre-SMA to the basal ganglia (66). As such, the striatum plays a critical role in modulating movement, and its activity is largely dependent on movement state. A functional dichotomy exists between caudate and putamen activity during IG movements. The caudate is linked to motor learning and sensory processing of proprioceptive input in this context (52, 67). Vaillancourt et al. (39) reported on IG force production at different levels during fMRI (39). Interestingly, the caudate was selectively activated only during the process of identifying and selecting the proper force to produce. Later, the group exquisitely showed that the caudate was selectively engaged in the processing of force for production but did not activate during force production (35). As such, the caudate is likely directly involved in the mapping of higher cortical motor process [input to the gray bridges is through premotor, SMA and M1 connectivity in IG movements (68)]. However, the caudate does not appear to assist in timing of IG movement execution. Instead, this function appears to be fulfilled by the putamen.

Many imaging studies have probed the function of the putamen and its role in timing of IG movements. For years, this structure has been reported as active during execution of IG tasks using fMRI (9, 35, 39, 49, 67, 69). Recently, these reports have been informed by alternate modalities indicating the neural timing of activity in the putamen with respect to regulation of IG movement. Using local field potential (LFP) recordings in macaques, Bartolo et al. recently investigated tuning of spiking potentials respective of movement state, either EG or IG cued (70). While the putamen is involved in both types of activity, the waveforms of the spike potentials tuned differentially to the locus of cue. These waveforms were characterized by frequency consisting of beta (13–30 Hz) and gamma frequencies (30–70 Hz) and were compared during the execution of IG and EG tasks. IG movements were strongly aligned with the beta band of the LFPs. Alternately, during EG movements, LFP activity was characterized by gamma frequency oscillations. Importantly, when sampling across larger distances the beta frequency coherence was high indicating a possible functional coordination of the putamen during IG tasks. Conversely, the spatial coherence of gamma frequencies was very low, which could be interpreted to indicate local, single event processing (70, 71). The authors concluded that the gamma frequency activity potentially indicated local computations directly related to external stimulus processing. In contrast, the beta band preference in the putamen was the result of a larger scale entrainment of CBGT circuits as these were associated with IG movement timing. The authors provided additional evidence for this in a recent paper investigating LFP oscillations during different task types (reaction time vs. internally timed tapping) (72).

### Cerebellum

Although we have thus far focused on the STC circuit in its role for IG movement, the cerebellum has been implicated in IG movement as well. While this may be surprising given the literature recognizing cerebellar involvement in explicitly EG tasks (41, 45, 47, 60, 73), IG tasks have been reported to recruit cerebellar regions. fMRI studies involving IG movements of the upper extremity have reported regional recruitment of cerebellum, including lobulus V (anterior cerebellum) and lobules VIIB and VIII (inferior cerebellum) (74). During arm pointing movements (using both left and right arms), activations were found ipsilateral to movement in lobule V and lobule VIIA (34). While studies employing lower extremity movement during IG tasks are scarce, a study from Schlerf et al. (34), in which they attempted to characterize cerebellar activity during IG upper and lower extremity movements, is worth noting. The group investigated differential cerebellar involvement when modulating task difficulty of both the upper and lower extremities (fingertip and toe, respectively). Toe tapping activated more anterior areas of the cerebellum compared to finger movements regardless of movement difficulty. Interestingly, the group also found greater bilateral cerebellar activity during more difficult lower extremity movements compared to analogously difficult conditions during finger tapping. Other groups have further investigated the possibly differential cerebellar representations of lower limb movement as contrasted to upper extremity movement. A recent study by Kuper et al. (32) aimed to identify overlapping or distinct cerebellar activity when contrasting finger and foot tapping at the joints' maximum movement range (32). Interestingly, while overlapping cerebellar finger and foot activity were present [cf., Ref. (34)], Kuper et al. found activity appeared to follow a somatotopy with foot activation occurring more rostrally compared to finger movement.

However, the attribution of cerebellar recruitment with respect to IG vs. EG locus of cue is incomplete without proper conceptualization of timing within a movement context. In a recent consensus paper (75), Richard Ivry denoted an important qualifier of cerebellar activity respective of movement timing, by noting that much work using imaging of the cerebellum has difficulty teasing apart externally cued movement and its transition to emergent timing. The difficulty is partially explained by the challenge that lies in differentiating absolute time vs. perceived time and the role that the cerebellum plays in optimizing the coordination of the two. Taken in this light, the role of cerebellar recruitment during IG vs. EG movement becomes somewhat mottled in the imaging literature. What may be deemed "internally guided" may be continuously informed by an emergent timing within the participant when this individual is trying to synchronize internal timing to his/her perception of absolute timing.

### **INTERNALLY GUIDED MOVEMENT IN PARKINSON'S DISEASE**

As PD is caused by the loss of endogenous dopamine, the disease profoundly affects basal ganglia function. The most immediate impact of dysfunction in IG movement in PD relates to impaired cortico-basal ganglia communication driving initiation of movement. One such behavioral manifestation of this disruption is the presence of bradykinesia in PD during IG movement (76). As stated above, cortical initiation of IG movements relies on basal ganglia modulation (particularly at the striatofrontal interconnect at the post-commisural putamen) and signal augmentation that is highly sensitive to disruption when filtered through dysfunctional lentiform connections (77). However, this physiology is complicated by the importance of sensory feedback required for proprioception and kinesthetic integration. These afferent inputs are directly modulated by thalamo-cortical relays involving the subthalamic nucleus, ventrolateral, and centromedian thalamic nuclei (68, 78, 79). As such, disruption of the sensory integration back to motor planning cortex further complicates successful movement initiation by disruption of physical state monitoring required for movement.

Deficient gating of sensory signals in the basal ganglia (80) may lead to abnormal processing of proprioceptive input in motor regions, such as the SMA (81). In a related study, Goble et al. (82) examined how brain activity resulting from stimulation of proprioceptors is related to performance in a proprioceptorrelated task (82). Subjects lying in an fMRI scanner received vibrations on their foot, allowing for proprioceptive brain mapping via muscle spindle stimulation. Exposure to foot vibrations showed an association of the basal ganglia structures with structures involved in postural control. Movement studies show the basal ganglia play a significant role in motor learning. In a study by Jueptner and Weiller, positron emission tomography (PET) measurements of regional cerebral blood flow (rCBF) were used for studies of motor learning, visuomotor coordination, and sensory movement control (9). Furthermore, the basal ganglia have been shown to be involved in controlling ongoing movements, including feedback processing (83). Maschke et al. examined the role of the basal ganglia in kinesthesia, the conscious awareness of limb position, with a passive elbow movement task in participants with PD, participants with spinocerebellar ataxia (SCA), and age-matched healthy control participants. The PD participants showed significant kinesthetic deficit to control participants, whereas the SCA participants showed no kinesthetic deficit, allowing the conclusion that CBGT loops are important in kinesthesia (84).

Symptoms associated with PD provide some insight into the disease process on neural circuits and inform rehabilitation strategies of lower limb-related problems. FOG is a common and dangerous condition in individuals with PD whereby the person successfully initiates walking, but it is transiently unable to complete the gait cycle, frequently resulting in imbalance and increased fall risk. FOG is functionally distinct from bradykinesia and postural rigidity and is only conditionally affected by pharmacological treatment (85, 86). Although FOG is challenging to characterize as EG or IG, we will consider FOG a failure to continue a self-initiated movement, the primary manifestation of the condition. FOG very likely represents a disrupted sense of internal rhythmic timing (87). Peterson et al. (88) informed upon the brain regions that are involved in PD patients with and without FOG. In this study, PD patients imagined walking during fMRI acquisition. The results showed significantly lower activity in supplementary motor regions, globus pallidus, and cerebellum in individuals with PD who experience FOG compared to those who did not report FOG. Furthermore, the authors reported decreased activity in the mesencephalic regions associated with postural stability (88). These widespread activity differences between groups indicate that individual discrete neural circuits do not easily account for the motor dysfunction exhibited by FOG. Rather, cortical, subcortical, and cerebellar loops are all affected in FOG. This finding may explain the large variability of pharmacological and surgical outcomes relating to FOG in PD (85, 86, 89). Importantly, this study indicates rehabilitative strategies that are focused on IG movements need to be holistic in approach because no single sub-circuit or structure is preferentially involved in these movements.

### **EXTERNALLY GUIDED NEURAL CIRCUITRY – CORTICAL, CEREBELLAR, AND PD**

Externally guided movements involve both overlapping and discrete brain regions for successful task completion. While studies vary widely in terms of the methodology employed, externally cued movement of differing types involves similar cortical, subcortical, and cerebellar substrates. We will explore these areas as they are associated with EG movements and discuss how PD may disrupt proper execution with implications for rehabilitation programs to follow.

### **Cortical Involvement in EG Movement**

Prior to modern neuroimaging, the initial cortical origin of IG movements was thought to be SMA (57), as cortical activity in EG was viewed in the context of the motor planning and, as such, focused on the relationship between premotor areas with primary motor cortex (90–92). This work in the non-human primate model still provides invaluable insight into cortical function, but it has been updated in the human model with functional neuroimaging. The following text details the differential recruitment of cortex in EG as contrasted to IG in light of modern imaging techniques.

In 2006, Elsinger and colleagues published an important paper differentiating IG and EG cortical dynamics during an upper extremity button press task in fMRI (60). Indeed, this group did show that lateral premotor areas were more active during EG response; however, this activity was likely part of a spatial reaction network associated with the monitoring of external stimuli. This network involves the right hemisphere parietal–premotor–frontal eye field regions (93). This monitoring network can be interpreted exquisitely due to the task-selective nature of the regions recruited. The frontal eye fields are critical in processing saccades and act as a visual motor integration respective of higher-level visual input. The spatial processing of object location is strongly associated with right parietal regions, which would necessarily inform premotor areas to prime selection and then execution of accurate motor response. It has been shown that increased task complexity corresponded to an increase in network coherence (60). Because the task was mapping of specific finger movements to external stimuli, the increased activation of this circuit as a result of increased complexity may be interpreted as multisensory integration of proprioceptive, visual, and kinesthetic information.

Cortical recruitment in EG tasks can also be described in the context of movement entrainment or sequence learning. Much attention has been paid to the role of EG in movement training to automaticity (94–97). Using PET, Jenkins et al. (95) was one of the first groups to use neuroimaging techniques to probe motor skill acquisition and the neural correlates that underlie the stages of motor learning (novel stimulus, movement entrainment, and automatic movement). Overall, the cortical representation pattern was characterized as moving from anterior frontal regions to more posterior areas after continued task practice. Specifically, the group denoted that novel motor learning was the only task condition to show activity in the prefrontal cortex. The entrainment phase of motor practice was associated with consistent activity within the SMAs [see also Ref. (98)], while at movement automaticity (maximum performance accuracy), the dominant cortical activity was in the parietal lobe (95). Work involving fMRI has also shown this anterior to posterior shift in cortical recruitment as tasks become well practiced (96) with an overall reduction of volume of cortical regions recruited after automaticity is achieved.

Importantly, the paradigms denoted thus far have only employed upper extremity movements. Much less is known about cortical execution of EG movements in contrast to IG respective of lower extremity movement. Bruce Dobkin's laboratory has published much of the rehabilitation-focused imaging research on the lower extremity. In a series of studies involving the use of lower extremity EG movement, this team denoted reliable recruitment of left SMA, bilateral sensorimotor areas (M1S1), and right parietal lobules (24, 99). Other labs have also approached lower limb movement with considerations for rehabilitative outcome. For example, Trinastic et al. (26) attempted to differentiate cortical recruitment in motor cortex during EG plantarflexion as compared to dorsiflexion. Findings showed that although dorsiflexion recruited additional cortical areas as compared to plantarflexion, the tasks commonly recruited medial SMA (26). Kapreli et al. (100) approached lower limb movements relating to motor overlap of the somatosensory network shown in upper extremity movements. In this study, the group reported similar regions of activity between tasks; however, lower limb movements were much less lateralized and tended to recruit both hemispheres during motor activity (100). Laterality differences between upper and lower extremities have been reported during visual monitoring (101) and motor tracking tasks (102) comparing movements of the hand vs. the foot. This increased bilateral response [also reported by Trinastic et al.(26) during dorsiflexion as compared to plantarflexion] in lower extremity tasks as compared to those involving the upper extremity has likely been noted because locomotion requires bilateral coordination to maintain balance (26, 103). This differentiation also provides an opportunity for rehabilitation specialists to design interventions that take advantage of the greater cortical recruitment of lower limb movements.

### **Subcortical Recruitment in EG**

Although there exists regional differentiation of cortical recruitment respective of EG vs. IG movements, the neural system does not function as wholly discrete neural compartments as characterized thus far. Understanding the integration of subcortical and cerebellar structures in light of movement circuits is crucial to properly inform rehabilitation using IG and EG strategies in movement.

As stated earlier, sensory integration during EG task execution is a central component to neural activity in this modality. Without continuous state monitoring (postural, positional, visual, auditory, etc.), coordinating movement with external cues would be impossible. As such, the basal ganglia and thalamic sensory relay centers are critical components to successful EG task execution. To this end, in healthy individuals performing EG tasks, a complex interaction of thalamic (pulvinar, ventroanterior lateral, and ventroposterior lateral), cerebellar, and cortical regions form sensory circuits, which allow for selected task actions to coordinate with cues.

Taniwaki et al. (104) used fMRI to characterize discrete corticosubcortical circuits associated with IG vs. EG movements in the upper extremity. The group employed a structural equation modeling (SEM) approach with *a priori* network structures identified to test for path strength between ROIs during EG compared to IG. The group identified discrete ROIs of active structures regardless of movement type before entering the regions into a confirmatory structural equation model. In reference to previous literature and task-based activation in the study (left hand movements), the group identified IG movements with right putamen, globus pallidus, ventrolateral thalamus, dorsomedial thalamus, bilateral ventroposterior lateral thalamus, cerebellum, and SMA. However, the EG movements were associated with right ventral premotor, left ventroposterior lateral thalamus, right dorsomedial thalamus, and bilateral cerebellum. The group tested output path coefficients respective of whether each movement type was associated with CBGT loops or cortico-cerebellar loops. IG involved stronger path links with the CBGT structures listed above. However, EG was associated with stronger cerebellar connectivity to ventral premotor cortex via the thalamus. The cortico-cerebellar connectivity did not involve mediation by the striatum in EG, while IG was strongly associated with putamen activity of the right (contralateral) side. So, although EG tasks appear to be less dependent on the CBGT loops, they rely strongly on cerebellar input (104). Additional work since Taniwaki et al. (104) has largely confirmed this conclusion (43, 73).

### **Cerebellum**

Unsurprisingly, the cortico-cerebellar system is perhaps the most cited neural pathway to be associated primarily with EG movements (41, 43, 45–47, 60, 73, 104). The importance of cerebellar feedback during EG tasks likely indicates the role of the cerebellum as a modulator of complex motor dynamics and proprioception. The cerebellum acts as a *de facto* servo system to modulate gross motor action into controlled and coordinated movement. This is also reflected in the cortical networks upstream of cerebellar activity in EG, as the frontal eye fields perform an analogous role in saccades and transitions to smooth pursuit. Additionally, these structures serve in concert with the semicircular canals and the cerebellum (particularly the flocconodular lobe) to regulate postural stability and vestibular state (105).

Imaging studies have expounded upon our understanding of cerebellar function in cued movements that had previously been largely derived from literature related to cerebellar damage (106). Functional MRI of upper extremity movement has indicated discrete lobes of the cerebellum that appear responsible for facilitating movement in response to external cues. For example, 10 healthy right-handed subjects, while fixating on a visual cross, were cued to press a button in response to hearing a sound, causing activations in lobules V and VI in the right anterior cerebellum (36). With a finger-tapping task similar to the one described above, activations were found in right lobules IV–V and right lobules VIIIA and VIIIB (37). Sauvage et al. (23) compared the neural substrates involved in execution vs. mental imagery of sequential movements (fast and slow) of the left foot in 12 volunteers. Overt movement execution and motor imagery shared a common network: premotor, parietal and cingulate cortices, striatum, and cerebellum (23). Motor imagery recruited the prefrontal cortex, and motor execution recruited the sensorimotor cortex. Slow movements recruited frontopolar and right dorsomedial prefrontal areas bilaterally in execution and motor imagery. Fast movements strongly activated the sensorimotor cerebral cortex. However, the anterior vermis, lobules VI/VII and VIII of the cerebellum were activated in fast movements, in imagery and execution (23). Fast movements are similar to ballistic movements, which have also been implicated in cerebellum imaging studies (107). These findings indicate regional functional specificity potentially exclusive to the execution of EG tasks in the upper and possibly in the lower.

Bostan et al. (108) recently showed that in the cebus monkey, the cerebellum is connected disynaptically with the subthalamic nucleus via the pontine nuclei. While previous work by the group had insinuated structural isolation of the cerebellum from the basal ganglia and STN (68), the findings of this paper clearly show a bidirectional connectivity of these regions in higher order primates. As such, modulation of the STN via cerebellar control is likely implicated in pathophysiology like PD. As described above, the STN exerts powerful effects on the basal ganglia. Given the critical involvement of the cerebellum in EG movements, it is possible that additional cerebellar input accounts for a portion of the behavioral differences of PD patients when comparing IG vs. EG task performance.

Unfortunately, few studies compare EG and IG movements in the lower extremity, which could help delineate the cerebellar components involved with each movement type. Currently, findings indicate EG movements of the lower limb, when employing comparable cues to those used to cue upper limb movements, tend to recruit both overlapping and discrete cortico-cerebellar neural structures (30, 34). Just as the cerebellum is involved in IG movements (in addition to the STC circuit), subcortical striatal structures are involved in EG foot movements. Sixteen healthy subjects performed dorsi-plantar flexion of the foot actively (responding to auditory cues at 1.25 Hz) and passively. Passive movements activated cortical regions similar (but reduced) to those activated by the active task. Activations during active and passive movement were found not only in the contralateral M1 and S1 cortices but also in the premotor cortical regions (bilateral rolandic operculum and contralateral SMA) and in subcortical regions (ipsilateral cerebellum and contralateral posterior putamen) (27). Additionally, differential activation has been noted in the cerebellum depending on whether movements are IG or EG, regardless of presence or absence of visual feedback and activation related to proprioceptive input (29). Clearly, this area of inquiry requires additional carefully planned studies to identify cerebellar circuits and auxiliary structures that can be preferentially targeted by EG interventions.

### **PARKINSON'S DISEASE AND EG**

Parkinson's disease affects both IG and EG movements. However, EG movements may be less impaired in early stages of the disease as compared to IG movements (44). Interestingly, PD is associated with alterations in recruitment of cortico-cerebellar networks despite only mild overt performance differences. For example, Elsinger et al. (48) used the paced finger-tapping task (PFT) (with the right hand) and observed that PD participants had decreased accuracy and increased variability on the task compared to controls. Whether the PD participants were on or off dopamine supplementation did not affect task performance. However, decreased activation in the left sensorimotor cortex, cerebellum, and medial premotor system was noted in PD subjects compared to controls (48). In another study, PD participants (tested both on and off anti-parkinsonian medication) and agematched controls participated in an EG sinusoidal force task with visual cues and varying speeds while gripping a squeeze bulb in their right hand. The group reported that off-medication as opposed to on-medication PD participants recruited the bilateral cerebellum and primary motor cortex as compared to onmedication PD participants and controls (40). Cerasa et al. (47) described findings from a right hand, finger-tapping paradigm of IG and EG movement with visual cues in PD patients. Both PD and healthy subjects engaged somewhat similar neural networks in both EG and IG movement, yet the PD group showed greater activity in sensory and associative cortices. For example, in the EG condition, PD subjects showed increased activation of the calcarine cortex bilaterally, potentially indicating an increased reliance on visual input for task performance. In the IG condition, the cerebello-thalamic pathway was shown to be involved to a greater degree in PD subjects, possibly denoting a compensatory modality shift to EG mapping, which is perhaps more robust to failure in PD than neurotypical IG pathways (47).

With respect to lower extremity function, recent findings have been mixed regarding cerebellar changes in PD. For example, Schwingenschuh and colleagues, using an EG ankle dorsiflexion task, found that people with PD activated lobules I–V in the ipsilateral cerebellum during EG movement, and similar cerebellar activations were found for healthy controls. After oral administration of levodopa, the PD participants showed increased activity in subcortical structures (contralateral putamen and thalamus), compared to control participants who showed no alteration of function. These findings suggest that the cortico-subcortical motor circuit in PD is sensitive to exogenous dopamine administration (25). Externally cued motor imagery has also been employed during fMRI to probe changes in corticocerebellar structures in gait. Cremers et al. (109) reported that people with PD compared to healthy controls had decreased activity in cerebellar vermis and SMA. Importantly, individuals with PD who exhibited greater gait disturbance were less likely to recruit cerebellar and cortical regions characterized by the healthy control group during gait visualization (109). This finding is in contrast with results of Spraker et al. who noted increased recruitment of cerebellar structures and pathways with disease progression (110). Interestingly, what might account for the varied results between studies is the relative difficulty with which studies using fMRI can quantify modulation of cortical structures by the basal ganglia. Recently, Cagnan et al. (111) demonstrated using LFPs with people with PD that phasic synchronization of basal ganglia structures is more associated with tremor and motor dysfunction in PD. Interestingly, when on dopaminergic treatment, the phase locking in beta waveforms in STN and globus pallidus abated to a more dynamic oscillation and, importantly, the individuals with PD exhibited improved motor function. Possibly, were LFPs in STN to be acquired in concert with EEG, alteration of cerebellar and cortical activity may indeed reflect oscillatory activity in the

beta frequency range. In turn, combining the three techniques, while methodologically challenging, may help account for the varying reports of cortical involvement using fMRI.

At present, due to the dearth of research involving lower limb movement and imaging, drawing any conclusions about the effect of PD on neural circuits respective of EG movements in the lower extremity is challenging. **Table 1** lists a summary of the relevant imaging studies in the context of IG vs. EG movement in healthy controls and individuals with PD, which were considered in the text above. Clearly, there are many questions to be answered, making this investigation important for consideration for motor and rehabilitation scientists.

### **REHABILITATION PROGRAMS FOCUSED ON RHYTHMIC MOVEMENT**

Until this point, this review has attempted to elucidate the neural mechanisms that are involved with IG and EG movements respective of the hand and foot respective of both healthy individuals and individuals with PD. We now turn to modern rehabilitative programs that may select for a movement modality or their interaction. These programs selectively engage and optimize movement using internal or external cues, or in many cases, both.

Therapy programs that include external guidance through consistent rhythmic auditory stimulation (RAS) (e.g., a metronome or music) can facilitate movements and are recommended for people with PD. RAS has been used to improve gait in those with PD via external sensory cues consisting of metronome beats. Studies have shown the positive effects of RAS on FOG and gait parameters (112). Although other sensory cues such as visual and proprioceptive cues have been examined, auditory cues appear to be most effective in improving gait in PD (113). Coupling gait to rhythmic auditory cues may rely on a neural network engaged in both perceptual and motor timing in individuals with PD (114). In fact, some individuals with PD may have an impaired perception of beat timing. Leow et al. (115) examined the impact of beat salience in effectively improving gait cadence and other parameters by comparing "high groove" (i.e., music with a strong underlying beat) to "low groove" music. Individuals with poorer perception of beat timing were helped by high groove music because of the salience of the beat. Such musical support might help facilitate gait in those with PD. This finding is highly relevant to dance- or music-based rehabilitation because poor or good beat perception affects gait performance when synchronized to music (115). Leow et al. also showed that more familiar music elicited less variable strides and faster stride velocity and better synchronization with the music (116). Salience of a beat (as mentioned above) and familiarity with music are therefore considerations for rehabilitative purposes.

Indeed, music therapies may have some utility in ameliorating some function in individuals with PD. Recently, Bella et al. trained PD participants on musically cued gait therapy, consisting of synchronizing movement to familiar folk music without lyrics (a bell cued participants' movement). Findings included not only increased gait speed and stride length but also strong gains in motor synchronization (tapping) and perceptual awareness on just noticeably different tasks (117). Findings from stroke literature and the application of music-supported therapy (MST) might also shed light on possible beneficial effects of rhythmic movement with auditory support for people with PD. MST uses musical instrument playing to treat paresis of the upper limb and adheres to four principles: massive repetition, audio-motor coupling, shaping, and emotion–motivation effects. After 4 weeks of MST in combination with usual care, chronic stroke participants assigned to MST showed improvements on the Wolf motor function test in comparison to a control group (118). Additionally, a case study revealed audio-motor coupling when a patient was exposed to a passive listening task with unfamiliar and trained melodies. Before MST, only the auditory cortices were activated; after MST, motor regions were also activated (119). Given that participants actually play an instrument, MST is an intriguing form of musical therapy that can exploit benefits of both EG and IG strategies, and their accompanying neural circuitry.

A meta-analysis recently demonstrated that music-based therapy, including dance, positively affects PD gait and gait-related activities (120). Recently, dance has indeed gained attention as a music-based therapy that may be able to effectively address impairments related to PD. An understudied aspect of dance interventions is the interplay of external vs. internal guidance across multiple sensory modalities. Proprioceptive and kinesthetic inputs based on tactile cues are crucial for motor adaptation and dance performance. Visual cues no doubt play a role in postural control, navigation, and emotional understanding, as well as having a curious positive effect on FOG. However, auditory cues (e.g., percussion or other musical rhythms) play a strong role in guidance of movement and can be EG (e.g., bass percussion) or IG [e.g., fermata pause (notes held longer than music's tempo)] or even delivered as disruptive asynchrony (e.g., syncopation). We believe rehabilitation regimens like dance and other rhythmic training likely provide a synergistic multisensory adjuvant to motor skills training in both aging and disease models. However, in consideration of the discussion of the neural pathways associated with IG and EG motor training strategies, answering the question why dance may be effective is helpful. At this point, it is unclear to what extent external musical auditory or visual (and tactile, in the case of any sort of partnered/contact dance) cues play to elicit therapeutic effects vs. the improvements gained by increased attention and cognitive engagement used to plan and enact movement. Studies are needed to answer these questions when considering the research that has accumulated supporting beneficial effects of rehabilitative methods involving rhythmic training.

### **Dance Therapy**

In the last 10 years, a series of studies have investigated the effects of adapted Argentine tango dance (adapted tango) for individuals with PD. Participants experienced significant gains in mobility, balance, and QOL (121–124). These improvements were maintained 1 month later, (123) and up to 3 months later (125, 126). After participating in 1 year of tango classes offered in the community, participants with PD also demonstrated decreased disease severity (127). Recently, a study demonstrated a 12-week adapted tango program, which was disseminated to several novice instructors and offered in the community, improved spatial cognition, as well as disease severity in participants with mild-moderate PD (126). Other forms of dance have been investigated for efficacy for those with PD. A study that investigated the feasibility of Irish set dancing, in comparison to standard physiotherapy, found the dancing safe and feasible. Furthermore, participants tended to improve more in gait, balance, and FOG after dancing, than after the standard care (128). Dance may have an immediate effect on mobility in those with PD as improvements have been found in as little as 2 weeks of tango (129) and contact improvisation training (130). The very popular "Dance for PD" method has been investigated for its efficacy, and it was found to improve the motor subscale of the UPDRS after 16 sessions (20 h of treatment) in an uncontrolled study (131).

Because several studies have been conducted on the efficacy of tango, it will be examined and considered for its qualities, to serve as an exemplar with qualities that can relatively easily be identified within the IG/EG dichotomy. Argentine tango has steps, patterns, music, and importantly, partnering that may address specific impairments associated with PD. Partner dancing is a sophisticated, yet accessible system of tactile communication that conveys motor intentions and goals between a "leader" (planner of movement) and "follower" (externally cued mover). An "embrace" or "frame" between the leader and follower is the position maintained by the arms throughout all steps in adapted tango. In adapted tango classes (121, 123, 125), participants consistently both led and followed all dance steps with healthy partners, and therefore alternated between two motor training approaches (a) leading, consisting of internally guiding movement plans and (b) following, consisting of responding to external guidance. Thus, qualities of effective rehabilitative programs are found in both leading and following within the context of adapted tango. While in the role of leader, participants practice self-directed, internally generated movements; while, in the role of follower, participants practice responding to external cues from the partner. There are key differences between leading and following that may address specific needs and result in distinct training gains in mobility, because as we have outlined above the neural circuitry that drives leading and following movement likely differs.

Individuals who perform the leading role in dance are thought to adopt a world-centric reference frame. To lead a dance successfully, these individuals need to multi-task by focusing on environment, follower, music, and both current and future motor plans. Leading, which should be using IG cognitive and motor skill, is thought to involve employing a "movement strategy" that demands increased focus on movement plans and mentally rehearsing and/or preparing for movement. Leaders in partnered dances must determine precise spatiotemporal movement parameters of a dance sequence, e.g., amplitude, direction, timing, and rotation. As such, leading may pose a challenge for individuals with PD, given that many have impaired executive control, specifically in cognitive processes involved in planning and executing complex, goal-directed behavior (11, 132). Importantly, the individual who follows in adapted Argentine tango is not required to plan precise spatiotemporal parameters of movement (e.g., direction, length of step, timing, and amount of rotation). From moment to moment, the follower receives movement guidance regarding the afore-mentioned parameters from the leader via tactile cues. Because followers are not devoting attentional resources to planning movement, potentially they can attend more to their postural control, which becomes more and more necessary as a person ages or contends with a neurodegenerative movement disorder.

Although the adapted tango dichotomy of leading and following roles provides a convenient analog to the "ideal" EG and IG motor training vehicle, it must be admitted that underlying strongly rhythmic movements that characterize dance forms continuously employ both EG and IG strategies. Whether these rhythms are internally created or manifested from external guidance (from music and tactile cues of a leader), dance movements obey their inherent timing. As such, rhythmic cues may be very responsible for any gains seen in PD rehabilitation as a result of dance participation. However, in practice, it can be challenging to both train a person with PD (or even a "healthy" individual) in complex shapes of a dance form and also teach them the sophisticated rhythms that make up most dances. As such, the consideration of shape vs. timing must be acknowledged.

The "shape" of a movement sequence is the simple biomechanics of the steps, irrespective of speed or timing. Rhythm usually refers to a repetitive pulse that is repeated in cycles through a musical or movement form. Even IG movement, which does not appear to "obey" a particular repeated rhythm has an intrinsic timing, and occupies a temporal space [see Ref. (75)]. In a dance class for people with motor challenges, there are a number of stages through which a dancer may go in order to achieve dance mastery, which includes a mastery of coordinating movement to music. This movement entrainment may or may not be the same as reports from Wu et al. (96) investigating finger movement training to automaticity, as coordinated rhythmic movements involve a much more complex interplay of IG and EG timing. During rhythmic training, in the first/novice stage, the individual begins to understand a dance pattern, and puts their body through novel motions that will occupy some sort of temporal space, but may not align precisely with a dictated rhythm/timing given by an instructor.With practice the dancer enters a second stage, in which he/she has the motor control to coordinate his bodily timing to the musical timing. As the steps become more complex, there is vacillation between the first and second stages. But whether or not the dancer (a) benefits from precisely moving to the beat of the music, or the dance rhythm (b) from listening to the music, and/or – thinking – they are moving precisely to the beat and obeying a dance rhythm, or (c) benefits mostly from concentrating on the shape of the movement, while creating their own internal rhythmic timing, is unclear.

The role of the instructor is extremely important with respect to properly implementing adapted rhythmic training programs, as the trained instructor becomes the model by which the individual gages performance. During initial instruction, the student models performance externally by visual approximation of his or her own movement to that idealized by the instructor. As the student progresses and achieves additional kinesthetic feedback and motor flow during the rhythmic movement, motor adaptation translates from externally derived imitation to IG motor flow. However, despite this transition, the role of the instructor to provide the movement template is central to successful training. In these regards, the student's interpersonal relationship with their instructor has great import on their training trajectory; however, this relationship necessarily introduces variability due to social interaction between student and teacher. Interestingly, recent work has approached the interpersonal dynamics involving adapted tango. A research group at Emory University and Georgia Tech has been investigating the ability of robots to act as leaders and followers in a simple tango step pattern. The robots are able to maintain a stable distance from their human partner, characteristics of a human counterpart.While experiments using these robots are ongoing, a recent report demonstrated that expert dancers have indicated reasonable ecological validity of the leading and following performance of the robots (133). This line of work offers unique insights as to the effects of interpersonal dynamics on rehabilitative outcome using dance therapy. Furthermore, these investigations could serve as an interesting platform upon which to test ideas about IG and EG movement schemas.

### **RHYTHMIC MOVEMENT IN REHABILITATION OF INTERNALLY AND EXTERNALLY GUIDED MOVEMENT**

At this time, few studies offer evidence of neural changes as a result of focused training in people with PD or healthy controls. A study utilizing PET showed improved vocal intensity after training in the Lee Silverman voice training (LSVT) LOUD program for speech improvement. These motor improvements were correlated with modification in motor, auditory, and prefrontal areas, but there was no effect on the basal ganglia (134). However, in healthy participants lying supine, increased activity in the putamen was noted using PET when tango movements were performed with a single limb to a metered beat (135). In a related finding, after a week of tango lessons, healthy adults exhibited increased activity of supplementary motor (SMA) and premotor cortices during imagined tango-style walking. In participants who had completed a week of locomotor attention training involving physical and mental practice, activation was examined during an overt foot motor task consisting of ankle dorsiflexion. Posttraining the foot task showed reorganization of sensorimotor areas, in keeping with other studies on lower limb motor learning, suggesting that functional connectivity of the sensorimotor network may be modulated by focusing attention on the movements involved in ambulation (136, 137). Dobkin et al. (24) assessed how ankle dorsiflexion could be utilized as an fMRI paradigm to measure the efficacy of a rehabilitative strategy – body weight-supported treadmill training – for hemiparetic subjects. During voluntary ankle movement, the subjects completed two sets of five isolated movements of 10°. The study observed reorganization in the brain with training; specifically, decreased activity in the ipsilateral primary sensorimotor cortex (S1M1) as therapy gains increased and by 2–6 weeks of training, three of the four subjects saw increased activity in the contralateral S1M1 (24). More studies are clearly needed to examine neural changes in combination with observed clinical motor and cognitive changes, particularly with respect to particular motor training strategies.

### **LIMITATIONS**

As stated earlier, there is a paucity of research investigating the neural correlates of lower limb movements in aging and disease. As such, much of the research concerned with initiation of movement from an IG vs. EG perspective is derived from work in the upper extremity. While we have noted overlap of cortical, subcortical, and cerebellar structures when comparing IG against EG movements of the hand and foot, considerable variation has also been reported particularly regarding laterality of movement. While this variation might, in part, limit interpretation in the current review, its presence provides a direct opportunity for inquiry in both basic science and applied rehabilitation investigations. In addition, a challenge that Dobkin et al. have described in imaging-related literature in rehabilitation is the problem of describing neural activity using BOLD imaging respective of time of data acquisition – both post-onset of pathology and progression through the rehabilitation regimen (24). The group has provided evidence that neural activity during early training tends to be larger in volume. However, with increasing exposure, despite similar task performance, neural activity appears to be consolidated to smaller cortical volumes. This presents a challenge in the review of rehabilitation literature as simple static statistical threshold comparisons of activity volumes across studies conflate the neurological variability of the substrates under consideration due to both age and the rehabilitation regimen. Again, due to the low number of rehabilitation studies comparing IG vs. EG in the imaging literature, we lack the statistical tools to describe this variability properly in the current review. Perhaps most importantly, in the current review, we have not discussed the critical aspect of dosing of rehabilitative regimens engaging in EG and IG movement strategies. The identification of an optimized amount of treatment has been overlooked in rehabilitation research. This may have been of necessity given the field has been engaged in identifying specific regimens for efficacy. However, given recent success in identifying programs, it is appropriate to begin to ask the question, "how much?" and to probe for differences between IG and EG motor training with respect to dose. Furthermore, this review has only very briefly outlined the investigation into musicbased and dance-based therapies within the PD population and has likely not thoroughly covered the rhythmic and cognitively driven (IG) aspects of other forms of exercise (e.g., spinning, Tai Chi, walking, swimming, and boxing), which deserve attention.

### **CONCLUSION**

Underlying mechanistic commonalities may exist among therapies that effectively target symptoms of individuals with PD (138). As a jumping off point, one can consider the effects of levodopa and deep brain stimulation, the current standard of care treatment

## **REFERENCES**


for today's population with PD. As with the case of deep brain stimulation, suppression of abnormal downstream network activity produced by the malfunctioning basal ganglia may result from stimulating the subthalamic nucleus (139). Rehabilitation may also create changes downstream of basal ganglia structures. If the mechanism of improvement resulting from IG motor training is similar, there may be a reduction in abnormal neural activity along the STC circuit, which likely mediates IG motor tasks and includes the putamen, ventral anterior thalamus, rostral SMA, and primary motor cortex. An alternative possibility could be increased activity in the basal ganglia, which have been demonstrated to be hypoactive in drug-naïve individuals in early stages of PD (140). Moreover, when examining twins discordant for PD performing right hand, finger sequencing tasks for task specific influences on the STC and CTC pathways before and after levodopa administration, it was noted that levodopa corrected hypoactivation in the contralateral STC, but over-corrected activation in the ipsilateral STC and bilateral CTC pathways; therefore, standard PD pharmacology affects compensatory changes (51) and effective IG or EG motor training as well. Nevertheless, currently, insufficient evidence exists to determine the mechanisms by which IG and EG motor training are efficacious, and future work is necessary to do so. Furthermore, to understand the mechanisms underlying impairments and training effects in whole-body balance and mobility tasks, lower limb neural activity must be investigated first within the context of IG and EG tasks in individuals with and without PD. Knowledge about neural changes that may occur after repeated and targeted training with IG or EG tasks will allow us to develop better rehabilitation training strategies for those with PD and supplement pharmacological and surgical developments.

### **AUTHOR CONTRIBUTIONS**

MH and KM drafted the manuscript, performed literature review, and edited the final manuscript for submission. HL and JB contributed literature search and review and tabular organization. BC performed final review and critical appraisal of the manuscript.

## **FUNDING**

This work is supported by the US Department of Veterans Affairs' Rehabilitation Research and Development grants: #0870-01A1- MH and E-0956W-KM. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government.


study between short- and long-term effects. *Stereotact Funct Neurosurg* (2009) **87**(4):241–8. doi:10.1159/000225977


processing with ankle movements. *Exp Brain Res* (2005) **166**(1):31–42. doi:10. 1007/s00221-005-2335-5


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Hackney, Lee, Battisto, Crosson and McGregor. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Effects of Auditory Rhythm and Music on Gait Disturbances in Parkinson's Disease

*Aidin Ashoori1 , David M. Eagleman2 and Joseph Jankovic3 \**

*1Columbia University College of Physicians & Surgeons, New York, NY, USA, 2Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA, 3Department of Neurology, Parkinson's Disease Center and Movement Disorders Clinic, Baylor College of Medicine, Houston, TX, USA*

Gait abnormalities, such as shuffling steps, start hesitation, and freezing, are common and often incapacitating symptoms of Parkinson's disease (PD) and other parkinsonian disorders. Pharmacological and surgical approaches have only limited efficacy in treating these gait disorders. Rhythmic auditory stimulation (RAS), such as playing marching music and dance therapy, has been shown to be a safe, inexpensive, and an effective method in improving gait in PD patients. However, RAS that adapts to patients' movements may be more effective than rigid, fixed-tempo RAS used in most studies. In addition to auditory cueing, immersive virtual reality technologies that utilize interactive computer-generated systems through wearable devices are increasingly used for improving brain–body interaction and sensory–motor integration. Using multisensory cues, these therapies may be particularly suitable for the treatment of parkinsonian freezing and other gait disorders. In this review, we examine the affected neurological circuits underlying gait and temporal processing in PD patients and summarize the current studies demonstrating the effects of RAS on improving these gait deficits.

*Cathy Craig, Queen's University Belfast, UK*

#### *Reviewed by:*

*Edited by:* 

*Graziella Madeo, University of Rome Tor Vergata, Italy Maria Stamelou, University of Athens, Greece*

> *\*Correspondence: Joseph Jankovic josephj@bcm.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

> *Received: 17 September 2015 Accepted: 22 October 2015 Published: 11 November 2015*

#### *Citation:*

*Ashoori A, Eagleman DM and Jankovic J (2015) Effects of Auditory Rhythm and Music on Gait Disturbances in Parkinson's Disease. Front. Neurol. 6:234. doi: 10.3389/fneur.2015.00234*

Keywords: Parkinson's disease, gait, freezing, music, rhythm

### INTRODUCTION

Gait disorders, particularly freezing of gait (FOG), are among the most disabling features of Parkinson's disease (PD) (1). Rhythmic auditory stimulation (RAS), such as listening to marching music, has been used to ameliorate this motor abnormality (2). The observation that sensory input, such as RAS, can help overcome freezing suggests that the motor program for gait is relatively intact in patients with PD but cannot be appropriately accessed without the sensory input (3). In this review, we will examine the role of auditory rhythm and music on parkinsonian gait.

…there's something about the temporal structure of the music, the emotional content of the music, that arouses areas of the brain that are still functioning and allows a lost ability to become present as they participate in the music. – Dr. Concetta Tomaino, Executive Director/Co-Founder, Institute for Music and Neurologic Function (4)

A recent success story for music therapy from a neurological perspective is the speech recovery of Congresswoman Gabrielle Giffords after suffering a gunshot wound to the head in 2011. Giffords was unable to speak due to severe damage in the left hemisphere of her brain. However, remarkably, Giffords was able to sing parts of songs. After working for several years with music therapy, she slowly regained the natural rhythm of speech through vocalizing musical phrases (5). In Giffords's words, "music therapy was so important in the early stages of my recovery because it can help retrain different parts of your brain to form language centers in areas where they weren't before you were injured" (6). Through singing, Giffords's undamaged brain regions were able to rewire themselves to recover her ability to speak. Indeed, research has shown that music not only helps patients recover from stroke but may improve gait in patients with PD, and learning to play a musical instrument may induce neuroplastic changes that may translate into improved motor and cognitive function (7). This was emphasized by the late Oliver Sacks in his book *Musicophilia*, entirely devoted to this topic (8). Rarely, however, playing a musical instrument may uncover underlying motor abnormality which becomes manifested as task-specific dystonia (9, 10).

Rhythmic stimulation through music and sound has been shown to improve motor deficits in a variety of movement disorders. Rhythm is defined as the time-based pattern of music or sound that consists of perceptible groupings of notes, beats, accents, and phrases (11). Beat is the unit of rhythmic pulse (11).

Tasks requiring melody perception and production recruit both the auditory and the motor areas of the brain (12–17). Passively listening to rhythmic stimuli, even in the absence of motor actions or intent, recruits the auditory systems as well as the mid-premotor cortex (PMC) and the supplementary motor area (SMA) (18). Through a process called rhythmic entrainment (19), humans naturally move in synchrony to external rhythmic cues. The evidence of rhythmic entrainment can be observed when humans spontaneously move or dance to the beats of music, even without being consciously aware of their action. However, rhythmic entrainment is not limited to auditory cues. As humans walk side by side, they naturally synchronize their footsteps without instruction or conscious intent (20–22). This bipedal locomotion relies on our innate internal timing, which may control our conscious and subconscious abilities to extract rhythm from the external world (23).

The strong connections between gait, innate internal timing, and rhythmic perception are demonstrated by humans' rhythmic preference in music. Although humans' perceptible temporal range is 40–300 bpm (24–27), the preferred musical tempo is at 120–130 bpm (28). This preferred tempo is at the middle of the range of the average gait cadence of males (103–150 strides per minute) and females (100–149 strides per minute) across different age groups (29). Accordingly, humans' natural musical rhythmic preferences may have been influenced by their natural spontaneous gait rhythm. This powerful connection between rhythm and locomotion has led rhythmic entrainment to be clinically employed for gait rehabilitation in patients with neurological disorders including stroke, traumatic brain injury, cerebral palsy, and PD (7, 19). Rhythmic entrainment through music tempo has also been used to improve running cadence (30), which may be beneficial in preventing injuries in runners with PD or in athletes with "runner's dystonia" (31).

### GAIT IMPAIRMENTS IN PARKINSON'S AND CURRENT THERAPIES

Parkinson's disease, the second most common neurodegenerative disorder (32), is a complex neurological disorder that negatively impacts both motor and non-motor functions (33). The disease is caused by the degeneration of dopaminergic (DA) neurons in the substantia nigra associated with neuronal inclusions called Lewy bodies, leading to DA deficiency in the basal ganglia (BG) (34). This deficiency results in four cardinal symptoms of PD that can be remembered by the tremor at rest, rigidity, akinesia (or bradykinesia), and postural instability (33–35). These symptoms are often accompanied by gait impairments (36) that are particularly prominent in the postural instability gait difficulty (PIGD), in contrast to the tremor dominant, subtype of PD (37). Gait abnormalities become also more severe in the late-stage PD (38).

Gait disorders in PD are characterized by stooped posture, shuffling steps, flexed knees, narrow base, reduced arm-swing, turning *en bloc*, and FOG, which is one of the most debilitating features of PD (1, 39). While walking, patients suddenly lose the ability to lift their feet and become stuck in place for several seconds or even minutes despite their efforts to initiate forward movement (40). FOG can be provoked by perceived obstructive environmental cues, such as attempting to walk through narrow doorways. Compared to healthy adults, PD patients have a shorter stride length, slower velocity, and more unpredictable fluctuations between consecutive strides (1, 38, 41–46). Indeed, FOG has been shown to be associated with marked disruption to internal rhythmic timing (47). **Table 1** lists and summarizes the basic parameters used to measure the quality of gait.

Emergence of gait abnormalities often indicates a poor prognosis for PD patients as they correlate with bradykinesia, rigidity, and cognitive impairment associated with cortical Lewy bodies (36, 48) and leads to more frequent falling, a major cause of death among patients with PD (1). Several studies have shown that FOG in patients with PD correlates with poor quality of life, disease severity, apathy, and exposure to anticholinergic drugs; it may, but not always, improve with DA therapy (49).


The mechanisms of PD-related gait disorders, and FOG in particular, are not well understood. Impaired functional connectivity between the BG and the dorsolateral prefrontal cortex and the posterior parietal cortex has been suggested by recent connectivity studies (50, 51). Although DA deficits clearly play an important role in gait disturbances associated with PD, FOG often does not respond well to DA therapy, suggesting extranigral pathology in this particular gait disorder. In a cross-sectional study involving 143 PD patients using positron emission tomography imaging, patients with FOG had lower DA striatal activity, decreased neocortical cholinergic innervation, and greater neocortical deposition of β-amyloid compared to non-freezers (52).

Conventional therapeutic interventions for PD, such as pharmacotherapy and deep brain stimulation (DBS), can be effective for treating the cardinal motor symptoms but have shown limited efficacy in gait abnormalities (53). Levodopa, a DA precursor and one of the main pharmacotherapies of PD, has limited therapeutic effects on balance and gait disturbances (40). Furthermore, anti-PD medications may produce side effects including lightheadedness, drowsiness, and dyskinesias which can exacerbate gait abnormalities (1). Although DBS typically improves tremor, rigidity, bradykinesia, and levodopa-related motor complications (54), this therapeutic modality results in only minimal benefits in patients whose primary symptoms are PIGD (1, 55, 56).

### NEURAL MECHANISMS OF CUED GAIT TRAINING

In recent years, there have been numerous studies demonstrating the therapeutic efficacy of RAS in gait abnormalities associated with PD. An increasing body of research suggests that PD involves a deficit in temporal processing (57) and that internal rhythmic timing is more disrupted among PD with gait deficits than among patients without gait deficits (47). It has been proposed that internal timing is dependent on striatal DA levels (58), and that timing problems may be a potential marker for frontal and striatal dysfunctions in PD (59). Accordingly, we hypothesize that the temporal deficits in PD are a major contributor to gait impairments. This is supported by the finding that DA replacement therapy reduces the timing deficits in PD (60), and that timing deficits are induced by changes in the expression levels of striatal D2 receptors (61). Furthermore, timing deficits are also found in other DA-related disorders including schizophrenia (58, 62, 63).

To understand temporal dysfunction, one must consider the two fundamental modes of timing: explicit and implicit timings. Explicit timing is required to make deliberate estimates of duration and relies on internal sense of time (64). Implicit timing utilizes external cues and relies less on conscious time-based judgments, engaging automatic timing systems. An example of an implicit timing task is the serial prediction task, which requires the subject to use a regularly timed stimulus to make temporal predictions about future stimuli (64, 65). Patients with PD have greater difficulty with explicit timing than with implicit timing. More specifically, PD patients have problems with explicit temporal discrimination tasks involving tactile, visual, and auditory stimuli, and explicit timing performance decreases as disease severity increases (66–69). The underlying neural networks of implicit and explicit timing are distinct. While implicit timing mainly recruits the cerebellum and is less dependent on the BG and the SMA (70–72), explicit timing recruits the BG, the SMA, the PMC, and the cerebellum (73).

The BG–SMA–PMC network is directly involved in rhythm perception in the presence or absence of motor actions (18, 74, 75). In this network, the dorsal striatum (caudate and putamen) of the BG serves the most crucial role since it generates the internal pacing required for time estimation (73, 76). Thus, the BG is directly involved in perceptual and motor timing (77–79). The D2 receptors in the striatum mediate the DA signaling that controls the speed of this internal pacing (80–85). The lack of DA innervation to the BG in PD causes slower internal pacing (76), which leads to impairments in motor and perceptual timing abilities (17, 69, 72, 86, 87). In further support of the BG's crucial role in timing, non-PD patients with focal lesions in the BG have similar difficulty with motor rhythmic synchronizations and have difficulty adapting to tempo changes (88).

Given that gait and other motor deficits in PD are strongly associated with timing impairments, RAS is a promising strategy for gait rehabilitation. Although PD patients have impairments with external timing due to internal pacing dysfunction, patients still have the ability to make temporal predictions through implicit timing. In other words, PD patients can still use external rhythmic cues to inform temporal-based decisions, such as when the next footstep should occur. Since implicit timing is still mostly intact in PD patients, they compensate for the disruption in the BG–SMA–PMC (explicit timing) by recruiting the cerebellum (89) (essential for implicit timing).

Although internal pacing is disrupted in PD patients, this timing alteration can be corrected and recalibrated through motor–sensory interaction with the world (3, 90). Cued gait training utilizes the implicit timing abilities still present in PD patients to recalibrate the internal clock. In RAS, PD patients are instructed to walk while synchronizing their footsteps to the salient beats of the music or metronome. RAS can be combined with visual cues such as patterned tiles or stripes placed along the walkway for multisensory cueing.

The schema in **Figure 1** summarizes the basic neural pathways involved in gait training. In the absence of external cueing, internal cueing signals generated by the BG–SMA–PMC circuit feed into the motor programs, which are carried out in the medial motor areas comprised of the SMA and the cingulate motor area (91). During locomotion, the spinocerebellar, the spinothalamic, the spinoreticular, and the spinohypothalamic tracts carry somatosensory information, such as proprioception back to the brain (3, 92). The information carried by the somatosensory feedback modulates the internal clock of explicit timing (62) in the BG–SMA–PMC circuit and helps plan and predict future cued motor tasks.

The motor programs of gait appear to be relatively intact in PD patients, but due to impaired internal timing, the programs cannot be easily accessed without external cues (1, 3, 33). External rhythmic cues include visual and auditory sensory stimuli and can serve as surrogate cues for the impaired internal timing (93, 94). Accordingly, auditory and visual stimuli can bypass the

damaged BG and help the patients improve their gait by inducing motor–sensory feedback signals that recalibrate internal pacing. After the correct temporal scheme is re-established with RAS and potentiated through the BG–SMA–PMC circuit, patients can sustain improved locomotion for a period of time in the absence of external cueing. Gait rehabilitation through RAS has been recognized to benefit PD patients for almost two decades. In 1996, Thaut et al., using renaissance style instrumental music as the rhythmic cues, 3-week gait training with RAS significantly improved gait velocity, cadence, and stride length in PD patients (44). One year later, a similar study showed that RAS with cues that were 10% faster in tempo than the patients' baseline cadence had even a greater improvement on gait deficits (95). Since then there have been numerous reports on the effect of music- or metronome-based gait training in PD patients. Below, we will discuss some of the recent key studies on cued gait training to better understand the challenges of gait therapy and to formulate a future direction for RAS in PD.

### Optimal Auditory Cues for Gait Training

Gait-training studies in PD patients have used either music or simple isochronous sounds, such as a metronome, as cues for RAS. Cue type can affect gait parameters differently depending on factors, such as the participants' health and age. Although there has not yet been a published direct comparison between music and metronome in gait rehabilitation in PD patients, several studies have done this with healthy participants. One study reports that healthy young adults walked faster with music than with metronome cues (96). Another similar study in healthy older adults (age >65) demonstrated that both music and metronome cues significantly increased cadence, but that only music significantly increased stride length and gait velocity (97). Contrary to these results in healthy participants, Huntington's disease patients walked faster when cued by the metronome rather than with music (98).

While the studies in healthy subjects suggest that cues with music are more effective than with a metronome at increasing gait velocities, a study by Leow et al. (99) reports that cues with a metronome rather than with music elicit better gait synchronization in healthy young adults. The same study further compares the effects of two types of music on gait: high-groove music (high beat salience) and low-groove music (low beat salience). Between these two types of musical cues, high-groove music elicited better gait synchronization and faster gait velocity. Low-groove music was not as effective, and even had a detrimental effect on gait in weak beat-perceivers (99). Music familiarity is also an important factor in RAS. RAS with familiar songs results in faster gait velocity and less stride variability than with unfamiliar songs. This is likely due to the fact that synchronizing footsteps to a familiar beat structure require less cognitive demand. Enjoyment of familiar music may also have had a role in eliciting a faster gait (100).

A variety of devices have been developed to provide customized fixed-temp RAS. Recently, a research group in Madrid, Spain (Brainmee™) developed Listenmee®, an intelligent glasses' system, that employs RAS to improve gait (101). The glasses are portable and contain built-in headphones that allow the user to listen to isochronous (metronome-like) auditory cues while walking. The sounds are customizable to various styles, such as ambient, percussive, electronic, and vocal. The user controls the device via Bluetooth with the Listenmee® smartphone application. The research groups plan to turn the device into an auditory feedback system by integrating feedback to spatial movements. The device will include a built-in video camera and a laser emitter to assess motion in the visual field and provide responsive visual cueing. The group has yet to publish the results of the efficacy of this integrated visual and auditory feedback system.

An experiment showing the efficacy of the non-feedback device involved 10 PD patients between the ages of 45 and 65 years (101). Inclusion factors consisted of a history of frequent FOG and falling as well as failure to respond to medication and physical therapy. Five of the patients received DBS with minimal gait improvement prior to the study. In this study, patients were instructed to walk while off DA therapy. Cadence, stride length, and walking speed were measured with and without RAS. Patients showed significant improvement for all three gait parameters while listening to auditory cues.

### Musically Cued Gait Training: Sustained Benefits Beyond Gait Rehabilitation

A recent study by Benoit et al. (102) shows that musically cued gait training significantly improves multiple deficits of PD, including in gait, motor timing, and perceptual timing. The study consisted of 15 non-demented patients with idiopathic PD (Hoehn and Yahr stage 2). The patients had no prior musical training and maintained their DA therapy regimen during the trials. There were three 30-min training sessions per week for 1 month. During each session, the participants walked to the salient beats of German folk music without explicit instructions to synchronize their footsteps to the beat. Compared to pretraining gait performance, the PD patients showed significant improvement in gait velocity and stride length during the training sessions. The gait improvement was sustained for 1 month post-training, indicating a lasting therapeutic effect for uncued gait.

This RAS training also significantly improved motor and perceptual timing. Pretraining, immediately post-training, and 1 month post-training, patients participated in a battery of motor and perceptual timing tests of duration discrimination, beat alignment, paced tapping, and adaptive tapping. Prior to training, 73% of the patients displayed timing deficit that decreased to 67% immediately post-training and only 40% 1 month post-training. Thus, in addition to gait, RAS improves perceptual timing with continued therapeutic effect even in the absence of auditory cueing. This study in the context of the previously mentioned study by Leow et al. suggests a circular relationship between rhythm perception and gait performance: improved beat perception increases the efficacy of gait training (99) and improved gait training increases beat perception ability (102) (**Figure 2**).

### Interactive Cueing Systems

Although the efficacy of gait training with RAS has been proven, the rigid, fixed-tempo of the cues implemented by most studies has limited applications to PD patients. Fixed-tempo RAS requires increased demand for attention to synchronize footsteps with auditory cues, thus invoking higher-level cognitive processes (103). This can be problematic for PD patients, in whom multitasking while walking can trigger or exacerbate their gait difficulties (104–106). Even in healthy participants, fixed-tempo RAS can result in random and unpredictable stride intervals (107). Therefore, attempts have been made to improve RAS by integrating an adaptive system that provides feedback from human rhythm to determine cueing rhythm. A cueing system that aligns to the patients' movements would demand less attention, which may lead to greater gait improvements than with fixed, non-adaptive cueing (108).

WalkMate, an interactive RAS device developed by Yoshihiro Miyake and colleagues, was tested on 20 PD patients undergoing

DA therapy and on 16 healthy controls (109). The device utilizes pressure sensors in the shoes that feed gait timing data into a computer system, and adjust the metronome cueing tempo in real-time. The efficacy of WalkMate on gait was compared with fixed-tempo RAS and a silence-control condition. Gait dynamics were analyzed using the detrended fluctuation analysis (DFA) fractal-scaling exponent, which is associated with gait adaptability and one of the best measures of predicting falling (46, 109, 110). In a silent-control condition the PD patients had significantly lower fractal scaling (higher variability) in stride than the healthy subjects. During fixed-tempo RAS, PD patients' stride had even lower fractal scaling than during the silent-control condition, consistent with past findings (107). With WalkMate, PD patients' fractal scaling became significantly better than the silent-control condition and reached the DFA baseline of healthy subjects in the silent-control condition. Furthermore, gait improvement persisted in the absence of the adaptive WalkMate cues 5 min after the training sessions (**Figure 3**) (109).

More recently, a similar device named D-Jogger was tested on healthy subjects to study the synchronization of gait to adaptive rhythmic cues (111). D-Jogger is a music player that adjusts the musical tempo to the listeners' gait rhythm (**Figure 4**) (112). In the most effective adaptive strategy (out of the four adaptive strategies tested), the participant initially begins walking in the absence of music. The music then begins by the first beat matching the footfall and continues with a tempo equal to the average gait tempo sampled from the previous 5 s. The results from healthy participants motivate further testing of D-Jogger on patients with PD or other movement disorders.

### Virtual Reality: A Potential for Combined Visual and Auditory Cueing

In PD patients, locomotion and postural control have an increased dependence on perceptual vision (113, 114) that can

from Hove et al. (109).

be corrected using visual cues (115, 116). Multiple studies have shown that matching footsteps to visual cues such as equidistant horizontal lines along a walkway improves gait and reduces FOG in PD patients (117–119). Although visual cueing can be beneficial, replicating clinical scenarios would be unfeasible for patients who wish to train at home in a daily basis. Furthermore, as with auditory cueing, fixed walkway strips may be less effective than an interactive system that adjusts cueing based on the patient's movement and gait parameters. Instead, an ideal cueing system would involve adaptive feedback and include both visual and auditory stimuli. Immersive virtual reality (VR) technology could fill this gap by optimizing visually cued gait training. VR is an immersive and interactive computer-generated environment that simulates the real-world experience (120) and can be operated using a custom-made or commercially available headmounted display. The use of VR with visual cueing for clinical rehabilitation is still in its infancy, though multiple studies have found that in chronic stroke patients VR-based training improves cadence, step length, stride length, symmetry, and other gait parameters (120–123).

Recently, immersive VR was shown to be effective for gait rehabilitation in PD (124). The study uses a pair of VR glasses that projects a virtual checkered tile floor into the user's visual field. The user is instructed to walk along the floor, and the VR floor adapts to the user's body movements by simulating the visual effect of walking. Twenty PD patients with a mean age of 71.25 participated in the study. While wearing the device, the patients tried to match their steps with the adjacent tile to regulate their gait via the VR visual feedback. When cued by the VR display, the patients showed significant improvement in walking speed (*P* = 0.002) and stride length (*P* = 0.002) compared to baseline. Fifteen minutes post-training and without the device, the patients showed even greater improvements in walking speed (*P* = 0) and stride length (*P* = 0) compared to baseline (**Figure 5**).

Although these findings are promising, more well-controlled studies are needed to demonstrate the efficacy of VR-based therapies for PD. A potential expansion of VR gait training should involve adaptive, multisensory visual (e.g., virtual tiles or strips) and auditory (e.g., metronome and music) cues. Simultaneous multisensory cues could have a stronger combined effect than each cue alone. VR systems can be portable, enabling patients to train their gait in the comfort of their home. VR devices already have the computing capacity required for the integration of simultaneous adaptive cueing and can be internally processed or remotely processed in a smartphone connected to the VR

device via Bluetooth. Furthermore, VR devices are capable of measuring the users' performance via tracking technology (125), which would allow VR systems to provide feedback of the users' improved gait performance during and following training (126). Thus, a multisensory and adaptive VR device with performance tracking should be explored as a superior gait-training therapy.

### CONCLUSION

Similar to how the metronome helps musicians maintain a steady tempo during a musical performance, RAS provides an effective approach for reducing gait impairments in PD patients. The efficacy of RAS reflects the overlapping neurological domains involved in gait and beat perception. Importantly, RAS is safe (127), inexpensive, non-invasive, and free of adverse health effects. One major limitation to most RAS methods is the fixedtempo design that requires increased cognitive demand and can negatively impact gait. However, an interactive cueing system that adapts to the patients' gait parameters may be able to resolve this limitation and maximize gait improvement from RAS. For RAS to be successful, the intervention should be initiated early in the progression of PD to maximize a participant's ability to adapt to the demands of the training before the development of cognitive impairment.

Further investigation of mechanisms of gait impairment in various parkinsonian disorders is needed. For example, an unresolved question is whether lower body parkinsonism, which is frequently associated with FOG, is a subtype of PD (37) or whether it represents a separate entity, such as vascular parkinsonism (128, 129), cortical Lewy Body disease (48), or atypical parkinsonism such as progressive supranuclear palsy or normal pressure hydrocephalus (1). The long-term impact of RAS on gait impairment and other motor and cognitive deficits should be objectively assessed by randomizing subjects to either participate in RAS by a trained therapist at least once a week for 6 months or participate in routine gait training. Novel methods and instruments, such as quantitative stepping-in-place with a concurrent mental task using a fourth generation iPod Touch sensor system (130), are needed to assess the effects of RAS on gait and mental function. The type of music and rhythm needed to optimize response to RAS should also be further evaluated. For example, in one study of healthy individuals' strikingly prominent (salient) commercially available music increased measures of cadence, velocity, and stride length, but simple music tempo did not (131). We suggest that different types of music, rather than the traditional rhythmic auditory cues, are carefully evaluated in patients with PD to determine which music most effectively improves PD-related gait disorders.

Another approach to gait rehabilitation is the use of VR for PD. While initial research to this immersive approach is promising, further studies are required and should integrate RAS. VR technology holds the potential to deliver more effective rhythmic cues by combining RAS and visual cueing, which we term rhythmic auditory and visual stimulation. With modern technology, VR-based rehabilitation could be made portable, and smartphones could be programed to process adaptive cue algorithms. Portability and

Badarny et al. (124).

ease of use could increase the frequency of gait-training sessions and improve compliance. Adaptive auditory and visual cueing could also be combined with tactile stimulation as a more salient gait therapy for PD patients. Concepts of tactile stimulation could be informed by recent innovations, such as the versatile extrasensory transducer (VEST), a non-invasive, low-cost vibratory VEST developed by Novich and Eagleman (132). Thus, RAS is a promising therapy for the gait impairments in PD and other movement disorders, and combining adaptive RAS with visual

### REFERENCES


and tactile cues in a VR device could further enhance the efficacy of this therapy.

### ACKNOWLEDGMENTS

We would like to thank Amanda M. Buch, a research scientist at Columbia University Medical Center, for her editing and valuable comments. We also thank the National Parkinson Foundation for their support of the Baylor College of Medicine Center of Excellence.


*J Neurol Neurosurg Psychiatry* (2012) **83**(10):979–85. doi:10.1136/ jnnp-2012-302263


motivation and interval timing. *J Neurosci* (2007) **27**(29):7731–9. doi:10.1523/ JNEUROSCI.1736-07.2007


clock speed. *Neuropsychologia* (2013) **51**(2):284–92. doi:10.1016/j. neuropsychologia.2012.09.014


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Ashoori, Eagleman and Jankovic. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# A Technological Review of the Instrumented Footwear for Rehabilitation with a Focus on Parkinson's Disease Patients

*Justyna Maculewicz\*, Lise Busk Kofoed and Stefania Serafin*

*Sound and Music Computing Group, Department of Architecture, Design and Media Technology, Aalborg University Copenhagen, Copenhagen, Denmark*

In this review article, we summarize systems for gait rehabilitation based on instrumented footwear and present a context of their usage in Parkinson's disease (PD) patients' auditory and haptic rehabilitation. We focus on the needs of PD patients, but since only a few systems were made with this purpose, we go through several applications used in different scenarios when gait detection and rehabilitation are considered. We present developments of the designs, possible improvements, and software challenges and requirements. We conclude that in order to build successful systems for PD patients' gait rehabilitation, technological solutions from several studies have to be applied and combined with knowledge from auditory and haptic cueing.

*Edited by:* 

*Marta Bienkiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Yoshihiro Miyake, Tokyo Institute of Technology, Japan Vladimir Ivkovic, Massachusetts General Hospital and Harvard Medical School, USA*

*\*Correspondence:*

*Justyna Maculewicz jma@create.aau.dk*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 30 September 2015 Accepted: 04 January 2016 Published: 20 January 2016*

#### *Citation:*

*Maculewicz J, Kofoed LB and Serafin S (2016) A Technological Review of the Instrumented Footwear for Rehabilitation with a Focus on Parkinson's Disease Patients. Front. Neurol. 7:1. doi: 10.3389/fneur.2016.00001*

Keywords: instrumented footwear, rhythmic rehabilitation, Parkinson's disease, gait rehabilitation, auditory feedback, haptic feedback

### 1. INTRODUCTION

The purpose of this review article is to present a context of usage of instrumented footwear in Parkinson's disease (PD) patients' auditory and haptic rehabilitation. We present developments of the design, possible improvements, and software challenges and requirements. We summarize existing technological solutions and applications. Literature on rhythmic auditory rehabilitation provides design requirements for the hardware and software, which is necessary to build successful rehabilitation wireless systems for PD patients with instrumented footwear as a data collector and feedback device.

### 2. A CONTEXT OF INTERACTIVE SHOE USAGE

Interactive shoes with embedded sensors have been used in many different scenarios. Gait analysis is the prominent one. Tao et al. (1) reviewed gait analysis challenges and presented a selection of wearable sensors, which can be used for gait analysis. In the context of PD, it is necessary to collect information about patients' balance (2), gait cadence, velocity, and stride length (3). Interactive shoes give opportunities for collecting data about users' current state. This can be followed by notifying users or supervising persons (e.g., doctors and physiotherapists). Data gathered by sensors can send information to rehabilitation systems when a specific problem or need occur and activate cueing stimuli in the form of auditory signals or vibrations.

## 2.1. Gait Impairments in PD

Gait impairments have received a lot of attention in recent years since they are a common cause of disability in people with PD (4). Various aspects of gait have been found to be affected by PD, but they can be influenced and improved through rehabilitation based on auditory or haptic cueing. The most common ones are freeze of gait (5), balance (6), gait velocity, cadence, stride length (3), increased spatio-temporal variability (7), and difficulties with gait initiation (8). These disturbances are lead to restricted mobility, weakened balance, and consequently to increased risk of falling (9). Such disturbances have been found to influence patients' general quality of life (10).

### 2.2. Rehabilitation through Auditory Stimulation

### 2.2.1. Metronome-Like Rhythmic Stimulation

It has been shown that following a rhythmic auditory cue helps gait performance in patients with PD (11–15). The PD patients are usually provided with an auditory metronome, or markedly rhythmic music, and asked to match consecutive footfalls with the onset of each beat (16). External rhythms presented by auditory cues may improve gait characteristics (13–15) and can also be used to identify deficits in gait adaptability (17). Spaulding et al. (3) in their review pointed that the auditory cueing elicited positive changes in gait cadence, velocity, and stride length.

### 2.2.2. Mutual Entrainment

The aforementioned way of stimulation lacks the interactivity component where the system could adapt to the user. In this case, mutual entrainment between system and user happens. Miyake (18) proposed the Walk-Mate to implement the mutual entrainment for the rehabilitation of PD and hemiplegic patients. Baram (19) suggested that gait rehabilitation must be performed in a closed-loop system to avoid constant vigilance and need of attention strategies to prevent reversion to impaired gait patterns caused by repetitive stimuli. Hove et al. (20) reported that random, disconnected stride times (low fractal scaling) predicts falling for PD patients. Fixed rhythmic auditory stimulation lowers fractal scaling and requires attention. Gait rehabilitation should lead to achieving the more stable and not random stride times structure, which can be observed in healthy gait (20). Systems based on mutual entrainment principles can emergently respond to unpredictable changes in human behavior (21). The studies by Hove et al. (20) and Uchitomi et al. (22) showed that the gait fluctuation of the patients gradually returned to a healthy stride times fluctuation level in the interactive conditions. This effect did not occur in fixed tempo and no-cue conditions.

### 2.2.3. Improvements through Ecological Stimuli

Rhythmic sounds (metronome-like) only specify step duration of gait, with no information relating to spatio-temporal properties of walking actions. Rodger et al. (16) recently proposed the use of ecological signals as a new approach to auditory rehabilitation. Ecological signals, defined as those stimuli, which are encountered in everyday life, have the potential to convey richer information. Listeners, based on footstep sounds, can determine gender and mood of the walking person (23). Complex walking sounds, such as footsteps on gravel, may convey both temporal (step duration) and spatial (step length) properties of gait (16).

### 2.3. The Advantages of Haptic Stimulation

Little research has focused on foot-based vibrotactile systems. The sensitivity of the sensory system of the feet is sufficient for vibrotactile guidance (24). Signals from mechanoreceptors in the foot are one of the main sensory sources for gait generation and modification (25). It is likely that mechanoreceptive afferents in the sural nerve provide rich information about contact patterns between the foot and the environment during stance and locomotion (24). When exposed to audio and haptic stimulation, subjects are able to best recognize different materials delivered haptically or as a combination of auditory and haptic feedback (26). Both auditory and haptic feedback are represented as temporal variations, which can be simulated with similar patterns, at different frequency ranges. Since most of the pedestrians wear shoes when walking it makes it an excellent platform for mounting actuators (27).

## 3. THE EXISTING TECHNOLOGIES

This section presents studies describing foot plantar measurement systems and their usage in auditory and haptic rehabilitation and focuses on main advantages coming from their possible usage.

### 3.1. Foot Plantar Measurement Systems

In this subsection, we will mention a few interesting instrumented insoles which followed by a review by Abdul Razak et al. (28).

Amjadi et al. (29) presented a flexible foot pad containing force-sensitive resistors arrays for the foot sole distributed force detection. Stassi et al. (30) described an easy and cost-effective approach used to fabricate the conformable insole based on a piezoresistive material. It measures both the pressure distribution under 64 nodes arranged in the main plantar regions and the mean plantar pressure during walking activity with a sampling frequency of 20 Hz. While developing instrumented insoles, Tamm et al. (31) focused mostly on accuracy, long-term stability and reproducibility, and time resolution. They proposed a thin, light weight self-contained platform for mobile wireless pressure sensing insole system with 24 separated points of measurement on the foot. Suresh et al. (32) demonstrated a proof-of-concept of a new high-resolution plantar pressure monitoring pad based on fiber Bragg grating (FBG) sensors. Motha et al. (33) introduced a unique approach to measure applied pressure. The change in capacitance is entirely led by variation of relative permittivity of the surrounding dielectric medium with applied pressure. Tan et al. (34) presented another low-cost design for plantar pressure measurements. They proposed a system based on carbonembedded piezoresistive material sandwiched between two layers of electrodes to form a pressure sensing insole.

### 3.2. Foot Plantar Measurement Systems with Its Tested Applications

Redd and Bamberg (35) presented a simple system of two forceresistive sensors per insole, connected to a mobile application that delivers feedback. Their tests showed that the feedback system is capable of influencing the gait of the user, without the need for direct supervision by a rehabilitation specialist. The system proposed by Santoso and Setyanto (36) consists of a sensing unit and a signal processing unit. The sensing unit is based on piezoelectric stress sensor module (two in each insole) and data acquisition module both wirelessly connected. The system is able to differentiate running from walking in athletes performance. Madavi and Giripunje (37) proposed a wearable device for the diabetic person of sensory neuropathy. It identifies the ulcerous condition, which may be created in foot plantar surface area. Temperature sensors or pressure sensors are used to detect the infected area. The output data can be transmitted wirelessly to the hospital system. Grenez et al. (38) described the development of a hardware system simulating a shoe, which consists of three pressure sensors, two bending sensors, an accelerometer, an Arduino mini, and a Bluetooth module. The developed prototype is able to differentiate between healthy gait and imperfect gait. Holleczek et al. (39) described the development of textile pressure sensors, which are more comfortable in usage than standard ones. The textile pressure sensors were developed using the principle of a variable capacitor. These sensors were attached to the socks (three in each sock at relevant positions under the heel and the ball of the foot) of a snowboarder for the monitoring of the in-shoe pressure distribution.

### 3.3. Instrumented Footwear Systems with Vibrating Stimulation

Hijmans et al. (40) described a technology, which could be used in the future to improve balance in healthy young and older people and in patients with a stroke or diabetic neuropathy. The system uses cork insole covered with a leather layer. A C2 electromechanical actuator and a piezo actuator or the VBW32 skin transducer, activated by a custom-made noise generator, were chosen to provide tactile stimulation to the feet.

The goal of the application called Gilded Gait is to simulate the perception of a range of different ground textures and serve as the navigation in the city (41). The system contains six vibrations panels to present the feedback patterns, push-down switch, and an accelerometer to detect user's steps. Three different patterns of vibrations were designed to simulate different ground textures. Recognition of the patterns was possible only if they were asked to choose from the list rather than recall a material.

Velázquez et al. (42) described the development of a tactile communication system (16 actuators embed into a shoe insole) and a pilot study on recognition of different information (direction, pattern, emotion recognition, and language learning) assigned arbitrary to vibration patterns, which as was shown can be easily learnt and understood.

### 3.4. A System Based on Accelerometer

Walk-Mate (43) is a system mainly used as a gait compensation device and as a gait rehabilitation training device by analyzing improvements in locomotion before, during, and after rehabilitation in hemiparetic patients and comparing it with a previous gait training method. Walk-Mate generates a model walking rhythm in response to the user's locomotion in real time, and by indicating this rhythm using auditory stimuli, provides a technology that supports walking by reducing asymmetries and fluctuations in foot contact rhythm.

### 3.5. More Complex Systems

Watanabe and Ando (27) introduced a system called Pace-synch shoes. Pressure sensors embedded in a shoe sole served as step detectors and provided data for a vibration motor to be activated. The users reported that when the vibration was presented at heelstrike timing, it was perceived as natural, while the vibrations at the other timings caused odd feelings. They also reported that, when they walked matching their step cycles to the vibration at the heelstrike timing, their way of walking did not subjectively change. As the authors claimed, their method would be applicable for training and coaching in sports and for rehabilitation in health care.

The system described in Zanotto et al. (6) allows for synthesizing continuous audio-tactile feedback in real time, based on the readings of piezoresistive and inertial sensors embedded in the footwear. The system contains 4 piezoresistive sensors, a 9-degreeof-freedom inertial measurement unit, and five actuators in each shoe. All information is stored and processed in the belt unit. The results of this preliminary experiment indicated that ecological underfoot feedback may alter the natural gait pattern of healthy subjects.

The prototype proposed by Tajadura-Jiménez et al. (44) allows for the dynamic modification of footstep sounds, as people walk, and measures changes in walking behavior. This sandals-based system captures sounds of a person's footsteps via a microphone attached to the sandals. Two force-sensitive resistors are attached to the front and the rear part of the sandal insole that detect the exerted force by feet against the ground as well. A triple axis accelerometer was attached to the walker's left ankle. Augmenting the high frequencies of the sound leads to the perception of having a thinner body and enhances the motivation for physical activity inducing a more dynamic swing and a shorter heel strike.

### 3.6. Existing Application for PD Patients

The work presented by Winfree et al. (45) is one of the most interesting studies in the context of this review. These authors described a prototype of a shoe-based system, which contains FSR sensors and an actuator, which are activated in the certain situations. The system was used in a short intervention study with PD patients. The most crucial aspects of the system are its portability, wireless communication, low-cost development, adjustable automatic software, ease of learning, and presentation of appropriate audio and haptic signals. The core of the system operates in a way that if only the ball or toe of the foot is in contact with the ground, the toe actuator vibrates. When both the heel and ball or toe are concurrently in contact, both tactors vibrate. This condition is met during stance phase of ambulation. The Berg Balance Scale (46), Timed Up and Go (47) performance tests, and the FOG questionnaire (48) were used to obtain measurement data during pre-and post-treatment. It was shown that this stimulation provoked significant changes to all measures except time on toe sensor and step duration.

Bächlin et al. (49) aimed to develop a system, which overcomes the limitations of previous systems, such as the continuous nature of the cueing intervention, manually triggered cueing or provided only during training sessions, but not provided at the time of episodic gait disturbances. The challenge set by these authors is to detect freeze of gait episode and apply automatically interactive rhythmic auditory stimulation to overcome the problem. This system consists of a wearable computer, a set of acceleration sensors and headphones. The study was the first one in which FOG is automatically detected, and the results are very promising. The system detected FOG with high sensitivity and also received acceptance from the users.

### 4. EMERGING GUIDELINES FOR INTERACTIVE SHOES DESIGN FOR PD AND DISCUSSION

Identifying non-invasive treatments to alleviate the symptoms of PD is important to improving PD patients' life quality (45). Several studies exploring rhythmic auditory stimulation (RAS) and its modified interactive versions in PD rehabilitation showed improvements in gait cadence, stride length, and gait velocity (3). It is promising to use all above mentioned knowledge and technology to build an instrumented footwear system for the PD rehabilitation based on data collected from the instrumented footwear and feedback presentation through auditory and tactile channels. This kind of a system gives a lot of opportunities for a remote communication between a patient and a doctor or therapist. System interactivity allows for not only presenting feedback but also giving cues for patients based on their performance. The system could detect the events such as FOG, loss of balance and impaired gait patterns and subsequently present cues to correct it and help patients overcome these issues.

### 4.1. Hardware

Our literature review shows that from the hardware perspective pressure sensors, actuators and accelerometer need to be embedded into the instrumented footwear system. Pressure sensors will allow for step detection, and for monitoring balance, and pressure applied to selected parts of the feet. Important is the choice of the pressure sensors with adequate accuracy and durability. The aforementioned studies exhibit that it is possible to use a wide variety of pressure sensors and materials from which they are made. Moreover, they present high diversity in the number of data collecting points, ranging from 2 (35) to 75 (34). It is possible to find low-cost solutions in both categories. However, the number of the sensors embedded in a shoe sole depends on the available calculating power of each system. These data can be easily used as a basis for calculating velocity, stride length, cadence, and temporal variability.

The actuators present haptic feedback or cues to the users. Haptic feedback has three main advantages: it can be hidden in the shoe, can motivate users to perform a step by detecting FOG or gait initiation, and can increase the perceived naturalness of the auditory stimuli, which can serve as a higher motivation for rehabilitation. There are no sufficient studies to indicate the best placement of actuators in a shoe sole; based on the study by Watanabe and Ando (27), we believe that a heel is the best potential candidate to be stimulated by an actuator. Although Kennedy and Inglis (50) indicated that the ball and the arch are the most sensitive areas to vibrotactile stimulation.

The placement of an accelerometer is quite optional, but it should be hidden in the shoe and well protected from the displacement. Accelerometers collect data about feet acceleration in three-dimensional space. They give more precise information about feet movement than FSR sensors, and they are crucial in detecting balance problems. An accelerometer was successfully used for FOG detection in Bächlin et al. (49).

### 4.2. Software/Application

Few studies consider the use of ecological signals, despite their richness of information and acceptance from the users' perspective. Bächlin et al. (49) demonstrated the need for a context-aware system. The ideal system should be adaptive to the participant's speed (19, 20) and able to present cueing signals constantly or in *ad hoc* manner (49). The overall goal is to design a system, which patients would like to wear everyday and feel comfortable with it. For example, auditory cueing should be used by patients only during short sessions every day. The haptic stimulation could serve constantly, especially when a person would like to go out of their home. According to the patients' needs, it should be able to choose constant or *ad hoc* stimulation by choosing program on the main computer unit. Each system should be personalized and be programed for specific needs such as FOG, loss of balance, or slowed pace, to be mentioned among others.

### 5. SUMMARY

In this review, we summarized systems for gait rehabilitation based on instrumented footwear. We focused on the need of PD patients, but since only a few systems where made with this purpose, we went through several applications used in different scenarios when gait detection and rehabilitation is considered. Future designs could benefit from this knowledge. We outlined the hardware and software needs to run rehabilitation with the use of haptic and auditory cueing and feedback. There is still work to be done, but since technology for foot plantar measurement and feedback presentation is developing very fast, we should focus on specific applications and build customized systems for everyday use. The future trends outlined as well by Bächlin et al. (49) are: (1) miniaturization of the system and the main operating unit, which could be a part of the patients everyday clothing and hidden to make user feel comfortable; (2) specific calibration and customization of the rehabilitation programs based on patient-specific problems; (3) possibility of outdoor usage, so patients will be more secure and independent.

### AUTHOR CONTRIBUTIONS

JM is the most responsible person for this article. Since the type of article is mini review, she was responsible for literature search and selection and writing part. Both LK and SS were consulted at each step of article preparation, writing and corrections.

## REFERENCES


*3rd IEEE International Workshop on Robot and Human Communication, 1994. RO-MAN'94 Nagoya, Proceedings*. Nagoya: IEEE (1994). p. 118–23. doi:10.1109/ROMAN.1994.365945


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Maculewicz, Kofoed and Serafin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Effects of Physical Rehabilitation Integrated with Rhythmic Auditory Stimulation on Spatio-Temporal and Kinematic Parameters of Gait in Parkinson's Disease

*Massimiliano Pau1 \*† , Federica Corona1 , Roberta Pili2 , Carlo Casula3 , Fabrizio Sors4 , Tiziano Agostini4 , Giovanni Cossu3 , Marco Guicciardi5 and Mauro Murgia4,5†*

*1Department of Mechanical, Chemical and Materials Engineering, University of Cagliari, Cagliari, Italy, 2Department of Surgical Sciences, University of Cagliari, Cagliari, Italy, 3AOB "G. Brotzu" General Hospital, Cagliari, Italy, 4Department of Life Sciences, University of Trieste, Trieste, Italy, 5Department of Pedagogy, Psychology, Philosophy, University of Cagliari, Cagliari, Italy*

#### *Edited by:*

*Marta Bien´kiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Matthew Rodger, Queen's University Belfast, UK Ana Kleiner, Politecnico di Milano, Italy*

#### *\*Correspondence:*

*Massimiliano Pau massimiliano.pau@dimcm.unica.it*

*† Massimiliano Pau and Mauro Murgia contributed equally.*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 02 March 2016 Accepted: 27 July 2016 Published: 11 August 2016*

#### *Citation:*

*Pau M, Corona F, Pili R, Casula C, Sors F, Agostini T, Cossu G, Guicciardi M and Murgia M (2016) Effects of Physical Rehabilitation Integrated with Rhythmic Auditory Stimulation on Spatio-Temporal and Kinematic Parameters of Gait in Parkinson's Disease. Front. Neurol. 7:126. doi: 10.3389/fneur.2016.00126*

Movement rehabilitation by means of physical therapy represents an essential tool in the management of gait disturbances induced by Parkinson's disease (PD). In this context, the use of rhythmic auditory stimulation (RAS) has been proven useful in improving several spatio-temporal parameters, but concerning its effect on gait patterns, scarce information is available from a kinematic viewpoint. In this study, we used three-dimensional gait analysis based on optoelectronic stereophotogrammetry to investigate the effects of 5 weeks of supervised rehabilitation, which included gait training integrated with RAS on 26 individuals affected by PD (age 70.4 ± 11.1, Hoehn and Yahr 1–3). Gait kinematics was assessed before and at the end of the rehabilitation period and after a 3-month follow-up, using concise measures (Gait Profile Score and Gait Variable Score, GPS and GVS, respectively), which are able to describe the deviation from a physiologic gait pattern. The results confirm the effectiveness of gait training assisted by RAS in increasing speed and stride length, in regularizing cadence and correctly reweighting swing/stance phase duration. Moreover, an overall improvement of gait quality was observed, as demonstrated by the significant reduction of the GPS value, which was created mainly through significant decreases in the GVS score associated with the hip flexion–extension movement. Future research should focus on investigating kinematic details to better understand the mechanisms underlying gait disturbances in people with PD and the effects of RAS, with the aim of finding new or improving current rehabilitative treatments.

Keywords: rhythm, sound, RAS, Parkinson, gait, kinematics, spatio-temporal parameters, gait analysis

### INTRODUCTION

Parkinson's disease (PD) is a neurodegenerative disorder traditionally attributed to the progressive degeneration of dopaminergic neurons in the substantia nigra and, more recently, of other nondopaminergic systems of basal ganglia and of other regions of the central nervous system (1–3). Although PD patients report both motor and non-motor symptoms, the former (tremor, rigidity, bradykinesia, postural instability, and gait disturbance) have a huge impact on daily activities and

may severely reduce the patients' quality of life. In particular, the management of gait disorders, which are frequently encountered in PD, is of crucial importance because, as the disease progresses, they result in immobility (which causes loss of independence) and risk of falling (4).

Individuals with PD typically exhibit a gait pattern characterized by short stride length, increased cadence, and reduced velocity (5), which tends to further deteriorate with the progression of the disease (6). For this reason, pharmacological therapies are not sufficient to adequately deal with gait impairments and physical therapy is essential to cope with the deterioration in motor functions. Within the physical therapy domain, in the mid-1990s the efficacy of a therapy associated with rhythmic sounds, called Rhythmic Auditory Stimulation (RAS) (7), proved to be successful.

The rationale underpinning the effectiveness of RAS interventions lies in the origin of the gait disturbance in PD. The simultaneous activation and relaxation of many muscles in a coordinated way with very high temporal precision is necessary to perform a fluent gait. In healthy humans, this process is generally performed automatically. In PD patients, the cognitive mechanisms responsible for automatically processing the temporal coordination of movements – which typically involve basal ganglia – are somehow impaired (8, 9). Indeed, empirical evidence suggests that the "internal clock" that regulates both perceptual and motor processes is affected by PD (10, 11). As a consequence, patients affected by PD generally perform poorly in cognitive tasks involving temporal processing and in the execution of automatic cycling movements, such as walking. To cope with this impairment, interventions based on RAS provide patients with an auditory temporal guidance, which facilitates the regulation of their movements while walking (12).

In one of the first studies of RAS by Thaut and colleagues (7), the researchers randomly assigned patients to one of three conditions: RAS training, internally self-paced training and no training. Even though the analysis of spatio-temporal parameters revealed improvements in both training conditions, the patients assigned to the RAS condition had significantly better results in gait velocity, stride length, and step cadence compared to the other two conditions. In the subsequent years, researchers manipulated important parameters of the original training protocol [for recent reviews, see Ref. (9, 12, 13)]. For instance, some studies investigated the immediate effects of RAS in real-time imitation tasks [e.g., Ref. (14, 15)], while other studies manipulated the duration of the training program (i.e., number of weeks, number of sessions, duration of each session), the stimuli (i.e., tempo and type of sounds), and exercises [e.g., Ref. (8, 16–23)]. Overall, the majority of these studies confirmed the efficacy of rehabilitation accompanied by RAS, in particular in terms of spatio-temporal parameters of gait (18, 19, 22, 24–26).

It is noteworthy that the effects of RAS on gait patterns of people with PD were usually assessed by analyzing changes that occurred within spatio-temporal parameters, such as velocity, cadence, and stride length (9), while other important aspects, such as kinematic parameters (i.e., joint angular displacements at ankle, knee, hip, and pelvis districts) remained mostly unexplored. The only exception is represented by the study carried out by Picelli et al. (27) who investigated the effects of cued walking at different cadences on spatio-temporal and kinematic parameters of gait, finding that auditory cues are able to improve gait through modifications of motor strategies. The fact that kinematics has been rarely investigated is quite surprising, considering that previous studies recognized the importance of investigating the kinematic profiles of gait patterns in people with PD (28). In fact, this analysis allows the identification of a number of distinctive features (i.e., flat foot contact, reductions in the range of hip extension in mid-stance, knee flexion in swing, and plantarflexion at toe push-off) (28) which are crucial when the effects of neurosurgical, pharmacological, and rehabilitative treatments must be assessed (13).

The literature reports few attempts to investigate the effectiveness of rehabilitative treatments integrated with RAS through kinematic analysis of gait in other kinds of neurological diseases, such as stroke or cerebral palsy (29–31). In particular, two studies (30, 31) assess the overall deviation from a physiologic gait pattern from a kinematic point of view using the gait deviation index (GDI), a multivariate measure of overall gait pathology based on a set of features extracted from kinematic data (32). In both cases, RAS was found to have a beneficial effect on kinematic as well as on spatio-temporal gait patterns.

Thus, on the basis of the aforementioned considerations, this study aimed to assess the effect on gait patterns of 5 weeks of rehabilitative treatment that included gait training assisted by RAS. We hypothesized that a rehabilitative protocol integrated with RAS would improve not only the spatio-temporal parameters of gait, but also the kinematics. Moreover, to investigate the possible persistence of training effects, we performed a follow-up assessment 3 months after the end of the treatment.

### MATERIALS AND METHODS

### Participants

In the period from October 2014 to March 2015, 50 outpatients with PD admitted to the G. Brotzu General Hospital (Cagliari, Italy) for rehabilitation treatment were informed about the study. Assessment was carried out by a neurologist (Giovanni Cossu) experienced in PD, when patients were in "ON" state 60–90 min after intake of the usual morning l-DOPA dose. All screened patients met the PD UK Brain Bank criteria (33). The inclusion criteria for the study were as follows: ability to walk independently with no assistance; hearing capacity sufficient to perceive the auditory cues; absence of significant cognitive impairment (e.g., Mini-Mental Status Examination (MMSE) > 24; Frontal Assessment Battery (FAB) > 13); absence of psychiatric or severe systemic illnesses; mild-to-moderate disability assessed by means of the modified Hoehn and Yahr (H&Y) staging scale (1 ≤ H&Y ≤ 3). Patients were excluded if they were engaged in any training or rehabilitative program in the 3 months prior to the beginning of the study. At the time of enrollment, all participants were treated with l-DOPA and five of them were also taking dopamine agonists.

After the medical examination and an interview to establish the motivation level of potential participants, 31 individuals were included and scheduled for the treatment. The local ethics committee approved the study and all participants signed an informed consent form after a detailed explanation of the purposes of the study and of the methodology used for the experimental tests.

### Rehabilitation Protocol

Participants performed 5 weeks of supervised rehabilitative treatment (articulated in 2 × 45-min sessions/week) as outpatients at the hospital's Physical Medicine and Rehabilitation Department. Each of them was supervised by a physical medicine specialist (Carlo Casula) and individually assisted in the training by a certified physical therapist. The typical training session included a set of exercises aimed to enhance mobility, balance, and posture as well as specific gait training (see Appendix A for details). In particular, 20 min of each session were dedicated to continuous level walking, while participants equipped with a portable MP3 player and headphones listened to the auditory cues (RAS). During this period, participants were also instructed to perform at their homes (at least three times a week) a subset of the same exercises as used at the hospital, including 30 min of gait with RAS. Patients were provided with a diary in which they selfreported both the duration and type of activities performed at home. The diary was monitored by the physical therapists twice a week.

The RAS consisted of auditory beats whose pace (beats per minute – bpm) was personalized for each participant on the basis of the first gait assessment performed before the beginning of the study. The pace, that reflects on imposed gait cadence during the training, was set on the basis of the difference between the cadence of each patient and of healthy individuals of the same age range as reported in previous studies (34, 35). In particular, for participants whose cadence at the beginning of the study was:


At the end of the 5 weeks of supervised training, participants were instructed to perform, on a daily basis, the same exercises at home for the subsequent 12 weeks. They were invited to perform 30 min of exercises 5 days a week. This training was completely unsupervised. After this period, they were called to the laboratory for the follow-up assessment. In the follow-up, patients were interviewed by a physical medicine specialist and in general they confirmed their adherence to the training program during the unsupervised period.

### Measurement of Spatio-Temporal and Kinematic Gait Parameters

The acquisition of both spatio-temporal and kinematic gait parameters was performed at the Laboratory of Biomechanics and Industrial Ergonomics of the University of Cagliari (Italy) before the beginning of the study (T0), after its conclusion (+5 weeks, T5) and after 3 months follow-up (+17 weeks, T17) using an optoelectronic system composed of eight infrared Smart-D cameras (BTS Bioengineering, Italy) set at a frequency of 120 Hz. After anthropometric data collection, 22 spherical retroreflective passive markers (14 mm in diameter) were placed on the skin of the individual's lower limbs and trunk at specific landmarks, following the protocol described by Davis et al. (36). Participants were then asked to walk barefoot at a self-selected comfortable speed in the most natural manner possible on a 10-m walkway for at least six times, allowing suitable rest times between the trials. The raw data were then processed with the Smart Analyzer (BTS Bioengineering, Italy) dedicated software to calculate:


Kinematic data were summarized using the Gait Variable Score (GVS) and the Gait Profile Score (GPS). These concise measures of gait quality were recently proposed by Baker et al. (37) as a simplification of the GDI approach previously formulated by Schwartz and Rozumalski (32): in fact, using GPS instead of GDI has some advantages, such as the reduced set of parameters considered (9 vs. 15) and the fact that GPS can be decomposed into individual joint and plane scores (GVS). This approach was found effective in characterizing gait alterations in individuals with PD (38, 39) as well as in those affected by other neurological and non-neurological diseases, thus demonstrating general validity and a broad spectrum of applications (40–42). Specifically, the GVS represents the root mean square (RMS) difference between the tested subject's curve for a certain movement (e.g., knee flexion–extension) and a reference curve calculated as the mean value of tests performed on the unaffected subjects. The GPS combines the nine GVS values in a single score, which indicates the degree of deviation from a hypothetical "normal" gait (i.e., the larger the GPS, the less physiological the gait pattern); values for healthy individuals lie in the range of 5–6° (41). In the case of the present study, the reference data were obtained from a database of healthy individuals, of the same age range of the subjects here tested, available from the Smart Analyzer software.

### Statistical Analysis

Spatio-temporal and kinematic variables of gait were assessed before treatment with RAS (T0), at the end of treatment, i.e., 5 weeks after the baseline (T5), and 3 months after the end of treatment, i.e., 17 weeks after the baseline (T17). When different measures were available for the left and right limbs, a preliminary *t*-test was carried out to assess possible differences between them. Given that no significant differences were found for any of the investigated parameters, the mean value calculated across the two limbs was considered representative of each participant and was used for the subsequent analyses.

The independent variable was time (T0, T5, T17) and the dependent variables were the nine GVS scores plus the GPS index, the dynamic ROM of hip, knee and ankle joints in the sagittal plane, and the seven spatio-temporal parameters previously listed. To evaluate possible differences in the dependent variables across time, a set of repeated-measures analyses of variance (RM ANOVAs) was applied. When the sphericity assumption (calculated with the Mauchly's test) was violated, data were corrected with the Greenhouse–Geisser formula. When the normality distribution assumption (calculated with the Shapiro–Wilk's test) was violated, Friedman's test instead of the RM ANOVA was used. When the omnibus values of RM ANOVAs and Friedman's tests were significant, the contrasts using the paired-samples *t*-test and the Wilcoxon's test, respectively, were calculated. The alpha level was set at 0.05 for the omnibus tests and was adjusted with the Bonferroni formula for the contrasts (0.05/3 comparisons = 0.017). The analyses were performed using the IBM SPSS Statistics v.20 software (IBM, Armonk, NY, USA).

### RESULTS

Of the 31 patients who entered the study, 26 (20 males, 6 females) completed the training program and underwent the three gait assessments. Five participants were forced to leave the study due to musculoskeletal injuries (not related to the rehabilitative program, four cases) or chemotherapy (one case). The main anthropometric and clinical features of the 26 participants are given in **Table 1**.

The effects of the physical therapy with RAS across time are separately reported for spatio-temporal and kinematic variables.

### Spatio-Temporal Parameters

The spatio-temporal parameters calculated for the three experimental conditions are shown in **Table 2**, while **Table 3** provides the details of the cadence values for each participant at the


baseline and after the rehabilitative treatment. **Figure 1** shows the values of the spatio-temporal parameters of the participants compared with those calculated for an age- and gender-matched group of healthy individuals tested in the same laboratory.

#### Table 2 | Comparison between spatio-temporal parameters assessed before and after rehabilitation.


*Values are expressed as mean* ± *SD.*

*T0, baseline; T5, after 5 weeks of supervised rehabilitation; T17, 3-months' follow-up. a denotes statistical significance with respect to baseline.*

*bdenotes statistical significance with respect to T5.*


#### Table 3 | Cadence values for each participant before and after rehabilitative treatment.

All parameters revealed a significant omnibus value for time: gait speed [*F*(2, 50) = 8.402; *p* < 0.001; η<sup>2</sup> = 0.252]; cadence [χ<sup>2</sup> (2) = 7.462; *p* < 0.05; *W* = 0.143], step width [χ<sup>2</sup> (2) = 20.356; *p* < 0.001; *W* = 0.391], step length [*F*(2, 50) = 20.775; *p* < 0.001; η<sup>2</sup> = 0.454], and percentage of swing phase [*F*(2, 50) = 6.171; *p* < 0.005; η<sup>2</sup> = 0.198], double support phase [*F*(2, 50) = 7.370; *p*< 0.005; η<sup>2</sup>= 0.228], and stance phase [χ<sup>2</sup> (2) = 13.000; *p*< 0.005; *W* = 0.250]. Thus, we were able to calculate the contrasts for all variables.

The contrasts revealed augmented gait speed [*t*(25) = 2.839; *p*< 0.005; *d*= 0.433], cadence (*Z*= 2.845; *p*< 0.005; *r*= 0.394), and percentage of swing phase [*t*(25) = 3.172; *p* < 0.005; *d* = 0.628], and reduced percentage of double support phase [*t*(25) = 3.206; *p* < 0.005; *d* = 0.604] between T0 and T5. Moreover, the data showed that these improvements were kept constant between T5 and T17, and in T17 were still significantly different from T0: gait speed [*t*(25) = 3.580; *p* < 0.001; *d* = 0.591], cadence (*Z* = 2.222; *p* < 0.05; *r* = 0.308), percentage of swing phase [*t*(25) = 2.505; *p* < 0.01; *d* = 0.483], and percentage of double support phase [*t*(25) = 2.967; *p* < 0.005; *d* = 0.618].

As for the percentage of the stance phase, we found a pattern of results similar to that of previous analyses, with significant improvements between T0 and T5 (*Z* = 2.502; *p* < 0.01; *r* = 0.347) and no difference between T5 and T17. However, in this case, the

Figure 1 | Spatial–temporal parameters of the participants with PD. Gray lines indicate the reference values calculated for a sample of healthy individuals. Error bars indicate SD.

difference between T0 and T17 (*Z*= 2.007; *p*< 0.05; *r*= 0.278) was no longer significant after the Bonferroni correction. Conversely, we found that step length significantly improved between T0 and T5 [*t*(25) = 3.423; *p* < 0.001; *d* = 0.573], and had still improved in T17, both compared to T0 [*t*(25) = 6.634; *p* < 0.001; *d* = 0.871] and T5 [*t*(25) = 2.805; *p*< 0.005; *d*= 0.333]. Finally, we also found higher values in T17 compared to both T0 (*Z* = 3.672; *p* < 0.001; *r* = 0.509) and T5 (*Z* = 3.213; *p* < 0.001; *r* = 0.445) for step width. However, in this case, the difference between T0 and T5 did not reach a significant value (*p* < 0.09).

### Kinematic Parameters

Kinematic changes due to physical therapy with RAS were evaluated through the GPS, GVS, and dynamic ROM values.

### GPS and GVS Values

Higher GVS values indicate a large deviation from physiologic conditions for a specific movement of the nine previously listed; the GPS combines all the nine GVSs in a single value to summarize with a single value the overall quality of the gait pattern. The GPS and GVS scores calculated for the three experimental conditions are shown in **Table 4**, while **Figure 2** shows the GVS calculated in the sagittal plane for hip, knee, and ankle joints and the GPS values compared with those calculated for healthy individuals of the same age range.

A significant omnibus value was found for the GPS [χ<sup>2</sup> (2) = 8.615; *p* < 0.05; *W* = 0.166] and GVS of hip flexion– extension [χ<sup>2</sup> (2) = 10.272; *p* < 0.01; *W* = 0.198] and ankle dorsi–plantarflexion [*F*(2, 50) = 4.759; *p* < 0.05; η<sup>2</sup> = 0.160]. The contrasts revealed lower GPS scores in T17 compared to both T0 (*Z* = 2.109; *p* < 0.05; *r* = 0.292) and T5 (*Z* = 2.502; *p* < 0.01; *r* = 0.347). Similarly, the GVS of hip flexion–extension was lower in T17 compared to both T0 (*Z* = 3.565; *p* < 0.001; *r* = 0.494) and T5 (*Z* = 2.299; *p* < 0.05; *r* = 0.330). In this case, the difference

Table 4 | Comparison between kinematic parameters of gait assessed before and after rehabilitation.


*Values are expressed as mean* ± *SD.*

*T0, baseline; T5, after 5 weeks of supervised rehabilitation; T17, 3-months' follow-up. a denotes statistical significance with respect to baseline.*

*bdenotes statistical significance with respect to T5.*

between T0 and T5 appeared to be significant (*Z*= 1.740; *p*< 0.05; *r* = 0.241), but this value was no longer significant after the Bonferroni correction. Finally, it was found that the GVS of ankle dorsi–plantarflexion in T17 was higher than in T0 [*t*(25) = 2.726; *p* < 0.05; *d* = 0.746].

### Dynamic ROM

Differently from GVSs, higher ROM values indicate a better functionality of a certain articular joint. The ROM calculated for the three experimental conditions are shown in **Table 5**. The omnibus analyses revealed significant values for ROM of knee flexion–extension [χ<sup>2</sup> (2) = 13.000; *p* < 0.005; *W* = 0.250] and ROM of hip flexion–extension [*F*(1.2, 30.2) = 20.058; *p* < 0.001; η2= 0.445]. The values for ROM of knee flexion–extension significantly improved from T0 to T5 (*Z* = 3.048; *p* < 0.001; *r* = 0.423), remained stable between T5 and T17, and were still significant at T17 compared to T0 (*Z* = 3.213; *p* < 0.001; *r* = 0.446). The hip flexion–extension ROM significantly improved from T0 to T5 [*t*(25) = 3.943; *p* < 0.001; *d* = 0.498]. In T17, the values were significantly higher compared to both T0 [*t*(25) = 5.209; *p* < 0.001; *d* = 0.620] and T5 [*t*(25) = 2.622; *p* < 0.01; *d* = 0.132].

### DISCUSSION

The main goal of the present study was to assess the effectiveness of 5 weeks of rehabilitative treatment that included gait training assisted by RAS. The major novelty of the research is represented by the use of state-of-the-art technologies for quantitative human movement analysis to verify possible changes introduced by the treatment in the gait patterns of tested participants, especially in terms of kinematics. We also aimed to verify, after a 3-month follow-up, whether the positive effects of the training were maintained or not.

Our results confirm previous reports as regards the positive effects of RAS on spatio-temporal parameters of gait, whose results all (except step width) significantly improved at the end of the supervised treatment. In particular, in four cases out of seven (i.e., step length, gait speed, cadence, and double support phase duration), such changes were maintained at the 3-month follow-up. It is also noteworthy that the increase observed for gait speed (0.14 m/s) can be considered a large clinically meaningful effect (43). Moreover, the results obtained here show that the training resulted in a recovery of functionality characterized by post-rehabilitation/follow-up values similar to those calculated in previous studies for healthy individuals of the same age range (34, 35, 44–48) as shown for the cases of speed, step length, and cadence in **Figure 1**.

By contrast, it was quite surprising to observe a significant increase in the width of the base of support as a sort of "side effect" of the treatment; in fact, higher values of this parameter are usually associated with reduced stability and fear of falling (49, 50). We hypothesized that such apparently negative effects are actually due to increased speed, meaning that the individuals appeared to adapt their gait strategy to the new speed they were able to achieve by enlarging the base of support, as they felt more confident. This phenomenon was previously observed by Helbostad and Moe-Nilssen (51), who reported the existence of

Table 5 | Comparison between dynamic ROM assessed before and after rehabilitation.


*Values are expressed as mean* ± *SD.*

*T0, baseline; T5, after 5 weeks of supervised rehabilitation; T17: 3-months' follow-up. a denotes statistical significance with respect to baseline.*

*bdenotes statistical significance with respect to T5.*

a "u-shaped" relationship between gait speed and step width in elderly subjects. However, this issue should be further investigated with specific tests on gait at different speeds in individuals with PD, to verify whether the same kind of trend remains in presence of the pathology.

As regards gait kinematics, the significant decrease in the GPS value (7.57° at the follow-up vs. 8.34° of baseline) indicates that after the training the kinematics of the gait pattern appeared closer to the physiological condition. As previously mentioned, there are indeed few studies that have investigated the effects of rehabilitative treatments on gait kinematics in individuals with PD (52–54) and none of them employ RAS as a tool to support gait training; thus, it is difficult to find data for comparisons. Moreover, the only two existing attempts to characterize the effects of RAS on kinematic patterns (30, 31) involved diseases different from PD (i.e., stroke and cerebral palsy). However, it is noteworthy that in both cases a significant reduction in the overall index of gait quality and, thus, a general improvement of the gait pattern was found, similar to what was observed in the present study.

Examining the data of the present study in detail, it is interesting to observe that the major contribution to the improvement in the kinematic pattern of gait was essentially originated by a marked reduction in the hip flexion–extension GVS value and, to a lesser extent, in the knee flexion extension, as shown in **Figure 2**. In particular, the comparison between the hip flexion–extension angle during the gait cycle at the baseline and at the T17 follow-up (**Figure 3**) shows that the regularization of this movement is associated with a generalized decrease in flexion at heel contact and at the end of the swing phase, and with a correspondent increased extension at terminal stance.

Abnormal hip joint movements are quite common in neurological disorders as a compensation strategy for the lack of movement of the ankle joint (12), thus, it is likely that the positive effect of training integrated with RAS on the whole lower limb kinematic chain acts to recover a more physiological synergy between hip, knee, and ankle joint action. Moreover, our participants' GVS score, associated with hip flexion–extension at baseline, was the largest in comparison with normality and, thus, it is likely that such a movement underwent more beneficial effects with respect to other joints which at the baseline resulted less impaired.

The large improvements originated in hip kinematics by the rehabilitative treatment probably represent one of the main factors responsible for the changes observed in the spatio-temporal parameters, particularly as regards step length and gait speed. In fact, previous studies observed that increases in hip ROM consequent to physical training are associated with increased step length (55) and that reduced peak hip extension is associated with a range of gait alterations, including reduced step length in both healthy subjects (45) and individuals affected by neurological diseases (56).

It was also surprising to observe that the GVS of ankle dorsi– plantarflexion was almost normal at the baseline, slightly (but not significantly) increased after rehabilitation, but significantly worse at the follow-up assessment, thus indicating a relevant deviation from normality. A possible explanation of this phenomenon can be found in the way that gait training is administered by physical therapists; to reduce the impact of existing (or future) shuffling gait and its negative consequences (i.e., slips, trips, and falls), the patient is stimulated to accentuate dorsiflexion at heel contact and the plantarflexion at toe off phase, thus making the movement a bit more "unnatural" as a preventive measure.

Other signs of improved gait kinematics come from the analyses of the dynamic ROM, which show that hip and knee ROM in the sagittal plane significantly increase after the training, while at ankle level no relevant changes were observed. These results are partly consistent with those of Kim et al. (30) who detected an increase in hip ROM of 6.4° after 9 sessions of gait training assisted by RAS (in our case 4.4° at the follow-up) and no significant changes as regards the ankle (similar to what was found in the present study). By contrast, after rehabilitation we found increases of ROM of the knee joint similar to those found by Kim et al. (2.1° vs. 3.2° in our study); however, they failed to achieve statistical significance, probably due to the limited size of their sample, which was composed of only 13 participants.

From a broader perspective, the present study further supports the efficacy of rehabilitation accompanied by RAS as a strategy to improve the gait parameters of Parkinson patients, thus confirming what was previously found in the literature [see Ref. (9, 12, 13)]. The major innovation of our study is that for the first time we report the effects of RAS not only on spatiotemporal parameters but also on gait kinematic variables. The original data reported herein are particularly important in gaining a better understanding of how the mechanics of gait are affected by auditory cues in this particular category of patients. Like every empirical work, the present study certainly has some limitations. The most important of these is the absence of a control group. Owing to the limited number of patients available in the hospital and to lack of space, we were able to test only one group of patients before and after the treatment. This prevents us from generalizing the results of the study and limits the possibility to assess whether the proposed treatment, which includes the gait training integrated by RAS, is superior in comparison with other kinds of rehabilitative approaches. Thus, even though we reported original information on patients' gait kinematics, there is a need for future studies specifically focused on investigating the effects of RAS on gait kinematics in a randomized controlled trial.

Finally, it is noteworthy that although optoelectronic systems represent the most sophisticated option available for human movement analysis, they are expensive, require a dedicated laboratory (i.e., the whole equipment is not easily portable), and data acquisition and processing is time-consuming and can be performed only by specialized personnel. Future studies should consider other emerging techniques, such as wearable inertial sensors, which have been already successfully employed to characterize spatio-temporal parameters of gait in individuals with PD in clinical settings (57) but that can also be used to obtain data regarding the whole kinematic pattern (58).

### CONCLUSION

The overall analysis of gait patterns in individuals with PD before and after rehabilitation integrated with RAS – carried out taking into account not only the spatio-temporal parameters of gait but also the kinematic trends of lower limb joints – supplied new evidence about the effectiveness of such an approach. In particular, this technique appears capable of restoring several gait aspects to acceptable levels, thus making the ambulation function very similar to that of healthy individuals of the same age. In fact, it not only regularizes the cadence but also acts to increase speed and step length and creates a more physiological subdivision between stance and swing phase. As a plus, it is now recognized that this tool can influence gait kinematics, as the overall quality of gait pattern results significantly increased. However, from this point of view, some aspects still remain unclear. In fact, the positive kinematic effects of the training integrated with RAS appears

### REFERENCES


basically restricted to the hip joint and are not always immediately visible after the end of the supervised treatment, but they rather tend to become evident after a longer period during which participants performed home-based gait training on a daily basis. The program also demonstrated limited effectiveness on knee joint functionality, which improved only in terms of dynamic ROM and, as regards the ankle, a slight (though significant) worsening of functionality was detected. As it appears quite consolidated that RAS has a positive effect on spatio-temporal aspects of gait, research should now focus on investigating kinematics in greater detail (but also kinetics and EMG variables that were also analyzed in a few studies) to better understand the mechanisms underlying gait disturbances in people with PD and, thus, establish new or improved rehabilitative treatments.

### AUTHOR CONTRIBUTIONS

MP, MM, MG, CC, RP, FS, and TA designed the study; CC and RP performed the physical capacity assessment and prepared the rehabilitation protocol; GC performed the neurological evaluations; MP and FC collected and processed the data; MM and MG performed the statistical analyses; MP and MM wrote the manuscript; and TA, FS, GC, CC, and RP revised the manuscript.

### ACKNOWLEDGMENTS

This study was partly supported by the Autonomous Region of Sardinia (grant CRP-78543 L.R. 7/2007). The author Mauro Murgia was supported by the Autonomous Region of Sardinia, Master&Back Programme 2013 (PRR-MAB-A2013-19330). The authors wish to express their gratitude to the physical therapists Marilena Fara, Giovanna Ghiani, Alessandra Pani, Elsa Sau, Gino Sedda, and Mauro Usala for their valuable support during the training program.


comparison of two physical therapy protocols. *Mov Disord* (2000) 15(5):879– 83. doi:10.1002/1531-8257(200009)15:5<879:AID-MDS1018>3.0.CO;2-9


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Pau, Corona, Pili, Casula, Sors, Agostini, Cossu, Guicciardi and Murgia. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## APPENDIX

## A Rehabilitation Protocol


# **Dynamic high-cadence cycling improves motor symptoms in Parkinson's disease**

*Angela L. Ridgel <sup>1</sup> \*, Robert S. Phillips <sup>1</sup> , Benjamin L. Walter 2,3 , Fred M. Discenzo<sup>4</sup> and Kenneth A. Loparo<sup>5</sup>*

*<sup>1</sup> Department of Exercise Physiology, Kent State University, Kent, OH, USA, <sup>2</sup> Movement Disorders Center, University Hospital Cleveland, Cleveland, OH, USA, <sup>3</sup> Department of Neurology, Case Western Reserve University, Cleveland, OH, USA, <sup>4</sup> Department of Industrial Automation, Rockwell Automation, Mayfield Heights, OH, USA, <sup>5</sup> Department of Electrical Engineering and Computer Science, Case Western Reserve University, Cleveland, OH, USA*

**Rationale:** Individuals with Parkinson's disease (PD) often have deficits in kinesthesia. There is a need for rehabilitation interventions that improve these kinesthetic deficits. Forced (tandem) cycling at a high cadence improves motor function. However, tandem cycling is difficult to implement in a rehabilitation setting.

**Objective:** To construct an instrumented, motored cycle and to examine if high cadence dynamic cycling promotes improvements in motor function.

### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

### *Reviewed by:*

*Elizabeth B. Torres, Rutgers University, USA William Richard Young, Brunel University, UK*

*\*Correspondence:*

*Angela L. Ridgel, Department of Exercise Physiology, Kent State University, 350 Midway Dr, Kent, OH 44242, USA aridgel@kent.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

> *Received: 22 July 2015 Accepted: 20 August 2015 Published: 02 September 2015*

#### *Citation:*

*Ridgel AL, Phillips RS, Walter BL, Discenzo FM and Loparo KA (2015) Dynamic high-cadence cycling improves motor symptoms in Parkinson's disease. Front. Neurol. 6:194. doi: 10.3389/fneur.2015.00194* **Method:** This motored cycle had two different modes: dynamic and static cycling. In dynamic mode, the motor maintained 75–85 rpm. In static mode, the rider determined the pedaling cadence. UPDRS Motor III and Timed Up and Go (TUG) were used to assess changes in motor function after three cycling sessions.

**Results:** Individuals in the static group showed a lower cadence but a higher power, torque and heart rate than the dynamic group. UPDRS score showed a significant 13.9% improvement in the dynamic group and only a 0.9% improvement in the static group. There was also a 16.5% improvement in TUG time in the dynamic group but only an 8% improvement in the static group.

**Conclusion:** These findings show that dynamic cycling can improve PD motor function and that activation of proprioceptors with a high cadence but variable pattern may be important for motor improvements in PD.

**Keywords: movement disorders, exercise, rehabilitation, neuroplasticity, bradykinesia, motor function**

## **Introduction**

Approximately 630,000 people in the US were diagnosed with Parkinson's disease (PD) in 2010 and it is estimated that PD prevalence will double by 2040 (1). As PD progresses, the combined motor and non-motor symptoms often lead to decreased independence and quality of life. The economic impact of PD, including treatment, social security payments, and lost income from inability to work, exceeded \$14.4 billion in 2010 (1). The degenerative nature of PD results in progressive deterioration of motor skills along with reduced sensory and cognitive function. The current treatment for PD is medication (levodopa, dopamine agonists) and surgical intervention (deep brain stimulation). These treatments only partially treat the symptoms and do not slow progression of the disease. Furthermore, they often have undesirable side effects, such as dyskinesia (2). In light of projections of increased prevalence of PD, there is a need for innovative new treatments to improve symptoms and delay disease progression.

Our studies, and those of several others, have presented strong evidence that certain exercise interventions promote changes in neural drive in PD (3–9). Although the exact mechanisms are still unknown, it has been suggested that increases in sensory input or feedback resulting from these interventions may play a role in this motor improvement. Individuals with PD often have deficits in kinesthesia (conscious awareness of limb and body position in space) (10, 11). Sensorimotor integration may be dysfunctional in PD and has been implicated in the etiology for bradykinesia and atypical movement in PD. Kinesthesia is likely to be the key modality affecting this dysfunction (12). However, levodopa does not appear to improve kinesthetic deficits in PD (13, 14) and has been associated with suppression of sensitivity to joint position (15). Therefore, there is a great need for rehabilitation interventions that improve proprioceptive deficits in PD.

Animal model studies have shown that high-intensity exercise can promote neural plasticity and neuroprotection against dopaminergic cell loss (16). Several reports in humans have shown that high-intensity treadmill training (17, 18) and high-cadence cycling (3, 6, 9, 19, 20) promote functional improvement in PD but there are still several unanswered questions: (1) How does motor function change immediately after high-intensity exercise, (2) What features of exercise (speed, intensity) optimize motor function, and (3) What are potential mechanisms of function improvements after high-intensity exercise?

To begin to address these questions, we have developed a novel rehabilitation approach called dynamic cycling. This work builds upon our original "forced exercise" paradigm that used a stationary tandem bicycle and an able-bodied trainer to assist individuals to pedal with a rapid cadence (80 rpm) (6). Highcadence tandem cycling resulted in a 35% reduction in PD motor symptoms (UPDRS scores), whereas individuals who cycled at a self-selected cadence (60 rpm) showed no improvement. Despite these remarkable results, large-scale use of the tandem cycling paradigm is not feasible in a rehabilitation or home setting. Furthermore, it has proven difficult to reproduce the dynamics of tandem cycling using currently available motorized cycles. The dynamic cycling paradigm that we developed uses a motorized stationary cycle to assist individuals with PD to pedal at a cadence faster than they can (or would) pedal on their own. In addition, this rehabilitation paradigm is unique because the motor rotates the pedals at a high speed with a slight, but prescribed, variation. These dynamic changes in cadence appear to be an important component of tandem cycling (21). Therefore, we hypothesize that dynamic cycling will promote greater improvements in PD symptoms and motor function than cycling at a lower cadence (static cycling). Findings from this study will provide important data to support future research examining long-term rehabilitation benefits as well as role of afferent input during dynamic cycling in the reduction of PD motor symptoms.

## **Materials and Methods**

### **Participants**

Inclusion criteria were as follows: 50–79 years of age, diagnosis of idiopathic PD, and no contraindications to exercise, including uncontrolled cardiovascular disease or stroke. Exclusion criteria included history of heart attack, any surgical procedure for treatment of PD, including deep brain stimulation, pallidotomy, or thalamotomy. All potential study subjects were pre-screened over the telephone with the American Heart Association/American College of Sports Medicine exercise pre-participation questionnaire (22). Individuals with greater than or equal to two risk factors for coronary artery disease (moderate risk) were required to obtain physician clearance prior to exercise. Fifty individuals with idiopathic PD qualified and agreed to participate in this study. This study was carried out in accordance with the recommendations of the Kent State University Institutional Review Board with written informed consent from all subjects.

### **Study Design**

This study was a randomized two group pretest–posttest design. Each participant visited the lab for four sessions. Individuals were randomized into either: (1) dynamic cycling or (2) static cycling. During the first session (Friday), baseline motor function was assessed and individuals completed the first cycling session. During the next two sessions (Monday/Wednesday), each participant exercised for 40 min on the instrumented bike. During the last session (Friday), post-intervention motor function was assessed. There was at least 48 h between the last exercise session and the post-intervention assessments. Each cycling bout began with 5 min of warm up (low resistance pedaling at 40–50 rpm). Participants then completed 30 min of dynamic or static cycling and ended with 5 min of cycling at 40–50 rpm. Rating of perceived exertion (RPE) and heart rate (HR) was monitored by a research assistant during each session. Participants were encouraged to maintain their HR within 50–80% of their estimated HR reserve. Data from the cycle were collected continuously during each session. All exercise sessions were completed while individuals were "on" anti-Parkinson's medications. Participants served as their own controls from pre-cycling to post-cycling testing to account for performance variability that is often present in PD.

### **Intervention**

During dynamic cycling, motor output speed varied between 75 and 85 rpm. Motor torque was adjusted to accommodate changes in the rider force exerted on the pedals. The motor did the majority of the work to turn the pedals but individuals were encouraged to push on the pedals and to not be passive. During static cycling, individuals cycled on the instrumented bike, at a self-selected speed, without the motor assist. Speed was not controlled but the rider experienced an inertia load on the pedals, similar to what they would experience on a typical stationary bike. Individuals were directed to choose their own pedaling speed. HR was collected with a Polar Wearlink<sup>+</sup>™ Coded Transmitter worn on the chest, which transmitted to a HR monitor interface board. The control platform was a commercially available programmable logic controller (PLC). The PLC determined the appropriate motor speed and load (torque) values, and sent motor control information to the motor drive. The motor drive implemented a high-speed inner loop controller that provided the appropriate voltage and current to the motor. Motor feedback was used as feedback for the drive to maintain motor speed and torque. Data from the controller box were downloaded and archived onto a laptop computer. Additional details on the design of the control program and the cycle can be found in the paper by Mohammadi Abdar (23).

### **Outcome Measures**

All assessments were completed while the individuals were "on" anti-Parkinson's medication. The primary outcome measure was the UPDRS Part III Motor Exam. UPDRS Motor III was administered by a blinded movement disorders specialist prior (preintervention) to the three cycling sessions and 2 days following the last exercise session (post-intervention). The UPDRS Motor III has universal acceptance as a rating scale for PD patients and it has been shown to be reliable and valid (24, 25). The total score, scores for each primary symptom (i.e., tremor, bradykinesia) and scores for upper and lower extremity were analyzed. The secondary outcome measure was the Timed Up and Go (TUG). This test is used primarily as a measure of mobility but is also useful as a measure of bradykinesia during walking (26). To complete the TUG, participants were asked to stand up from a standard chair and walk a distance of 3 m, turn around and walk back to the chair and sit down again. The time to complete the task was recorded with a stop watch. Each participant performed three trials and the average was calculated.

### **Statistics**

Demographic variables between the two groups were compared using an independent samples *t*-test. Comparison of pre-cycling and post-cycling changes in UPDRS motor scores and TUG time were performed using paired-samples *t*-test in each group (dynamic, static) independently. All statistical analysis was completed using SPSS V. 22 and the alpha level was set to 0.05.

### **Results**

A detailed description of the design and controller parameters for the motorized cycle was described in a previous paper (23). In summary, the bike chassis used for this study consisted of a commercial exercise bike frame (Motomed Viva 2, Reck, Germany) that was modified to include additional sensors to monitor bike operation and rider condition (cadence, torque, power, HR). In addition, a high performance servomotor and variable speed drive were coupled to the pedals and a programmable controller with custom control algorithms, data acquisition, network capability, real-time display with operator controls, and data archiving were provided (**Figure 1A**). The electronic components integrated on the bike chassis were the operator display, emergency stop button, HR monitor interface board, TTL to serial level converter board for the HR monitor. All other electronic components, such as the drive, programmable controller, network adapter, and power supplies, were mounted in an enclosure that was external to the bike but connected via cables for motor power, display power, motor feedback, and communications (**Figure 1B**).

and the research assistants. Fifty individuals were randomized to either the dynamic or

static cycling group (**Figure 2**). Seven females and 16 males with mean age 67.3 *±* 0.9 years completed three 40-min static cycling sessions and 11 females and 13 males with mean age 67.2 *±* 1.6 years completed three dynamic cycling sessions (**Figure 2**). Two individuals from the static group and one person from the dynamic did not complete the intervention due to the reasons outlined in **Figure 2**. There were no significant differences

group. Two individuals from the static group did not complete the intervention due to DBS (not reported in prescreening) and a diagnosis of a non-Parkinson's movement disorder (PSP). One person from the dynamic did not complete the protocol due to hip pain during the cycling. Data from 23 individuals in the static group and 24 individuals in the dynamic group were analyzed.



#### **TABLE 2 | Cycling and physiological variables**.


*rpm, revolutions per minute; bpm, beats per minute; RPE, rating of perceived exertion. Mean ± SD, independent t-test was used to compare the two groups.*

*Mean ± SD, independent t-test was used to compare the two groups.*

in any of the demographic variables (age, H&Y, height, weight, body mass index, disease duration, and levodopa equivalent dose) between the dynamic and static cycling groups (**Table 1**).

Dynamic and static cycling modes resulted in similar individual assessments of RPE but there were significant differences in cadence, power, torque, and HR between the two groups (**Table 2**). Specifically, individuals in the dynamic cycling group showed a higher cadence (78.6 *±* 1.1 versus 66.0 *±* 3.2 rpm, *p <* 0.001) but a lower power, torque, and HR than the static cycling group.

The overall UPDRS III score (**Figure 3A**) showed a significant 13.9% (4.0 pts) improvement in the dynamic group (*t* = 2.676, df = 23, *p* = 0.013) and only a small 0.9% (0.2 pts) change in the static group (*t* = 0.189, df = 22, *p* = 0.85) after just three cycling sessions. Analysis of the individual UPDRS III components showed that lower extremity (*t* = 3.8, df = 23, *p* = 0.001) and rigidity scores (*t* = 2.6, df = 23, *p* = 0.013) also improved significantly in the dynamic group but there were no significant changes in the static group in any of the UPDRS Motor III scores. Interestingly, UPDRS scores in the upper extremity (**Figure 3B**) showed a significant 18% (2.6 pts) improvement after dynamic cycling (*t* = 2.54, df = 23, *p* = 0.018) compared with a 7% (0.9 pts) improvement in the static group (*t* = 1.32, df = 22, *p* = 0.19).

In addition, there was a 16.5% improvement (2.1 s, *t* = 1.7, df = 23, *p* = 0.10) in Timed Up and Go test (TUG) time in the dynamic group but only an 8% improvement (0.87 s, *t* = 1.3, df = 21, *p* = 0.19) in the static group (**Figure 3C**). Although this change was not statistically significant due to variability in responses among individuals, it is interesting that the dynamic group showed a twofold improvement in the TUG compared to the static group.

Although the baseline mean scores of UPDRS between the two groups were not the same, the majority of individuals in the dynamic group (15/24, 62%, **Figure 4A**) showed improvements

in motor symptoms and a much smaller percentage of the static group showed a positive change (9/23, 39%, **Figure 4B**).

### **Discussion**

Dynamic high-cadence cycling for only three sessions resulted in a 4-point reduction in motor symptoms of PD as measured with the UPDRS Motor III test. These findings were similar to a 1 month high-intensity treadmill training program (body weight supported treadmill training, BWSTT) which resulted in a 3-point improvement in UPDRS Motor III (18) and an 8-week highintensity BWSTT intervention which showed a 2.8-point improvement. In addition, an intensive 4-week LSVT® BIG intervention reported a 5-point change in UPDRS Motor III (27). By contrast,

our UPDRS improvements were less than that reported after a single session of forced cycling (as evaluated "off " medication) (3). However, a 4-point change is well within the minimum clinically important difference (CID) of 2.3–2.7 points (28), suggesting that this difference is recognized and valuable to individuals with PD. UPDRS Motor III scores were evaluated while individuals were "on" medication in this study (see Discussion below), so we expect that future studies with a longer intervention and with "off " medication evaluations would yield even greater improvements. The significant changes seen in upper extremity UPDRS scores after dynamic cycling (2.6 points) are in agreement with previous high-cadence cycling papers (3, 6). The enhanced function of the upper extremity with a lower extremity intervention further supports the view that dynamic cycling could promotes changes in neural drive in PD.

Timed up and go is a widely used measure of mobility, balance, and fall risk in PD. Time to completion in the TUG test improved by 2.1 s after dynamic cycling but variability among subjects resulted in a non-significant difference. The minimal detectable change (MDC), difference in scores which reflect true change and not error, for PD is reported as 3.5 s (29). However, Latt and colleagues (30) reported that individuals who complete the TUG in *≥*12 s have an increased fall risk. Individuals in the dynamic group had a baseline TUG of 12.7 *±* 11.6 s that was decreased to 10.5 *±* 7.1 s after only three dynamic cycling sessions. Furthermore, it is likely that a longer-term intervention would promote greater improvements in balance and mobility.

### **Dynamic Cycling Versus High-Intensity Exercise**

Dynamic cycling promoted similar improvements in motor function to that reported in other high-intensity (or vigorous) exercise interventions (5, 18, 27), but the HR (91.1 *±* 2.5 bpm) and ratings of perceived exertion (12.7 *±* 1.1) values recorded during dynamic cycling sessions are defined as light to moderate intensity (22, 31). During dynamic cycling, the motor does the majority of the work to turn the pedals and less effort is required by the individual. The significant decrease in power in the dynamic group (31.2 *±* 4.1), compared to the static group (8.0 *±* 4.3), reflects this effort. The value of this rehabilitation paradigm is that it promotes significant improvement in PD symptoms with a reduced the risk of injury, excessive fatigue, and non-compliance. All of the individuals in the dynamic group were able to successfully complete three sessions without any unusual fatigue or injuries.

### **Possible Mechanisms**

The improvement in Parkinson's motor symptoms in the dynamic group was intriguing because individuals in the static group were working harder (higher HR and power) but showed no improvement in symptoms. These results suggest that motor improvement after dynamic cycling is not driven by purely cardiovascular or metabolic mechanisms (3). We propose that complex and variable sensory input during dynamic cycling increases sensory feedback from the periphery and subsequent activation of the basal ganglia circuits. Activation of these circuits could enhance central motor processing. Accurate voluntary movement requires somatosensory input from the periphery. Peripheral receptors, such as joint receptors, golgi tendon organs, muscle spindles, and cutaneous receptors, send information from the limbs to the cortex. Several studies have identified proprioceptive impairment in PD, specifically in muscle spindle responses, load sensitivity, and kinesthesia (12, 32–35). This suggests that deficits in peripheral afferent input or sensorimotor integration likely contribute to abnormal motor output in individuals with PD.

During dynamic cycling, proprioceptors measuring joint angles, muscle length and force, and cutaneous receptors on the bottom of the foot (36) would be activated. Improvements in motor function and mobility after bouts of cycling in individuals with PD could be due to increases in afferent input to the cortex. Several EEG studies in healthy individuals have shown that significant sensorimotor processing is present during active pedaling (37) and that high-cadence training promotes neural efficiency as defined with EEG spectral power analysis (38). This indicates that activation of proprioceptors with a high frequency but variable pattern may be important for symptom improvements in PD.

Bradykinesia, one of the most central cardinal symptoms of PD, may have significant origins in the alteration of scale perception as it relates to movement and may point to a possible underlying dysfunction in sensorimotor integration (12, 39). Our data with dynamic cycling suggest that the combination of (1) high-cadence cycling and (2) the introduction of variable cadence improve symptoms in PD, most notably rigidity and bradykinesia. The idea that dynamic cycling could invoke the retuning and integration of kinesthesia, as it relates to motor programing, is compelling. Naito has shown that kinesthetic input illusion activates primary motor cortex, as well as other related motor areas, including cingulate motor area and supplementary motor area (40), in healthy individuals. They also suggested that sensorimotor integration could occur directly in these motor regions. Thus, exploration of this mechanism by studying sensory changes in individuals with PD through the course of the adaptive dynamic cycling intervention has a high likelihood of yielding illuminating results regarding mechanisms of improved motor function.

Several studies have shown that bradykinesia and gait in PD can be improved with dynamic sensory cues (41–43). The theory of paradoxical kinesia, which suggests that motor action triggered by sensory stimuli circumvents damaged basal ganglia pathways (41, 44, 45), has been proposed as a mechanism for these improvements. In addition, research investigating the benefits of dancing in PD has suggested that the strong musical rhythms and asymmetrical movements in tango provide important sensory feedback cues that promote improvements in balance and gait (46–48). However, additional research examining the changes in proprioceptive sensitivity after dynamic cycling is necessary.

### **Limitations**

There are a few limitations to this study. We chose to exercise and test individuals in the "on" medication state in an effort to examine a true functional state. Individuals with PD would not exercise while "off " medication on their own. In addition, there is an increased risk of fall, injury, or discomfort during the "off " medication state. However, a recent exercise study by Prodoehl et al. (49) suggested that testing while "on" medication is adequate, as long as the timing of the last dose of medication relative to testing is controlled. In this study, we completed the pre-intervention and post-intervention testing at the same time of day and recorded when the last medication dose was taken in an attempt to minimize this variable. A second limitation of this study is a small sample size, which led to significant variability in responses within the groups. We did not want to limit our pool of participants by narrowing the inclusion criteria and, as a result, we had a wide range of disease severity and symptoms in our study. The pre-intervention UPDRS Motor III scores ranged from 11–40 in the static group (out of 108 possible) to 4–55 in the static group. Although participants were randomized into either dynamic or static cycling, the baseline UPDRS scores were different between the two groups. However, our statistics analyzed the baseline and post-intervention scores in each group independently to minimize the effect of this difference. Lastly, despite our hypothesized sensory-based mechanism of improvement, we did not measure sensory function directly in this study. Future studies will measure changes in proprioceptive sensitivity using a passive joint repositioning test (50).

### **Conclusion**

We believe that dynamic cycling provides variable sensory input to the basal ganglia that promotes improvements in motor speed and quality. The dynamic nature of this paradigm will allow for optimization of the therapy per individual through adaptive control mechanisms and over time. Due to the variation in responses to this therapy, additional work is needed to determine how dynamic cycling can be individualized for people with varying degrees and severity of symptoms. Future studies will test this theory by examining both motor and sensory function throughout the long-term dynamic cycling intervention.

## **Author Contributions**

AR was the primary designer of the study, oversaw data collection, data analysis, preparation of the manuscript, and agreed to be accountable for all aspects of the work. RP was primarily responsible for data collection and analysis and assisted in manuscript preparation. BW assisted in study design and data collection (UPDRS). FD and KL assisted in study design and were responsible for design and development of the dynamic

### **References**


cycle. All authors have given final approval of the version to be published.

### **Funding**

This work was funded by a National Institutes of Health Grant R21 HD068846 to Angela L. Ridgel.

### **Acknowledgments**

The authors thank Hassan Abdar and Jay Rao for their work to develop the cycle and the controller. Their technical expertise was invaluable. Kayla Wilson assisted with data collection and subject recruitment. The authors are also grateful for the time and support of the Parkinson's disease community.


**Conflict of Interest Statement:** Provisional patent filed through KSU on 12/2014; full application has not yet been submitted; no royalties have been distributed.

*Copyright © 2015 Ridgel, Phillips,Walter, Discenzo and Loparo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# **Cues and Attention in Parkinsonian Gait: Potential Mechanisms and Future Directions**

*Daniel S. Peterson1,2 \* † and Katrijn Smulders <sup>2</sup>†*

*<sup>1</sup> Veterans Affairs Portland Health Care System, Portland, OR, USA, <sup>2</sup> Department of Neurology, Oregon Health & Science University, Portland, OR, USA*

**Keywords: cues, dual tasking, walking, Parkinson's disease, cognition**

## **INTRODUCTION**

Gait in people with Parkinson's disease (PD) is arrhythmic, small in amplitude, and variable (1–3). In addition, people with PD often exhibit reduced automaticity of movement (4), resulting in increased attention directed toward gait. This can be observed empirically when they have to perform a secondary task in addition to gait, so-called dual-task walking. In dual-task conditions, people with PD show larger impairments in gait than their healthy peers (5–7); for review, see Ref. (8). One strategy to improve gait in people with PD is cueing. Cueing is a well-established rehabilitation technique for improved locomotion in people with PD (9). In the clinic, auditory cueing is typically used to improve consistency and rhythmicity of steps. In individuals with PD who freeze, visual and auditory cues can also be used in a transient manner to break freezing events [for review, see Ref. (10)]. However, the mechanisms through which cueing improves gait are incompletely understood. The purpose of the current manuscript is to present proposed mechanisms of action of cueing. Further, we highlight the importance of cognition and, specifically, attention, in the efficacy of cueing. Finally, we present several possible directions for future research in the field.

Attention plays an important role in the efficacy of cueing. For example, as reduced movement automaticity may contribute to poorer gait function (e.g., smaller, more variable steps) in people with PD (4), external cues may act as pace-makers, taking the place of this additional cognitive control and reducing the amount of attention needed to maintain stable gait. This would mean that cued gait would allow more attention to be devoted to other secondary tasks, and one would expect lower dual-task costs (11). Alternatively, external cues may help to focus attention on gait. This should be particularly helpful in conditions that require more attention, such as walking while negotiating obstacles. If this were true, then one would expect to see a *prioritization* of the gait task over other tasks while using cues. Finally, it could be true that in specific circumstances and subgroups, external cues represent an additional cognitive task to walking, also requiring attention (12). Thus, cues may compete with gait for attentional resources and reduce gait quality during complex or attention-demanding environments.

Research has provided clues regarding the role of attention in cued gait. In a sub-analysis of the RESCUE trial (11), the effect of cueing was tested while completing either simple walking (no secondary task) or a complex secondary motor task – carrying a tray with glasses filled with water. Interestingly, results showed that gait speed improvements through cueing were only apparent while completing the complex motor task; a detrimental effect was observed during simple walking. In other words, the cue prevented gait slowing even while carrying the tray. These results suggest that cueing improves dual-task ability, and seem to support the idea that cues reduce attentional demands, thus freeing up attentional resources to secondary tasks. However, while this conclusion is plausible, it is also possible that cueing forced allocation of attention toward gait, potentially to the detriment of the secondary motor task performance. To distinguish between these two explanations,

#### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Nicolas Farrugia, Goldsmiths University of London, UK Colum Doan MacKinnon, University of Minnesota, USA*

#### *\*Correspondence:*

*Daniel S. Peterson petedani@ohsu.edu, dspeterson8@gmail.com*

*†Daniel S. Peterson and Katrijn Smulders have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 14 September 2015 Accepted: 23 November 2015 Published: 08 December 2015*

#### *Citation:*

*Peterson DS and Smulders K (2015) Cues and Attention in Parkinsonian Gait: Potential Mechanisms and Future Directions. Front. Neurol. 6:255. doi: 10.3389/fneur.2015.00255*

it would be necessary to capture the performance on the secondary task (i.e., what happened to the glasses on the tray) in the complex gait plus cueing condition. If cueing reduces the required amount of attention for gait, we would expect the cues not to impact performance carrying the tray. In contrast, if cueing led to allocation of attention specific to gait (prioritization of gait over secondary task), then deterioration on the tray-carrying task would be expected.

In a more recent study (13), a secondary cognitive task was used to evaluate the effects of attentional or "internal" cueing (i.e., "think about taking larger steps"), and "external" auditory cueing during attention-demanding situations. Results showed that attentional cueing, but not auditory cueing, resulted in improved gait velocity, possibly due to the specific focus on length of steps in that condition. Interestingly, the benefits of attentional cueing were retained during dual tasking. Although not formally analyzed, the authors also reported no differences in the cognitive task during the different conditions. This lends some support to the idea that cueing, in particular, "internal" attentional cueing, may reduce the amount of attention needed for gait. Additional indirect evidence for this hypothesis comes from a study where PD subjects had to respond to auditory stimuli that were presented regularly, while walking. A concern in this dual-task study was that presentation of these secondary-task stimuli at a fixed interval (1 or 2 s) could act as external cues, thereby improving gait. To test this, an additional condition was added in which auditory stimuli were presented at random intervals (1–3 s). Interestingly, the more difficult 1s condition (higher time pressure) yielded less dual-task costs on gait than the 2 s or variable interval conditions, possibly reflecting a cueing effect. However, during this 1 s condition, we also observed higher dual-task costs on the auditory cognitive task (i.e., slower and less accurate responses). Thus, although the cognitive stimuli may have worked as a cue to help to keep gait speed up, this came at a price, namely, drawing attention from the cognitive task. Given these partially conflicting results, additional work directly assessing performance on both gait and cognitive tasks with and without cues and dual tasks will be necessary to elucidate these interactions.

The concern that external cues might be detrimental for complex gait due to attentional costs to attend to the cues is not supported in the abovementioned studies. This concern was also directly addressed in a study looking at obstacle avoidance performance while walking on a treadmill (12). This challenging task has previously been shown to demand considerable attention (14– 16). In line with previously mentioned studies, this complex motor task was not affected by external cues, providing further evidence that external cues do not add additional attentional costs during walking, thus worsening gait performance.

Together, these findings confirm that cueing improves gait even in complex or cognitively challenging environments. However, one should note that not all gait parameters change in a similar fashion through cueing. For example, spatial parameters such as gait speed and stride length are typically improved during visual, attentional, and rhythmic cues, while temporal cues such as cadence are unaffected (11, 17, 18). These results are not fully consistent however, as Lebold and Almeida recently showed that visual cues increased stride length, but reduced cadence in people with PD (19). Further, recent work has begun to investigate the effects of cues on gait stability parameters, such as spatiotemporal variability. However, there is currently a paucity in high-quality research on the effect of cues on these stability metrics that may be more reflective of quality of gait.

### **ROLE OF COGNITIVE IMPAIRMENTS**

A deeper understanding of shifts in attentional load and prioritization during cued gait is particularly important for individuals with limited cognitive resources. This population often exhibits more pronounced gait and postural decline (20–22), including freezing of gait (FoG) (23), and evidence suggests that these deficits are at least partially due to limited attentional capacity (22). Indeed, cognitively impaired elderly exhibit larger dual-task costs on gait than their cognitively intact peers (24–26).

However, previous research on cued gait in PD has been focused primarily on people who were cognitively intact. To effectively use cues, one must have sufficient cognitive capacity to (1) not be overwhelmed by multiple, perhaps distracting stimuli, and (2) be able to deploy the cues at the necessary moment. Thus, people with cognitive dysfunction may have difficulty utilizing cues. This specific concern was addressed in a study evaluating cueing effects in PD with mild-to-moderate cognitive impairment (MMSE 15–26) and found that cueing was feasible and effective (27). However, the performance on the secondary task was not reported, raising the possibility (as noted above) that cueing may have shifted prioritization to the gait task, resulting in poorer performance on the secondary gait task. Interestingly, Willems and colleagues (28) showed that PD patients who freeze benefited from cueing only when the frequency of the cues matched their step frequency, whereas non-freezers also benefited from all cue presentations. Further, PD patients who freeze have been shown to exhibit a dramatic performance loss while dual-tasking despite the presence of cues. This was not the case in PD without FoG (29). These findings could be attributed to the poorer executive and attentional functions in PD with FoG (30–34). However, it cannot be ruled out that other, motor-related differences between PD patients with and without FoG underlie this lack of cueefficacy in people who freeze. Given these results, additional work administering cues with and without secondary task in individuals with more severe cognitive difficulty is warranted.

### **FUTURE DIRECTIONS**

Given current evidence regarding gait and attentional impairments in people with PD, certain considerations could be made to improve the efficacy of cues in the face of secondary cognitive challenges. Several possible approaches are outlined below.

First, different cue modalities can affect different gait characteristics. Rhythmical auditory cues were shown to be more effective than rhythmic visual or proprioceptive cues to improve gait speed, possibly due to a higher degree of integration of auditory rhythmical cues into stepping (27). Alternatively, visual cues including projecting lines on the ground have previously been shown to be more effective than auditory cues at improving step length (35, 36); however, a recent meta-analysis reported that visual cues did not result in additional improvement in stride length over auditory cues (37). Proprioceptive cues may also be beneficial for people with PD. Recent studies have suggested that walking on a treadmill improves gait in people with PD (38), and these improvements are maintained during subsequent over ground walking (39). Although the mechanism of gait improvements during treadmill walking is not fully understood, proprioceptive cues from the treadmill may contribute to these improvements. Such proprioceptive cues require little to no cognitive processing or attention (40, 41), and therefore may be particularly beneficial for individuals with cognitive challenges. However, there is a relative lack of literature investigating proprioceptive cuing, and additional work will be necessary to fully understand the effect of this type of cueing on gait in people with PD. In particular, treadmill locomotion may be different than overground walking in a number of ways (42, 43), including, but not limited to proprioceptive cues. Therefore, studies investigating the specific effects of proprioceptive cues, controlling for other treadmill related sensory inputs, are warranted.

A final perspective on cue modality choice is the personal preference. For rhythmical cues, auditory cueing seem to be preferred over other modalities. Nieuwboer and colleagues showed that when given the choice of three different rhythmic cues, proprioceptive (vibrations on the wrist), visual, and auditory cues, 67% of participants chose auditory, 33% chose proprioceptive, and 0% chose visual (44). However, no studies have compared personal preference of spatial (e.g., lines on the ground) to rhythmic cues.

Another aspect of cueing is whether to limit it to training circumstances and hope for transfer to situations without cues or to broaden its use to everyday life. Extending the use of cues outside training situations, like walking over a busy sidewalk, will introduce a number of challenges, but may be necessary given the relatively limited success of cue training over longer followup periods (44). Technological advances may help to reduce these challenges. For example, cues that are automatically initiated by arrhythmic gait (captured by body-worn sensors) may represent an important tool for cue utilization (45). Instead of continuous cueing, this "just in time" or "on demand" cueing might be more practical to use during daily activities. In addition, automatically initiated cueing could be helpful for individuals with cognitive dysfunction who may not be able to initiate cueing

### **REFERENCES**


at the appropriate moment. Incorporation of visual cues which are integrated in the environment, so-called "augmented reality cueing" is also being developed. For example, Espay and colleagues designed a system that projects a tiled floor pattern over the real environment, with the optical flow of the tiles adapted to the individual's walking speed (46). Another example is the use of Google glass to present visual optical flow as cues (47). While these forms of cueing are promising, additional testing will be necessary to identify which approaches are beneficial for patients.

### **CONCLUSION**

Cues can improve gait in people with PD. Different hypotheses have been put forward to explain these improvements, but further research is necessary to understand how cues improve gait. In particular, capturing changes in performance in both gait and cognitive tasks during cued dual-task will help elucidate underlying mechanisms. These interactions are particularly important to understand how individuals with reduced cognitive capacity utilize cues in distracting environments. Further, recent research has provided insight into how and when to choose cue modality, as well as how technology can integrate cues into real world environments, further reducing structural or cognitive interference. Continued investigation of these topics will improve our ability to utilize cues to improve gait in people with PD.

### **AUTHOR CONTRIBUTIONS**

Both authors (DP, KS) contributed equally to the drafting and revision of this manuscript.

### **FUNDING**

This work was supported by a grant from the United States Department of Veteran's Affairs Rehabilitation Research and Development Service (Career Development Award-1: #I01BX007080; PI: DP), and the Medical Research Foundation of Oregon (Early Investigator Awards; PI: DP, KS). The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.


47. Zhao Y, Heida T, Nonnekes JH, van RJA. Rhythmic cueing with Google glass for patients with Parkinson's disease [abstract]. *Movement Disorders* (2015) **30**(Suppl 1):350.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Peterson and Smulders. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Beyond the Metronome: Auditory Events and Music May Afford More than Just Interval Durations as Gait Cues in Parkinson's Disease

Matthew W. M. Rodger\* and Cathy M. Craig

*School of Psychology, Queen's University Belfast, Belfast, UK*

Keywords: Parkinson's disease, auditory cues, gait, music, affordances

## INTRODUCTION

Among the most apparent and adverse symptoms of Parkinson's disease (PD) are disturbances in gait. These include shuffling (small amplitude steps), instability (asymmetry and variability between steps), freezing of gait (cessation of movement and difficulty with initiation), and general disfluencies in walking movements and posture (Morris et al., 1996; Bloem et al., 2004; Grabli et al., 2012). Limitations of pharmacological interventions to alleviate gait disturbances (Lord et al., 2011), have led to interest in exploring non-pharmacological means of enhancing walking in PD, to complement drugs-based approaches. Sensory cueing, in which perceptual guides for movement are presented visually, acoustically, or haptically, is one such approach. While sensory cueing, in particular rhythmic auditory cueing, is a viable and promising approach to enhancing gait in PD, it is our opinion that this approach could be expanded by developing a more action-focussed framework for understanding the information available to patients in sound (cues) and how this information influences gait.

#### Edited by:

*Elvira Brattico, Aarhus University, Finland*

Reviewed by: *Floris Tijmen Van Vugt, McGill University, Canada*

\*Correspondence: *Matthew W. M. Rodger m.rodger@qub.ac.uk*

#### Specialty section:

*This article was submitted to Auditory Cognitive Neuroscience, a section of the journal Frontiers in Neuroscience*

Received: *01 December 2015* Accepted: *31 May 2016* Published: *14 June 2016*

#### Citation:

*Rodger MWM and Craig CM (2016) Beyond the Metronome: Auditory Events and Music May Afford More than Just Interval Durations as Gait Cues in Parkinson's Disease. Front. Neurosci. 10:272. doi: 10.3389/fnins.2016.00272* AUDITORY CUEING AS INTERVAL-SPECIFICATION

One form of sensory cueing for gait is to present a person with rhythmic sound, such as a metronome or beat-based music, and ask them to walk in time with the sound. This Rhythmic Auditory Stimulation (RAS) as gait cues for PD has been shown to lead to improvements in different aspects of gait, such as step length, duration, speed, and variability (e.g., Thaut et al., 1996; De Icco et al., 2015). An explanation for these improvements is that the cerebellar-thalamocortical circuits in the brain that support detection and synchronization to regular perceptual events are relatively preserved in PD, whereas the basal ganglia-thalamocotrical network that supports actions to one's own internal beat are impaired by the disease (Dalla Bella et al., 2015). Because the timing for action is externalized in RAS, each movement (step) can be matched to each perceptual cue (beat), resulting in a more stable gait pattern with larger steps (Nombela et al., 2013).

Although RAS has shown promising positive effects on gait in PD, this approach seems predicated on a limited view of how auditory events can specify information for action. According to the model, inherited from research on sensorimotor synchronization (SMS) more generally, successfully cued performance is defined as the adjustment of intervals between movement boundaries such that each movement begins/ends at the same moment in time as the onset of the perceptual event (beat). A paradigm of this is finger-tapping to a metronome: accurate tapping is achieved when successive tap events (minimum vertical displacement of the finger) coincide temporally with beat events (peak intensity of sound onset). The translation of this to RAS cueing of gait is represented in **Figure 1A**. Importantly, this view supports the use of a metronomic (isochronous) beat as the simplest (purest) depiction of interval durations; a listener must simply hop between evenly-spaced islands of sound in a sea of silence (Rodger and Craig, 2014). This model is highly problematic for PD patients, who struggle with synchronizing movements to discrete auditory cues, especially as the disease progresses (Bieñkiewicz and Craig, 2015). Although comparison of synchronization performance to discrete versus continuous sounds has not been directly tested in PD (to our knowledge), recent evidence has shown that continuous sounds lead to better step rhythmicity in patients than a metronome (Young et al., 2016). Hence the idealization of gait cueing as SMS may place an insurmountable burden on PD patients, and indeed is likely an unrealistic picture of what actually occurs during RAS gait cueing, given that most reported research is vague about how participants are instructed to "walk to the sounds" and generally analyse effects on global measures of gait (e.g., average step cadence) rather than step-to-beat synchronization.

### FROM INTERVALS TO EVENTS: CONSIDERING THE ACTION-RELEVANCE OF SOUNDS

The interpretation of RAS gait-cuing as discrete interval-matching not only implies a challenging demand for PD patients, it also draws a limited picture of sensorimotor

FIGURE 1 | Different models of auditory cueing of gait. (A) Walking to a beat is depicted as discrete movement-to-discrete sound mapping. The task for participants (often implicitly instructed) is to temporally match a particular instance within the gait cycle (i.e., heel contact) to the onset of each beat, whether represented by a metronome or by the down-beat of musical rhythm. (B) walking to a repeated sound is depicted as the coupling of the gait cycle to dynamic information in the auditory event, in this case the continuous sound made by footsteps on gravel, with the intensity envelope of the sound highlighted in the representation.

behavior. Temporal experience is not discretized, but rather is a continuous unfolding flow, populated by meaningful happenings (Gibson, 1975). Moreover, our actions are sequenced, prospectively aimed and ongoing in continuous lived experience (Dewey, 1896; Gibson and Pick, 2000; Merleau-Ponty, 2002). These philosophical concerns have practical consequences: by focusing on cues purely as representations of intervals we may overlook information relevant for action within the perceptual events themselves. In SMS, repeated sounds with identical interval durations can have substantially different effects on synchronized movement timing and trajectory, depending on whether the sound events are discrete or continuous in nature (Rodger and Craig, 2011, 2014). In particular, discrete sounds lead to greater accuracy in coinciding movement end-points with interval onsets, while continuous sounds lead to greater consistency in movement durations and smoother trajectories. These findings can be understood from the perspective of the mappings afforded between sensory and movement events discrete sounds specify clear event boundaries to map movement boundaries onto, while continuous sounds provide information throughout the movement to couple onto and produce regular repetitive actions, but which are not anchored as tightly to pre-set interval boundaries. This description frames auditory cues for gait as affordances, defined as opportunities for action specified in the sensory array, detectable by an agent in its environment (Gibson, 1979). In thinking of cueing sounds as auditory affordances, one must take seriously the way the event is specified in perception, the action-possibilities afforded in the event, and the capabilities of the perceiver to detect and act upon such an event (Steenson and Rodger, 2015).

An illustration of how auditory cues can be considered as action-specifying events is provided by research using footstep sounds as gait cues. Although rhythmic sounds can be artificially generated to be as simple as possible, such as short-tone metronomes, many rhythmic sounds in our environments are more richly patterned by the actions which produce them. Examples include the slosh of oars through water when rowing, the clunks of a chain gang working on a rail line, and the sounds made by footsteps during walking. As listeners, we are particularly tuned into the actions (and actors) that make such sounds and can make accurate judgments about these (Li et al., 1991; Kennel et al., 2014). From footstep sounds, listeners can recognize the gender or emotion of the walker (Li et al., 1991; Giordano and Bresin, 2006), and can discern the spatial properties of the walking action, including step length (Young et al., 2013). Furthermore, listeners can make use of spatial properties of footstep sounds in their own actions. Young et al. (2013) showed that when asked to walk as though they were generating the sounds of recorded footsteps, participants spontaneously adjusted their own step-length in accordance with the step lengths that produced the recorded sounds. Hence, walking sounds influenced spatial as well as temporal properties of gait in listeners. This is represented in **Figure 1B** as a repeating but continuous movement event (foot plantar rotation across the ground) mapped to a continuous auditory event (gravel crunch). Importantly for present purposes, this effect was also found for people with PD: walking to the sounds of long strides on gravel led to increases in patients' step length (Young et al., 2014). Interestingly, these effects were comparable to a condition in which participants were asked to walk with long steps in time to a metronome, but the gravel footsteps led to the added benefits of reduced variability in both step length and duration. Hence, the action-relevant gravel sounds specified parameters of walking beyond merely step-to-step intervals which was useful in guiding stepping movements in people with PD.

### MUSIC AS RICHLY-STRUCTURED, ACTION-RELEVANT SOUND EVENTS

One rich source of action-relevant auditory events is music. A number of research programmes investigating auditory cueing in PD have used music as RAS, either pre-recorded songs with patient-tailored tempo adjustments (e.g., De Bruin et al., 2010; Benoit et al., 2014), specially synthesized musical tracks (e.g., Cancela et al., 2014), or augmented pre-existing music to highlight beat events (e.g., Thaut et al., 1996). This research has demonstrated benefits of music as a stimulus for gait improvement, and the benefits of music can exceed those of metronomic cues (Wittwer et al., 2013). However, interpretations of such benefits are often couched in the above-described model of SMS; walking to music is taken to involve matching step events to the down-beats or "pulse" within the music. Hence, music is considered a special example of beat-based cueing, a "metronome+." It is still considered essentially the presentation of time intervals, but with better signposted interval boundaries (being embedded in the musical structure), and/or being a metronome which is emotionally engaging or motivational. This view still makes it hard to understand why music as gait-cue can lead to spatial / velocity benefits (Wittwer et al., 2013) when compared to metronome cueing. A view of music that takes into account the action-relevance of musical sounds may be more explanatory.

Music is a complex and highly structured concatenation of events, the result of skillfully sequenced actions by musicians. As such, music contains multiple, nested opportunities for action which include mapping movement event boundaries to down beats, but may also include other cues to movement. Music is rarely discrete, but rather a continuous stream of sound sculpted across different time-scales to include edges (e.g., beats), contours (e.g., melodies) and landscapes (e.g., chord progressions). Different parameters of continuous gait cycles may map to these different events in the music, not necessarily just heel-strike to down beat. This affords flexibility to adapt the ongoing gait pattern to the auditory events, without the rigidity of a one-to-one mapping implied in metronomic cueing. For example, a patient may either lift his/her toe off with a beat, place the heel with a chord, or swing the leg with part of a melody, and still have the veridical experience of being in time with the sound. They may also switch unconsciously between these mappings mid-walk, and still be adaptively matching the walking pattern to the sound, such is the nested nature of multiple musical events. Hence, music continuously provides multiple opportunities within each cycle to prospectively map ongoing movements to. Moreover, music is often described as conveying a sense of motion above the individual auditory events that comprise it (Eitan and Granot, 2006; Zhou et al., 2015). This may afford a mapping for the experience of forward propulsion that continuous walking generates, which in turn may explain why walking to music can lead to greater overall gait velocity than a metronome of the same tempo (Wittwer et al., 2013). These possibilities indicate the importance of conceptualizing music as cuing more than step intervals.

The promise of music as a special form of gait-cueing for PD is further enhanced by the creative nature of music in that many parameters of musical sound may be sculpted and composed to afford desired actions. With a metronome, tempo and interval variance (e.g., 1/f noise) are about the only degrees of freedom available to the experimenter/clinician, but within the constraints of a particular tempo music is free to vary in a multitude of ways. Harmonic structure of chords, and balance between bass and melody have been shown to have influences on movement (Hove et al., 2014; Komeilipoor et al., 2015). Stylistic choices, such as instrumentation, may be adapted to suit the listener's musical preferences, with the intention of enhancing the invitations to action afforded (Whitagen et al., 2012; Schiavio and Altenmüller, 2015). Moreover, composing music for walking allows auditory affordances to be adapted to the capabilities of the listener-walker, which may be particularly relevant for PD (Cancela et al., 2014). Given the pre-existing challenges for PD patients to generate "healthy" styles of gait, one can imagine compositions of music which affords cyclical step patterns and forward propulsion, but without the regimentation of marching music, for example. Such suggestions are speculative, but are warranted within the proposed framework as an underexplored approach to musical action sounds for gait cueing.

### CONCLUSION

Our opinion is that by considering the affordances that auditory events such as footstep sounds or music specify for action, that go beyond the metronomic indication of temporal intervals, auditory cues for gait in PD can be better understood and further expanded. It is important to recognize that this view is not in opposition to existing approaches, but rather offers a reinterpretation of previous findings. Metronomes may afford intercepting a regular beat (or reacting to it) through movement, while action sounds and music afford this and more. Moreover, these opportunities are mediated by the actor's capabilities to detect and act on such information in the sound, which should inform the tailoring of cues to the characteristics of people with PD. By considering auditory cues for walking in PD along with the mapping of action capabilities to temporally patterned sound events, a richer, more flexible understanding and development of cueing approaches to gait in Parkinson's is possible.

### AUTHOR CONTRIBUTIONS

Both authors MR and CC conceived of the ideas presented in the manuscript, MR wrote the main draft which CC edited critically to revise intellectual content. Both authors approve the

### REFERENCES


final manuscript and agree to take responsibility for content contained.

### ACKNOWLEDGMENTS

The authors wish to thank Dr Marta Bieñkiewicz for some useful discussions during the preparation of this manuscript.


Merleau-Ponty, M. (2002). Phenomenology of Perception. London: Routledge.


temporal components of gait in Parkinson's disease patients. Neuropsychologia 57, 140–153. doi: 10.1016/j.neuropsychologia.2014. 03.009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Rodger and Craig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Effect of Dopaminergic Medication on Beat-Based Auditory Timing in Parkinson's Disease

#### *Daniel J. Cameron1 \*, Kristen A. Pickett2 , Gammon M. Earhart <sup>3</sup> and Jessica A. Grahn1,4*

*1Brain and Mind Institute, University of Western Ontario, London, ON, Canada, 2Occupational Therapy Program, Department of Kinesiology, University of Wisconsin-Madison, Madison, WI, USA, 3Program in Physical Therapy, Department of Neuroscience, Department of Neurology, Washington University School of Medicine in St. Louis, St. Louis, MO, USA, 4Department of Psychology, University of Western Ontario, London, ON, Canada*

#### *Edited by:*

*Oscar Arias-Carrión, Hospital General Dr. Manuel Gea González, Mexico*

#### *Reviewed by:*

*Mirta Fiorio, University of Verona, Italy Mohamed Mosaad Salama, Mansoura University, Egypt*

> *\*Correspondence: Daniel J. Cameron dcamer25@uwo.ca*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 18 September 2015 Accepted: 08 February 2016 Published: 22 February 2016*

#### *Citation:*

*Cameron DJ, Pickett KA, Earhart GM and Grahn JA (2016) The Effect of Dopaminergic Medication on Beat-Based Auditory Timing in Parkinson's Disease. Front. Neurol. 7:19. doi: 10.3389/fneur.2016.00019*

Parkinson's disease (PD) adversely affects timing abilities. Beat-based timing is a mechanism that times events relative to a regular interval, such as the "beat" in musical rhythm, and is impaired in PD. It is unknown if dopaminergic medication influences beat-based timing in PD. Here, we tested beat-based timing over two sessions in participants with PD (OFF then ON dopaminergic medication) and in unmedicated control participants. People with PD and control participants completed two tasks. The first was a discrimination task in which participants compared two rhythms and determined whether they were the same or different. Rhythms either had a beat structure (metric simple rhythms) or did not (metric complex rhythms), as in previous studies. Discrimination accuracy was analyzed to test for the effects of beat structure, as well as differences between participants with PD and controls, and effects of medication (PD group only). The second task was the Beat Alignment Test (BAT), in which participants listened to music with regular tones superimposed, and responded as to whether the tones were "ON" or "OFF" the beat of the music. Accuracy was analyzed to test for differences between participants with PD and controls, and for an effect of medication in patients. Both patients and controls discriminated metric simple rhythms better than metric complex rhythms. Controls also improved at the discrimination task in the second vs. first session, whereas people with PD did not. For participants with PD, the difference in performance between metric simple and metric complex rhythms was greater (sensitivity to changes in simple rhythms increased and sensitivity to changes in complex rhythms decreased) when ON vs. OFF medication. Performance also worsened with disease severity. For the BAT, no group differences or effects of medication were found. Overall, these findings suggest that timing is impaired in PD, and that dopaminergic medication influences beat-based and non-beat-based timing differently. Judging the beat in music does not appear to be affected by PD or by dopaminergic medication.

Keywords: beat perception, rhythm, timing, Parkinson's disease, dopamine

## INTRODUCTION

Parkinson's disease (PD) causes progressive motor and cognitive deficits, including deficits in timing (1, 2). Timing deficits in PD are likely related to dopaminergic dysfunction in the basal ganglia as the striatal dopaminergic system is known to be involved in timing (3, 4). One particular type of timing, beat-based timing, involves the timing of events relative to a regular interval, or a "beat," such as in musical rhythm. In music, the beat is the regular, perceived emphasis to which listeners tend to synchronize their movements (e.g., by clapping their hands or tapping their feet). Beat-based timing activates the basal ganglia, among other cortical and subcortical regions of the motor system (5–9). There is mixed evidence regarding whether the basal ganglia have a *specific* role in beat-based timing compared to other types of timing. One study found that both beat-based (relative) and non-beat-based (absolute) timing was worse in participants from three clinical populations (not including PD) with impaired basal ganglia function, compared to control participants (10). This result was interpreted as supporting the "unified model" of timing (11), in which the basal ganglia play a central role in all types of timing, and not specifically in beat-based timing. However, another study found that patients with PD had a selective deficit in beat-based timing (12). Although both patients and controls discriminated metric simple (beat-based) rhythms more accurately than metric complex (non-beat-based) rhythms, patients with PD were significantly less accurate than controls for metric simple rhythms, but not for metric complex rhythms. This finding suggested a selective role for the basal ganglia in beat-based timing. However, all patients in the study were tested while ON dopaminergic medication, thus the influence of dopamine and dopaminergic medication on beat-based timing is not well understood.

For timing in PD generally, there is mixed evidence for dopaminergic medication's influence. Dopaminergic medication improves timing production of participants with PD in a task using intervals in the range of 30–120 s, but not in the range of 250–2000 ms (13). In addition, participants with PD perform more similarly to control participants (with less timing variability) on a set of timing tasks while ON medication than while OFF (14). A study using behavioral and positron emission tomography (PET) found no effect of medication on PD patients' ability to synchronize their tapping with an isochronous tones, although dopaminergic denervation was related to tapping accuracy (15). Another neuroimaging study (16) found that, although dopamine replacement therapy did not improve performance of patients with PD in a motor timing task, neural activity increased toward the level of controls in the dorsal putamen and supplementary motor area [regions associated with beat perception; (5)] during task performances. One study investigated the influence of dopaminergic medication on how well participants with PD detected beat structure in rhythms (17). Participants with PD and controls decided whether rhythms (that were either beatbased or non-beat-based) had a beat. Participants with PD did not significantly differ from controls, for either type of rhythm, although numerically participants with PD were worse than control participants at the task, and showed less difference in performance between beat-based and non-beat-based rhythms (control participants were better at recognizing that beat-based rhythms indeed had a beat than they were at recognizing that non-beat-based rhythms did not have a beat). As both groups had a small sample size (*n* = 9) and there was high variability between subjects, a real group difference in beat-based timing could have been missed. The study did find a small effect of dopaminergic medication: responses were faster when ON vs. OFF medication, and those responses were more accurate (though not statistically significantly). The task required explicit detection of beat structure in rhythms, similar to the Beat Alignment Test (BAT) in the present study. This explicit nature differs from the rhythm discrimination task used previously (12) and in the current study, for which an *implicit* influence of beat-based timing is expected: metric simple (beat-based) rhythms should elicit better performance than metric complex (non-beat-based rhythms), but no explicit awareness of the beat is required or assessed. Thus, there is mixed evidence for the influence of dopaminergic medication on timing in PD and limited evidence for its influence on beatbased timing in particular.

The uncertainty regarding dopaminergic medication's influence on timing is partly related to the uncertainty regarding the extent to which cognitive deficits in PD (including timing) are associated with dopamine and would thus be modulated by dopaminergic medication. Besides deficient dopamine, other factors also contribute to cognitive deficits in the disease, including structural changes to the brain (18, 19), and accumulation of amyloid plaques and tau protein (20, 21). Moreover, the role of dopamine, and influence of dopaminergic medication, in cognition is variable, as previous studies show both improvement and worsening of different cognitive functions by medication, depending on task demands and individual differences in baseline dopamine levels (22, 23), as well as side of motor symptom onset (24).

The current study investigated the role of dopaminergic medication on beat-based timing in individuals with PD. We tested participants with PD on two beat perception tasks in two sessions: OFF and ON medication. We also tested control participants in two sessions, but did not give them medication, to assess practice effects. The two tasks were a rhythm discrimination task and the BAT [from the Goldsmiths Musical Sophistication Index; (25)]. In the discrimination task, participants decided whether two rhythms were the same or different. In several studies, the discrimination task has elicited better performance for metric simple rhythms (beat-based), compared to metric complex rhythms (non-beat-based) (5, 12, 26–28). This "beat-based advantage" is thought to depend on the beat-based timing (or relative timing) mechanism, which is thought to, in turn, depend on basal ganglia function (12). The second task, the BAT, presents excerpts of real music clips with a sequence of regular tones added to the music. The tones are either aligned or misaligned with the beat of the music, and participants decide whether the tones were on or off the beat of the music.

Both tasks assess beat perception; however, beat perception in the discrimination task arises solely on temporal information, without the rich variety of acoustic cues present in real music. The discrimination task also requires a comparison of two separately presented rhythms, introducing a working memory component. We hypothesized that if beat-based timing depends on basal ganglia function, and is thus impaired in PD, dopaminergic medication should improve discrimination of metric simple rhythms (but not metric complex rhythms). In contrast to the discrimination task, the BAT assesses beat perception in the context of real music, meaning that there are numerous musical features, unrelated to timing, that emphasize the beat (e.g., bass timbres or certain chord changes are more likely to occur on the beat). Beat perception in the BAT therefore does not rely solely on timing cues. The BAT has, to the best of our knowledge, not been used in the context of PD. If participants with PD perform worse than controls, it would provide converging evidence for a deficit in beat perception and suggest that other musical cues to the beat do not sufficiently compensate for that timing deficit. Similarly, if dopaminergic medication improves BAT performance in participants with PD, then beat perception in real musical contexts is likely dependent on basal ganglia function. Alternatively, if the groups do not differ, and/or there is no effect of medication on BAT performance, then beat perception arising from non-temporal cues likely does not rely on intact dopaminergic function of the basal ganglia.

### MATERIALS AND METHODS

### Participants

Participants with PD were recruited from the St. Louis, MO, USA region to participate in an ongoing study investigating interventions for improving gait in PD. The data from these participants (*n* = 72, 30 female, mean age 66.8 years, SD = 9.4, mean of 2.18 years of musical training) are the baseline data, collected before any intervention, and only from participants who were taking dopaminergic medication at the time of testing. Controls (*n* = 70, 50 female, mean age 67.6 years, SD = 9.0, mean of 3.37 years of musical training) were recruited from the London, ON, Canada region. All participants scored at least 27 on the Mini Mental State Examination. Group differences in age and mean years of musical training were not statistically significant (*p* > 0.05). Participants' highest achieved education levels were scored from 1 (high school) to 4 (advanced degree), and the groups did not differ on this measure (mean PD education = 2.79, SD = 1.05, mean control education = 2.90, SD = 1.07, *p* > 0.05). All participants provided informed, written consent in accord with procedures approved by the respective ethics boards at Western University and Washington University School of Medicine.

### Stimuli

For the rhythm discrimination task, two types of rhythms were used: metric simple and metric complex (see **Figure 1**). The stimuli are described in detail elsewhere (12). Both types are composed of intervals that are related by small integer-ratios, and were presented in one of three tempi, corresponding to the shortest interval duration (i.e., rhythms consisted of interval durations of 1, 2, 3, and 4, in which "1" was equal to 200, 233, or 267 ms, "2" was twice the duration of "1," etc.). The trials used were a subset of the trials used in a previous study (12), and are listed in **Table 1**.

#### Table 1 | Trials for rhythm discrimination task.


*Empty cells indicate that the second rhythm was the same as the first. Numbers indicate relative duration of intervals (1* = *200, 233, or 267 ms).*

For the BAT [from the Goldsmiths Musical Sophistication Index; (25)], stimuli were 17 excerpts of music in which an isochronous tone sequence was embedded. The tone sequence could either be aligned with the beat of the music ("on beat," four trials), faster or slower than the musical beat (period-shifted, eight trials), or at the same beat rate, but misaligned to the musical beat (phase-shifted, five trials).

### Procedure

Testing consisted of completing each task twice in the same day, with between 30 and 90 min separating the two testing sessions. Participants with PD were OFF medication in the first session (participants were asked to withhold all anti-Parkinson's medication for at least 12 h prior to the session) and ON medication in the second session [all participants with PD were regularly taking levodopa (l-DOPA), typically in combination with carbidopa, except one participant who was taking rasagiline and pramipexole, one who was taking amantadine and pramipexole, and another who was taking ropinirole]. Between testing sessions, participants completed the MMSE and a questionnaire about their musical training. Additionally, participants with PD completed the Movement Disorders Society Unified Parkinson's Disease Rating Scale [MDS-UPDRS; Ref. (29)].

Participants completed both tasks on laptops. Auditory stimuli were presented via headphones, and instruction text was presented on the laptop display. For the rhythm discrimination task, participants heard three consecutive rhythms (see Stimuli) and responded whether they thought the final rhythm was the same as the first two (which were always identical). During the two presentations of the first rhythm, the text "Original rhythm: First listen" and "Original rhythm: Second listen" were displayed, respectively, in white text. During presentation of the final, comparison, rhythm, the text "SECOND rhythm" was displayed in red text. Following presentation of the second rhythm, "Was the SECOND rhythm the same or different? If same, press (S). If different, press (D)" was displayed in white text. Participants then indicated whether they thought the second rhythm was the same as or different from the original rhythm. Four practice trials were completed before testing.

For the BAT, participants completed 17 trials (see Stimuli) in random order. Participants were given verbal instructions to listen to each music excerpt and to respond whether the embedded tone sequence was "ON" or "OFF" the beat of the music. During listening, the laptop display read "Please Listen," and following each excerpt, it read "Are the tones on or off the beat? Press "y" for YES or "n" for NO on the keyboard." Three practice trials were completed before testing.

### Analyses

Rhythm discrimination scores (proportion of correct trials) were initially analyzed in a 2 × 2 × 2 mixed analysis of covariance (ANCOVA) with the between-subjects factor of group (PD vs. control), and the within-subject factors of session (first vs. second session, also corresponding to OFF vs. ON medication for the participants with PD) and metricality of rhythms (metric simple vs. metric complex), and including the covariates musical training (years) and education level (both mean-centered separately for the two groups). Analyses were repeated without covariates that were non-significant and/or did not interact with other factors in the initial analysis.

As our primary research interest was the relationship between beat perception and dopaminergic medication [known to influence cognition in PD; (30)], and because we were unable to test the healthy controls ON medication, we conducted a separate 2 × 2 ANCOVA on the PD patient data alone. This ANCOVA included the within-subject factors medication (OFF vs. ON) and metricality (metric simple vs. metric complex), and the covariate of MDS-UPDRS (subscale III, off medication, mean-centered).

Furthermore, as only a subset of trials from the previous 2009 study (12) were used in the discrimination task (due to limitations of testing time), data from the 2009 study were reanalyzed to include only the subset of trials that were used in this current study. Results from the 2009 study and current study (ON and OFF medication, separately) were compared using independent samples *t* tests.

Beat alignment test scores (proportion of correct trials) were analyzed in a 2 × 2 mixed ANCOVA with the between-subjects factor of group (control vs. PD) and within-subject factor of session (first vs. second session, also corresponding to OFF vs. ON medication for the participants with PD), and musical training and education (both mean-centered, separately for the two groups) as covariates.

### RESULTS

### Rhythm Discrimination

Participants with PD were worse than control participants at discriminating rhythms [main effect of group, *F*(1,136) = 10.86, *p* = 0.001, η*<sup>p</sup>* <sup>2</sup> = 0 0. 74]. In the second session, control participants performed better than in the first session, and participants with PD did not show this improvement [a statistical trend toward an interaction of group and session, *F*(1,136) = 3.31, *p* = 0.071, η*p* <sup>2</sup> = 0 0. 24], confirmed by follow-up paired *t* tests comparing average scores within each session for healthy controls, [*t*(68) = 2.22, *p* = 0.030, and for participants with PD, *p* > 0.05], as shown in **Figure 2**. Overall, discrimination was better for metric simple rhythms than for metric complex rhythms [main effect of metricality, *F*(1,136) = 36.50, *p* < 0.001, η*<sup>p</sup>* <sup>2</sup> = 0.212 ]. This effect of metricality was present for both groups in both sessions (*p*< 0.05, in all cases), as shown in **Figure 3**. Regarding musical training, there was a statistical trend toward those more training being better at discriminating rhythms [*F*(1,136) = 3.77, *p* = 0.054, η*p* <sup>2</sup> = 0 0. 27]. Although education level was not a significant covariate, there was a trend toward education interacting with metricality [*F*(1,136) = 3.63, *p* = 0.059, η*<sup>p</sup>* <sup>2</sup> = 0 0. 26]. Follow-up comparisons of these data show that trends were in opposite directions: those with more education tended to do better with metric complex rhythms and slightly worse with metric simple rhythms. The three-way interaction between group, session, and metricality did not reach significance (*p* > 0.1).

Analysis of only the data from participants with PD shows that metric simple rhythms were discriminated better than metric

groups, in both sessions. Error bars indicate ± 1 SEM.

training and education) for both groups, in both sessions, and for both metrical types of rhythms.

from participants with PD (discrimination scores are collapsed over

OFF and ON medication conditions).

medication) and from a previous study (ON medication only, and recalculated to include only the trials used in the current study).

complex rhythms [main effect of metricality, *F*(1,69) = 29.81, *p* < 0.001, η*<sup>p</sup>* <sup>2</sup> = 0 0 .3 2], as is evident in **Figure 3**. The effect of metricality was larger when participants were ON medication compared to OFF medication [interaction of metricality and medication, one-tailed, as the direction of differences was hypothesized, *F*(1,69) = 2.77, *p* = 0.050, η*<sup>p</sup>* <sup>2</sup> = 0 0. 39]. A paired *t*-test of the difference scores (metric simple minus metric complex, adjusted for UPDRS) confirmed a significantly greater difference while ON vs. OFF medication [*t*(70) = 1.69, *p* = 0.048]. Follow-up *t* tests indicated that metric simple scores numerically increased and metric complex scores numerically decreased from OFF to ON sessions, although neither change was statistically significant (*p* > 0.05).

Movement Disorders Society Unified Parkinson's Disease Rating Scale scores (off medication) significantly covaried with overall discrimination performance [*F*(1,69) = 11.49, *p* = 0.001, η*p* <sup>2</sup> = 0.143], as shown in **Figure 4**. MDS-UPDRS scores and mean discrimination scores (averaged over sessions and metricalities) negatively correlate (*rSp* = −0.32, *p* = 0.007). Although there was no significant interaction between MDS-UPDRS, medication, and task performance, exploratory analysis showed that performance negatively correlated with MDS-UPDRS scores both OFF and ON medication (*rSp* = −0.28, *p* = 0.020 and *rSp* = −0.36, *p* = 0.002, respectively).

We compared the results from our current participants with PD to the results of the 2009 study using the same task (with previous data recalculated to include only the same trials used in the present study). The 2009 sample was participants with PD, ON medication with one session of testing [*n* = 15; (12)]. The current ON medication sample was numerically more similar to the 2009 sample (also ON medication) than to the current OFF medication sample. Overall, however, independent samples *t* tests comparing the 2009 sample and the current sample ON and OFF medication (separately) show no statistically significant differences (*p* > 0.05) between discrimination scores for either metric simple or metric complex rhythms (see **Figure 5**). Thus, when including identical trials between the 2009 sample and the current sample, the two groups of patients did not differ in performance.

### Beat Alignment Test

Performance on the BAT did not differ between groups, or between sessions (see **Figure 6**). BAT scores were not associated with musical training or education, or MDS-UPDRS scores in participants with PD. Furthermore, BAT scores did not correlate with mean discrimination scores (*p* > 0.05). When analyzing data from participants with PD alone, we found no effect of medication or interactions involving medication (*p* > 0.05).

### DISCUSSION

Overall, the rhythm discrimination task was sensitive to timing deficits in PD: participants with PD performed worse than control participants, and participants more severely affected by PD (those with higher MDS-UPDRS scores) did worse than those less affected. Furthermore, control participants improved over the two sessions while participants with PD did not, suggesting that in addition to a deficit in timing, they may be less able to

learn from repetition of the timing task. Although we did not expect this result, and do not consider it a primary finding of the study, it is consistent with previous evidence of learning deficits in PD. For example, patients with PD show less consolidation than controls after learning a motor control task (31), and slower re-learning after disruption of a previously learned motor task (32). Moreover, dopaminergic medication can change the nature of learning in PD (33–35), possibly contributing to the lack of improvement in discrimination task performance from session 1 (OFF medication) to session 2 (ON medication) in PD. However, as the repetition of the task is confounded with medication status in participants with PD (the first session was always OFF medication, followed by ON medication in the second session), we cannot fully disentangle effects of repetition and medication in participants with PD.

The BAT was not sensitive to timing deficits in PD, as performance on the task by control participants and those with PD did not significantly differ. The additional information in real music may give listeners with PD sufficient cues regarding the beat, such that beat-based timing deficits do not impair task performance. In addition to the other musical cues to the beat that are present in the BAT, participants compare simultaneously presented sequences (the beat in the musical stimulus, and the overlaid tone sequence). The discrimination task, by comparison, requires a memory-based judgment, which requires the encoding, rehearsal, and retrieval of a rhythm. These processes involve the internal generation of rhythms, which is supported by the presence of temporal structure, such as the beat. Thus, the difference in cognitive processes required by the BAT and the rhythm discrimination task may underlie the difference in findings between tasks regarding the particular nature of timing deficits in PD. Consistent with this interpretation, performance on the BAT and discrimination task did not correlate, suggesting that these tasks are indeed sensitive to different aspects of beat perception, and rely on different underlying cognitive processes. Although BAT performance was numerically better for control participants than participants with PD, the large sample sizes suggest that any potential difference between patients and controls is small.

Regarding the discrimination task, we found mixed results. Contrary to our expectation, we did not see a clear deficit in beat perception associated with PD. Participants with PD performed better on the task for metric simple (beat-based) rhythms than for metric complex (non-beat-based) rhythms, similar to control participants. This is at odds with the finding from a previous study that used the same two types of rhythms in the same rhythm discrimination task, showing that the beat-based advantage (superior discrimination of metric simple rhythms) was significantly reduced in PD compared to controls. However, when results from the previous study were recalculated to include only the exact same trials used in the current study, the difference in performance between metric simple and metric complex rhythms increased, and closely resembled that of the current data (see **Figure 5**), suggesting that our current study did indeed replicate the previous results. Moreover, the previous study's recalculated data most closely resembles the ON medication data from the current study. This is notable as participants in the previous study were ON medication. The dependence of performance on the specific trials included demonstrates a potential limitation of discrimination tasks: performance is dependent not just on condition differences (rhythms are easier to encode and maintain when participants perceive a beat, therefore performance is generally better for beat-based rhythms) but also the specific nature of the discrimination being made. The task required participants to detect whether a change occurred in the rhythm, and the change, when present, was always a transposition (or swapping) of two time intervals. For example, the rhythm 211314 could become 211134. Some transpositions are easier to detect than others. Changes to the beginning or end of a rhythm are easier to detect than those in the middle because of primacy and recency effects (36). In addition, transposition of disparate intervals (e.g., 3 and 1) may be easier to detect than transposition of more similar intervals (e.g., 3 and 4). Thus, by reducing the number of trials selected for the current study, the results could be more influenced by these trial-specific differences that are not related to beat perception, but are instead related to the specific nature and location of the change in the rhythm. However, as the current results do not differ from the previous results (which used a much larger set of trials) when reanalyzed to include only the same trials, we feel it reasonable to interpret the current findings as replicating the previous finding that beat-based timing is impaired in PD, although trial selection influenced the exact pattern of results. Further support for a beat-based timing impairment needs to be acquired. This may be best accomplished by using different tasks, such as rhythm reproduction that do not have the limitations present in discrimination tasks.

For participants with PD, the difference in discrimination performance between metric simple and metric complex rhythms increased when ON vs. OFF dopaminergic medication. The data therefore suggest that medication influences beat-based timing in PD, although the pattern of the influence is complex. In particular, the worsening of performance for metric complex (non-beat-based) rhythms was unexpected. The improvement of performance for beat-based rhythms may be due to dopaminergic medication improving basal ganglia function, as they are thought to play a critical role in beat perception. Another possibility is that dopaminergic medication biases participants to search for a beat structure. This bias would improve performance for beatbased rhythms, in which a beat structure can be detected, and therefore searching for it is beneficial, but the same bias would worsen performance for metric complex rhythms, in which the beat structure is difficult to find, and attempting to search for it distracts from using another, better, strategy to remember the rhythms. As mentioned above, task repetition (first vs. second session) is confounded with medication (OFF vs. ON); however, the metricality-dependent change in performance (improved performance for metric simple rhythms and worsened performance for metric complex rhythms) is less likely due to repetition than to medication. An expected effect of repetition would be improved performance for one or both types of rhythms, but not worse performance. As such, we interpret the overall lack of improvement at the task as a deficit in learning the task (compared to control participants' overall improvement from the first to second session), but the different direction of performance change for metric simple and metric complex rhythms as an effect of medication on beat-based timing.

Overall, these data present further evidence that timing is impaired in individuals with PD (and worsens with severity of

### REFERENCES


the disease) that beat-based timing may be particularly impaired in PD, as we replicated findings from Grahn and Brett (12), which used a more complete set of trials (less subject to trial-specific effects) to show a lack of beat-based advantage, and that, consistent with previous work (17), dopaminergic medication may improve beat-based timing in PD.

### AUTHOR CONTRIBUTIONS

All authors contributed to design. DC and KP collected data. DC analyzed data, and all authors contributed to writing.

### ACKNOWLEDGMENTS

This work was supported by a grant from the Parkinson Society Canada to JG and NIH grant R01NS077959 to GE. Additional support was provided by the Greater St. Louis American Parkinson Disease Association (APDA) and the APDA Center for Advanced Research at Washington University in St. Louis. The authors would like to acknowledge and thank Esther Lachance, Ryan Duncan, Martha Hessler, Rich Nagel, and Marie McNeely for their help testing participants.

and off medication in Parkinson's disease. *Brain Cogn* (2008) **68**(1):30–41. doi:10.1016/j.bandc.2008.02.121


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Cameron, Pickett, Earhart and Grahn. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Parkinson's Is Time on Your Side? Evidence for Difficulties with Sensorimotor Synchronization

*Marta M. N. Bien´kiewicz1 \* and Cathy M. Craig2*

*<sup>1</sup> ISM, Faculté des Sciences du Sport, Université d'Aix-Marseille, Marseille, France, 2School of Psychology, Queen's University Belfast, Belfast, UK*

There is lack of consistent evidence as to how well PD patients are able to accurately time their movements across space with an external acoustic signal. For years, research based on the finger-tapping paradigm, the most popular paradigm for exploring the brain's ability to time movement, has provided strong evidence that patients are not able to accurately reproduce an isochronous interval [i.e., Ref. (1)]. This was undermined by Spencer and Ivry (2) who suggested a specific deficit in temporal control linked to emergent, rhythmical movement not event-based actions, which primarily involve the cerebellum. In this study, we investigated motor timing of seven idiopathic PD participants in event-based sensorimotor synchronization task. Participants were asked to move their finger horizontally between two predefined target zones to synchronize with the occurrence of two sound events at two time intervals (1.5 and 2.5 s). The width of the targets and the distance between them were manipulated to investigate impact of accuracy demands and movement amplitude on timing performance. The results showed that participants with PD demonstrated specific difficulties when trying to accurately synchronize their movements to a beat. The extent to which their ability to synchronize movement was compromised was found to be related to the severity of PD, but independent of the spatial constraints of the task.

### *Edited by:*

*Ryuji Kaji, Tokushima University Hospital, Japan*

#### *Reviewed by:*

*Maria Trinidad Herrero, University of Murcia, Spain Angelo Quartarone, University of Messina, Italy*

#### *\*Correspondence:*

*Marta M. N. Bien´ kiewicz marta.bienkiewicz@univ-amu.fr*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 21 July 2015 Accepted: 13 November 2015 Published: 27 November 2015*

#### *Citation:*

*Bien´ kiewicz MMN and Craig CM (2015) Parkinson's Is Time on Your Side? Evidence for Difficulties with Sensorimotor Synchronization. Front. Neurol. 6:249. doi: 10.3389/fneur.2015.00249*

Keywords: basal ganglia, temporal control, sensorimotor synchronization, Fitts' law, event-based timing, index of difficulty, PD, motor sychronisation

## INTRODUCTION

Parkinson's disease (PD) is a neurodegenerative disease of the substantia nigra pathways in the brain that affects 4.5 million Europeans, a number that is estimated to double by 2030 (3). As decreased dopamine levels debilitate normal motor function, people with PD tend to move 30–40% slower (bradykinesia) than healthy adults with a movement range that is often underscaled (4). The underscaled (hypometria) movement is characterized by decreased amplitude when compared to the movement of a neurologically healthy adult (i.e., shorter stride, smaller handwriting). This is also accompanied by an irregular pattern of force unfolding over time, with patients needing twice as much time to achieve peak force compared to healthy adults. All of these symptoms associated with the disease combine together to cause particular difficulties with performing everyday actions such as walking (characterized by a shuffling gait and difficulty with turning), dressing, handwriting (5), or using a computer mouse (6).

An important question is how well patients can adjust their movements to meet the spatio-temporal demands of a given task, in particular being at the right place at the right time. In the early 1960s, Draper and Johns (7) reported that people with PD have the same velocity for their movements regardless of movement amplitude. This was later questioned in a study by Teasdale et al. (8) and also by Mazzoni et al. (9) who both showed that although patients move slower, they can modify their movement times (MT) and maintain similar levels of accuracy to that observed in healthy adults. Furthermore, Sanes (10) noted that patients with PD struggle with high velocity movements over longer distances; with a notable breakdown in the ability to perform the task when the accuracy demands increase being observed. In a similar vein, Rand (11) reported particular difficulties in PD participants with the temporal aspects of motor control (the deceleration phase and higher movement variability) when a spatial accuracy constraint was introduced in a pointing task.

In general, healthy adults tend to move faster in order to intercept the target when the distance between the hand and object is greater (12). The velocity of every aiming movement is partially defined by Fitts' law that describes a movement based speed accuracy trade-off (13, 14). Interestingly patients do not show differences in the time they need to initiate movement when compared to healthy adults in reaction time aiming tasks (15). However, in reciprocal aiming longer MT and dwelling time were found for patients in late stage of PD compared to controls (16). Importantly, none of these studies considered the spatial and informational constraints of the task in the context of the temporal accuracy of the movement, something that is very important in sensorimotor synchronization.

Controlling a movement in many instances also requires that it is controlled with respect to external events in the future. A plethora of research suggests that dopamine plays an important part in temporal processing and prediction [see Ref. (17) for review]. Studies on rats demonstrate that lesions of the hippocampus result in increased dopamine release to the striatum, which disrupts timing in both second and minute scales (18). In humans, administration of a dopamine agonist (haloperidol) to healthy adults affects our ability to accurately discriminate temporal durations under 500 ms (19).

For decades, research based on the finger-tapping paradigm has provided strong evidence that patients are not able to accurately reproduce an isochronous interval (1, 20). Spencer and Ivry (2) have shed new light on those results in their study that uses both finger-tapping and circle-drawing. In both tasks, when patients executed movements in an intermittent, discontinuous manner (with a pause after each motor response), they performed as well as healthy adults. However, if patients with PD performed the same circle-drawing movement in a continuous manner, they exhibited increased temporal variability (2). This suggests that the timing of movements, which emerge and require intrinsic temporal control may be controlled differently from the timing of movements that are linked to a sensory event, such as coupling a movement to the sounding of a beat. These findings gave rise to the idea that event-based timing (moving to a beat) operates independently from basal ganglia structures and relies on preserved cerebellar functions instead (21). However, a report by Diedrichsen et al. (22) suggests that PD patients have difficulty with accurate synchronization, as a result of problems with error correction processes. In a similar vein, Grahn and Brett (23) reported impaired ability of PD patients to discriminate complex rhythmical structures. Other studies measuring perception of time intervals without using a motor response, i.e., where patients verbalize whether two time intervals are different or not, failed to find evidence for differences between people with PD and healthy adults (24). This contradictory evidence leaves many questions unanswered about the temporal control of the movement in PD.

In this study, we aimed to investigate whether people with PD are able to synchronize an aiming movement toward a target to the sounding of an external acoustic beat. We wanted to see how the spatial-temporal control of the movement is influenced by both the spatial demands of the task (e.g., cover the distance between two target zones) and the informational load of the task [i.e., the index of difficulty (ID) for accurate interception], as characterized by Fitts' law. Previous research also suggests that a 2.5-s inter-beat interval is the upper threshold for accurate timekeeping, with longer durations causing a breakdown in temporal control (25). To investigate this further with PD participants, we deliberately used longer inter-beat intervals (1.5 and 2.5 s) than the standard metronome frequencies used in finger-tapping studies [usually under 1 s; (1, 20, 22)]. Performance of PD patients was compared to the performance of a group of healthy adults (26). This group was deemed appropriate as a study conducted by Drewing (27) demonstrated no significant difference in synchronization ability (error correction of phase relation) across the life span. Furthermore, the temporal accuracy in synchronization was found to be stable from adolescence (approximately age 15) into old adulthood (59–88 years). In addition, Elliott et al. (28) demonstrated that elderly adults (63–80 years) matched young controls in their synchronization performance when moving in time with an isochronous metronome.

### MATERIALS AND METHODS

### Participants and Medical Assessment

Seven right-handed, idiopathic PD participants (one female and six males; average age *M* = 63.4, SD = 5.9) volunteered to participate in the experiment. All of the participants were righthand dominant and had normal or corrected to normal vision. Participants were recruited through the out-patient clinic at Belfast City Hospital. Ethical approval was granted by the Office for Research Ethics Committees Northern Ireland.

All participants were tested in the morning or early afternoon depending on their optimal functioning and mobility levels during the day. There was no change in their medication schedule to avoid any hazards in a non-medical setting (Psychology Laboratory, Queen's University of Belfast). Before the experiment began, the medical condition of each patient was assessed by a qualified PD nurse. The assessment comprised of: the Hoehn and Yahr scale (29), the Unified PD Rating Scale (30), an Objective Dyskinesia Rating Scale (31), and the Mini-Mental State Examination (MMSE) (32). The Hoehn and Yahr scale (H&Y) rating scale classifies severity of the disease starting from 0 (no signs of the disease), 1 – unilateral signs, 2 – mild stage with bilateral signs, 3 – moderate symptoms with postural instability, 4 – severe disability, ability to walk preserved, up to 5 – the most advanced stage, where patients are unable to move without assistance and are usually using a wheelchair (29). The Unified PD Rating Scale (UPDRS) is an additional scale, applied to measure disability and severity of the symptoms in PD patients, in both clinical and research settings (30). It consists of four parts, examining various spheres of functioning and well-being (Cognition, Mood, and Behavior; Activities of Daily Living; Motor Examination; Complication of disease and Therapies). In this study, patients were examined using all parts of the UPDRS. The higher the score in UPDRS, the more advanced the symptoms of the disease. In addition, the state of all patients was assessed by the nurse as either being "On or Off." "On" describes a phase of the day when the symptoms of the disease are partially alleviated by the dopaminergic treatment, as opposed to "Off " where they experience a "wearing off " of the medication's effect. All patients were assessed as being "On" before testing. Severity of dyskinesias was evaluated using the Objective Dyskinesia Rating Scale (31). Patients were asked to perform three motor tasks: lift a cup, put a coat on and walk. Their performance was assessed on a 5-point scale (0 – dyskinesia absent during the motor task, to 5 – violent dyskinesia debilitating motor task performance) and summed for all three tasks. Additionally, the MMSE was used to screen for cognitive impairment and confirm patients' ability to understand the task and consent to the study. The MMSE is a widely used scale measuring orientation to time and place, working and short-term memory and language (32). A score of over 25 points (with maximum of 30) is interpreted as normal (preserved cognitive functioning). Lower scores suggest increased probability of cognitive impairment or dementia of a mild level (21–24 points), moderate level (10–20 points), or severe level (below nine points) (33). Clinical features of patients are presented in **Table 1**.

Participants were of a varied medical state according to the PD motor disability scales (H&Y, UPDRS) and years of the disease. The participant coded as PD7 was at the most advanced stage of the disease (H&Y stage 4, UPDRS 89) and was the only patient who had a MMSE score 1 point below the norm (lower range of mild cognitive impairment). In this case written consent was taken from the spouse of the patient.

The control group consisted of eleven right-handed, healthy adults (4 females and 7 males; average age *M* = 24.8 years, SD = 2.5 years), who volunteered to participate in the experiment [same participants as in Ref. (26)]. All participants were right-hand dominant with normal or corrected to normal vision and no neurological history.

### Apparatus

The setup was identical to that outlined in the study of Bieńkiewicz et al. (26) with the only difference being two inter-beat durations used (1.5 and 2.5 s) instead of three. Inter-beat intervals were displayed using two 50 ms sine tones: "beep" (500 Hz) and "bop" (700 Hz) synthesized using Adobe Audition and timed using a MP3 player (providing 1 ms precision of replays). Signals were presented through noise-isolated headphones at fixed volume levels. To explore the effects of spatial accuracy on the temporal control of movement, four sets of spatial targets were presented on laminated A3 printout displays. The width and distance between targets varied across sets to control for the ID, represented by Fitt's law as:

$$\text{ID} = \log\_2 \text{(2D/W)}\tag{1}$$

where W is the width of the target and D is the distance between targets for one interceptive movement (13). Spatial conditions are summarized in the table below (**Table 2**). Participants wore a thimble on their index finger that had a reflective marker placed at the end. The thimble minimized the friction between the finger and the laminated targets and ensured that the finger could move easily back and forth between targets (see **Figure 1**). Motion data were recorded using 8 Oqus 300 cameras (Qualisys Motion Capture System) sampling at a frequency of 200 Hz with a spatial accuracy close to ±0.1 mm. Before the start of each experimental condition coordinates of the target zones were recorded to calibrate the motion capture data with respect to the target position. An analog input from the MP3 player was recorded through a Qualisys Analog Board to allow for the temporal alignment of the timing of the sound events and the movement data between targets.

### Procedure

Participants were seated comfortably at a desk and were asked to move their index finger back and forth between the two targets displayed in front of them in such a way that the arrival of the finger in the target zone coincided with the sounding of the beat. A total of 10 beep interceptions in the target on the left side and 10 bop interceptions in the target on the right side were recorded per condition. The recording began on the eleventh pointing movement and stopped after a further 20 movements. A relatively


*M, male; F, female; UPDRS, Unified PD Rating Scale; MMSE, The Mini-Mental State.*

low number of trials was selected to avoid fatigue in patients. Participants were not instructed on how fast they should move between the target zones. The order of the spatial conditions was randomized using the Latin squares method.

### Data Analysis

Data analysis and processing was consistent with our previous study (26). Selected aspects of temporal control were analyzed and included temporal variability (success rates, asynchrony, spread of error and central timekeeping variance), movement organization: including movement strategy, time and velocity. These measures allowed us to investigate in detail the spatio-temporal aspects of task performance and compare behavior between patients and controls used in a previous study. Positional data were filtered in MATLAB (34) using a second-order low-pass Butterworth filter at a frequency of 8 Hz from which the subsequent first derivative was

Table 2 | An overview of the target sizes and widths and corresponding IDs used in the experiment.


*Note how for the same amplitude of movement (distance between the targets) conditions differed in terms of ID as a result of the different target sizes. However, conditions I and III had the same ID but different distances between the targets [reproduced from Ref. (26)]. With kind permission from Springer Science+Business Media: Bien´kiewicz MMN, Rodger MWM, Craig CM. Timekeeping strategies operate independently from spatial and accuracy demands in beat-interception movements. Exp Brain Res (2012) 222:241–53. Table 1.*

taken for the velocity profile. Time stamps demarcating the end of the finger movement were computed as the first frame in which the velocity fell below 5% of peak velocity for each interceptive movement and where the finger was located within the boundaries of the target zone. We classified a trial as accurate if a participant stopped in the target zone within the temporal window of the sound event (50 ± 10 ms error). The spread of error measure was calculated as the absolute difference between the temporal errors (with relation to beat onset) made in consecutive trials. A detailed description of this measure is included in a previous study (26). For statistical analysis, mean values for each variable were calculated for each trial/participant and then analyzed using (2 × 4) Repeated Measures ANOVA, followed by *post hoc* comparisons.

### RESULTS

### Success Rates

The majority of participants found it challenging to synchronize their movement to the sounding of the beats. Those difficulties were expressed through early or late arrival in the target zones. In some trials, participants demonstrated a consistent pattern of initiating movement at beat onset instead of attempting to control their movements prospectively such that they would stop in the target zone at the same time as beat onset (see **Figures 2A,B**). Participant PD7 (with the most advanced PD symptoms in the tested sample) had to stop after just two trials as he reported that he could not get "into the swing" of the experiment and could not anticipate beat onset (see **Figure 2C**).

As the distance between the two target zones determined movement amplitude, temporal accuracy could be explored by manipulating the spatial requirements of the movement needed to complete the task. We found that PD patients were significantly

strategies operate independently from spatial and accuracy demands in beat-interception movements. Exp Brain Res (2012) 222:241–53. Figure 1.

Figure 2 | Illustrations of the irregular synchronization patterns observed in patients at different stages of the disease compared to performance (A) of a healthy adult – PD2 [H&Y 1.5–2 (B)] and PD7 [H&Y 4 (C)]. (B) shows the trajectory of the movement of patient PD2 who treated the sound as a cue for starting to move instead of a signal to stop. Patient PD7, at a more advanced stage of the disease (C) demonstrated a striking inability to synchronize the movement to the beats. Depicted trials were for the 2.5 s condition and ID3, 20 cm distance. Shaded rectangles denote target zones. Black musical denotes demarcate the onset of the beat.

#### Table 3 | Percentage of high accuracy trials within the temporal window of 70 ms (50 **±** 10 ms).


*A summary of the percentages of high accuracy trials for each participant compared to the average healthy adult performance (n* = *11; bottom line). Participants are listed according to their UPDRS rating (from the lowest score – to the highest). Participants PD7 was not able to proceed with the rest of the trials. Note how with longer interval duration – 2.5 s, patients had lower success rates than the healthy adults\* [sample from Ref. (26)].*

less successful than the controls at synchronizing their movements to the onset of the beats [Repeated Measures ANOVA Between-Subject Effect *F*(1,13) = 55.002, *p* < 0.001, η<sup>2</sup> = 0.786] [see **Table 3** and Bieńkiewicz et al. (26)], but kept their movement spatially accurate by reaching target zones.

Problems with synchronization, revealed by poor success rates in temporal accuracy, were further investigated by analyzing synchronization errors. Participants struggling to synchronize with the beat could either arrive in the target zone too early, underestimating the duration of the inter-beat interval, or too late, overestimating the duration of the inter-beat interval. Therefore, if they arrived and stopped in the target zone more than 10 ms before the sound event (less than −10 ms), the trial was classified as negatively asynchronous. If they arrived and stopped in the target zone 10 ms after the occurrence of the beat (i.e., >60 ms – duration of the sound 50 + 10 ms error) the trial was classified as positively asynchronous. The cut-off for temporal accuracy was set arbitrarily based on previous literature (25).

### Negative Asynchrony

Arriving and stopping in the target zone before the sound event implies that the participant's could not accurately anticipate when the sound event was going to happen. Our tested sample of PD patients had 40% more negative asynchronies than a group of healthy controls Bieńkiewicz et al. (26) (see **Table 4**).

The difference between patients and healthy controls was statistically significant [between subjects effects repeated measures ANOVA of ratio of negative asynchronies on two time intervals – *F*(1,15) = 14,434, *p* = 0.002, η<sup>2</sup> = 0.490 – patients (time interval 1.5 s, *M* = 0.47, SD = 0.33, 2.5 s, *M* = 0.58, SD = 0.38) vs. healthy controls (time interval 1.5 s, *M* = 0.06, SD = 0.04, 2.5 s, *M* = 0.20, SD = 0.12)]. However, the differences between the 1.5 and 2.5 durations in the patient group were not statistically significant, *p* > 0.05.

In the patient group, we observed a significantly higher frequency of negative asynchrony errors over 300 ms before the occurrence of the beat in both interval durations than in healthy controls [Repeated Measures ANOVA Between Subjects Effects – patients vs. healthy controls on the ratio of negative asynchronies over −300 ms for two interval durations – *F*(1,14) = 5.498, *p* = 0.034, η<sup>2</sup> = 0.282)]. An increase in the magnitude of negative asynchronies for longer interval durations supports our hypothesis that temporal control is compromised for longer interval durations.

### Positive Asynchrony

Arriving and stopping in the target zone after the sound event suggests that the inter-beat interval representation was too long and compromised the temporal control of the movement. Patients made fewer positive asynchronies than what was found with healthy controls in our previous study – [Repeated Measures ANOVA Between Subjects Effects *F*(1,15) = 21.43 *p* < 0.001, η<sup>2</sup> = 0.588 – patients (time interval 1.5 s, *M* = 0.47, SD = 0.31, 2.5 s, *M* = 0.23, SD = 0.28) vs. healthy controls (time

interval 1.5 s, *M* = 0.84, SD = 0.08, 2.5 s, *M* = 0.67, SD = 0.14) on the ratio of positive asynchronies per condition]. Nonetheless, at the more advanced stages of the disease we observed a tendency toward an increase in the ratio of positive asynchronies (see **Table 5**).

The profiling of positive asynchronies revealed significantly higher values of error in the patient group – [Repeated Measures ANOVA Between Subjects Effects *F*(1,15) = 5.53 *p* = 0.03, η<sup>2</sup> = 0.269 – patients (time interval 1.5 s, *M* = 0.38, SD = 0.37, 2.5 s, *M* = 0.51, SD = 0.31) vs. healthy controls (time interval 1.5 s, *M* = 0.11, SD = 0.09, 2.5 s, *M* = 0.27, SD = 0.15) on the ratio of positive asynchronies over 350 ms per temporal condition]. In other words, if patients overestimated the duration of the interval the magnitude of their error (in ms) was significantly greater than that of healthy adults. There was no difference in the patient group between the 1.5 and the 2.5 s interval durations and the ratio of positive asynchronies over 350 ms after the occurrence of the beat, *p* > 0.05.

### Temporal Variability

In Ref. (26), we demonstrated that the spread of error is a novel and robust measure of temporal variability for this task. In this study, we expected to find a larger spread of error for PD patients, showing a breakdown in their temporal control when compared to that of healthy adults. In healthy adults (26), we observed a pattern of increased temporal variability when synchronizing with longer interval durations compared to shorter 1.5 s durations. PD patients had significantly higher values in the spread of error than healthy adults [Repeated Measures ANOVA *F*(1,15) = 4.743, *p* = 0.003, η<sup>2</sup> = 0.444]. We found a trend toward a higher spread of error values with more advanced stages of the disease (**Figures 3** and **4**).

The UPDRS total score strongly correlated with the measure of spread of error for the 1.5 s interval Spearman's one-tailed ρ= 0.85, *p*= 0.007, and with the 2.5 s interval (ρ= 0.67, *p*= 0.05). **Figure 4A** illustrates the spread of error for a patient at a mild stage of the disease. For almost all of the display, the spread of error is higher than the average found for healthy controls, oscillating between


*Percentages were calculated as the percentage of trials classified as negatively asynchronous with the inter-beat interval in each condition. In comparison to healthy adults\* (n* = *11) (26) patients had a higher ratio of negative asynchronies. Participants are listed according to their UPDRS rating (from the lowest score – to the highest).*

#### Table 5 | Summary table of positive asynchronies made by patients.

#### Interval duration


*Percentages were calculated as the percentages of trials classified as an overestimation of inter-beat intervals in each condition. In comparison to healthy adults\* [bottom* 

*row – sample taken from Ref. (26)] patients had a lower ratio of positive asynchronies. Participants are listed according to their UPDRS rating (from the lowest score – to the highest). Note how patients with more advanced stage PD have higher ratios of positive asynchronies.*

100 and 300 ms. With the severity of the disease (See **Figure 4B**) we observe an increase in the spread of error from 150 to 450 ms, where the increased variability in synchronization errors means greater difficulty with task performance. Interestingly, no effect of the informational load and/or the amplitude of movement was found in the spread of error, *p* > 0.05 (See **Figure 4C**). Although we expected that for more challenging spatial conditions (i.e., higher IDs), we would observe a larger spread of error, this was not found to be the case. The increase in the spread of error in PD was attributed to the temporal demands of the task.

### Movement Strategies

Preliminary analyses of MT revealed two different strategies of timekeeping that were consistent with our findings from the study on healthy controls. Our assumption that patients might reveal different movement strategies compared to those of healthy adults was not upheld. Three patients adopted a MT strategy and adjusted MT to the interval duration, while the other three varied their Waiting Time (WT strategy) in the target zones according to the interval duration. Patients in both groups were at varying stages of the disease. We identified MT in PD3, PD5, and PD6; and WT in PD1, PD2, and PD4. We were unable to identify a pattern of movement strategy in PD 7 – our most advanced patient.

**Figure 5** illustrates the differences in mean MT for the different interval durations in the MT group. Adjusting MT to the duration of the interval suggests that, similar to healthy adults, patients filled the inter-beat interval with movement. The MT was adjusted to the interval duration, and not the required movement amplitudes or the informational loads of the task (different IDs). MT was found to be significantly different for the two interval durations – 1.5 s (*M* = 1.30 s; SD = 0.05) and 2.5 s (*M* = 2.20 s, SD = 0.11) [repeated Measures ANOVA Effect on Interval Duration on MT *F*(1,2) = 218.95, *p* = 0.005, η<sup>2</sup> = 0.99]. Patients did not show any differences in the mean MT when compared to healthy controls who used the same strategy (*M* = 1.23 s, SD = 0.02; *M* = 2.12 s, SD = 0.07 respectively to the interval duration) [between subjects effects repeated measures ANOVA – group healthy controls vs. patients *p* > 0.05]. No effect of IDs or movement amplitude was observed on the mean MT, *p* > 0.05.

The second group of patients was found to scale their waiting times to the duration of the interval (see **Figure 6**). Waiting time was significantly longer for interval duration 2.5 s compared to 1.5 s [repeated measures ANOVA, *F*(1,2) = 56.99, Pillai's Trace = 0.96, *p* = 0.02, η<sup>2</sup> = 0.96]. The time of the inter-beat interval was filled by both waiting in the target zone and then moving toward the other target on the opposite side. Again, no effect of ID or movement amplitude was observed on the mean waiting times, *p* > 0.05.

There was no difference between the means of the waiting times for the different conditions between the patients and the healthy control group [1.5-s interval PD group *M* = 0.94 s, SD = 0.18 s, healthy controls – *M* = 0.75, SD = 0.06; 2.5 s interval *M* = 1.91 s, SD = 0.18, healthy controls respectively *M* = 1.76, SD = 0.02, Between Subjects Effects – group Repeated Measures ANOVA *p* > 0.05]. Again, the lack of any differences between patients and healthy controls suggest that their ability to use waiting time or MT strategies to fill time between the inter-beat intervals is preserved and cannot explain their difficulties with synchronization to the beat.

### Movement Time and Velocity

In the previous section, we demonstrated that patients (strategy MT) maintained constant MT across a given interval duration. Our results show that to achieve this across different spatial and informational constraints of the task, they varied the speed of their movement across conditions (**Figure 7**). We observed a significant effect of interval duration on the peak velocity [repeated measures ANOVA *F*(1,2) = 22.74, *p* = 0.04, η<sup>2</sup> = 0.92] and spatial and informational constraints of the movement [*F*(3,6) = 16.26, *p* = 0.003, η<sup>2</sup> = 0.89]. Bonferroni pairwise comparisons revealed the most pronounced differences (*p* = 0.05) between the speed of the movement within the same level of difficulty – ID3 and different amplitudes of the movement – 10 and 20 cm. In comparison to healthy controls,

patients moved slower in both the 1.5-s interval (23% slower), and the 2.5-s interval (15% slower) conditions, yet these differences did not reach statistical significance [*p* > 0.05 for Between Subjects Effects Repeated Measures ANOVA group vs. mean peak velocity]. Patients adjusted the speed of their movement to meet the demands of the task and despite moving more slowly than healthy adults; those differences did not reach statistical significance.

### Central Timekeeping Variance

In our study in healthy controls (26), we found an increase in central timekeeping variance [calculated with Vorberg and Wing (35) model for synchronous finger tapping] with longer interval durations (1.5 vs. 2.5 s and 3.5 s), only for those participants who used their MT as a timekeeping tool. Healthy controls who kept their MT constant and adjusted waiting time in the target zones (strategy WT), did not show an increase in central timekeeping variance.

In this study, we found an identical increase in central timekeeping variation with the longer interval duration for PD patients who also used a MT strategy. However, the most interesting finding was that timekeeping variance was five times higher for the 1.5-s interval and three times higher for the 2.5-s interval in the patient group compared to that found for healthy controls (see **Figure 8**).

The effect of the group on central timekeeping variance was found to be significant [Repeated Measures ANOVA for Between Subjects Effects – across two time intervals conditions *F*(1,8) = 11.4, *p* = 0.01, η<sup>2</sup> = 0.58, patients (time interval 1.5 s, *M* = 0.048, SD = 0.03, 2.5 s, *M* = 0.119, SD = 0.09) vs. healthy controls (time interval 1.5 s, *M* = 0.009, SD = 0.01, 2.5 s, *M* = 0.039, SD = 0.03)].

The difference between central timekeeping variance for the two interval durations in the patient group did not reach statistical significance, *p* > 0.05. A comparison of patients using the waiting time (WT) strategy revealed that patients had higher values of central timekeeping variance than healthy adults, but the difference was not significant [repeated measures ANOVA for between subjects effects, *p* > 0.05]. A comparison of motor variance between two interval durations did not show significant differences (1.5 s *M* = 0.05, SD = 0.06, 2.5 s *M* = 0.03, SD = 0.03) [repeated measures ANOVA *p* > 0.05]. The Lag 1 values were in the limits of (−0.5 to 0) as posited by the original Wing–Kristofferson model (duration 1.5 s, *M* = −0.03, SD = 0.03, 2.5 s *M* = −0.07, SD = 0.12). The average correction parameter α was equal to 1.17, representing the proportion of the asynchrony corrected on the adjacent motor responses (*t*-tests did not reveal any differences across two interval durations with the healthy controls sample, *p* > 0.05). Therefore an increase in variability can only be attributed to

central timekeeping processes excluding factors involving the execution of the movement.

### DISCUSSION OF THE EXPERIMENTAL FINDINGS

### Main Findings

Our primary aim was to explore how PD patients deal with a synchronization task. Using measures of temporal accuracy, we found that with the severity of the disease, the ability to accurately synchronize movement to a sound event decreased. PD patients made greater synchronization errors when compared to healthy adult controls and their synchronization performance was more variable as measured by the spread of errors (26). In addition, patients demonstrated a tendency to underestimate the duration of the interval, rather than overestimate it, as was the case with a group of healthy adult controls. Underestimating the duration means that they tended to arrive in the target zone before the occurrence of the beat. Those results are in line with reports from other finger-tapping studies, which also showed that patients underestimate the interval duration and present a higher magnitude of temporal errors (20, 22, 36).

Difficulty with synchronization was more pronounced for the longer interval duration (2.5 s). The patient with the most advanced UPDRS score (89) was not able to coordinate movement with respect to the sound events, and as a result, the testing session was stopped after only two experimental conditions. This patient reported that it was too difficult to get into the "swing" of the movement, thus implying an inability to produce a rhythmical movement.

The second aim of our study was to explore temporal aspects of movement organization and compare findings to previous research. We wanted to investigate whether patients are able to adjust their movement to different informational, spatial and temporal demands of the task. We expected patients might organize their movement in a different way to healthy adults. Instead, all patients showed the same movement strategies as healthy adults and adjusted MT to the duration of the interval or waited in the target zone for a period of time that matched the inter-beat interval duration. In this respect, MT or waiting

time was used as a tool for timekeeping. We expected that PD patients will have prolonged movement and waiting time as demonstrated by Onla-or and Winstein (16), but the difference with healthy adults was statistically not significant. We also did not find evidence that PD patients have difficulty with adjusting their movement velocity to the amplitude of the movement or target size. Patients modulated their movement velocity to meet the spatial and informational constraints of the task in the same way as healthy controls did. In line with previous research (9), we also observed that patients moved with lower velocities when compared to healthy adults, although those differences did not reach statistical significance. Neither informational load of the task, nor the amplitude of the movement had an impact on the successful performance of the task. We expected that patients might show increased difficulty when moving toward a display with a high informational load (ID = 4, two 2.5 cm × 4 cm (W × H) targets placed in 20 cm distance), but this was not the case. The informational load of the task, in line with Fitts' speedaccuracy trade off, had an impact on the velocity of the movement between the target zones. Lower IDs triggered faster movement between target zones while higher IDs (smaller targets) forced participants to move slower between targets even though they were the same distance apart. We imposed a movement range on patients by using pre-designated target zones for intercepting the sound event. Patients did not show differences in the movement amplitude with regard to different interval durations nor did their performance differ from that found for healthy adults.

The third aim of this study was to model the difficulties with the timing of the movement. When temporal variance of the movement was split into the two components of the Vorberg and Wing (35) model of synchronization timekeeping, we found that patients had significantly higher central variance when compared to controls (five times higher for the interval duration of 1.5 s and three times higher for the interval duration of 2.5 s). This increased variability with longer interval durations is characterized by patients adjusting their movements to the interval duration. Increased variability of central timekeeping was previously reported in studies using the finger-tapping paradigm (1, 20, 37). Similarly to what our results show, the severity of the disease led to greater variability in interval reproduction (22, 38). We did not, however, observe an increase in motor variance with longer interval durations, confirming that increased variability can only be attributed to timekeeping mechanisms.

An unexpected finding was that patients at advanced stages of the disease tended not to synchronize their movement so that it ended with the beat, but instead, used the beat as a trigger to start the movement. Each time the experimenter observed this, patients were reminded that this was not the purpose of the task and were encouraged to synchronize their movement so that it ended with sound events. In PD, coupling a movement onto an external temporal framework (such as a metronome) is known to improve movement, such as gait, and is referred to in the literature as cueing (39–45). However, we explicitly asked patients to stop at the occurrence of the sound event, not initiate their movement. Our interpretation of this behavior is that as patients suffer from difficulties with movement initiation they used the beat as a cue to start moving. Patients claimed they switched to synchronizing the start of the movement with the beat unconsciously.

### Interpretation of Findings

Our results suggest that degeneration of basal ganglia circuitry might undermine the temporal prediction ability i.e., anticipating when something is going to happen in the near future. Those difficulties seem to be independent of the range of movement and increase with stretching the time interval before the occurrence of the event. To our knowledge, this is the first time that a study has addressed sensorimotor synchronization in PD patients in the context of a beat interception task based on aiming movement. Previous work by Diedrichsen et al. (22) has demonstrated a decreased ability of patients' to synchronize finger-tapping movements to a metronome. Our results do not support the dissociation of timing proposed by Spencer and Ivry (2). Our task was event-based, yet patients demonstrated an increase in variability in both synchronization errors and timekeeping when compared to healthy adults. Spencer and Ivry (2) argued that the basal ganglia plays a minimal role in event-based timing, but is highly involved in continuous timing. Our results agree with Wing (46) proposal that both structures like the cerebellum and basal ganglia might play an important role in motor timing processes. Other authors also posited that the basal ganglia are involved, not only in the generation of "internal beats" as suggested by fingertapping studies (47, 48), but also in the perception of complex rhythm (23). Activation of the basal ganglia during discrete timing tasks has also been reported in neuroimaging studies (fMRI), but in contrast to Spencer and Ivry (2) findings, this has not been found during temporal prediction tasks that involve continuous timing (49). Diedrichsen et al. (22) posited that basal ganglia structures are involved in error correction processes. We did not find significant differences between the error correction parameter in the group of patients and healthy adults using the Vorberg and Wing (35). Therefore, the role of the basal ganglia in the temporal control of movement and consequences of the neurodegeneration in PD on timing remains open to debate.

It is also important to note that we did not manipulate in any way the patients' medication schedule. All of our patients who participated were on normal doses of their medication. Previous research shows that patients show higher variability in timing when tested 24 h after a break in their medication schedule (20). Deprivation of dopamine supplementation may compromise basal ganglia function to an even greater extent. In spite of a regular medication scheme, our study showed that patients with PD performed significantly worse in our synchronization task compared to healthy adult controls. Many other studies exploring event-based timing in PD were based on testing patients on medication (22, 47, 48, 50). Testing patients "off " medication would allow us to further explore the pattern of synchronization difficulties in PD. This however, would have to be conducted in a clinically supervised setting.

There are two major limitations to this study. First of all, we tested a relatively small sample of patients, which does not allow us to draw final conclusions about the causal relationship between the severity of the disease and ability to synchronize. We aim to replicate this study with a large sample of patients to validate reported findings here. Second, a limitation is the comparison of PD performance to the group of young, healthy controls, not to elderly matched controls. Although there is a convincing evidence that ability to synchronize is preserved and not significantly different in healthy aging (28, 51), we aim to include elderly controls in the replication of this study. We would suggest that a variation of this type of task could be employed by healthcare practitioners to monitor the severity of PD symptoms and be one of a large number more objective behavioral markers that look at disease progression.

## CONCLUSION

We found preliminary evidence that patients suffer from specific difficulties with event-based timing, namely synchronization with an external acoustic beat. This type of task requires prospective motor control (i.e., coupling movement to neural based dynamic information that helps anticipate when the beat is going to sound) and also efficient error correction processes that help tune the unfolding movement so it is in synchrony with the sounding of the next beat. By imposing a range of movement for participants, we have employed a different experimental paradigm to that used in other discrete timing experiments [e.g., Ref. (20, 22, 36, 52, 53)]. Although all of the participants moved more slowly than the healthy adult controls, they did tend to use similar strategies when performing the task and also showed that movement amplitude remained uncompromised. This enabled us to explore how temporal control varied within the controlled spatial parameters of the movement. Indeed, we found that a decrease in the temporal control of the movement seems to be independent of a decreased scaling of the movement as observed in bradykinesia, or information load of the task, but links to impaired ability to predict when something is going to happen.

### AUTHOR NOTES

Correspondence concerning this paper should be addressed to MB, UMR 7287 CNRS & Université d'Aix-Marseille, Faculté des Sciences du Sport, CP 910, 163, av. de Luminy F-13288 Marseille Cedex 09, France.

### ACKNOWLEDGMENTS

Authors would like thank: Ms. Sarah Mason and Dr. David Craig for conducting a medical assessment of the patients and help with recruitment in Belfast City Hospital; Dr Mark T. Elliot, SyMoN Lab, School of Psychology, University of Birmingham, UK for providing MatTap toolbox and PD participants who volunteered to take part in the study. This research was funded by the European Research Council TEMPUS\_G project (210007 StIG).

### REFERENCES


in individuals with Parkinson's disease. *Neuropsychology* (1994) **8**(2):218–26. doi:10.1037/0894-4105.8.2.218


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Bieńkiewicz and Craig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Regional Interplay for Temporal Processing in Parkinson's Disease: Possibilities and Challenges

*Michael Schwartze1,2\* and Sonja A. Kotz1,2*

*1Department of Neuropsychology and Psychopharmacology, Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, Netherlands, 2Department of Neuropsychology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany*

Parkinson's disease (PD) is primarily associated with two dominant features: cardinal motor symptoms and the loss of cells in the substantia nigra pars compacta of the basal ganglia. Consequently, these aspects are major foci in PD-related research. However, PD is a neurodegenerative disease, which progressively affects multiple brain regions outside the basal ganglia and leads to symptoms outside the motor domain. Much less is known about the individual contribution of these secondary regions, their interplay and interaction with the basal ganglia, and the respective network dynamics in the overall manifestation of PD. These regions include classical motor structures such as the cerebellum and the supplementary motor area (SMA). However, just as the basal ganglia, these regions display a fine-grained microarchitecture, which supports sensory and sensorimotor functions. One such function is temporal processing, which has been ascribed to a network comprising all of these regions. On the one hand, pathological changes in this temporal processing network may be part and parcel of motor and non-motor symptoms in PD. On the other hand, a better understanding of the role of each network node may offer a novel perspective on compensatory mechanisms, therapeutic interventions, as well as the heterogeneity and individual differences associated with PD. We unfold this perspective by relating the neural foundations and functional implications of temporal processing to pathophysiological and neurofunctional changes characteristic of PD.

### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

### *Reviewed by:*

*Katja Kornysheva, University College London, UK Deborah Lynn Harrington, VA San Diego Healthcare System, USA*

*\*Correspondence:*

*Michael Schwartze michael.schwartze@ maastrichtuniversity.nl*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 19 October 2015 Accepted: 21 December 2015 Published: 18 January 2016*

#### *Citation:*

*Schwartze M and Kotz SA (2016) Regional Interplay for Temporal Processing in Parkinson's Disease: Possibilities and Challenges. Front. Neurol. 6:270. doi: 10.3389/fneur.2015.00270*

Keywords: Parkinson's disease, temporal processing, timing, compensation, network

## INTRODUCTION

The basal ganglia system is a complex of several nuclei deeply embedded into the vertebrate brain. It has extensive connections to other subcortical and cortical regions, and via these connections, the basal ganglia system contributes to a wide range of motor and non-motor behaviors (1–5). One of the most fundamental functions attributed to the basal ganglia and associated brain regions is the patterning and chunking of behavior from simple motion sequences to complex cognitive sequences (6, 7). Motor patterns are expressed in physical motion, whereas cognitive patterns are not. They may be conceived as routines or "habits of thought," reflecting a dominant processing mode, which emerges from the repetitive future-oriented sequencing of cognitive actions (6). These mechanisms reflect a common theme across behavioral domains, and likewise, a general basal ganglia function,

which may guide both the iterative build-up and the evaluation of repetitive motor and cognitive behaviors (8).

In this context, the term "sequencing" may be broadly characterized as a function that pertains to the sequential order of motor and cognitive actions as well as to the specific temporal structure of these sequences. In other words, sequencing not only defines the order of successive actions in time but also their specific temporal relations. Motor and cognitive patterns are typically expressed as entire sequences of actions, i.e., they are "packaged as a unit ready for expression," and monkey research indicates that accented neural activity marks the boundaries of such packages (8, 9). These accentuations introduce a temporal non-linearity, as "time can be decoded with higher resolution at the beginning and the end of the movement sequences than during them" (8). Considering that motor and cognitive patterns may not only be expressed in isolation, the continuous chunking of actions may effectively give rise to a combination of slower and faster dynamics in ongoing sequential behavior. Although it may be difficult to explicitly decode the temporal fine structure of actions that constitute the fast dynamics, it may be possible and functionally relevant to track the temporal structure of accented boundary markers, which constitute the slow dynamics. It has been suggested that the alternation of accentuation and de-accentuation reflects low attentional demands required to process actions between the markers (8). The questions arise if such temporal non-linearities have a perceptual equivalent, and if they can be exploited to optimize the allocation of attention in time for both, the production and the perception sequential behavior (10).

Habitual behaviors unfold in the hundreds-of-milliseconds to seconds range that is central to research into interval timing, another fundamental behavioral function that has been ascribed to the basal ganglia system, and a wider network of sensorimotor regions, which includes the supplementary motor area (SMA) and the cerebellum (11–13). In analogy to the above terminology, an interval may be defined as the temporal quantity between two successive markers. Interval timing tasks indicate that dopaminergic neurons show activity patterns, which consist of a burst at the beginning of a trial and a second burst at the expected time of reward, with sustained activity throughout the interval (14). Interval timing is typically conceptualized as a general activity that is inherent to the production and the perception of temporally structured behavior. As such, interval-based temporal processing is a crucial component of non-motor and motor aspects of activities as diverse as walking, speaking, or playing music, and, conversely the perception and the evaluation of the temporal structure that arises from the same activities (14). Due to the fundamental nature of basal ganglia contributions to sequencing and interval timing, the conceptual and structural overlap between these functions has widespread implications, especially if the patterning and chunking of motor and cognitive action sequences and temporal processing in cortico-striatal circuits reflect aspects of an even more general sensorimotor sequencing capacity, which guides production and perception. In the following, we will reflect on some of the potential consequences of this overlap in Parkinson's disease (PD), focusing on the role of the basal ganglia and its interaction with associated temporal processing regions in this particular context.

### PARKINSON'S DISEASE AND (DYSFUNCTIONAL) TEMPORAL PROCESSING

Impaired sequencing of motor actions is one of the hallmarks of PD, a neurodegenerative disease that leads to a loss of dopamine releasing neurons in the substantia nigra pars compacta of the basal ganglia. PD is commonly diagnosed on the basis of these primary motor symptoms, most characteristically the slowing of movements, rigidity, and resting tremor. However, PD is also a progressive disease, and these characteristic motor symptoms are asymmetric and may surface only after internal mechanisms fail to compensate for the impact of the disease (15). Although motor symptoms are most striking, PD typically also has a deteriorating impact on numerous cognitive functions that is apparent in the early "pre-motor" phase of the disease [for a recent review see Ref. (16)]. This most likely reflects the extensive structural and functional connectivity of the basal ganglia system. Consequently, non-motor symptoms can precede motor symptoms in early non-medicated patients (17).

In line with the rationale discussed above, damage to the basal ganglia system should also lead to impaired temporal processing as well as a temporally specific dysfunction of sequencing behavior. Moreover, the latter may be a consequence of the former as impaired temporal processing may factor into other motor and non-motor cognitive aspects of the disease, expressed in suboptimal timing in production and perception. The structural and functional characteristics and the interplay of brain regions that engage in temporal processing may therefore provide a novel perspective on some of the dynamic pathological and compensatory changes in regions such as the cerebellum and the SMA. Such changes have been observed in the progression of the disease but their specific role in the pathogenesis of PD remains unclear.

Seminal work in the temporal processing domain has established that the cerebellum is involved in precise and automatic discrete event-based (salient-feature) temporal processing, whereas the basal ganglia and associated cortico-striato-thalamocortical circuits engage in attention-dependent interval-based (continuous-event) temporal processing, thus creating an explicit representation of the temporal relation between successive events (12, 14, 18). However, apart from these primarily discussed systems, temporal processing most consistently activates the SMA and prefrontal regions (19). At least three important aspects should be emphasized: (i) these regions can engage in temporal processing also when very little or even no movement or movement preparation is involved (18), (ii) they can interact across different timescales, thereby potentially forming an integrative subcortico-cortical temporal processing network (10, 20), and, (iii) the same regions are affected by PD and change their activation patterns during the progression of the disease.

Parkinson's disease patients display compromised performance in various temporal processing tasks spanning production and perception, which has been attributed to various components of a dysfunctional "internal clock" mechanism (21–23). Some aspects of this performance are reminiscent of primary motor symptoms. For example, PD patients tend to speed up during repetitive self-paced finger-tapping tasks, comparable to the typical phenomenon of gait festination (22, 24, 25). However, there is a considerable degree of heterogeneity in many results, which may reflect specific task characteristics, different stages of the disease, patient subgroups, as well as the differential engagement of compensatory mechanisms (26–28). With respect to compensation, it may be relevant to consider the relation of action selection, ordering, and implementation as separate from the specific temporal structuring of these processes as independent components of a general behavioral sequencing capacity. Thus, a better understanding of the role of other brain regions in temporal processing, most critically the cerebellum and the SMA, may pave the way to personalized and specifically targeted manipulations of temporal structure.

### HYPO- AND HYPERACTIVITY: DYSFUNCTION OR COMPENSATION?

Post-mortem PD brain tissue analyses have revealed a selective loss of pyramidal neurons in the pre-SMA (29). However, neuroimaging studies that explicitly targeted the role of associated regions such as the cerebellum and the SMA in PD and also applied typical sequential temporal processing tasks are relatively rare. The existing evidence confirms complex patterns of interactions that change in the progression of the disease. Early studies observed that cerebellar hypoactivity in synchronization and continuation phases of a finger-tapping task at a base tempo of 600 ms that was partially normalized by medication (24). There are reports of hyperactivation of the pre-SMA in *de novo* PD patients relative to healthy controls, most likely reflecting the contribution of the pre-SMA to the temporal sequencing of self-initiated opening and closing movements of the (right) hand at approximately every 1000 ms (30). In the same study, this finding was paralleled by bilateral hyperactivation of the superior cerebellum (mainly ipsilateral) and hypoactivation of the ipsilateral inferior cerebellum. The authors hypothesized that these changes in the cerebellar activation pattern may reflect compensation for a dysfunctional cortico-striatal motor loop. Similarly, PD patients off levodopa medication showed hyperactivation of the left cerebellum and the SMA during a synchronization-continuation finger-tapping task using a base tempo of 750 ms. These results were also interpreted as an indication of compensation for cortico-striatal dysfunction (31). However, in another group of patients off medication, patterns of concurrent hypoactivation of the basal ganglia and the pre-SMA were observed during the performance of sequential hand movements (32), as well as of the basal ganglia and the cerebellum during a target interception task that required predictive motor timing (33). Other studies have shown hypoactivation of the pre-SMA and the caudal portion of the SMA paralleled by hyperactivation of the ipsilateral cerebellum in patients performing repetitive paced button presses (34), as well as an augmented recruitment of cerebello-thalamo-cortical circuits in PD patients performing continued finger tapping following a pacing sequence with a base tempo of 500 ms (35). While these examples are clearly selective, they serve to illustrate that the experimental tasks applied in PD research vary widely. The heterogeneity of the results reflects this variability, thus highlighting the need for individualized approaches, potentially on the basis of individual temporal processing profiles.

### STRATEGIES FOR INTERVENTION

Therefore, one goal could be to improve the efficiency and usability of interventions relying on rhythmic auditory cueing in order to provide patients with wearable devices that are customized to meet their individual therapeutic needs [for recent reviews, see Ref. (36–38)]. For example, a clearly distinguishable event structure in the hundreds-of-milliseconds range should be optimally suited to trigger automatic salient-feature cerebellar temporal processing mechanisms. Subliminal changes in this event structure could be used to compensate for impaired sequencing abilities due to basal ganglia pathology by pushing the dysfunctional system toward a more stable state over an extended period of time, e.g., by working against the tendency to accelerate movement rates. This strategy may assist or even circumvent the impaired build-up and evaluation of repetitive motor and cognitive action sequences in patients by assigning part of the sequencing task to potentially less affected brain regions such as the cerebellum. Accordingly, this perspective may shift the focus from basal ganglia pathology to the function of a distributed system, in which basal ganglia contributions to sequencing and temporal processing have to be interpreted relative to the function of the other network nodes. A potential starting point in this endeavor may be to obtain individual measures of basic temporal processing capacities such as "spontaneous motor tempo" (the ability to generate a temporally regular sequence of events) or "preferred perceptual tempo" (the preferred tempo of sensory events), which have been found to be correlated and linked to the ability to exploit temporal structure in a sensory task (39, 40). If combined, such basic measures may be indicative of some of the characteristics of the temporal processing network, i.e., the dysfunctional "internal clock" of a patient.

Due to the focus on the basal ganglia, cerebellar contributions to pathological and compensatory mechanisms are often overlooked (15). For example, although resting tremor is perhaps the most characteristic of the Parkinsonian symptoms, its origins are unknown. Research into this phenomenon has targeted multiple structures in the basal ganglia system, but it is of note that the tremor can be abolished via stimulation of thalamic targets of cerebellar output, which renders the cerebellum a potential source of the tremor-inducing pathological oscillations (15, 41, 42). Speculations of this kind also have to consider the progressive character of the disease, which may not only manifest in dynamic network changes but may also affect the causality underlying these assumptions. Overall, cerebellar compensatory mechanisms may be most efficient during the early stages of the disease, but they may fail once the pathological changes become more severe (15). However, the partial neglect of cerebellar contributions to PD also stands in contrast to the substantial evidence for reciprocal short-latency direct connections next to cortically mediated connections between the two systems (42–46). These structural connections suggest a tight coupling between the cerebellum and the basal ganglia and associated regions but their functional significance remains unclear. One candidate of particular interest in the context of sequencing and interval timing may be the cerebellar triggering of dopaminergic activity in prefrontal areas or the ventral tegmental area through projections via the thalamus that marks the beginning of a trial in interval timing tasks (11, 14, 47).

In addition to more systematic investigations of the cerebellum and the SMA in PD by means of temporal processing tasks that are known to activate these regions, the differentiation of their primary or compensatory engagement may be addressed on the basis of recent evidence for a more fine-grained structural differentiation of cerebellar, SMA, and also basal ganglia subregions. On the one hand, the distinct functional connectivity patterns revealed in this context may provide the opportunity to dissociate between primary pathological and secondary compensatory aspects and allow further exploration of specific network functions (5, 44, 48). On the other hand, these findings may be used to refine existing approaches, including the application of neurostimulation techniques such as repetitive transcranial magnetic stimulation (rTMS). rTMS has been used to target particular PD symptoms, patterns of hypoactivity and hyperactivity in specific regions, or fluctuations in behavior associated with long-term drug administration (49–53). For example, lateralization of the SMA and its differentiation into rostral and caudal subregions are reflected by different aspects of temporal processing tasks such as the temporal range or the engagement of sensorimotor as opposed to sensory processing (54, 55). Depending on the type and the temporal structure of the behavior of interest, such dissociations may be useful to identify the most promising target for the stimulation.

Task-dependent temporal processing characteristics may also partly explain the differential responses in specific regions to particular stimulation frequencies. For example, the rightlateralized SMA activity in temporal processing tasks in the suprasecond range may dominate the response to 1-Hz rTMS stimulation. 1-Hz stimulation has been found to impact the timing of anticipatory postural adjustments in PD patients if it is applied over the SMA but not over the dorsolateral premotor cortex, and to improve motor but not non-motor symptoms (56, 57). Obviously, further research is necessary to entertain these possibilities but the fundamental nature of temporal processing may bear the potential to improve the principal effectiveness of these methods, which are typically considered a promising form of treatment for PD (58, 59).

### REFERENCES


### CONCLUSION

Although the cerebellum and the SMA are affected in PD and should hence be considered in a more encompassing view of the disease, their contribution to the overall pathogenesis is a matter of debate. Perhaps most importantly, it is still unclear if changes in cerebellar and SMA activity are aspects of the primary pathology and/or secondary compensatory mechanisms (60). However, their contribution may be entirely secondary to the cell loss in the substantia nigra, which is the focus of basic research and therapeutic intervention. Further, a better understanding of their interaction with the basal ganglia deems necessary to account for the complexity of the disease and to open potential new directions for therapeutic interventions.

The overlap between the sequencing and the temporal processing functions ascribed to the basal ganglia and corticostriatal circuits offer one such direction. Moreover, the relatively specific concepts that have been developed with respect to the neural mechanisms underlying temporal processing in the basal ganglia and the cerebellum may allow improvement of existing strategies such as cueing and stimulation techniques. In this context, knowledge about the interplay of the basal ganglia with other regions engaged in temporal processing may offer a means to improve behavioral compensatory strategies via the informed manipulation of temporal structure or the identification of promising targets for interventions targeting the neural level.

An interesting open question concerns the transfer of therapeutic effects from basic temporal processing tasks to more complex behavior or from motor to non-motor processing and vice versa (37). Such transfer effects may reflect the essentially sensorimotor nature of the overarching network, as well as the general behavioral function of the basal ganglia in the sequencing of actions in both domains. Accordingly, therapeutic intervention in PD may aim to balance the dysfunction of this overarching network by targeting specific network functions to improve performance in several domains rather than focusing only on the most prominent motor symptoms reflecting to an extent the increasing interest in non-motor features of the disease.

### AUTHOR CONTRIBUTIONS

MS and SK have written the manuscript.

meta-analytic connectivity modeling with behavioral filtering. *Neuroimage* (2012) **60**:117–29. doi:10.1016/j.neuroimage.2011.12.010


fMRI during the performance of pre-learned sequential movements. *Brain Res* (2007) **1147**:265–71. doi:10.1016/j.brainres.2007.02.046


literature. *J Neurol Neurosurg Psychiatry* (2005) **76**:1614–23. doi:10.1136/ jnnp.2005.069849


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Schwartze and Kotz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Toward Precision Psychiatry: Statistical Platform for the Personalized Characterization of Natural Behaviors

*Elizabeth B. Torres1,2,3\*, Robert W. Isenhower1 , Jillian Nguyen1 , Caroline Whyatt1 , John I. Nurnberger4 , Jorge V. Jose5,6 , Steven M. Silverstein7 , Thomas V. Papathomas2,8 , Jacob Sage9 and Jonathan Cole10*

*1Psychology Department, Rutgers University, New Brunswick, NJ, USA, 2Rutgers Center for Cognitive Science, Rutgers University, New Brunswick, NJ, USA, 3Computer Science Department, Center for Biomedical Imaging and Modeling, Rutgers University, New Brunswick, NJ, USA, 4Department of Psychiatry, Institute of Psychiatric Research, Indiana University School of Medicine, Indianapolis, IN, USA, 5Department of Physics, Indiana University, Bloomington, IN, USA, 6Department of Cellular and Integrative Physiology, Indiana University, Indianapolis, IN, USA, 7Department of Psychiatry, Rutgers University Robert Wood Johnson Medical School, New Brunswick, NJ, USA, 8Department of Biomedical Engineering, Rutgers University, New Brunswick, NJ, USA, 9Movement Disorders, Rutgers University Robert Wood Johnson Medical School, New Brunswick, NJ, USA, 10Poole Hospital and Bournemouth University, Poole, UK*

#### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Yuhui Li, University of North Carolina at Chapel Hill, USA Kristina Denisova, Columbia University, USA*

> *\*Correspondence: Elizabeth B. Torres ebtorres@rci.rutgers.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 15 August 2015 Accepted: 18 January 2016 Published: 02 February 2016*

#### *Citation:*

*Torres EB, Isenhower RW, Nguyen J, Whyatt C, Nurnberger JI, Jose JV, Silverstein SM, Papathomas TV, Sage J and Cole J (2016) Toward Precision Psychiatry: Statistical Platform for the Personalized Characterization of Natural Behaviors. Front. Neurol. 7:8. doi: 10.3389/fneur.2016.00008*

There is a critical need for new analytics to personalize behavioral data analysis across different fields, including kinesiology, sports science, and behavioral neuroscience. Specifically, to better translate and integrate basic research into patient care, we need to radically transform the methods by which we describe and interpret movement data. Here, we show that hidden in the "noise," smoothed out by averaging movement kinematics data, lies a wealth of information that selectively differentiates neurological and mental disorders such as Parkinson's disease, deafferentation, autism spectrum disorders, and schizophrenia from typically developing and typically aging controls. In this report, we quantify the continuous forward-and-back pointing movements of participants from a large heterogeneous cohort comprising typical and pathological cases. We empirically estimate the statistical parameters of the probability distributions for each individual in the cohort and report the parameter ranges for each clinical group after characterization of healthy developing and aging groups. We coin this newly proposed platform for individualized behavioral analyses "*precision phenotyping*" to distinguish it from the type of observational–behavioral phenotyping prevalent in clinical studies or from the "one-size-fits-all" model in basic movement science. We further propose the use of this platform as a unifying statistical framework to characterize brain disorders of known etiology in relation to idiopathic neurological disorders with similar phenotypic manifestations.

Keywords: precision phenotyping, sensory–motor noise, autism spectrum disorders, Parkinson's disease, schizophrenia, deafferentation

## INTRODUCTION

Precision medicine is a new approach to acquire and integrate knowledge from biomedical research and clinical practice (1). It is a computation-enabled platform poised to radically transform the ways in which we currently conduct biomedical research and patient care by integrating personal information across many layers, from genes to behavior (**Figure 1A**). The personalized approach has been successful in areas such as cancer research and treatment. In contrast, the disciplines of mental health and social sciences tend to follow a "one-size-fits-all" approach and rely primarily on the bottom layers of the knowledge network – selfreports and clinical ratings and their interpretation – but not as much on objective physical measurements tailored to the individual.

In recent years, the need to shift from symptom- and interpretation-based approaches in neurological disorders and mental illnesses to more objective methods has been voiced in a various ways. One such method is the Precision Psychiatry initiative of the National Institute of Mental Health (NIMH), where various task forces have been created to achieve more objective science that unveils biological signatures of core *dimensions* of functioning (e.g., cognition, positive valence systems, and arousal), as they are expressed across neuropsychological and neurological disorders on a spectrum. More explicitly, the Research Domain Criteria (RDoC) initiative (2) is one of the various attempts to bring basic research on mental illness to a new level of rigor that also helps accelerate progress. Surprisingly, however, the current RDoC matrix does not include a dimension of sensory–motor function (3). Here, we argue that movement and its sensation could be a great ally in tailoring research and treatment to the patient's needs and inherent predispositions. As such, it may be a good idea to include sensory–motor function as a dimension within RDoC and to incorporate objective and movement-based outcomes into research on brain disorders. For the remainder of this paper, we demonstrate a novel quantitative method that we believe is particularly useful in this regard.

A simple experimental paradigm is presented, with a new statistical method for individualized behavioral analyses and a new kinematic data type (explained in Section "New Data Type and Different Assessment of Motor Variability"). Both form the basis of a unifying platform to help personalize research and patient care within the field of disorders of the central and peripheral nervous systems. More specifically, we address data analyses differently from traditional approaches (**Figure 1B**) and use the new platform (**Figure 1C**) to empirically estimate the individualized stochastic signatures of the moment-by-moment fluctuations in performance across several clinical and nonclinical populations. These populations range from typical controls of various age groups (children, young college students, middleaged, and elderly participants) to patients of various types. The latter include Parkinson's disease (PD) at mild and severe stages, schizophrenia (SZ) patients of different ages, and individuals with a diagnosis of autism spectrum disorder (ASD), from various age groups. In addition, we include parents of individuals affected by ASD to investigate whether their motor patterns fall within the signatures uncovered in the normal control groups. Finally, we describe the data from a patient who lost peripheral sensory inputs from touch, pressure, and movements, resulting in loss of proprioception from the neck down, but whose motor nerves are unaffected. We provide an overview of the statistical parameters that are empirically estimated from the movement kinematics of all participants and demonstrate that these reveal fundamentally different features across disorders, as well as overlapping features. Results are discussed in the context of Precision Medicine and Precision Psychiatry. In particular, we emphasize that these analytic methods produce fine-grained variables that are well-suited to bridging the gap between coarse behavioral descriptors and genetic factors, which may underlie some sensory–motor noise signatures across disorders.

## METHODS

### Subjects

To empirically estimate the ranges of statistical parameters underlying kinematics data, we use data from various subject groups. All subjects provided written informed consent on forms approved by The Rutgers University Institutional Review Board (IRB) or Indiana University IRB. All protocols were approved by the IRB committees, in compliance with the Helsinki Act. Clinical records were obtained in compliance with the Health Insurance Portability and Accountability Act (HIPAA). Parents of the subjects with ASD signed the IRB approved consent on behalf of their child/adult participant with ASD. Table S1 in Supplementary Material summarizes the demographic characteristics of the 176 participants.

The control subjects were subdivided into four broad groups including young children (CT1), young college students (CT2), middle-aged subjects (CT3) and the elderly subjects. In line with previous research indicating maturity of pointing kinematics after 4 years of age (4), we further subdivided the CT1 group. Within CT1, we examined individuals between 3 and 4 years of age (CT1a) and those between 5 and 10 years of age (CT1b). Another control group was the parents of a subset of the children affected by ASD. The latter control group had no ASD diagnosis but their movements were visibly different from those of other control middle-aged individuals. This prompted us to perform this comparison, despite a lack of clinical diagnosis. From a subset of the parents, we estimated that the age of the cohort ranged between 32 and 39 for mothers and 32 and 44 for fathers.

The demographic information and clinical scores for the ASD group are shown in Table S2 in Supplementary Material. Within this group, ASD1 was composed of 3–12-year-old participant and ASD2 was comprised of 13–25-year-old participants.

A PD cohort was also included. They were recruited from the PD support group of the Rutgers-Robert Wood Johnson Medical School's Movement Disorders Center. The PD group was subdivided into 9 subjects with mild-to-moderate disease severity (PD1) who were ambulatory, independent, and with some visible resting tremor but without visible action tremor at the time of the visit, and 17 subjects (severe PD2) who had very impaired mobility, some of whom were ataxic and some with freezing of gait. The latter group generally needed assistance to walk and

individual is rather characterized by a family of PDFs with individualized rate of accumulation and change of these stochastic parameters as a function of treatment and disease progression.

had visible action and resting tremor. Table S3 in Supplementary Material shows the demographics of both groups.

A group of eight elderly individuals (ages 75–77 years old) with no formal diagnosis of a movement disorder was included as a control group for the PD groups comprising a broad range of ages (46–77 years old). These subjects were part of an earlier study (4) where we had aimed to statistically characterize action tremor during pointing behavior in typically aging individuals.

show fluctuations in performance (e.g., amplitude and timing) that accumulate information toward an expected value, then shift signatures in non-stationary fashion [bar indicates snapshot of behavior in (B) along the continuous stream]. The probability distribution function (PDF) is continuously empirically estimated. A given

A group of 23 patients with SZ was included in the study to ascertain their motor signatures in relation to the other cohorts and to age- and sex-matched controls. SZ patients were recruited from Rutgers University Behavioral Health Care clinics. Patients were either enrolled in a daily partial hospital program (PHP) or were outpatients who only required biweekly or monthly visits to health-care providers. There were 10 patients enrolled in the PHP and 13 patients in the outpatient program. Table S4 in Supplementary Material reports demographics and Table S5 in Supplementary Material reports the Frontal System Behavior Scale (FrSBE) self-rating scores for executive dysfunction.

Finally, a special subject (IW) without proprioception, secondary to deafferentation (5) was included. This subject suffered a lack of proprioception and touch from the C3 level down due to acute sensory neuronopathy syndrome. This syndrome led to irreversible sensory nerve destruction at the dorsal root ganglia level of fibers conducting touch, pressure, and movement information from the periphery to the central control centers of the brain. The motor nerves and the movement output of the deafferented subject were unaffected. This particular participant has learned to move in a controlled manner again using mental concentration on movements with visual supervision to help close the feedback loops. The signal that we are capturing in most subjects contains a blend of motor and sensory noise. This participant provides an example of motor noise in the absence of sensory afferents conducting movement information. We use the locations of his estimated signatures of speed profile-dependent variability during visual feedback and in the dark. These points located on the Gamma parameter plane (see Methods below) serve as a reference to anchor the data from other patients.

Due to the individualized nature of this approach, whereby each participant is their own control [i.e., examined with respect to its own empirically estimated family of probability distribution functions (PDFs) from the motion parameters], it is not necessary to match the number of participants in each clinical category to the exact number of controls of a given age/sex. However, the controls included a broad range of ages from both sexes, spanning the age and sex range of the participants with a clinical diagnosis. In this sense, the aim of the age- or disease-stage subdivisions was to use the new methods to show how to obtain the statistical summaries for the kinematic parameters of interest for representative cross-sections of the typical population. However, we emphasize that the individualized scatterplots that we provide can also be examined blindly (i.e., without *a priori* imposed clinical labels).

### Experimental Setup

**Figure 2A** depicts the basic experimental paradigm consisting of a full pointing motion forward to the target and back to rest. The forward segment toward the target is instructed and goal directed. As such, it results in a deliberate pointing action. In contrast, the retraction away from the target, after the pointing action ended, is spontaneously performed without instructions. Pointing accuracy was not required, as the continuously periodically alternating nature of this motor action was more relevant to our analyses than the accuracy of the pointing act. The experiment took place under conditions of visual feedback.

In the case of the deafferented subject, we studied the pointing movements across several conditions. These included pointing in complete darkness while relying only on the memory of the target (flashed for a second); pointing in the dark with a light-emittingdiode (LED) attached to the moving finger but no visual feedback from the target; and pointing in the dark with no LED on the finger but with continuous visual guidance from the target ON throughout the motion. We separated this subject's performance according to visual feedback conditions: dark vs. vision.

### Instructions to the Participants

Participants sat in a chair facing the target location at a comfortable distance for reaching (i.e., they did not have to completely stretch the arm; see schematics in **Figure 2A**). They were instructed to touch the target when it was presented. The forward motions toward the target were explicitly instructed with the words "Touch the target when it appears." In marked contrast, the retracting motions from the target back to rest were not instructed – participants spontaneously performed these movements. In this sense, we underscore that the retracting motions were automatically performed without any explicit visual target. Our previous work had demonstrated striking differences between the kinematics of the instructed forward motions and the spontaneous retractions. These differences manifest in families of reaching motions such as pointing (6–8), reach-to-grasp (9), and also in martial arts routines requiring forward and back motions (10, 11). Building on these previous results, we examined the spatiotemporal features of these two separate movement types in order to assess how such differences may manifest across neuropsychiatric and neurological disorders.

### New Data Type and Different Assessment of Motor Variability

Since Bernstein's work on the importance of motor variability (12) to central control of self-produced movements, many studies have assessed the variability of kinematic parameters. In the reaching domain, these have included end point error (13, 14), speed (15), and joint angles (16, 17), among many others. In all cases, the noise-to-signal balance has been examined under the assumption of normality. Variability is thus described relative to a central value (the assumed mean). Often, only a small number of trials are used to determine the fluctuations of a given parameter around that mean. To this end, the average of that parameter is obtained, and the ±deviations around the central value are computed assuming the symmetric (theoretical) Gaussian PDF. This is illustrated for the case of the speed profile taken as a set of movement parameters in **Figure 1B**. The assumptions of Gaussian PDF extend to stochastic models of motor control (13, 18) and Bayesian estimation-based models (19). To the best of our knowledge, the PDFs most likely underlying kinematic parameters of hand movements across disorders of the nervous systems have not been empirically estimated. Furthermore, estimations of such PDF in crosssections of the normal population as a function of age groups have not been performed either. Such estimations are necessary to assess the noise-to-signal ratios of movement parameters

FIGURE 2 | Basic experimental paradigm, sample raw data and methods. (A) Full forward and back pointing loops continuously measured as they unfold in a forward segment deliberately intended toward the target and a spontaneous (uninstructed) retraction away from the target. Touch screen is used to automatically register the end of the goal directed motion (when the hand speed is at near zero velocity and its position is at near zero-distance to target). At this point, the target-hand distance and the speed increase as the hand starts retracting away from the target. This position-speed hand configuration marks the beginning of retraction movement segments. The ending of those segments are determined by near zero-velocity criteria marking pauses in the continuous motions. (B) Trajectories and corresponding speed profiles from continuous motions of a young child naturally performing the task without constraints. Data extraction relies as previously described on hand positional distance and speed criteria. Sample trajectories for two sample forward (black) and backwards (blue) movement trajectories that were automatically detected using these criteria are shown. Notice that end point errors can be large in young children, particularly those in the spectrum. Therefore, no restrictions are imposed on target accuracy. The focus is rather on the spread of the moment-by-moment peak velocity (i.e., fluctuations in performance) during a full pointing loop. (C) Discrete segments of speed profiles from the 3D hand trajectories are color coded to identify the ones marked on the 3D plot in (B) and on the continuous speed plot in (C) (black forward, blue backwards). Numbers mark these segments, aligned at movement onset. Landmarks of the continuous motions are the peak speeds (stars) between local minima (circles). (D) Examples of speed profiles from retractions aligned to movement onset in a child with ASD (top) and a CT1 child (bottom) with inset histograms showing the corresponding normalized frequency of the time to reach the local maxima. (E) Sample speed profiles continuously taken from a representative elderly participant with landmarks (speed minima, speed maxima, and screen touches) zoomed in for clarity. (F) Hand trajectories forward to the target from a young representative control (CT2 group) and corresponding speed profiles aligned to movement onset. Black dots mark the spread of the speed maxima along the forward trajectories and their temporal speed profile from trial to trial (for clarity, only forward segments are shown aligned to movement onset but backwards are similar in the opposite direction). (G) Analytical steps to empirically estimate the statistical signatures of these motions (see text for detailed explanation).

across the general population. This is in contrast to assuming a theoretical PDF *a priori* to describe the normal subset of the population without empirically estimating it. Specifically, we do not know how sensory–motor priors develop under normal or atypical conditions, how they may shift with typical aging, or how they may change with a degenerative disorder of the nervous system.

We assess the continuous time series of movement speed to empirically estimate the noise-to-signal ratios of velocitydependent parameters from hand movement trajectories. The raw data in this case are the speed profile (such as that depicted in **Figure 1C**) continuously tracked by motion capture sensors. The waveform of the speed temporal profile from point-to-point varies, as it depends on the curvature of the underlying positional trajectory, tied as well to internal parameters such as the rotation of joints and the changes in muscle states (20, 21). Yet, regardless of the shape of the hand path, the speed rises during the acceleration phase to reach a maximum value and then generally decays in the deceleration phase to a stop or pause en route to the target. In the motor control literature such submovements and their variability around the assumed mean are commonly studied [e.g., Ref. (22)]. For highly automated straight point-to-point hand movements, their shape is approximately symmetric (23, 24).

Our interest, however, is not in the variations of the hand's submovements around a mean value obtained under the assumption of a theoretical symmetric (e.g., Gaussian) distribution. Instead, we are interested in the accumulation (and the rate of change) of minute fluctuations in performance that occur from moment to moment in the parameters associated with the speed of motion. These include (among others) the fluctuations in maxima and those in the time to reach the velocity peak from the last pause or stop instance. As the motion of the arm-hand linkage is repeated, these minute fluctuations in speed accumulate and give rise to various frequency distributions. The shape and scale (dispersion – see Step 3 below) of these distributions can be estimated with high confidence to empirically approximate, along a continuum, the family of PDFs most likely describing the underlying random process. Once again, this is in contrast to assuming a theoretical PDF *a priori*.

To experimentally measure fluctuations in the speed amplitude from moment to moment, we accumulate the changes in the speed maximum [termed here peak velocity (PV)]. Since the speed waveform localized between two minima within the time series of speed (e.g., **Figure 1C**) may change the shape and amplitude from local speed minima to local speed minima, we must first normalize it (25). To this end, we obtain for each minima-to-minima segment the following index:

$$\text{nPVindex} = \frac{\text{PV}}{\text{PV} + \text{Average}(V\_{\text{min to min}})}$$

Here, PV denotes the peak velocity (speed maximum) and the denominator contains the sum of the PV and the average speed between two consecutive local speed minima. We term this the normalized PV index. This normalization process also avoids possible allometry effects due to differences in the sizes of the limbs of the subjects (e.g., children vs*.* adults) (26). Larger values of this index indicate slower movements on average, since smaller averaged speed values in the denominator result in higher values of the index. These would be expected in the PD population that suffers from bradykinesia but not in the typical controls (for example).

The fluctuations in the overall profile, as determined by the changes in amplitude and timing of each peak, provide information about the individual rate of change of these variables as the nervous system of the person generates and then experiences them. Examples of the accumulation of variations in amplitude and timing in the time series of speed profiles are depicted in **Figure 1C**. We emphasize that this treatment of the variability problem fundamentally differs from traditional approaches, whereby it is assumed that the speed parameters follow a Gaussian distribution with known mean and variance. Therefore, further statistical analyses typically involve testing shifts in the mean/ variance above chance and the use of parametric models assuming population statistics under a "one-size-fits-all" approach (**Figure 1B**).

### Analytical Techniques

In a series of papers, we have described these statistical techniques [e.g., Ref. (8, 11)]. A brief summary for the purposes of this report has four main stages, as detailed in **Figure 2G**:

 (*Step 1*) Acquire time-series data (e.g., kinematics) from continuous trajectories of unconstrained target-directed pointing movements in three dimensions. **Figure 2B** shows sample data from the naturalistic hand trajectories of a young child. **Figure 2C** shows temporal speed profiles and the main landmarks used to study some of the patterns of velocitydependent variability. These include the velocity peaks (meter per second) and the time (milliseconds) to reach those peaks from the local minima, among others. Sample speed profiles automatically extracted from the continuous data are also shown in **Figure 2D** for ASD and CT1 children of comparable age. Sample data from adults are shown in **Figure 2E** (elderly participants) and **Figure 2F** (young CT2).

 (*Step 2*) Plot the frequency histograms (**Figure 2G** step 1) of the parameter of interest (e.g., the normalized PV index) using optimal binning (27, 28) and estimate the underlying family of probability distributions of speed profile-dependent parameter that best characterizes the trial-to-trial fluctuations in performance for each individual (**Figure 2G** step 2). Besides individual estimation, this procedure can also be done for cohorts of participants with a neurological disorder or typically developing individuals.

 (*Step 3*) Use maximum likelihood estimation empirically to obtain – from the data – the values and ranges of the shape (*a*) and scale (*b*) parameters of the continuous Gamma family of probability distributions. The Gamma PD F is given by:

$$\mathcal{Y} = f(\boldsymbol{\mathfrak{x}} \mid a, b) = \frac{1}{b^{\ast} \Gamma(a)} \boldsymbol{\mathfrak{x}}^{\ast -1} \boldsymbol{e}^{\frac{-\mathbf{x}}{b}}$$

 in which *a* is the shape parameter, *b* is the scale parameter, and γ is the Gamma function (29). We then plot the estimated Gamma parameters for each participant with 95% confidence intervals on the (*a*, *b*)-Gamma parameter plane. Using this method, we localize the individual participant and can compare each subject's location to those of the other subjects (**Figure 2G** step 3). Here, we also look at the overall data to identify self-emerging clusters and patterns, particularly in relation to a participant with a disorder of known etiology. For example, one could use this methodology to identify clusters or patterns across participants with a disorder that is clinically diagnosed based on symptoms alone vs*.* participants with the same clinical disorder, but with known genetic origins. By color coding the scatterplot points based on clinical criteria, we may be able to help with interpretation. In such cases clusters of participants with idiopathic diagnosis and similar symptoms may be studied in relation to those whose symptoms are of known etiology.

The noise-to-signal ratio [i.e., the Fano Factor (30), FF = empirically estimated Gamma variance/empirically estimated Gamma mean] is also obtained. The Gamma mean is given by μ = *a* × *b* and the Gamma variance is given by σ = *a* × *b*<sup>2</sup> . Notice that the noise-to-signal ratio, the Fano Factor is also in this case the Gamma scale parameter

*<sup>b</sup> a b a b* = = <sup>×</sup> × σ µ 2 2 (29). This is important as we are assessing the

levels of noise in relation to the empirical estimation of the Gamma parameters from the data as a function of clinical group type. Higher levels of noise correspond to an increase in the *b scale* parameter along the vertical axis of the Gamma plane, whereas lower levels of noise correspond to lower values of the scale parameter.

It is also important to emphasize that when the *shape* parameter *a* of the Gamma family is equal to 1 (*a* = *1*), the data follow the memoryless exponential probability distribution, a special case in the Gamma family. This is the most random distribution, coined as "memoryless" because events in the past do not accumulate information predictive of events in the future (29). Larger values toward the right of the shape axis on the Gamma (*a*, *b*)-plane tend toward the symmetric distributions, with a various skewed distributions in between the two extremes (31).

In the text, we will refer to the level of randomness by examining the value of the empirically estimated shape parameter. When close to *a* = 1, the shape denotes the memoryless Exponential distribution. When increasing the shape value to the right of the horizontal axis, we will refer to the accumulation of information toward the prediction of an expected value, away from *a* = 1, toward the Gaussian range of the Gamma plane. Likewise, we will refer to higher or lower noise levels according to the empirically estimated *b* Gamma-scale parameter value, which is the FF.

 (*Step 4*) Repeat this estimation procedure to characterize the rate of change of the Gamma parameters' stochastic trajectory. This step can detect conditions and stimuli that accelerate the change in the parameters down and to the right (i.e., to the right along the shape axis) away from random regimes of the Gamma plane (i.e., when *a* = *1*) and down along the scale axis, away from high noise-to-signal ratio values. **Figure 2G** step 4 marked with stars indicates the largest step, which illustrates a stochastic trajectory that is moving in the abovementioned statistically desirable direction. Results featured in this panel demonstrate this evolution within an ASD vs. CT experimental intervention (32). By examining large incremental steps in the stochastic signatures that lower the noise and increase the shape value toward Gaussian models, we can infer a range of implications, including which context is most appropriate to improve motor output within sensory manipulations? What effect a psychotropic medication dosage may have on each person? or Which therapeutic exercise, among a set of routines, is the most beneficial to the persons' statistical motor patterns?

### RESULTS

**Figure 4** provides a color-coded map of the summary statistics of all participants. **Figure 5** summarizes the *p*-values from pairwise statistical comparisons in matrix form across all neuropsychiatric/neurological disorders within the study. Tables S6–S9 in Supplementary Material summarize the statistical results of this study. Each supplementary table is accompanied by a Figures S3–S6 in Supplementary Material that helps visualize the results for each patient subgroup. Moreover, Figures S1 and S2 in Supplementary Material examine ensemble temporal kinematics data as per clinical diagnosis in relation to healthy controls. Below we discuss each finding separately.

### Typical Controls May Shift Statistical Signatures across the Life Span

The cross-sectional data under examination revealed that participants with no clinical diagnosis had different statistical signatures across ages, suggesting that even during typical development these signatures of fluctuations in motor performance may shift. Table S6 in Supplementary Material reports the ranges of the estimated summary statistics (first, second, third, and fourth moments), for the normalized PV index corresponding to the control participants grouped by age. The results of the pairwise comparisons of the medians of each moment, using the nonparametric Wilcoxon rank sum test, are also reported in Table S6 in Supplementary Material. Here, the young children (CT1a) aged 3–4 years old showed the highest levels of noiseto-signal ratio, but this was not significantly different from those of children aged 5–10 years old (CT1b). The mean value of the normalized PV index did differ significantly between the two groups of children (*p* < 0.01), whereby the older children were significantly faster on average. **Figure 3A** (leftmost top panel) shows the estimated PDFs of the two groups superimposed and contrasts them to those of the young college participants (CT2 aged 18–25 years old) in the right panel. **Figure 3A** also shows the PDFs for the CT3 groups (aged 30–57) superimposed on those of the parents of children affected by ASD in this study (green traces). The latter group ranged in age from 32 to 44 years old, overlapping with those of members of the CT3 group, yet their estimated PDF's fall closer to those of the elderly group (75–77 years old). Indeed, the statistical comparisons reported in Table S6 in Supplementary Material revealed no differences between the ASD parents and the elderly group across all moments, despite the large gap in these groups' ages. Both the elderly groups and the ASD parents move at comparable speeds on average and have comparable levels in the accumulation of noise (Figure S3B in Supplementary Material). Yet ASD parents and the elderly groups do show differences in the time range of reaching the peaks (Figure S3C in Supplementary Material). Specifically, the probability plots of the time to reach the PV

CT3 is from 30- to 57-year-old participants. This panel also contains PDFs estimated from the parents of ASD children (green traces). Bottom-right panel is elderly participants between 75 and 77 years old. (B) Estimated PDFs from neuropsychiatric and neurological disorders: top-left is ASD (3–25 years old). Top-right is SZ (22–57 years old, see Table S1 in Supplementary Material for age break down). Bottom-left is mild PD and bottom-right is severe PD. All PDFs are plotted against those of the deafferented subject (yellow trace is from the condition with visual feedback and black trace from the dark condition.) (C) Power law relation (see text for fit exponent value and goodness of fit) on the log–log Gamma parameter plane between the estimated shape and the estimated scale parameters of the continuous Gamma family of PDFs. Each dot corresponds to an individual participant. (D) Notice the mean values taken across participants in each group whereby the ASD group falls apart with the highest noise and the value of the shape parameter tending away from the Gaussian distribution case.

shows these differences. When examining the normalized PV index, the parents line up with the deafferented subject under visual guidance and with the elderly participants. There in Figure S3B in Supplementary Material, one can appreciate that the elderly participants move slower than the other controls. Yet, when we examine their timing to reach the peak, the elderly participants fall within the ranges of the younger controls. This suggests that under similar timing scale, it is the noise in the distance traveled by the hand up to PV that most likely account for their bradykinesia. This is in contrast to ASD parents who move slower than controls under a timing scale that rather aligns with that of the subject without proprioception (Figure S3C in Supplementary Material). The ASD parents are as slow as the elderly participants, but in their case, both the distance traveled to the PV and the time to cover that distance are problematic.

The ASD parents significantly differed from age-matched controls in CT3 in terms of average movement speed and noise levels (*p* < 10<sup>−</sup><sup>6</sup> ). The summary statistics map for all control participants, along with their median values, are shown in Figure S3A in Supplementary Material accompanying Table S6 in Supplementary Material. Likewise, the probability plots comparing all control groups are shown in Figures S3B,C in Supplementary Material. Notice the overlap of the parents' signatures with those of the elderly controls (**Figures 3C,D**), as well as the separation between the deafferented subject and the CT3 control group who are of a similar age as this subject. CT2 is the ideal group in the sense that their distribution is normal (along the line of unity in the probability plot).

The lowest levels of noise-to-signal were registered in the young-to-mid age controls of the CT2 and CT3 groups. **Figure 3C** shows an emergent power relation between the estimated Gamma shape and scale parameters with model *f*(*x*) = *a* × *xb* common to all groups (fit with 95% confidence bounds), where *a* = 0.794 (0.747, 0.841)*; b* = −1.031 (−1.043, −1.019) and goodness of fit: sum squared error: 9.433e−07; adjusted *R*-squared: 0.9982; root mean squared error: 9.57e−05.

All participants fall on this line with CT2–3 having the lowest noise-to-signal (scale) levels and the largest shape

values – indicating distributions tending toward the Gaussian shape. The average parameter values per group are shown in **Figure 3D**. The middle-aged participants in CT3 showed the largest kurtosis values (see **Figure 4C**). **Figure 4D** summarizes the mean for each group, showing as well the shifts of these statistical signatures with normal development and aging. Along this map, the surprising finding was the overlapping of the signatures of the ASD parents with those of the elderly participants and away from those in age-matching CT3.

### Highest Speed-Dependent Noise in ASD Is Accompanied by Low Average Speed Values

Across all groups with neuropsychiatric/neurological disorders and the control groups, the ASD group generated the lowest values for the shape parameter and the highest values of the noise-to-signal ratio along the scale axis for the normalized PV index under examination.

**Figure 3B** shows the empirically estimated PDFs with those of the deafferented participant superimposed (yellow and black traces, see legend). **Figure 3C** shows that all ASD subjects have higher noise, even more so than the deafferented subject, whether pointing in the dark or using visual feedback. The ASD highest variability level is shown in the two-dimensional plot of **Figure 4A**. The averaged summary statistics of these parameters in **Figures 4B,D** also demonstrate this. These participants also move at the slowest rate regardless of age. Further information about this group can be seen in the four-dimensional plot, in which the skewness and kurtosis of their empirically estimated distributions are also shown. There the performance of the deafferented participant, while pointing in the dark, falls within the distribution ranges of the ASD participants, specifically closer to those in the younger ASD subgroup. This can also be seen in Figure S4 in Supplementary Material accompanying Table S7 in Supplementary Material with a focus on the ASD cohort. For clarity, this figure isolates the ASD participants in relation to the controls of similar ages. Very little overlapping between controls and ASD can be detected in this parameter space. Table S7 in Supplementary Material provides the ranges and outcome of statistical comparisons, whereas **Figure 5** provides a summary in matrix form of the *p*-values from pairwise comparisons.

Table S7 in Supplementary Material reveals significant statistical differences (*p* < 0.01) across all pairwise comparisons of noise levels between both ASD groups and controls subgroups with overlapping ages. This was also the case for the estimated mean with the exception of ASD1 and the elderly participants, and ASD2 and their parents.

**Figure 3B** also shows the largest dispersion for ASD across the cohort of patients and controls – comparable to the deafferented participant in the dark condition (black curve). However, the ASD group shows a mean value of the index that is shifted to the left, indicating faster movements on average than the deafferented participant (recall the denominator involving larger average speeds leads to a left shift of the index toward smaller values). In the ASD groups, as the average speed decreases (i.e., the normalized PV index increases), the variance increases.

pairwise across neurological conditions as divided by age (ASD and SZ) and severity (PD). (A) Empirically estimated mean. (B) Empirically estimated variance. (C) Empirically estimated skewness. (D) Empirically estimated kurtosis. Values of *p* > 0.05 are not statistically significantly different (ignore diagonal). Color bar are reflecting 10−number, the log of the *p*-values.

### Parkinson's Disease Patients Have Speed Statistics Comparable to Those of ASD Parents

A surprising result in the analyses of the typical controls was that ASD parents, who are young to middle aged, showed statistical features of the elderly group. Here, we also found unexpected similarities in the noise levels between the young ASD parents and the severe PD patients. The empirically estimated PDFs of the parents are seen in **Figure 3A** in the bottom-left panel, superimposed with the CT3 group of comparable age. Notice the disparity in dispersion between these groups; indeed the parents of children with ASD show results that are comparable to the elderly group of healthy adults. A further graphical view of this result is shown in **Figures 3C,D**, which illustrate the scatter across all participants and the groups' averaged shape and scale parameter planes respectively. **Figure 4A** shows the scatter of the summary statistics for the ASD parents, overlapping mostly with the elderly and the severe PD participants. In **Figure 4B,**

the average values taken across these three groups (elderly participants, ASD parents, and severe PD) are localized next to that of the deafferented participant pointing under visual feedback. **Figures 4C,D** show the 4D plots of the scatter and averaged values, with an additional plane, lifting the points according to the empirically estimated distribution of skewness and kurtosis (the size of the marker). This graphical illustration indicates that the signatures of the normalized velocity index of the ASD parents fall within the ranges of the elderly and severe PD patients, but farther away from the deafferented participant. In all plots, the ASD parents are not overlapping with the CT3 group who are within their age range. Figure S5C in Supplementary Material also shows a difference between the timing of the reaches of the ASD parents and the age-matched CT3. The probability plots of the time to reach the PV are shown in this figure for the ASD parents. They align with those of the mild PD patients and the deafferented subject under visual guidance. In contrast, the severe PD patients align their timing with that of the deafferented subject pointing in the dark.

## Summary Statistics Unambiguously Separate PD Subjects with Different Clinical Severity Levels

Table S3 in Supplementary Material reports the demographic and clinical information concerning the PD patients. Table S8 in Supplementary Material reports the ranges for the estimated statistical summary for PD patients of severe and mild-to-moderate stages according to clinical scores in Table S3 in Supplementary Material. These subgroups differ significantly in noise levels, estimated mean, and estimated variance of the normalized PV index (all rank sum tests on the medians *p* < 0.001) according to the empirically estimated distributions. The estimated PDF's for the severe and mild PD groups are shown in **Figure 3B** lower panels (see legend). Note the broad dispersion of the severe group in contrast to the sharper PDFs of the mild-to-moderate group. These differences in mean and variance are also visible in **Figure 3D** summarizing the mean values of the estimated Gamma shape and scale parameters.

The mild PD patients are closer in speed to the middle-aged CT3 group (nonsignificant differences in mean value) and to the younger CT2 group (nonsignificant differences in mean value and skewness level). Thus these patients are not yet bradykinetic, but their noise levels are significantly higher than younger controls in CT2 and CT3 groups. The mild PD significantly differ (*p* < 0.001) from the ASD parents (Table S8 in Supplementary Material) but surprisingly, the severe PD cohort has comparable statistics in noise, variance, and skewness close to those of the much younger ASD parents and to the elderly controls between 75 and 77 years old. Note here again that the ASD parents are closer in age to the middle-aged CT3 group, much younger than these PD and elderly participants. Further differences between groups can be observed in the 4D plots of **Figures 4C,D**. Here, one can see that the mild-to-moderate PD group falls closer to controls than to the severe PD. The latter falls closer instead to the deafferented participant when he moves aided by visual guidance.

The estimated PDFs for the severe PD in **Figure 3B** (bottom right) show the overlap with the PDFs of the deafferented participant without vision. Likewise, Figure S5 in Supplementary Material accompanying Table S8 in Supplementary Material shows the summary statistics for these two groups of PD patients in relation to age-matched controls and to the ASD parents. Notice as well the probability plots whereby the probability distributions of the controls tend to normal (close to the line of unity), while those of the patients deviate from the line of unity.

### Statistical Ranges of SZ Patients Show Heterogeneity Relative to Other Neurological Disorders

Patients with SZ did not significantly differ in stochastic signatures across ages. This is shown in matrix form in **Figure 5** for all moments. Furthermore, the younger SZ patients (22–30 years old) are closer to middle-aged CT3 adults than to young CT2 adults, despite overlapping ages with the latter group. Table S9 in Supplementary Material reports the ranges of all estimated signatures along with the statistical comparisons. Note that the estimated means of CT2 and CT3 do not significantly differ from SZ1, but the older SZ2 and SZ3 have noise levels comparable to those of the ASD parents and the elderly controls. Figure S6 in Supplementary Material, corresponding to Table S9 in Supplementary Material, shows the scatters of CT2, CT3, SZ1–3, and those of the ASD parents and the elderly group, all in relation to the deafferented participant pointing in the dark and pointing with visual guidance. Despite the overlap between some SZ and ASD parents, as well as the overlap of some SZ with the elderly participants, the average SZ groups stand on their own with the highest kurtosis in the estimated distributions. Figure S6 in Supplementary Material also shows the probability plots of SZ as a group (including all ages) in relation to CT3. Notice the deviation from the line of unity indicating departure from the normal distribution. Further, using the chi-square goodness of fit test for each individual age subgroup yielded significant deviations from normality (*p* < 0.05, *p* < 0.01, and *p* < 0.01 for young SZ1, middle-aged SZ2, and older SZ3 patients, respectively). The separation between SZ statistics and those of the deafferented participant in both pointing conditions is also evidence in this plot. The difference between the subjects with SZ and the ASD parents is also shown (parents divided by sex – red females and blue males). In summary, the statistics of SZ patients are atypical, highly heterogeneous, and different from those of other disorders.

### Temporal Differentiation between Forward and Retraction Movement Segments Differ across Neuropsychiatric/Neurological Disorders and Typical Cross-Sections of the Population

Examination of the normalized PV index, as a function of the proportion of time to reach the PV for each group within the study, revealed differences between the forward and the backward segments of the pointing loop. These differences for the control groups CT1, CT2, CT3, elderly participants, and ASD parents can be appreciated in Figures S1A in Supplementary Material (forward) and Figure S1B in Supplementary Material (retraction with the shift from the forward case). As the parents of individuals with ASD demonstrate the largest shift, we divided the group by sex into mothers and fathers. The legend provides information on the group type.

In this graph, along the *x*-axis, we plot the proportion of time to reach PV. Along the *y*-axis, we plot the normalized PV index. Typical ranges along the horizontal axis are about 0.5, where the peak tends to occur (midway between the two local minima). Lower values indicate reaching the speed maximum earlier, whereas higher values indicate reaching it later.

The graph in Figure S1A in Supplementary Material shows a separation between the younger CT1–3 groups and the remaining subjects, whereby the latter are slower on average (with higher normalized PV index). Along the time axis, these younger participants fall near 0.5, while the elderly participants and the parents vary. The elderly participants reach the peak earlier than half-way to the pause, and in the parents the mothers are similar to the younger controls but the fathers reach PV later. Indeed, the fathers of this cohort showed the largest difference in the proportion of time to reach the PV between the forward and the backwards reaches. The retractions peaked much earlier and the value of the normalized PV index dropped. This indicates a much faster retraction (as with the deafferented subject) whereby the end effector is retracted under less control. Specifically, the fathers were the slowest of the cohort to reach the velocity peak during the forward reach phase, but the retraction phase was almost a jerky, nonsmooth motion, suggesting poor motor control of these spontaneous, uninstructed reaches. The signatures of the deafferented participant are plotted for the cases of pointing in the dark and pointing when aided by visual feedback. In the dark, the deafferented participant reaches the PV earlier and moves faster on average than when guided by vision. This is expected and consistent with his description of feedforward strategies to initiate the motion and rely on visual feedback to compensate for the lack of proprioception. Evidence for feedforward control in IW was also shown in a mirror drawing task (33).

In the parameter plane of Figure S1B in Supplementary Material, we also plot the shifts comparing the timing in forward and back movements of the participants. Here, the ASD parents have the largest shifts in the retraction movements, toward the 0.5 values of the proportion of time and toward faster average speed along the *y*-axis. They perform more similar to controls in the retraction than in the forward segment. IW also shows changes in these patterns, with shifts in the opposite direction for the two conditions that he performed. Notice that the youngest group of children and the elderly participants show no discernible shift between forward and backwards reaches in these plots. Figure S1 in Supplementary Material also shows the distributions of these two parameters per group along each axis.

Figures S1C,D in Supplementary Material show similar plots for the patient groups under study in relation to the deafferented subject. All patients have slower average speeds than controls but the mild PD group shifts toward typical levels in the retractions. The severe PD group is, as expected, the slowest and tends to reach the PV earlier than the mild PD group. The distributions (color coded as in the legend) also show the differences across groups.

Further analyses of the temporal features for each group were performed on the actual time to reach the PV. The results are shown for each group in Figure S2 in Supplementary Material. Here, the axes are not adjusted, so as to allow for greater appreciation of the shapes and dispersions of the histograms within each group and in the forward and back condition. Note that the differences in scale along the *x-axis* of the time to reach the peak would prevent clarity if all graphs were set to the same scale. The Gamma plane summarizes the estimated Gamma parameters for each group and set of conditions under analysis (forward and retraction). The legend shows the color-coded data corresponding to the frequency histogram (each data set comprises 1000 data points randomly selected from the estimated Gamma parameters of the entire set). The ASD group is by far the most skewed, consistent with previous analyses of their temporal kinematics while performing other tasks (8, 32). The probability plot in Figure S2B in Supplementary Material (right panel) accompanying Table S7 in Supplementary Material captures the departure from normality of this temporal parameter in the ASD cohort. Note that in the Gamma parameter plane of Figure S2 in Supplementary Material, the temporal stochastic signatures of the ASD group (located at [2, 0.2]) in the shape, scale parameter plane, separate from the controls their age (located at [7, 0.1]), but close to the signatures from the elderly participants (located at [2, 0.1]) and the SZ patients (overlapping with the elderly participants).

As with the normalized PV index, this graph also shows that control groups CT2 and CT3 have the lowest noise-to-signal ratio and the largest shape values tending toward the Gaussian ranges of the Gamma plane. In stark contrast the temporal signatures of the subjects with severe PD and the deafferented participant are located far from the ideal controls.

### DISCUSSION AND SUGGESTED FURTHER STEPS

This work characterized the statistical ranges of velocity-dependent fluctuations in pointing performance across a large heterogeneous cohort of human participants. Besides characterizing the signatures of cross-sections of the typical population across the life span, such signatures were also empirically estimated for individuals with neuropsychiatric and neurological disorders. These included neurodevelopmental (ASD) and neurodegenerative (PD) disorders. We also studied the performance patterns of patients with SZ, a syndrome with onset of its characteristic set of features (i.e., psychotic symptoms) in early adulthood (in contrast to ASD and PD), and for which very little motor control research exists. Finally, we included a deafferented participant who lacks proprioception due to damage in afferent fibers conducting touch, pressure, and movement information from the periphery to the brain.

The main purpose of this work was to estimate statistical ranges across the population so as to initiate a path of change in statistical analyses from the current "one-size-fits-all" model, to a more personalized approach in line with current NIMH initiatives. Here, we have shown that the typical patterns differ across different groups within the population, with ideal Gaussian-like shape and the lowest noise-to-signal ratio in the control subjects from college to middle age. The younger children and the elderly participants at the extreme of the bell curve of the human life span have different distribution shapes and dispersion that give rise to different summary statistics. In contrast to the ideal statistics from the young typical controls, the ASD, PD, SZ, and the elderly groups had marked differences along at least one of the moments' axes that we empirically estimated here. Such differences were detected (Tables S6–S9 in Supplementary Material) above chance across the clinical groups.

In addition to the grouped data analyses, we also showed that the summary statistics spaces localize each individual with respect to other individuals. Without color coding the scatter by disorder or age, we can see a gradient of differences with clear separation between ASD and controls. Likewise, we can see a separation between mild and severe PD that is quite unambiguous. However, the SZ patients were more heterogeneous than the other groups with brain disorders, as can be seen in **Figures 4A,C**, where overlap with other groups is evident. They are mixed with the elderly participants, the middle-aged CT3, the ASD parents, the mild-to-moderate PD, and the severe PD groups. Two SZ patients fell at the tail of the ASD cluster. Of note, we know that other kinematic parameters unambiguously separate SZ from matched controls (Nguyen et al., under review1 ), which suggests that including more dimensions in the data representation across disorders would be more illuminating than projecting all information on one plane, or restricting our analyses to one set of parameters. Likewise, there is an imperative need to report psychotropic medication intake to researchers who study motor control in order to assess motor variability as a function of dosage, time under medication, and medication combinations, among other factors. It is broadly reported that psychotropic medications are known to have variable side effects on movement patterns. In the present cohort whether or not patients were on medication, motor noise signatures were different from those of controls. Yet the present personalized methods allow more detailed analyses based on medication status. This is an important additional dimension that needs to be explored at the individual level in future analyses.

The velocity-dependent parameters used here could unambiguously detect differences in levels of severity in PD and unique levels of noise in ASD. Regardless of age, sex, or medication status, these patterns were distinguishable in this cohort when labeling the locations by clinical condition. Likewise, taken as a group, the SZ had radically different distributions from the groups with overlapping ages, CT2, and CT3. At the individual level though, the data from the SZ patients also emphasize the need for a personalized approach to this devastating yet heterogeneous disorder. In this sense, the case of PD is relevant as it shows that the clinical diagnosis already does a good job at characterizing the emergence of relatively homogeneous subgroups as the disease progresses. The utility of these analytical tools in ASD has yet to be clinically confirmed as at present the diagnosis does not include motor symptoms at all. Likewise in ASD and SZ, it will be important to ascertain the effects of medication intake on motor patterns, a task that is now possible with this statistical platform.

A surprising result emerged from the data on ASD parents. Specifically, the signatures of ASD parents did not match those of typical controls their age. Instead, their signatures matched those of the elderly and the PD groups. There are many possible reasons and combinations of reasons for this result that we shall investigate in future work. One reason could be due to overall parental stress levels. Other reasons could include symptombased medication intake (e.g., antidepressants, stimulants, etc.) and/or genetic predisposition. Given their much younger ages than the elderly and severe PD groups, it may be useful to study the rate of change of these patterns in ASD parents. In particular, tracking the evolution of the motor signatures with age may illustrate whether the signatures of their motor patterns are in an atypical accelerated state of change.

These data sets provide insights into the general motor statistics of these populations, and underscore the need for a personalized medicine approach to psychiatric and neurological disorders. In particular, the case of PD, a disorder that is diagnosed and tracked based on visible changes in motor patterns, beautifully illustrates the potential utility of motor noise-based biomarkers to characterize each person relative to the rest of the population. The two cohorts of PD, mild-to-moderate and severe, could serve as anchors to reference other more heterogeneous disorders such as SZ. Some of the SZ patients had patterns comparable to those of severe PD while others fell closer to those of the mild-to-moderate PD group, and yet others were overlapping with CT3 participants. Statistical distance metrics based on this type of sensory–motor noise may help us discern alterations in motor feedback as a function of anxiety, dopamine receptor-blocking medications, and other factors in these populations.

One of the most striking features in the data from subjects with SZ, besides its heterogeneity in the motor domain, is the lack of similarity with the deafferented participant. Unlike the ASD and severe PD groups, who were close to the deafferented participant pointing in the dark, the SZ patients fell far from IW in the statistics parameter space. Subject IW lacks feedback from the fibers that transmit touch, pressure, and movement information but has temperature and pain channels spared. The result is noteworthy given the reports in the SZ literature of problems with thermoregulation (34–36) and higher thresholds for pain perception across patients relative to controls (37–40). It will be interesting to investigate velocity-dependent motion parameters as a function of those autonomic signals in SZ. The present results demonstrate that proper statistical analytics applied to continuous recordings are required to provide more meaningful answers to basic research questions and establish the nature of the relationships between these afferent inputs and specific sensory–motor deficits in anticipatory behavior and volitional control.

In the context of internal models for action (IMA), it is possible that across these neurological disorders there are differentiable and selective disruptions in various components of forward planning linked to different levels motor noise. Among brain areas that are thought to be important for forward computations within the framework of IMA are the cerebellum (41, 42) and the posterior parietal cortex (PPC) (43). The cerebellum is known to be a problematic brain structure in all of these disorders (44–51). Likewise, connectivity issues between parietal and motor cortices have been reported in all these patient types. In SZ, this has been the case (52, 53). In ASD, connectivity problems are also reported (49), and in PD, striatofrontal regions seem to be affected (54) possibly impacting parietofrontal loops involved in forward planning and decisions. Notwithstanding issues with imaging studies (55), in the light of problems with the velocity-dependent signals that we have quantified here at the motor output level, it is possible that communication between these key nodes of the brain and the periphery may be corrupted by excess motor noise partly impeding the continuous afferent and reafferent flows from the periphery and possibly disabling predictive coding.

Velocity-dependent peripheral input signals from selfproduced, goal-directed motions are an important source of guidance to the brain. They help compensate for synaptic transductions and transmission delays. In the context of visually

<sup>1</sup> Nguyen J, Majmudar U, Papathomas TV, Silverstein SM, Torres EB. Schizophrenia: the micro-movements perspective. (under revision, Neuropsychologia, 2015).

guided reaching, areas in the PPC are known to be important for the planning and execution of such actions. Regions in the PPC receive eye position and velocity afferent inputs via ascending prepositothalamocortical pathways (56). Proprioceptive inputs required for proper visuomotor geometric transformations for reaches (57) have also been found to converge to the PPC (58) from the dorsal column nuclei and the postcentral somatosensory cortex. Given the putative roles of the PPC in forward prediction (43, 59, 60), trajectory formation (61–63), and geometric visuomotor transformations (64, 65) along with its cerebellar inputs to the lateral and medial intraparietal (LIP and MIP) cortical areas (66), we suggest that the motor–PPC–cerebellar networks may be selectively disrupted across these disorders, and that part of this disruption is due to poor continuous updating involving afferent sensory guidance from more than one sensory–motor channel.

Afferent sensory channels convey reafferent kinesthetic signals from mechanoreceptors involving touch, pressure, and ongoing self-produced movements. They also convey pain signals from nociceptors and temperature-related signals from thermoreceptors (67). The present work identifies interference with kinesthetic reafference from ongoing movements, but it will be important to examine afferent deficits concerning thermoregulation and pain perception, as these contribute to corporeal self-awareness. Corporeal self-awareness is critical in forward computations and geometric transformations bound to be disrupted in the face of excess motor noise found here in all disorders.

The new analyses reveal striking statistical differences between mild-to-moderate and severe stages of PD. In particular, these two cohorts have selectively overlapping features with the deafferented subject IW. In the case of the mild PD, the signatures overlap with those of IW under conditions of visual guidance. Interestingly, an overreliance on visual feedback has been reported in PD (7, 68–70), along with a new view that proprioceptive coordination may become impeded as the disease progresses (71, 72). In a previous study, an egocentric frame of reference for visual guidance (anchored at the moving finger), but not an allocentric frame of reference (anchored at the external target), helped mild PD patients improve many aspects of their pointing trajectories (7). Vision alone is not useful to the patients with mild PD, but vision aligned with self-generated motion shifts their movement statistics to typical ranges (7). Given the statistical similarity of the mild PD and IW with vision, it is possible that in mild PD patients, the signatures of reafferent minute motor fluctuations that we found to be corrupted by noise and randomness may improve when guided by vision. In the case of severe PD, their signatures rather overlapped with IW as he pointed in the dark. This is also interesting as timing in their bradykinetic motions was comparable to those of the ASD parents. This was a rather surprising finding given the age disparity and the lack of any kind of neurological diagnosis. This result suggests further study of familial ASD.

Traditional studies of motor control assume normality in the distributions of kinematic parameters. This work shows that there is a range of skewed distributions, from the memoryless exponential to the symmetric Gaussian in the velocity-dependent code of pointing behaviors. This result underscores the importance of providing an empirical characterization of the statistical properties underlying human movements. By assuming normality and smoothing out as noise the motor output fluctuations, we miss important information in the data from both typical and pathological conditions. This work also highlights the significance of individualized statistical assessments that may enable the discovery of self-emerging patterns inherent in the data. Analyzed as an ensemble using clinical labels, the data are very revealing when empirical statistical estimation is used, rather than theoretical assumptions and homogeneous treatment of the data. The clinical literature of motor control makes a number of assumptions that may blur the true features of the kinematics data from neuropsychiatric and neurological disorders. This work emphasizes the importance of reconsidering those traditional practices and researchers teaming up with clinicians to better inform datadriven approaches.

### Implications of the Characterization of Motor Noise for Genetics Research

We have demonstrated here the importance of providing empirical estimation of the statistical features underlying motor behaviors. Across different disorders of the nervous systems, we were able to characterize the ranges of statistical parameters that are traditionally treated as homogeneous under the assumption of normality in the movement data. The noise that is traditionally smoothed out through data averaging and the retracting movement segments that are often discarded as nuisances in the data revealed fundamental differences across neurological disorders that may be of use to genetics research. Specifically, classification of different types of sensory–motor noise may be possible and may aid in linking specific genetic factors that give rise to selectively different levels of synaptic noise with differentiable levels of sensory–motor noise. The specificity of these biometrics has yet to be tested, and better instrumentation discerning motor from sensory noise in electromyography combined with high dimensional kinematic signals will be required. We need to unveil the origins of synaptic noise in the first place before understanding different gradients of sensory–motor noise. Yet, the same personalized statistical platform presented here can be used to examine time series of other related signals. In disorders of known etiology, it should be possible to investigate these questions so as to build similar statistical maps to those presented here, whereby genetic factors and their resulting synaptic noise would be another data dimension. Such questions can be addressed using the present statistical platform.

As presented here with the deafferented subject in different contexts, we could assess the patterns of sensory–motor noise from individuals that go on to receive a diagnosis of ASD, SZ, or PD but for whom a genetic history is available. One such a group is the Fragile X-related disorders, where premutation carriers may receive an ASD diagnosis at an early age or a PD misdiagnosis at a later stage in life, or a diagnosis of mood and other psychiatric disorders in the case of female premutation carriers (73). We suggest a new research program linking these disorders and deafferentation whereby the same statistical platform that we term "precision phenotyping" in this work could be used to better characterize this family of disorders in the human population at large. In this sense, the present results may be an important step toward developing a new analytical platform for Precision Psychiatry.

### AUTHOR CONTRIBUTIONS

Given the diverse population and clinical expertise, the following breakdown of contributions is in order: ET conceived study, analyzed all data, and wrote paper. Autism: RI, ET, and CW collected data; RI, CW, JN, and JJ designed study/analyses; and JN performed clinical assessment. Parkinson: RI and ET collected data; JS performed clinical assessment. Schizophrenia: JN and TP collected data and designed study/analyses; SS performed

### REFERENCES


clinical assessment. Deafferented subject: JC collected data and performed clinical assessment. All authors participated in the editing and approval of the final version of the manuscript.

### FUNDING

The Nancy Lurie Marks Family Foundation to ET (Career Development Award).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fneur.2016.00008


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Torres, Isenhower, Nguyen, Whyatt, Nurnberger, Jose, Silverstein, Papathomas, Sage and Cole. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Exploring music-based rehabilitation for Parkinsonism through embodied cognitive science

*Andrea Schiavio1,2\* and Eckart Altenmüller3*

*1School of Music, The Ohio State University, Columbus, OH, USA, 2Department of Music, The University of Sheffield, Sheffield, UK, 3 Institute of Music Physiology and Musician's Medicine, University of Music, Drama and Media Hannover, Hannover, Germany*

Recent embodied approaches in cognitive sciences emphasize the constitutive roles of bodies and environment in driving cognitive processes. Cognition is thus seen as a distributed system based on the continuous interaction of bodies, brains, and environment. These categories, moreover, do not relate only causally, through a sequential input–output network of computations; rather, they are dynamically enfolded in each other, being mutually implemented by the concrete patterns of actions adopted by the cognitive system. However, while this claim has been widely discussed across various disciplines, its relevance and potential beneficial applications for music therapy remain largely unexplored. With this in mind, we provide here an overview of the embodied approaches to cognition, discussing their main tenets through the lenses of music therapy. In doing so, we question established methodological and theoretical paradigms and identify possible novel strategies for intervention. In particular, we refer to the music-based rehabilitative protocols adopted for Parkinson's disease patients. Indeed, in this context, it has recently been observed that music therapy not only affects movement-related skills but that it also contributes to stabilizing physiological functions and improving socio-affective behaviors. We argue that these phenomena involve previously unconsidered aspects of cognition and (motor) behavior, which are rooted in the action-perception cycle characterizing the whole living system.

Keywords: embodiment, music therapy, Parkinsonism, dynamic systems, brain plasticity, motor rehabilitation, well-being

## INTRODUCTION

Over the last three millennia, across different times, places, and cultures, music making, and music listening have been often associated with medicine (1), meditation (2), and well-being (3), serving a variety of functions deeply intermingled with everyday-life and social activities (4–9). In Ancient Greece, for example, human musical behaviors were not considered as contemplative or abstract practices, but were rather actively employed for education, religious ceremonies, celebrations, and, indeed, medical treatments (10, 11). More systematic therapeutic interventions involving music emerged after the Second World War – for example to help ex-soldiers or injured civilians recovering from stress and other related conditions (12, 13). Rehabilitative protocols adopted in this period were mostly based on models provided by the social sciences of the day – where the "cultural role

### *Edited by:*

*Cathy Craig, Queen's University Belfast, UK*

#### *Reviewed by:*

*Pedro Chana, Universidad de Santiago de Chile, Chile Matthew Rodger, Queen's University Belfast, UK*

> *\*Correspondence: Andrea Schiavio Schiavio.1@osu.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 24 July 2015 Accepted: 26 September 2015 Published: 19 October 2015*

#### *Citation:*

*Schiavio A and Altenmüller E (2015) Exploring music-based rehabilitation for Parkinsonism through embodied cognitive science. Front. Neurol. 6:217. doi: 10.3389/fneur.2015.00217*

of music was interpreted as an effective facilitator for therapeutic concepts of 'wellbeing'" [(14), p. 174]. As such, the focus was on exploring how the employment of music could alleviate pain, promote emotional expression and sociality, motivate patients, and enhance self-esteem (15, 16).

From the early 2000s, with the unprecedented development of brain sciences and neuroimaging techniques, the study of music therapy shifted to a new, highly stimulating, research focus. Mirroring the same reorientation witnessed in other disciplines devoted to the study of mind,1 agency, and behavior, many scholars started to explore in greater details the neurological aspects related to musical activities in clinical and non-clinical contexts [see Altenmüller and Schlaug (17–19), Janata and Grafton (20), and Thaut (21)]. Within this area, a wealth of empirical evidence has showed the high degree of functional and structural plasticity of the human brain when involved in the complex demands associated with musical activity (22–26). For example, it has been demonstrated that intense Melodic Intonation Therapy (27, 28) may elicit – in patients suffering from non-fluent aphasia after left frontal lobe damage – the reactivation of inhibited languagecompetent brain regions in the right frontal brain networks (29–34). Additionally, other findings have confirmed the benefits of music-supported therapy in motor rehabilitation: first, studies with stroke patients revealed significant behavioral improvements in a variety of tasks related to fine motor skills (35, 36), accompanied by impressive reorganization of cortical sensorimotor networks (37, 38); second, research with Parkinson's patients has shown that entrainment with a rhythmically rich auditory feedback may alleviate Parkinsonian gait by "increasing the excitability of spinal motor neurons via the reticulospinal pathway, thereby reducing the amount of time required for the muscles to respond to a given motor command" [McIntosh et al. (39), p. 25; see also Arias and Cudeiro (40)]. Increasingly, the clinical adoption of music-based paradigms seems to offer not only a valid non-pharmacological tool for intervention in diverse contexts [including for example pain treatment, see Bernatzky et al. (41)] but also innovative insights into the anatomy and physiology of the brain [e.g., Särkämö et al. (42)]. In general, a rich variety of empirical findings have demonstrated how musical experiences may improve the lives of patients suffering from various neurological diseases [e.g., Forsblom et al. (43) and O'Kelly et al. (44)], integrating neuroscientific and musical research in novel and fascinating ways (45, 46).

To this already fertile ground, we would like to add insights from the recent *embodied* trend, which has recently emerged in cognitive science and in philosophy of mind [e.g., Lakoff and Johnson (47), Shapiro (48), Stewart et al. (49), and Varela et al. (50)]. This framework has contributed a new and important perspective on the sciences of mind and (inter)subjectivity, with its central thesis2 being that cognition "depends on the kinds of experiences that come from having a body with particular perceptual and motor capacities that are inseparably linked and that together form the matrix within which memory, emotion, language, and all other aspects of life are meshed" [Thelen et al. (51), p. XX].

We argue that framing music-supported therapy within a paradigm inspired by this claim may offer useful new ways of interpreting results obtained in clinical settings, and in turn potentially improve specific protocols for interventions. Rehabilitative strategies for Parkinson's patients, in particular, may necessitate a more unitary, holistic, view to fully appreciate the potential of music and its relevance beyond movement recovery only. This perspective aligns with recent non-reductionist trends in critical neuroscience (52–56), which emphasize the deep continuity of mind, behavior, body, brain, environment, affectivity, perception, and action; it thus contrasts with more traditional approaches where such elements are usually studied as discreet (and causally related) categories [see Colombetti (57), Kiverstein and Miller (58), and Thompson (59)].

In what follows, therefore, we discuss the need to implement insights from embodied cognitive science in research on the brain's anatomical adaptation and for music-based motor rehabilitation. First, we introduce the embodied approach by analyzing its main tenets and its role in neuroscientific and musical contexts. Here, a brief overview of the '4Es' perspective – which, as represented in **Figure 1**, defines cognition as Embodied, Embedded, Enactive, and Extended – is offered. Subsequently, we shift our focus to Parkinsonism, asking whether (and how) established rehabilitative protocols may benefit from the adoption of such compelling perspective. Finally, we explore possible clinical applications that the '4Es' approach may inspire, showing how these may bring forth a richer understanding of the complex network of dynamical interactions between music, environment, body, brain, movement, and well-being.

### VARIETIES OF EMBODIMENTS

The traditional 'cognitivist' approaches that dominated cognitive sciences for more than 50 years developed a productive research agenda that focuses principally on the role of mental representations, computations, and specialized cognitive architectures (60– 62). However, analyzing how external information is acquired, processed, and represented3 'in the head' scholars within this framework are often accused to not adequately take into consideration the body and the ecological niche in which the cognitive system is embedded (50, 63–65). Classic cognitivism, it is argued, downplays the active and adaptive engagements that unite living bodies and niche for the constitution of lived experience; and thus, in human terms, it ignores the most fundamental aspects of our being-in-the-world (66–68). Accordingly, the cognitivist framework may be seen to support a strong dichotomy between the *inner* domain of mind – functionally realized 'in the head' thanks to relevant domain-specific cognitive modules – and the *outer* realm of the (social and physical) 'objective' world, including the system's own body (59, 69).

<sup>1</sup> 'Mind' and 'cognition' are used as synonyms in this paper.

<sup>2</sup>As we will see, this main claim inspired the development of different research programs.

<sup>3</sup>Representational structures are usually defined by computational procedures that operate in light of (are realised by) functionally autonomous cortical regions.

In contrast to this 'orthodox' (70) perspective, various theories of *embodied cognition* have recently emerged as new frameworks for the study of human mind. Such approaches are supported by a growing cross-disciplinary research agenda that integrates relevant contributions in theoretical biology (71), dynamic systems theory (72), linguistics (73), neurophysiology (74), phenomenological philosophy (75), cognitive science (59), and artificial intelligence (76). In general, these embodied frameworks emphasize the formative roles of bodies and environments in driving cognitive processes (59, 77–79), as well as the primacy of action over more 'intellectual' faculties to make sense of the world (50, 80, 81). As a consequence, cognition is now often described in terms of dynamic sensorimotor interactions between the entire body of a living system and its environment (49, 82).

While such core insights are widely endorsed by advocates of the embodied approach, its richly interdisciplinary agenda has resulted in a number of interesting formulations and interpretations (83). This growing variety of approaches to 'embodied cognition' has stimulated the discussion across diverse fields – promoting a highly fruitful exchange of knowledge, methodologies, and insights, while nevertheless preventing the development of a 'standard' framework. In talking about the *embodied approach*, therefore, we actually refer to different research programs: Embodiment, Embedment, Enactivism, and Externalism (usually labeled as '4Es'), which all aim to capture how bodies, brains, and environment successfully interact in real-time worldly conditions (84). These approaches hold that to understand mind we should consider how a living system *acts* in a social and physical environment (85) rather than focusing on what goes on 'within the skull' only. In order to get a very basic idea of the '4Es' perspective, we introduce the following key points:


It is beyond the scope of this article to discuss each of the '4Es' approaches in detail; thus we will draw from them selectively – adopting both overlapping principles and distinctive insights (in particular from the *enactive* view) when necessary. While the debate over these perspectives is still heated in philosophy of mind and cognitive science, the embodied paradigm (in its four 'E' instantiations) has received little discussion in the context of music-based rehabilitative paradigms. But before we focus on how specific therapeutic settings may integrate existing methodological and theoretical models with insights from the embodied perspectives (mainly with regard to Parkinson's treatment), it will be necessary to consider three basic principles associated with embodied cognition and analyze their role in neuroscience (86–88) and music cognition (89–94).

### TRACKING DOWN EMBODIMENT

Although the '4Es' define different research agendas, they all maintain that "cognition is embodied" as their starting assumption4 [see Hanna and Maiese (96) and Ward and Stapleton (97)]. But what does it really mean? Broadly speaking, it is important to understand *embodiment* not as a given category that may facilitate certain aspects of perceptual and cognitive activity, or as a label to attach whenever bodily aspects are somehow involved in certain cognitive tasks. Rather, 'embodiment' should be intended as the pre-requisite of an agent's being-in-the-world (98, 99). Listening to music, thinking of a good life event, feeling sad, sharing a drink with a friend, and every other possible activity we may have experience of, are all conceivable only through our *living* and *lived* body (67, 100, 101). As Di Paolo and colleagues argue: "to say that cognition is embodied is to express a tautology – it simply cannot but be embodied" [Di Paolo et al. (102), p. 42]. This passage is best understood when considering what Varela et al. (50) define as the "three dimensions of embodiment": bodily self-regulation, sensorimotor coupling, and intersubjective interaction. Analyzed by several contributors [e.g., Thompson (59)], these insights offer a thorough perspective on the embodied view, emphasizing the explanatory power of moving beyond cognitivism across different levels of analysis. Importantly, as we will see, these "three dimensions" are extremely relevant for our discussion on musicbased therapy for Parkinson's disease (PD) patients, which we offer below.

### Bodily Self-Regulation

*Bodily self-regulation* concerns the way an agent's biological structure contributes in regulating, modifying, and controlling its homeodynamic requirements. These processes of metabolic autonomy ensure that the agent is alive and that it maintains a stable interaction with the world. Importantly, as reported by Colombetti (103), there is no 'self ' in self-organizing organisms: no 'message' is exchanged in hierarchical fashion between different independent levels via top-down or bottom-up pathways (104). Instead, the chemical, thermodynamic, and metabolic activity of the system's sub-networks participates as a whole in maintaining the system's homeostatic adaptivity. The process leading to adaptive stability, in which the living system (i.e., a unicellular organism, a mammalian, etc.) strives to maintain its autonomous identity, is realized through self-producing all that is needed for its maintenance (105, 106). The process, in other words, is not led by a 'ghost in the machine,' but rather by homeostasis (103).

<sup>4</sup>Although there is certainly a continuity among these perspectives, it should be noted that some of the arguments used by proponents of the extended mind thesis stand in open contrast with an enactive characterisation of cognition, and might eventually collapse into a functionalist-like framework, where computations and representation would still play a key role in driving cognitive processes, even if coupled with external resources [see Thompson and Stapleton (95)].

Consider the role of emotions, for example: seen as self-regulative processes (107, 108), they emerge within the dynamical interaction of a number of neural and extra-neural components, and not simply via an input–output sequential chain of events (57). Indeed, although defining an operationally closed network,5 the self-regulating processes aimed at keeping the agent's conservation as auto-sufficient, do establish a meaningful dialectic with the environment: "whence the intriguing paradoxicality proper to an autonomous identity: the living system must distinguish itself from its environment, while at the same time maintaining its coupling; this linkage cannot be detached since it is against this very environment from which the organism arises, comes forth" [Varela (109), p. 85]. By this view, all living systems are "self-organizing thermodynamic systems with emergent truly global or inherently dominating intrinsic structure, and not mere mechanisms like a can-opener or a digital computer" [Hanna and Maiese (96), p. 20]. The integrity of self-regulative processes always involves world, body, and brain at multiple levels and timescales (110). With regard to human musicality, these insights have been recently explored by research on bodily self-regulation in joint improvisation (111) and by the development of an enactive theory of musical emotions (Schiavio et al., under review).

### Sensorimotor Coupling

The second dimension of embodiment is *sensorimotor coupling*, which may refer to (i) the integration of sensorial and motor information occurring in the human brain (112), and (ii) the embodied forms of mutual determination established by organism and environment (113). While (i) and (ii) should always be considered as mutually dependent [(114), and see discussion in the Section "Intersubjective Interaction"], for reasons of simplification we now briefly treat them separately.

Perceptual processes, traditionally, are identified with a unidirectional stream of data from the world 'out there' that is retrieved, codified, and represented 'in the head,' eventually leading to a behavioral output (movement) (115). This process is putatively made possible by an exchange of information proceeding from the associative cortex to the agranular frontal cortex – where information is integrated with more sophisticated (i.e., decision making) aspects of intelligence. Modern neuroscience, however, is well aware of the limitations of this traditional model. Consider, for example, the highly complex cytoarchitectonic organization of the frontal lobe's motor cortex: as Gallese (115) notes, a number of anatomical and neurophysiological findings have revealed a rich variety of anatomo-functional areas, each endowed with specific functional properties and related to each other forming distinct cortico-cortical circuits (116). This means that each of these parieto-premotor circuits continuously participates in integrating sensorial and motor information, contributing in redefining the role of the motor cortex – from a mere 'muscle controller' to a much more complicated system (74). Here, within the inferior frontal gyrus, the lower part of the precentral gyrus and the temporal, occipital, and parietal visual areas (117) the existence of a so-called 'mirror' system (116, 118, 119) has been posited to indicate a set of bimodal and trimodal neurons, which are elicited not only when doing a given action but also when observing (and/or hearing, in the case of trimodal neurons) another individual performing the same action6 (120, 121). Thus, it is argued that in the brain, perception and action are not separated entities somehow encapsulated in autonomous and independent modules. Rather, they are always mutually integrated through a complex web of sensorimotor connectivity, involving anticipatory mechanisms that enable the system to respond adequately to the demands of the environment (122, 123).

In league with this discussion, a number of empirical findings report the activation of neural circuits involved in motor activity and the planning of motor sequences during listening tasks (22, 124). In a well-known PET study, Halpern and Zatorre (125) demonstrated that when musicians listen to or imagine music, blood-flow significantly increases in the right supplementary motor area (SMA), a region which is implicated in motor control [see also Kristeva et al. (126)]. As Rodger et al. (127) comments, the involvement of SMA and other brain areas – i.e., basal ganglia, and cerebellum – in similar tasks (128, 129) is usually seen to support "hypotheses about the induction of a sense of beat or pulse in the listener" [Iyer (130), p. 392]. To this, we add that these findings also reflect more generally the ecological situatedness of the whole organism: listening to music involves an active, skillful, sensorimotor, exercise, which is intrinsically determined by the *sensorimotor expertise* (in terms of motor vocabulary of musical actions, for example) of the musical animal – i.e., its personal capacity to co-constitute (and act in) its niche, through the establishment of a repertoire of meaningful relationships by which it maintains its autonomous identity or a 'point of view' (131–134). This resonates closely with a main principle of the 'enactive approach,' namely, the idea that perception and action are radically entwined extraneurally in non-linear terms – and that this forms the basis for our being-in-the world (59). Put simply, from this perspective, it is not only the brain that is exposed to musical feedback. Rather, the entire living system – with his or her listening biography, body, affectivity, and history of structural couplings with the (sonic and cultural) environment – *participates* as a whole in musical experience (46, 93, 127, 135–138). We will further develop these insights when discussing of PD treatments.

### Intersubjective Interaction

The third dimension of embodiment – *intersubjective interaction* – aims to look beyond traditional 'mentalistic' approaches of social cognition, which often conceive of social understanding in terms of simulation-like mechanisms7 or through the

<sup>5</sup> Indeed, the internal self-regulative processes are theoretically sufficient to maintain the agent's homeostatic stability.

<sup>6</sup>As it is well known, these "mirror neurons" do not code the precise movement performed by another individual but, rather, the *goal* of the given action. As long as the goal of a given action is different, these neurons are selective enough to distinguish between different kinematic schemas (74).

<sup>7</sup>The theory holds that understanding the behaviours and the beliefs of others is a matter of simulating them internally with my own cognitive system – "as if " I was behaving that way or possessing that belief.

construction of theoretical, spectatorial, models8 (139–142). Embodied – enactive in particular – approaches to interactivity, instead, define the processes of mutual interactions and coordination as *self-regulative* and *sensorimotor* networks (143–146). These networks are based on recursive patterns of action and perception mutually shaping each other dynamically (147). Consider two (or more) individuals playing together: no matter how much they rehearsed jointly or how many times they played the same piece, there will always be a sense in which each performance is different from one another, as even one brief 'crescendo' by a musician (or a particular environmental setting, or audience, etc.) will affect the other and the overall performance in real time [(138, 148, 149), p. 40]. As cognition is a process that occurs in a domain of interactions, it is realized through the *biological morphology* of the body and its dynamical and *sensorimotor interplay* with the others, where these aspects represent different *typologies* of embodiment and not separated domains (150). The body is not a rigid and fixed object, but rather a flexible entity that acts in (it modifies and responds to) the world (151); it is an "imprint of social engagement" (ibid.). The *living* and *lived* body is what allows the meaningful interactions with its environment; it is, as we stated above, the pre-condition for being-in-the-world (67, 152). The brain, accordingly, can be seen as 'participating' in the action rather than 'controlling' it9 (122). If cognition is realized in the domain of the system's meaningful and embodied *interactivity*, it is not 'located' in any traditional sense; rather it bypasses the notions of 'internal' and 'external' (95, 114).

The notion of embodiment, therefore, encompasses all the processes connected to living subjectivity, shaping, and being shaped by the environment in which an agent is embedded. This does not only entail 'basic' processes – such as perception or interaction: there is growing agreement across a variety of domains in highlighting the body's crucial role for high-level skills – such as problem solving and reasoning (153). Along these lines, it has been demonstrated that visual and rhythmic perception are shaped by looking and by body movements in both infants and adults (154–157), that motor experience facilitates memory for musical excerpts (158), and that walking is crucial for an infant cognitive development (72). Put simply, if we reduce mental life to the activity of the brain and the central nervous system, we may lose an important chance to understand the organism as ecologically situated – where bodies are not reduced to representations in the somatosensory cortex but are instead seen as constitutive category of the system's being-in-the world. Embodied theories entail both "micro phenomena within the body, for example, the physiology of sight, the biochemistry of muscle cell contraction, [and] macro phenomena, for example, the evolution of ecosystems" [Krieger (159), p. 351]. While the relevance of these insights is recognized by different authors in the context of cognitive science and philosophy of mind, it remains partially unexplored within other domains. In the Section "A Network of Non-linear Interactions," thus, we will consider the challenge posited by the embodied approaches more in detail, discussing how they may help us reconsider the ways in which we look at brain science.

### A NETWORK OF NON-LINEAR INTERACTIONS

In the last few decades, a growing number of researchers became interested not only in analyzing the cognitive operations in play while performing a musically relevant task, but also in understanding how are these operations associated with particular (networks of) brain regions10 (161, 162). Although most neuroscientific research has moved from functional *segregation* to functional *integration* [see Friston (163, 164)], and a number of brain scientists expressed doubts toward both neural localization and models based on a mind-brain identity as legitimate explanatory tools [see Bennett and Hacker (165) and Fuchs (166)], the tendency to look for 'neural correlates of music processing' nevertheless remains within musical research. As Peretz and Coltheart admit, "musical abilities are […] studied as part of a distinct mental module with its own procedures and knowledge bases that are associated with dedicated and separate neural substrates" (2003, p. 688) (167). This view of music as functionally autonomous seems to contrast with a vast range of findings in the literature, which highlight the multimodal and plastic nature of brain processing mechanisms and the widely distributed neural networks in both hemispheres this involves (168, 169). "Brain anatomy reveals that brain regions are interconnected in a rich and dense pattern, both locally and in terms of long-range connections" [Pessoa (170), p. 198]. To put it in a different way, anatomical segregations of musical functions seem to disregard the role of overlapping cortical regions and interindividual differences in brain substrates (171), as well as the observed evidence of ontogenetically developed – and rapidly adaptive – cerebral networks (24, 72). Neurons themselves display dynamical properties: there is no simple mapping from neural activity to behavior as what the neurons code depend on various time and contexts (172). Cross-sectional approaches to the study of the brain, thus, may downplay the developmental and ecological aspects shaping the living being-environment relationship (87). The brain is dynamical, self-organizing,11 and massively distributed (104, 174): it mediates and enables the non-linear12 and reciprocal interactions between the body and the world.

<sup>8</sup>Accordingly, people manage to make sense of the others by developing a "common-sense" theory of mind consisting of sets of laws connecting inner states to external stimuli, other inner states, or behaviours.

<sup>9</sup>As Di Paolo and colleagues insist: "Embodiment means that mind is inherent in the active, worldful body, that the body is not a puppet controlled by the brain but a whole animate system with many autonomous layers of self-coordination and self-organization and various degrees of openness to the world that create its sense-making activity" [Di Paolo et al. (102), p. 42].

<sup>10</sup>From the focus on the *theoretical level* between physical implementation and behavioural output (as in classical cognitivism), the exploration of the brain's functional neuroanatomy led a number of scholars to maintain that different brain regions represent different musical functions, such that neural tissue at a specific location govern given cognitive features [e.g., Tan et al. (160)].

<sup>11</sup> "Systems are self-organized when there is a reciprocal relationship among local areas and behavior at the global state of the system" (173).

<sup>12</sup>These interactions are "non-linear" because in a system of reciprocal continuity there is no clear "input" or "output" (59, 104).

Information, by this view, is not passively retrieved from the 'outer word' but rather *enacted* through the meaningful and sensorimotor activity of the organism13 (50).

Thus, because both genetic and ecological factors influence the development of neuronal networks (177–180) a number of scholars have found it necessary to look beyond brain reductionism (59, 88, 171, 181) and integrate traditional neuroscientific research with the study of a wider organism-world nexus (50, 57, 182). For example, recent work by Kiverstein and Miller (58), and Pessoa (170), shows how 'structure–function' mappings are best understood in terms of dynamical sub-components of a larger network, where a given function is highly context dependent and may vary over time in its dynamical interplay with the environment, which offers the animal various possibilities for actions according to its degree of complexity (183). It is worth noticing that the insistence on large-scale dynamic networks resonates closely with the view that sees cognition as belonging to a 'relational domain' (184), in which the living system acts in ways that are relevant to sustaining itself under precarious conditions. To understand the global behavior of a living organism, then, we need to do more than simply analyze one of its sub-components (i.e., the pathways underlying autonomic and muscular responses to music), as none of the system's part controls and defines the system by itself (185). The relation between biological organization and cognitive functions is thus best understood as 'circular,' rather than 'linearly causal' (71). This is to say that an *embodied view* on human musicality – and human cognition more generally – replaces the classic input/output framework with a non-linear perturbation/response distinction, in which the brain does play a very important part, but is not the sole factor involved.

By understanding cognitive processes as widely distributed across the entire body of the animal, and into its niche, the embodied approach goes beyond brain reductionism and provides a welcome alternative to classic computational frameworks (50). In what remains, we apply these insights to clinical research, arguing that an embodied perspective may help us address some of the challenges that emerge within this context in new ways. Focusing on music-based rehabilitative paradigms for PD patients, we explore the possibility that music may not just act 'externally' – somehow causing relevant behavioral responses – but rather that involves the agents' whole embodied being-in-the-world in active engagement; that it becomes a part of the network of non-linear interactions that characterizes the brain-body-world nexus (182). In doing so we hope to offer new insights into some aspects of PD treatment, and thus stimulate discussion on the interpretation and development of new approaches to rehabilitation.

### PARKINSONISM AND THE EMBODIED MIND

Parkinson's disease is a degenerative disorder associated with the progressive loss of the nigrostriatal dopaminergic neurons in the Basal Ganglia, which triggers functional changes in the same cortical network (186, 187). Non-motor symptoms are frequently the first signs and affect sense of smell and sleep regulation. Histologically, a classic mark of PD is represented by the presence of fibrillar aggregates of proteins called 'lewy bodies,' which displace other internal components of the remaining neurons in the midbrain, but also in the brain stem, the olfactory bulb and – at later stages – the cerebral cortex (188). The severe loss of dopaminergic cell activity in the midbrain results in hypokinetic disorders such as *akinesia* (the inability in initiating a movement), *bradykinesia* (slowness of movements) or *freezing* (impossibility to move in any direction) [see Berardelli et al. (189) and Grabli et al. (190)]. Usually, one of the first symptoms associated with PD is represented by an involuntary 4–5 Hz resting movement (191); clinical observations suggest that this *tremor* may disappear in voluntary actions, but can worsen with ambulation and with 'Froment's maneuver' (contralateral motor activity) (192). As the condition progresses, tremor is often accompanied with muscle *rigidity*, which leads to resistance of externally imposed joint movements (193). While states of relaxation may help, patients who are asked to move the contralateral limb often exhibit – like with tremor – an aggravation of the symptom (194, 195). Other typical motor deficits (often, but not always, emerging in laterstage PD) are *postural instability* and *gait disorders*, which result in an increased risk of falls (a predictor of mortality) and in turn critically challenge independent living habits and quality of life more generally (196, 197). Symptomatically, treatments with dopaminergic agonists or deep-brain stimulation have been demonstrated to be partially effective with many of these motor disorders (198–200), and are thus often integrated with noninvasive techniques based on music and rhythmic engagement (201, 202).

### Extending the Loop

A growing wealth of evidence shows how the contribution of music-based interventions is important for improving symptoms such as Parkinsonian gait (203–205). By matching their walking to the musical beat, or to a metronome, PD patients normally exhibit considerable benefit in terms of velocity, cadence, and stride length (206, 207). Interestingly, auditory cues for this kind of treatment display advantages when compared to visual, somatosensory, or combined cues: not only is reaction time to auditory cues shorter when compared to visual and tactile ones, but 'periodicity' is also best captured in sonic contexts rather than through other sensory systems14 (210, 211). Indeed 'timing' and 'periodicity' are fundamental aspects for gait, ensuring adequate

<sup>13</sup>Consider, for example, the nervous system: its inner states are always stimulated by the environment, but not influenced causally through an input-output relation; "the nervous system does not receive information. It rather creates a world by defining which configurations of the milieu are stimuli" [Weber (106), p. 15]. Functional localisation, reductionism, and 'internalist' frameworks are unable to capture these aspects by definition, as self-organisation occurs at a larger ecological scale (175, 176).

<sup>14</sup>To clarify this point, consider the apparent universality and spontaneity of sensorimotor synchronisation and impulse to move with music: evolutionary, this phenomenon may have promoted coordination and emotional communication (208, 209), contributing in strengthening the link between musical behaviours and wellbeing in intersubjective contexts.

consistency in pace and stability. As basal ganglia-cortical circuitry is typically involved in time-related processes – with a series of structures depending on dopaminergic innervation – its malfunctioning in PD has a significant impact on timing and motor synchronization (152, 212–214).

This is not to say, however, that 'timing' can be understood as a high-level cognitive ability that is functionally autonomous and encapsulated in the brain. First, besides the basal ganglia, it is likely that other cortical regions contribute in timing processing, thus constituting a distributed network that includes the cerebellum, SMA, pre-SMA, inferior parietal cortex, and premotor cortex (215–217). Moreover, the basal ganglia itself is involved in the selection and inhibition of motor processes (218), highlighting the deep connectivity of categories such as action, body, and 'timing.' Second, such connectivity implies that we cannot understand what 'timing' and 'periodicity' entail if we do not look beyond the boundaries of skull and skin to consider how the whole embodied agent *participates* in gait. Walking and synchronizing with a beat do not happen 'in the head'; they occur in the concrete sensorimotor dynamics of the world in which we are embedded, a world that is meaningful and rich of affordative structures ready to be acted upon. Music offers such affordances (219) according to the history of structural couplings between music users and sonic environment(s) (91, 92, 94, 133, 137, 138, 220, 221). We shall return to this point in the Section "Beyond Motor Recovery." What we want to stress, here, is that the organism's body,15 in its 'motor resonance' with the beat, enables the fluidity of the gait's 'kinetic melodies' in a continuous dynamical process of action and perception. This means that 'timing processes' – as subcomponents of the distributed network enabling gait – involve the entire body, and the world, literally *extending* beyond skull and skin. Thus musical rhythm offers a new pathway to enact selforganization through sensorimotor coupling by compensating for the malfunctioning of one of the system's sub-networks. The hyperactivity recorded in the cerebellum and in the pre-SMA at the preclinical stage (223–225) seems to confirm these network's self-organizing properties, which tend to develop other processes to counterbalance the impaired sensorimotor circle dynamically. As pre-SMA will eventually become hypoactive, left and right cerebellum and contralateral motor cortex have been observed as hyperactive also at later stages (226). Moreover, the compensatory mechanisms emerging in PD's pre-clinical and clinical stages show that self-organization also occurs on an ecological scale, integrating resources internal and external to the patient. Therefore, positing a single brain-body-world nexus – instead of the classic model based on the separation between internal (brain-bound) and external (worldly) domains – may help us better capture and model the ways in which the reorganization of the nexus' sub-networks unfolds in terms of dynamical and continuous interplay with the environment (175, 227, 228).

This process of wordly self-regulation, in which patients aim to recalibrate their sensorimotor engagement with the world, should also comprehend the 'social dimension' of embodiment, as the world involves other agents by definition. Stressing the importance of social interactions in a patient's being-in-the-world, it would be thus interesting to see how PD patient would respond to the so-called 'perceptual crossing paradigm,' which has been recently developed to study real-time situations in non-individualistic terms (229). Its simple methodology, which involves only "two subjects, a one-dimensional space, and a yes/no answer" (230), makes it particularly suitable for clinical contexts, and may illuminate on how PD affects the patient's capacity to interact with others. In the original experiment, as reported by Auvray et al. (229), pairs of blindfolded subjects in different rooms are asked to interact with each others in a computer-generated space. Participants are asked to move a cursor in this virtual space, clicking a mouse button when they perceive the presence of another participant. But since subjects are blindfolded, they only receive a tactile stimulation on the free hand when their avatar crosses an object in the onedimensional space. There are three different types of objects to be encountered: (i) the moving avatar of another participant, (ii) an object placed in a fixed location by the experimenters, and (iii) the moving 'shadow image' of the partner's avatar, that is an object that reproduces at a displaced distance the same movements of (i). The only difference between (i) and (iii), thus, is that with (i) a dyadic interaction is possible. As Froese and Di Paolo comment:

The two mobile objects exhibit exactly the same movement, but only an overlap of the receptor fields of both participants gives rise to mutual sensory stimulation. Note that the difference between these three types of objects cannot be directly provided by the sensors, which in all cases can only produce a binary, all-ornothing response depending on whether something is overlapping their particular receptor field or not. Thus, if the participants are to be successful at distinguishing which of the objects is the other agent's receptor field, they must accordingly rely on differences in the kinds of interactions that these objects afford. The results of the psychological study show that, at least under the minimalist conditions of this experiment, the successful recognition of an ongoing interaction with another person is not only based on individual capacities. It is also based on certain properties that are intrinsic to the joint perceptual activity itself [Froese and Di Paolo (231), p. 49].

Indeed, participants displayed greater accuracy in clicking the button when meeting the partner's avatar (65.9% of the clicks ± SD of 13.9) when compared to meeting the shadow image (23.0 ± 10.4%) or the static object (11.0 ± 8.9%) [see again Auvray and Rohde (230)]. In the case of PD patients, we predict a significant decrease in correct answers, as their ability to interact with the world might be partially compromised by the condition. The results, however, might be improved by exposure to *motorically familiar* musical cues. Indeed, hypothesizing that a malfunctioning sensorimotor coupling with the world makes the body an 'obstacle' for the living system's being-in-the-world (232), listening to music one can play may help to re-establish

<sup>15</sup> It is important to note that here we refer to "body" not as an objective piece of the world – the German "Körper" – but also as "Leib," a living and lived body with its autonomous and dynamical layers of self-organizational adaptivity [see Gallagher and Zahavi (222) and Merleau-Ponty (67)].

the correct sensorimotor loop with the environment through a 'motor resonance' enabled by the mirror mechanism. In the Section "Beyond Motor Recovery," we will try to describe how such hypothesis could be tested adequately, generating predictions that involve the whole living system in its dynamic interplay with the environment – and not only movements' rehabilitation.

### Beyond Motor Recovery

It is likely that the 'motor resonance' in play during music based motor rehabilitation involves the mirror mechanism mentioned above, as it does not seem to be significantly altered by PD (233). The activation of sensorimotor networks during music listening is well known (234, 235) – with musicians and subjects who have a practical knowledge of the complex order actions required to obtain a particular music showing stronger activations in the front-parietal-temporal network (132, 236). While the interpretation of such work is still a subject of controversy (237–239), it may nevertheless be argued that a 'motor vocabulary' of musical actions is formed when learning music. However, the firing of the neurons that might constitute such a 'vocabulary' (or 'repertoire') during listening tasks need not be understood in terms of 'information processing.' Rather it may be seen as allowing the system to *prepare for action*, possibly underpinning "a non-articulated immediate perception of the other person's intentional actions" [Gallagher (240), p. 541; see also Gallagher (147)]. As preparation for action is indeed an important component of intersubjective contexts – both phylogenetically and ontogenetically – mirror neuron theories may help us understand some other aspects of PD rehabilitative strategies. For example, they can explain why simple rhythmic excerpts or metronomic beats are widely and successfully adopted in this type of clinical research: almost everyone possesses (i.e., acquires through development) the motor expertise necessary to produce a repetitive beat. In this sense, the relationship between music and living systems is literally shaped by the history of structural sensorimotor coupling between them. Thus, the (therapeutic) compensatory mechanisms resulting from musical exposure appear to work when listeners-patients possess the adequate (meaningful) motor expertise relevant to re-enact the goal-directed actions afforded by the auditory cues.

A way to test this hypothesis in PD-related contexts might involve *familiarizing*16 subjects at an early clinical stage with musical stimuli that are more complicated than a simple beatpulse, and then observing at a later stage of the rehabilitation whether the same stimuli are more beneficial for gait (and – as it will emerge later – for more general improvement) when compared to standard simple beat or to unfamiliar music. This is to say that patients are not passive 'responders'; rather they actively 'enact' their own meaningful vocabulary of musical actions during music-based interventions, bringing forth their 'autonomous identity.' Increased familiarity with music-making in both individual and collective settings could foster the development of intersubjective rehabilitative contexts, where the interactivity of patients may generate more efficient results – increasing demands in sensorimotor integration. Indeed, this approach might be taken further to involve patients in music improvisation and the co-creation of musically relevant stimuli. Put simply, we suggest that by encouraging patients to develop more complex rhythmicmusical understandings, which they then develop and apply in the course of their treatment, new clinical possibilities may emerge that involve patients more comprehensively across the range of their being*.* In this way, treatment that involves increasingly adaptive and creative interactions with the environment (musical stimuli and other patients), may foster ways of being-in-the-world that lead to improved self-regulation, as well as a renewed, and much needed, sense of agency. Along these lines, the use of more sophisticated musical cues, and more intersubjective settings, might also lead to beneficial results beyond the motor domain. This is important, if we consider that a cascade of other nonsomatic symptoms17 often accompanies the motor dysfunctions described in the Section "Extending the loop": half of PD patients, for example, are reported to develop depression (241). But how could embodied theories say something about depression? And

how could music-based motor rehabilitation help? Relevant applications in clinical settings stemming from embodied theories have been recently explored within neuropsychiatric and psychopathological research – for example in schizophrenia (242) and depression (232). Research on the latter, in particular, suggests that depressive patients display similar symptoms to those of PD patients, including slow gait and reduced stride length (243–245). Indeed, like the PD sufferer, the depressive subject experiences a loss in their dynamical relation with the world and "cannot retain a position outside of her body" [Fuchs (53), pp. 99–100]. This is important when considering that, as Kyselo and Di Paolo (246) report, without the bodily power of action (for example in case of global paralysis) a subject may also suffer a decrease of cognitive activities such as imagery and goal-directed thinking [see Kübler and Birbaumer (247)]. Consider the following passage, where Fuchs and Schlimme (232) describe depressive melancholia as a case of 'hyperembodiment.' The authors argue that the process of becoming separated from the living system's peripersonal space results from psychomotor inhibition (as in PD) and a loss of the conative dimension of the body – its "affective and appetitive directedness":

Normally, it is this [conative] dimension that opens up the peripersonal space as a realm of possibilities, "affordances" and goals for action. In depressive patients, however, drive and impulse, appetite and libido are reduced or lost, no more disclosing potential sources of pleasure and satisfaction. Confined to the present state of bodily restriction, depressive patients cannot transcend their body any more. The open horizon of possible experiences shrinks into a locked atmosphere, in which everything becomes permeated by a sense of lost possibilities. With growing inhibition, sensory– motor space is restricted to the nearest environment,

<sup>16</sup>That is, developing the sensorimotor skills required to perform the musical stimuli.

<sup>17</sup>Other symptoms may include hallucinations, disorders of sleep and behaviour, dementia, psychosis, decrease of attention and memory, and language impairment.

culminating in depressive stupor. Thus, melancholia may be described as a reification or "corporealization" of the lived body, or as a "hyperembodiment" [Fuchs and Schlimme (232), pp. 572–573].

By this view, therapeutic interventions can be seen as an attempt to re-establish the functioning of the agent-environment system as a whole. Integrating standard rehabilitative settings for motor recovery in PD patients with more complex stimuli and activities, in early and later clinical stages, may lead to more beneficial results in terms of reshaping the motor resonance with the environment. These results are not limited to the motor domain, but may cover also non-somatic aspects of the pathology, as in the case of depression. Art-based therapies in general, and music therapies in particular, have been widely employed in the treatment of unipolar depression (248, 249), leading to encouraging results. An example comes from dance therapies, which have been proven effective in improving physical fitness and well-being more generally (250–252).

Acting upon the conative dimension of the sensorimotor coupling with the world, we argue that developing more meaningful musical environments could help in stabilizing the patients' embodied being-in-the-world (in a better fashion than with unfamiliar music or rhythmical beats only) by engaging the interactivity of the entire living system. The mechanisms underlying this are to be found in the neural compensatory mechanisms elicited by musical participation, and by the active engagement of the body in the concrete dynamics of action (253). Without positing a clear input–output relation between music and patient, an embodied approach to PD treatment with music emphasizes the self-regulatory aspects of brains and bodies, conceived as unities inseparable from their niche. Also, it conceives PD as a disturbance of the subjective sensorimotor skills to engage with the world, rather than solely a neurological pathology. Music, here, does not only influence the excitability of given neurons, but offers a new affordative space to the recovering embodied agent, compensating for the malfunctioning action-perception loop that characterizes the disease. It is important to stress once again that this does not exclude affectivity but, on the contrary, highlights the conative dimension of the living body as integrative part of its perceiving, knowing, doing, and being – opening new and fascinating possibilities for health and well-being.

### CONCLUSION

In avoiding the twofold reductionism of anatomical specificity and information-processing generality, the embodied trend provides a considerable challenge to established theoretical frameworks concerning the nature of mind, behavior, and agency. 'Embodiment,' although declined differently through the '4Es' described above, embraces the centrality of self-regulation, sensorimotor coupling, and intersubjective interactions for understanding the complex nature of our being-in-the-world (98, 100). We are confident this general reorientation can stimulate the development of new conceptual tools and research methods that may enhance standard rehabilitative settings within clinical contexts. In this paper we focused on how this may occur in PD research, hypothesizing that music based therapeutic settings could become even more efficient if coherently informed by theoretical models inspired by an embodied account to cognition.

Empirically, the adoption of embodied insights emphasizes the need to develop new experimental methods that are able to capture the way in which possible perturbations (i.e., a malfunction of a given sub-network) destabilize the whole brain-bodyworld system. Strategies for intervention, by this light, should not focus only on the isolated symptom thought to provoke a desired behavioral output (104). Rather, an embodied approach to motor rehabilitation should also consider, for example, conative, agentic, creative, and intersubjective dimensions as fundamental for the treatment of the patient – perhaps manipulating the degree of mutual interaction and affective experience according to the motor knowledge of the patient. Indeed, a musical stimulus (beyond a mere beat) is not only a 'timekeeper,' but also an actual *tool for cognition*, a meaningful event that affords a variety of selfregulative, interactive, and sensorimotor processes depending of the agent-music interaction's degree of complexity. With this in mind, and drawing on insights from research on mirror neurons, we hypothesized that PD patients might benefit from familiarization phases with more complex stimuli beginning in the early stages of the disease. The compensatory mechanisms in play during exposure to musical rhythms might then be more widely effective in the recovery of other (i.e., depressive) symptoms. This is just one example of how embodied approaches may define a broader approach to the study of PD rehabilitation, and why it necessitates further discussion and testing.

Overall, what we want to emphasize is that, theoretically, this kind of non-reductionist approach may be fundamental in rethinking many taken-for-granted assumptions concerning health and well-being, neuroscience, and music research. While the operational domain of the system's internal (e.g., brain) states is certainly fundamental to the interactive processes of such interactions, these internal processes alone cannot be identified

with 'cognition': to do so "is to confuse levels of discourse or to make a category mistake (neurons do not think and feel; people and animals do)" [Thompson and Stapleton (95), p. 27]. In other words, the processes allowing the system to maintain itself as autonomous are realized in the sensorimotor, dynamic, affective, interplay between bodies (including brains) and environments. These processes, as a whole, are not strictly speaking "neural," but rather define a non-linear network constituted by *both* neural and extra-neural interactive sub-networks (50).

Consider, for example, research in music psychology: while it is diverse and interdisciplinary, incorporating both 'subjective' (i.e., introspective, qualitative) and 'objective' (i.e., quantitative) methodologies (254), a common tacit assumption in the field is that (musical) experience is inner, computationally implemented, and reducible to neural activation. Embodied approaches challenge this perspective, showing that human musicality is deeply *embodied* (being constantly implemented by sensorimotor feedbacks and real-time bodily activities), *embedded* (as it is always situated in specific sociocultural niche), *enacted* (relying on the history of structural couplings between musical agents and musical environment) and *extended* (as no clear boundaries between internal and external resources exist in driving cognitive processes). And likewise, embodied perspectives represent a call for a new kind of integrative and non-reductionist music therapy – one that explores the possibilities of human musicality from diverse perspectives; and that may transform motor rehabilitation into a participatory activity where motion, emotion, listening biographies, and neural networks are all involved in a complex recursive interplay (255).

### References


Data on brain and behavioral activities has contributed greatly to new perspectives on the audio-visuo-motor integration underlying musical experience. We wish to stimulate researchers to integrate this body of knowledge with a critical analysis of the theoretical models underlying rehabilitative contexts (i.e., information-processing). Thus, by moving beyond traditional input-output and stimulus-response paradigms, our approach identifies large-scale networks (inside and outside the skull) as a solid alternative to reductionist approaches – highlighting the explanatory role of embodied perspectives in describing how an autonomous system develops, stabilizes, and transforms according to the reciprocal influences of local and global factors. In this way, hypokinesia, tremor, rigidity – but also depression, and other non-somatic symptoms – may be understood in a new light: as affecting the patient's *being-in-the-world* in a way that requires an recalibration of the *whole* brain-body-world network. In this way, the development of richer, embodied approaches to music intervention for PD (and other disorders) not only offers possibilities for improving the general quality of life of patients, it may also help us better understand how therapeutic recalibration occurs providing additional insights for clinical, musical, and neuroscientific, research.

### Acknowledgments

We are grateful to Tom Froese, Nicolas Farrugia, and Dylan van Der Schyff for comments on early drafts of this work. We would like to thank also the editors and the reviewers. Any remaining error is ours.


clinical implications. *Front Integr Neurosci* (2011) **5**:86.1. doi:10.3389/ fnint.2011.00086


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Schiavio and Altenmüller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Music stimulates muscles, mind, and feelings in one go

Stefan Mainka\*

*Neurologisches Fachkrankenhaus für Bewegungsstörungen/Parkinson, Beelitz, Germany*

Keywords: music, gait, training, rhythmic auditory stimulation

Gait and gait related mobility are among the most important factors for the quality of life of patients suffering from Parkinson's disease (PD). In most patients gait and gait related mobility will first lead to patients having to undergo physiotherapy. For gait training in PD there are a variety of evidence-based and well-established therapies. Among such therapies are Nordic walking, treadmill training, dancing, and amplitude training. All of these have proven to be effective in randomized controlled studies (Tomlinson et al., 2013).

One approach that was among the first that had been studied clinically is rhythmic auditory stimulation (RAS). RAS is a therapeutic training technique derived from the concept of neurologic music therapy. In this methodology, music or acoustic stimuli are used and shaped in order to enhance bodily functions such as cognition, speech, or movement (Thaut and Hoemberg, 2014). In RAS rhythmical stimuli or music are used to address and improve gait and gait related issues (Thaut and Rice, 2014).

Music with embedded metronome pulsation has been shown to be most effective in cueing movements. In comparison to single-pulse rhythmic stimulation (like from a metronome) response variability and synchronization offset are markedly reduced. This applies to the frequency range 60–120 bpm, which is relevant for gait training and also other gross and fine motor functions (Thaut et al., 1997). In an experimental study PD patients exhibited spontaneous improvements in gait velocity, cadence, and stride length through the use of functional music with an embedded metronome. A close synchronization between rhythm and step frequency suggested that there was evidence of rhythmic entrainment (McIntosh et al., 1997).

Thaut and colleagues demonstrated in 1996 the effectiveness of a 3-week home based training protocol. Thirty-seven PD patients (at Hoehn and Jahr, 2.5) had been randomly assigned to either RAS with functional training music, self-pacing gait training or no training at all. RAS training consisted of 30 min of daily gait exercises (walking on a flat surface, stair stepping, stop-and-go) with functional music. Music tempo was increased twice within each session by 5–10%.The initial tempo which was derived from the patients normal cadence (100 steps/min = 100 bpm) was also set 5–10% higher every week. The limit for tempo increase was set to 130 bpm. The RAS group improved significantly in velocity (25%) and step cadence (10%) compared with the self-pacing training group (velocity increase of 7%; Thaut et al., 1996).

To this day the work of Thaut and colleagues, in which functional music was used for optimizing gait training in PD patients, has not been reproduced in a randomized and controlled study. Nonetheless in the Cochrane review from 2013 RAS is listed as one of the profoundly evaluated gait training techniques (Tomlinson et al., 2013). This counts mainly for the cueing intervention with metronome, as evidence for the use of music is still scarce.

A second randomized and controlled clinical trial with musically stimulated gait training that was included in the review by Tomlinson et al. reveals profound methodological differences to the RAS technique. In this study by de Bruin, the music was chosen according to patients preference rather than functional criteria. The exercises were not focused on specific therapeutic issues, and most crucially the tempo of music—in Thaut's work the main component of the intervention varied ±15 bpm from the habitual walking step frequency and was not systematically increased (de Bruin et al., 2010).

#### Edited by:

*Marta Bienkiewicz, Aix-Marseille University, France*

#### Reviewed by:

*Matthew Rodger, Queen's University Belfast, UK*

> \*Correspondence: *Stefan Mainka mainka@kliniken-beelitz.de*

#### Specialty section:

*This article was submitted to Auditory Cognitive Neuroscience, a section of the journal Frontiers in Psychology*

Received: *08 July 2015* Accepted: *24 September 2015* Published: *08 October 2015*

#### Citation:

*Mainka S (2015) Music stimulates muscles, mind, and feelings in one go. Front. Psychol. 6:1547. doi: 10.3389/fpsyg.2015.01547*

It can be stated that there is a lack of sufficient evidence regarding the gait training of PD patients using functionally optimized music. Besides the characteristics of the stimulus (music vs. metronome) the adjustment of tempo in the acoustic stimulation is critical for the therapeutic success.

The optimal stimulation tempo for gait training with RAS is described in several works as 10% above the habitual walking cadence. At this tempo stride length and gait velocity are instantly pushed to an optimal level, while stride-to-stride variability is reduced and dynamic equilibrium is enhanced (Freedland et al., 2002; Willems et al., 2006; Hausdorff et al., 2007; Arias and Cudeiro, 2008, 2010).

Contrarian to this clinical practitioners described that speeding up can lead to a worsened initial heelstrike and provoke a more unsafe walking pattern with smaller steps. This was observed in patients with habitual cadences higher than 114 spm (Mainka and Trebs, 2011). These patients in the clinical praxis of the author often find it difficult to sense and regulate their tempo of steps, and have difficulties positioning feet and torso. In consequence of that accelerated and propulsively declined gait patterns occur. This phenomenon is called festination of gait. The question is whether and how can RAS be applied effectively to this clientele.

Gait festination is described as sharing the same pathological mechanisms as oral festination and freezing of gait (Moreau et al., 2007). This phenomenon is likely to be associated with impaired timing perception, cognitive decline (Riederer and Sian-Hülsmann, 2012) and left-sided symptom dominance (Flasskamp et al., 2012). Thus, for PD patients with a tendency toward festination of gait it might be helpful to address the timing regulation deficit in the RAS training programme. Due to the practical experience of the author these patients benefit from more controlled and measured movements. This can be addressed and facilitated by use of functional music with reduced tempo—namely 95–105 bpm according to the patients habitual cadence (comp. Mainka and Trebs, 2011). This practice is in line with an experiment focused upon PD patients suffering from freezing of gait. Tempo reduction (−10%) through a metronome led to a greater stride length (Willems et al., 2006).

The first clinical study referring to the therapeutic effects of tempo reduction was presented by Benoit in 2014. Fifteen PD patients trained daily for 4 weeks with adjusted functional music. The study was carried out with an uncontrolled and non-randomized one-armed design. The musical tempo was set 10% plus or minus according to the optimal stride length during the initial assessment. Through this paradigm seven subjects practiced RAS with reduced cadence stimulation, while the remaining eight were set to speed up. The whole study sample showed significant improvements in stride length and gait tempo that still persisted in the 1-month-followup. Additionally patients also improved in specific perceptive (duration discrimination and beat detection in musical excerpts) and motoric (synchronization with isochronous sequences) timing abilities that had not been trained explicitly (Benoit et al., 2014). Further clinical studies, should highlight the connection between perceptive-cognitive and motor decline and its sensitivity to external tempo regulation.

It has been hypothesized that rhythm and music stimulation might enhance or bypass impaired timing generation by recruiting a cerebellar–thalamic–cortical network. The network also involves the supplementary motor area. This indicates that not only motoric, but also perceptive timing is activated during rhythmic stimulation (Nombela et al., 2013). Music with its integration of time organization (rhythm and meter), attraction of attention (melody, specific sound cues, structure), and emotional activation is likely to play a vital role in these therapeutic issues.

The metronome offers the possibility of a practical and easy tempo control. It is looked at as an objective technical device. This may give the therapist and the patient a sense of functional efficacy and task orientation. Would it be better to stick with the "clean-and-easy" metronome instead of stirring up unwanted feelings during the training?

No, for the majority of the patients it would not be better.

Only music affects human beings beyond timing related mechanisms. There is neuro-anatomical evidence for a dopamine release within healthy individuals as a response associated to the experience of thrills during listening to one's favorite music (Salimpoor et al., 2011). Music instantly stimulates mobility and coordination in PD patients (Bernatzky et al., 2004). The patients are themselves quite clear what best helps them in their walking. 95.5% of 47 patients given the training decided to continue RAS with functional music (German folk style) instead of metronome after having tried both (Mainka, 2014). Music involves mental processing, which can effectively be used for carry over effects (e.g., continuing to hum or sing the melody; Satoh and Kuzuhara, 2008).

There are structural musical criteria for an optimal stimulation of gait training (see Table 1 in Datasheet 1 of Supplementary Material; comp. Thaut et al., 1996). The musical style may be selected according to the patient's preference. This could hold for optimal training motivation and emotional activation. It is noteworthy that the musical genre is exchangeable and essentially independent from structural and physiological functionality. Thus, the music may be rock, pop, folk, country, hip hop, metal, jazz, or another genre as long as structural criteria are incorporated.

The RAS gait training with music should be executed 3–5 times a week for 10–25 min. When appropriate gait maneuvers, stair stepping, stop-and-go exercises, or exercising with walking sticks can be implemented (see Table 2 in Datasheet 2 of Supplementary Material for potential therapeutic goals). With a portable mp3 Walkman it can be carried out anywhere—on the street, on the way to the supermarket, in park or in the countryside. The combination treatment of RAS with Nordic walking or treadmill training may open up additional therapeutic indications, but should be administered carefully with regard to the right stimulation tempo.

It is also important: clinical observations suggest that most patients with atypical Parkinson syndromes do not benefit from RAS. In these diseases (namely progressive supranuclear palsy, multisystemic atrophy, normal pressure hydrocephalus, subcortical arteriosclerotic encephalopathy, cortico-basal degeneration, primary akinesia with gait freezing) impairment of audio-motor entrainment may be related to a more widespread pathology. Thus, before starting RAS therapy it is worth evaluating audio-motor entrainment and considering the likely uncertainty of diagnosis.

Music with its physiological, aesthetic, and emotional qualities may be a highly effective treatment tool for people with Parkinson's disease. Gait, and gait related aspects, can be remedied through training in a joyful and motivating manner, that stimulates muscles, mind and feelings simultaneously.

### REFERENCES


### ACKNOWLEDGMENTS

Special thanks to Rohan Berry Crickmar for brushing up grammar and formulations of this article.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpsyg. 2015.01547


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Mainka. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Music therapy interventions in Parkinson's disease: the state-of-the-art**

### *Alfredo Raglio\**

*Department of Public Health, Experimental and Forensic Medicine, University of Pavia, Pavia, Italy*

**Keywords: relational music therapy, neurological music therapy, rehabilitation, rhythm, motor symptoms, nonmotor symptoms**

Parkinson's disease (PD) is a neurological disorder involving the progressive degeneration of the dopaminergic system, which gives rise to movement-related dysfunctions (such as bradykinesia, tremor, and rigidity) as well as other symptoms, mainly of cognitive and psychological nature. In the latter case, mood disorders prevails frequently causing anxiety and depression in all phases of the disease, sometimes even before the motor symptoms occur.

Aarsland and colleagues (1) report that 35% of the patients affected by PD present depression, whereas Richard (2) states that anxiety is to be found in 40% of the cases.

The literature shows that playing and listening to music may modulate emotions, behaviors, movements, communication, and cognitive factors, modifying the activity of the brain areas involved in the perception and regulation of these aspects (3, 4).

#### *Edited by:*

*Marta Bienkiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Yi-Huang Su, Technical University of Munich, Germany Madeleine Hackney, Atlanta VA Medical Center, USA; Emory University, USA*

> *\*Correspondence: Alfredo Raglio alfredo.raglio@unipv.it*

### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

> *Received: 07 June 2015 Accepted: 10 August 2015 Published: 31 August 2015*

#### *Citation:*

*Raglio A (2015) Music therapy interventions in Parkinson's disease: the state-of-the-art. Front. Neurol. 6:185. doi: 10.3389/fneur.2015.00185*

Music can produce substantial effects on movement-related symptoms as well as psychological ones in PD treatment. Concerning the first aspect, rhythm has a crucial role in rehabilitation, enhancing connections between the motor and auditory systems (5). Literature showed how a rhythmic auditory cues-based training can produce a compensation of

the cerebello-thalamo-cortical network leading to beneficial effects, for example, improving not only speed and step length but also perceptual and motor timing abilities (6, 7).

Areas involving rhythm perception are closely related to those that regulate movement (such as the premotor cortex, supplementary motor area, cerebellum, and basal ganglia – especially putamen) (8–18). A study conducted with fMRI (19) shows that whereas a regular pulse (in contrast to an irregular one) generally activates basal ganglia in a significant way, this is not the case in PD. Other studies (7, 20) support the idea that external cues (in particular rhythmical cues) can modulate the activity within the impaired timing system. This may mean that a regular rhythmic pulse stimulates the putamen activity, facilitating movement and providing an input for sequential movements and impaired automatized processes. Moreover, this could compensate for the lack of dopaminergic stimulation.

Rhythm can be also perceived visually and through the tactile sense, but the reaction time of the human auditory system is shorter by 20–50 ms, when compared to visual and tactile stimuli; moreover, it has a stronger capacity of perceiving rhythm periodicity and structure (6). Therefore, rhythm influences the kinetic system (through synchronization and adjustment of muscles to auditory stimuli), facilitates movement synchronization, coordination, and regularization, and may even produce an internal rhythm that persists in the absence of stimuli (21–23).

Many studies report that musical rhythm in PD treatment can improve gait (speed, frequency, and step length), limbs coordination, postural control, and balance (7, 18, 24–36). In view of the above, Neurologic Music Therapy (NMT) – especially Rhythmic Auditory Stimulation, one of its techniques – characterizes this approach to the disease: NMT aims at enhancing sensory, cognitive, and motor functions (as in PD treatment, in which specific rhythmic techniques can strengthen and improve the rehabilitative process).

Studies pertaining to non-motor symptoms (e.g., psychological aspects, such as anxiety or depression) are fewer in number, and therefore results are less certain (24, 36–41). From this point of view, music therapy approaches focused on the relationship between music therapist and patient serve as interventions that can produce substantial psychological improvements, creating moments of empathetic relationship and emotional connection. Sound is indeed an important means of communication, especially in a non-verbal context, promoting emotional expression and regulation. In PD treatment, the music therapeutic approach described above, based on free improvisation, is most assuredly less used, or maybe less documented. Sound/music improvisation can be considered as the main technique in the active music therapy approaches (42). During the improvisation, music therapist and patient interact freely using musical instruments, generally without musical rules or themes.

Moreover, making, and listening to, music can be considered as strong stimuli from the emotional point of view, playing an important role in the activation of the limbic system and neuro-chemical circuits (i.e., of the reward system) (43, 44).

A distinction between the relational and rehabilitative approach is extremely important, not only to determine the type of intervention but also to define the specific identity of the music therapy approach. In the former, the music therapist aims at building a relationship with the patient by means of interaction with melodic and rhythmical instruments and singing (improvisation). This facilitates the expression and modulation of patients' emotions and promotes communication and empathetic relationships. In the latter, the music therapist proposes music-based exercises (in particular using rhythmical patterns) to improve motor, cognitive, and sensory functions, generally impaired by neurological damage. Although the two methods are often mutually influenced, there is a substantial difference in terms of goals, intervention models, techniques, and music therapy training. From this perspective, I believe that research should take these aspects into account, in order to focus on goals more accurately, while maintaining the possibility of overlapping the two approaches – making them complementary rather than antithetical. Often literature does not consider the distinction among these different music therapy approaches, defining every experience as "music therapy." In contrast to this, I believe that the achievement of an adequate definition regarding different types of intervention is necessary; it should also highlight different theoretical and methodological bases and their different areas of application. To this end, a recent article by Raglio and Oasi (45) reports the main characteristics of music-based interventions, deriving from an in-depth analysis of their related contents and of literature.

**Table 1** shows how clinical studies characterized by strong scientific implications (Randomized Controlled Trials and Clinical Controlled Trials) and based on rehabilitative treatments (not on evaluating tasks in the presence of a musical stimulus) are lacking in literature. The table summarizes the main effects produced on motor and non-motor symptoms and distinguishes the various interventions in terms of different musical contents and techniques.

In PD treatment based on NMT, it is important to understand what specifically acts on the disease: it is unclear, for example, whether the beat produces an effect by itself in neuromotor rehabilitation or, in contrast, the effects are produced by music


**TABLE 1 | Randomized Controlled Trials and Controlled Clinical Trials from PubMed database regarding Parkinson's disease, including clinical studies (based on rehabilitative treatments) in English**.

*Search criteria: ("music" OR "music therapy" OR "rhythmic auditory stimulation") AND "Parkinson."*

(that we can consider the sum and combination of rhythm, melody, harmony, and other parameters) with a marked pulse (6). In fact, if music is excluded from interventions as described in many studies, we have to ask ourselves what then is the role of music therapists; furthermore, should such rhythm-based interventions be considered as part of music therapy programs, or rather simply as rehabilitative methods supported by rhythm, given that music does not assume a crucial role.

Additional weak points of NMT studies are given by the small sample size, the short duration of treatments and by the frequent lack of medium and long-term assessment (follow-up).

Another consideration is connected with the possibility to prove whether an effect produced during rehabilitation can persist also in the absence of stimuli. In this sense, the aforementioned studies conducted by Benjamin (21), Jackendoff (22), and Palmer and Krumhansl (23) offer various starting points that should be discussed in-depth, concerning the potential of the retrieval of imagined rhythm/music components, even after the stimuli have ceased or are absent (46).

A limited number of studies take the psychological outcome into account; these studies propose techniques based on a relational approach, which aims to develop an empathetic communication with the patient through the medium of sound. In music therapy, the active relational approach is based mainly on improvisation (42), and, from a practical standpoint, emphasizes the chances of enhancing and synchronizing the patient's movements. When reviewing videotapes of the therapeutic sessions based on this approach, it becomes clear that there are significant changes in regularity and fluency of movement of the upper limbs that also

### **References**


positively influence the patient's sound/music production. The music therapy relational approach in PD treatment is, therefore, less usual but equally as important; it is, in fact, less performancecentered and takes the patient's personal expression into account, thus allowing a dynamic calibration of the stimulus, through a process of continuous regulation, combining and integrating the physical and psychological components. Considerations emerging from the relational music therapeutic experience in PD treatment emphasize the rhythmic component of improvisation, leading the patient into synchronization, at the same time maintaining freedom of expression and a non-prescriptive approach based on the free sonorous-music improvisation. This facilitates the integration of psychological aspects (the empathetic relationship, the regulation, and modulation of emotional expressions) with motor functions (speed and fluency of movement, in particular of the upper limbs).

Therefore, interventions that involve music can offer important starting points in PD rehabilitation, effectively acting on motor, as well as non-motor symptoms. In this sense, research should increase the number of studies based on strong methodological criteria, also including a clear description of the intervention – whether it be relational or rehabilitating – a consistent and numerically significant sample, in addition to more sensitive tools to evaluate motor and psychological outcomes. In conclusion, a stronger research methodology and a clearer definition of the exact medium or parameter in music related to specific output of rehabilitation are needed. This will allow the development of adequate, and increasingly specific and effective music therapy approaches.


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Raglio. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Partnered Dancing to Improve Mobility for People With Parkinson' Disease s

#### Miek J. de Dreu<sup>1</sup> , Gert Kwakkel <sup>2</sup> and Erwin E. H. van Wegen<sup>2</sup> \*

*<sup>1</sup> Department of Neurosurgery, Clinical Imaging Tilburg, Elisabeth Hospital, Tilburg, Netherlands, <sup>2</sup> Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, Netherlands*

Keywords: Parkinson's Disease, dance, music, recreational therapy, Salsa, Tango, exercise

### INTRODUCTION

Parkinson Disease (PD) is a neurodegenerative disease for which no cure is available yet. It is the second largest neurological disease affecting an estimated 571 per 100,000 people in Europe with rising prevalence due to the aging population (Pringsheim et al., 2014). To date, dopaminereplacement therapy (DRT) is the first choice of treatment to lessen the impact of motor and non-motor symptoms, however DRT does not prevent progressive disabilities and does not change the course of the disease (Chaudhuri et al., 2006; Jankovic and Stacy, 2007). Therefore, other types of therapy are needed to supplement DRT.

Recent evidence suggests that physical activity and vigorous exercise may have the potential to slow disease progression (for an overview see Hirsch and Farley, 2009; Ahlskog, 2011; van Wegen et al., 2014). These findings are promising, however they require further investigation. The effects of physical activity and regular exercise on reducing the chance of developing secondary problems e.g., diabetes or cardiovascular disease are quite established (Lee et al., 2012). Nevertheless, inactivity is a major problem as patients with PD are approximately one third less active than age matched controls (van Nimwegen et al., 2011). Being physically active may be more difficult for patients with PD because of physical impairments, fatigue, and apathy (van Nimwegen et al., 2011). Physical rehabilitation, containing a variety of exercise interventions (e.g., individual and in groups) is recommended for patients with PD (Keus et al., 2014).

### Partnered Dancing

The European guideline for Parkinson's disease recommends dance as a meaningful approach to improve functional mobility and balance (Keus et al., 2014). However, this recommendation is based only on three proof-of-concept trials that investigate Tango dancing (Hackney et al., 2007; Hackney and Earhart, 2009a; Duncan and Earhart, 2012).

Several reviews have been published that included more studies regarding music based movement therapy and dance (de Dreu et al., 2012, 2014; Sharp and Hewitt, 2014; Shanahan et al., 2015a). A recent meta-analysis including five randomized clinical trials suggests significant positive effects of dance therapy on motor impairment, balance, gait speed, and health-related quality of life (Sharp and Hewitt, 2014). A systematic review investigating multiple types of dance found significant positive effects on endurance, motor impairment, and balance (Shanahan et al., 2015a). Furthermore, our meta-analysis investigating several types of music based movement therapies (dance and gait-based interventions using music as an auditory rhythmic cue) found significant positive effects on balance performance, UPDRS-II, walking velocity, stride length, dual task walking velocity, 6 m walk test, and the timed-up-and go test (de Dreu et al., 2012, 2014).

Most studies on dance in patients with PD have investigated Tango dancing (Hackney et al., 2007; Hackney and Earhart, 2009a,b,c, 2010a,b; Duncan and Earhart, 2012; Foster et al., 2013; McKee and Hackney, 2013; Duncan and Earhart, 2014). In addition, one study on Ballroom dancing

#### Edited by:

*Marta Bienkiewicz, Aix-Marseille University, France*

#### Reviewed by: *Sara Houston,*

*University of Roehampton, UK*

\*Correspondence: *Erwin E. H. van Wegen E.vanWegen@vumc.nl*

#### Specialty section:

*This article was submitted to Auditory Cognitive Neuroscience, a section of the journal Frontiers in Neuroscience*

Received: *18 September 2015* Accepted: *06 November 2015* Published: *11 December 2015*

#### Citation:

*de Dreu MJ, Kwakkel G and van Wegen EEH (2015) Partnered Dancing to Improve Mobility for People With Parkinson's Disease. Front. Neurosci. 9:444. doi: 10.3389/fnins.2015.00444* (Hackney and Earhart, 2009a,b) and two pilot studies on Irish set dancing (Volpe et al., 2013; Shanahan et al., 2015b) were published. Salsa dance classes for patients with PD are available in the Netherlands and in Canada (Ottawa). Furthermore, there are organizations in several countries providing modified modern dance classes and/or ballet classes (not partnered dancing) for patients with PD, e.g., Dance for PD (New York), Dance for Parkinson's (London), Queensland ballet, and Dance for Health (multiple cities in The Netherlands). Some small pilot studies have investigated these types of dance (Westbrook and McKibben, 1989; Batson, 2010; Heiberger et al., 2011), however the effectiveness of Salsa dance classes remains to be established in methodologically well-conducted randomized controlled trials.

Partnered dancing combines exercise with cognitive challenges in an enriched environment with (somato) sensory cues from the music as well as from the dance partner (Bläsing et al., 2012). The sensory cues from physical contact with the partner are specifically important during Tango and Salsa dancing. While Ballroom and Irish set dancing often have a predefined routine that is executed from start to end in the way that people are required to learn the entire routine by heart, Tango and Salsa dancing do not necessarily have such a routine, providing more flexibility in performance. During Tango and Salsa classes, participants are taught several short steps with specific somatosensory cues (signals) for each step (e.g., with a length of 8 or 16 counts in the music). Subsequently, the couple can apply these steps in any sequence. A dancing couple consists of a leader (traditionally the man) and a follower (traditionally the woman). However, this format is sometimes changed, e.g., in the studies about Tango dancing men and women practiced both the leading and following roles (Hackney and Earhart, 2010b). The leader determines which routine comes next and the follower responds to the somatosensory cues of the leader. This requires clear communication. An example in this context is a right turn for the follower, this can be indicated by the leader by raising the hand of the follower gently above his/her head, indicating the direction of the turn by choosing a spot just right or left from the center of his/her head. We advise the follower to turn with small steps in their own tempo and the leaders to follow the tempo of the followers. Some of these steps in Salsa and Tango are similar to physiotherapeutic strategies and training for weight shifting, turning, and backwards walking (Kamsma et al., 1995; Earhart, 2009). Consequently, there is a relatively high demand of planning skills for the leader and the responsiveness to somatosensory cues for the follower. In line with these observations, McKee and Hackney found that spatial cognition and executive function improved after 10 weeks of Tango dancing classes (McKee and Hackney, 2013). This finding is important in light of the decline of spatial cognition in neurodegenerative disease (Possin, 2010). These interactions resemble those with caregiver-mediated exercises after stroke (Galvin et al., 2011; Vloothuis et al., 2014) and may improve not only the functional mobility of patients but also decrease feelings of caregiver burden through mechanisms of empowerment and self-management. However, the effect of partnered dance on caregiver burden needs further investigation.

## BALANCE

Balance instability is one of the cardinal signs of PD (Kim et al., 2013) that responds poorly to and may even be worsened by DRT (Konczak et al., 2009). Balance problems and the related fall risk affect the daily life of patients with PD to a large extent and may prevent patients from being active (Wielinski et al., 2005; Abendroth et al., 2012).

Partnered dancing interventions have consistently improved balance performance without focusing on balance deficits (Earhart, 2009; de Dreu et al., 2012, 2014; Sharp and Hewitt, 2014; Shanahan et al., 2015b). The use of multiple types of sensory information simultaneously (e.g., auditory, somatosensory and proprioception) has been indicated as a critical aspect of balance control for patients with PD (Konczak et al., 2009; Conradsson et al., 2012; Lefaivre and Almeida, 2015). The predefined steps with stops, starts, changes in direction, and backwards stepping may provide practice of motor agility, which is another critical aspect of balance control and gait initiation/termination affected in patients with PD (Conradsson et al., 2012). An improvement of balance performance may enhance activities of daily living (ADL; Tan et al., 2012), health-related quality of life (Ellis et al., 2011), and with that, well-being of patients with PD.

## MUSIC IN PARTNERED DANCE

Music is an integral and essential part of partnered dancing that provides a rhythm as well as an emotional context via a complex structure (e.g., loudness, pitch, timbre, harmony, melody, duration of the tone etc.; Krumhansl, 2000).

Usually, the type of music is specific for the type of dance. The rhythm of the music provides a timeframe, aiding in movement execution similarly as auditory cueing, provided that the patient with PD recognizes the rhythm (Keus et al., 2007; Nieuwboer et al., 2007; de Bruin et al., 2010; de Dreu et al., 2014; van Wegen et al., 2014). An important aspect of music in this context is the "groove." The groove has been defined as the property of music that compels the body to move (Janata et al., 2012). Salsa music to date has not been investigated for groove, however Samba music (also Latin music) was found to contain a high level of groove (Madison et al., 2011). The structure of music (especially high groove and familiar music) may aid in synchronization with the rhythm compared to the isochronous beat of a metronome (Thaut et al., 1997; Janata et al., 2012; Getz et al., 2014; Leow et al., 2014; Hove and Keller, 2015). This is consistent with general functional perspectives of rhythmic music enabling and facilitating entrainment and precise synchronization of movements (Madison et al., 2011) and may be specifically important for patients with PD because of the problems in sensory-motor timing (Lucas et al., 2013; Hove and Keller, 2015).

Patients with PD may have some more difficulty in detecting the beat, acknowledging that the beat-based rhythm perception is worse when compared to controls (Grahn, 2009). Initially the majority of people (either Parkinsonian or healthy) may find it difficult to synchronize their steps to the rhythm of music, which may be related to a low familiarity with the music (Leow et al., 2014) and the cognitive load of timing the newly learned dance routines to the music (McKee and Hackney, 2013). Another explanation is that motor learning is required for proper timing of movements to external stimuli. Impairments in timing may decrease with improved sequential movement performance as a consequence. Further research on this aspect is needed.

Finally, music provides an emotional context and may temporarily alter mood (Krumhansl, 2000; Laukka, 2006; Zentner et al., 2008) through activation of specific brain areas such as amygdala, nucleus accumbens, hypothalamus, hippocampus, insula, cingulate cortex, and orbitofrontal cortex (Blood and Zatorre, 2001; Koelsch, 2014). This activation includes the release of several biochemical mediators (e.g., endorphins, endocannabinoids, dopamine, and nitric oxide; Boso et al., 2006). These neurophysiological aspects of music may increase therapy compliance for long-term interventions and distract from sensations such as fatigue during exercise (Hayakawa et al., 2000; Lim et al., 2011; Stork et al., 2015).

### SAFETY

Directly related to implementing a challenging dance training in a community setting is the risk for falling during the intervention. Partnered dancing is potentially a safe intervention. Provided the partner is strong enough, he or she may be able to provide physical support when necessary. Several other measures that may prevent falling in a dance class have been described by Hackney and Earhart (2010b). Safety largely depends on the skill of the dance-teachers to adequately adjust the difficulty of the steps to the ability of the participants. A recent small feasibility study of Tango dancing (N = 6; 4 weeks of weekly dance classes) reported no adverse events during dance classes (Blandy et al., 2015). Feasibility studies of Irish set dancing (N = 22; 8 months of weekly dance classes) reported one single fall with no injury (Volpe et al., 2013; Shanahan et al., 2015b). Therefore, partnered dance can be regarded safe and feasible when following the guidelines of Tango dancing (Hackney and Earhart, 2010b). However, the backward steps during Tango may pose a larger risk for falls than other dances. The European PDguidelines for physiotherapy suggests that backward stepping

### REFERENCES


during Tango dancing may increase the risk of falling during the dance intervention and highlights the importance of an adequate selection of patients for this type of intervention (Keus et al., 2014).

### SOCIAL ASPECTS OF PARTNERED DANCE

Partnered dancing is an exercise regimen that requires substantial interaction between the dance partners and incorporates group dynamics as participants may switch dance partners during the class. Activities peripheral to the dance class such as drinks during breaks and peer-interactions before and after class provide additional possibilities for social interaction which may improve adherence (Rosa et al., 2015). This aspect of dancing may be specifically important when aiming for preventing social isolation.

To conclude, partnered dancing in a community setting seems a viable way to exercise. It is an attractive form of exercise therapy for patients with PD because it naturally combines evidence based aspects of music, cueing techniques, motor learning, balance exercises, and physical activity while focusing on interaction and enjoyment between partners and the group. Although theoretically several arguments exist regarding the mechanisms of action and putative effects of partnered dance, future research could compare different dance styles with regard to safety and effectiveness in order to further the field of music based exercise therapy. Furthermore, the effect of partnered dancing on the wellbeing of the partner needs further research.

### AUTHOR CONTRIBUTIONS

MD wrote the manuscript. GK and EV provided feedback and suggestions throughout the writing process.

### FUNDING

GK has received research-funding from ZonmW (89000001/630000012), ParkinsonFonds (20094), and EU (ERC-291339). EV received research-funding from ZonMw (630000012), Parkinson Vereniging (2013R16), ParkinsonFonds (20094), and the Hersenstichting (F2013141).


Parkinson's disease. Front. Aging Neurosci. 3:14. doi: 10.3389/fnagi.2011. 00014


disease. Top. Geriatr. Rehabil. 30, 46–57. doi: 10.1097/TGR.000000000 0000005


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 de Dreu, Kwakkel and van Wegen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Therapeutic Argentine tango dancing for people with mild Parkinson's disease: a feasibility study

*Laura M. Blandy, Winifred A. Beevers, Kerry Fitzmaurice and Meg E. Morris\**

*College of Science, Health and Engineering, School of Allied Health, La Trobe University, Bundoora, VIC, Australia* 

Background: Individuals living with Parkinson's disease (PD) can experience a range of movement disorders that affect mobility and balance and increase the risk of falls. Low health-related quality of life, depression, and anxiety are more common in people with PD than age-matched comparisons. Therapeutic dance is a form of physical activity believed to facilitate movement and therapy uptake. As well as being enjoyable, dancing is thought to improve mobility, balance, and well-being in some people living with PD. The primary objective of this study was to evaluate the feasibility and safety of a 4-week Argentine tango dance program for people with PD.

### *Edited by:*

*Marta Bienkiewicz, Technische Universität München, Germany*

### *Reviewed by:*

*Maria Stamelou, University of Athens, Greece Mariana Moscovich, Universidade Federal do Paraná, Brazil*

#### *\*Correspondence:*

 *Meg E. Morris, College of Science, Health and Engineering, School of Allied Health, La Trobe University, Bundoora, VIC 3086, Australia m.morris@latrobe.edu.au*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

> *Received: 05 March 2015 Accepted: 12 May 2015 Published: 27 May 2015*

### *Citation:*

*Blandy LM, Beevers WA, Fitzmaurice K and Morris ME (2015) Therapeutic Argentine tango dancing for people with mild Parkinson's disease: a feasibility study. Front. Neurol. 6:122. doi: 10.3389/fneur.2015.00122*

Methods: Six community dwelling individuals with mild to moderate PD were recruited from Parkinson's support groups, movement disorder clinics, and the PD association in Australia. To minimize falls risk, participants were required to be <75 years of age and physically independent (Hoehn and Yahr stages I–III). They were also required to speak English. Participants attended a 1-hour dance class at a dance studio twice per week for 4 weeks. A professional dance instructor led and choreographed the Argentine tango dance classes. Physiotherapists were present to assist participants during the class and served as dance partners as necessary. The primary outcome was feasibility, which was determined by measures of recruitment, adherence, attrition, safety (falls, near misses and adverse events), and resource requirements. Secondary measures included the Beck Depression Inventory and the Euroqol-5D, administered at baseline and post intervention. Therapy outcomes pre- and post-intervention were analyzed descriptively as medians and interquartile ranges and using Wilcoxon matched pair signed-rank tests.

Results: The Argentine tango dance intervention was shown to be safe, with no adverse events. Adherence to the dance program was 89%. Depression scores improved after intervention (*p* = 0.04). Some challenges were associated with the need to quickly recruit participants and physiotherapists to act as dance partners during classes and to monitor participants.

Conclusion: The 4-week, twice weekly Argentine tango dancing program was shown to be feasible and safe for people with mild-to-moderately severe PD.

Keywords: Parkinson's disease, tango, feasibility, depression, quality of life

## Introduction

Idiopathic Parkinson's disease (PD) is a progressive neurological disorder associated with reduced mobility, falls, and reduced quality of life (QOL) (1). International guidelines endorse exercise therapy to retrain balance and preserve physical capacity for individuals living with PD (2). Meta-analyses have also shown exercise therapy to have positive effects with regards to physical functioning, balance, and health-related QOL (3, 4). Despite the potential benefits of movement rehabilitation, long-term adherence to traditional exercise programs can be problematic (5). A systematic review evaluating exercise adherence in PD reported that reduced motivation was a common reason for reduced participation (6). This demonstrates a need for community-based physical activity programs that facilitate uptake and enjoyment (7).

Emerging evidence suggests that therapeutic dance may be an appropriate and enjoyable form of physical activity for some individuals with PD (8). Dance may address some of the physical impairments in PD through teaching movement strategies, challenging balance, and improving physical fitness (9, 10). The musical rhythm could become an auditory cue to engage cortical control of movement, which in turn might potentially enhance motor learning (11). Preliminary trials suggest dance can facilitate improvements in gait, balance, and motor impairment in comparison to exercise (12), physiotherapy (13), and control conditions (8, 14–16). It has been proposed that therapeutic dance may also facilitate QOL and well-being through enabling movement expression and building social connections (17).

The Argentine Tango dance genre is arguably one of the most suitable dance forms for people with PD (8, 14, 15). It has been proposed to target the movement impairments of PD with strong musical rhythms that trigger movement and enable greater amounts of physical activity. Compared to other dance genres, the choreography can be designed to train specific movement strategies such as walking backwards and turning (18). Furthermore, as a partnered form of dance, tango may facilitate interpersonal connections that positively affect QOL and mood (18).

While a growing number of pilot studies have explored the effects of dance on movement disorders in PD, there is a paucity of feasibility data and recommendations that allow researchers to design future protocols (19). Comprehensive exploration of the safety of specific dance genres is still required given that individuals with PD have a propensity to fall (20). This research also focused on QOL and depression outcomes, as there is little published literature on the effects of therapeutic dance on perceived QOL and mood for adults with PD.

The primary purpose of the current study was to evaluate the feasibility and safety of an Argentine tango dance intervention and to provide recommendations for future research. The specific aims were to: (i) determine if 4 weeks of twice weekly Argentine tango dance classes were feasible and safe for people with PD, allowing the development of recommendations for a future research protocol and (ii) measure the within-group change for depressive symptoms and health-related quality of life (HRQOL) following participation in the dance classes.

### Materials and Methods

### Design

This feasibility study adopted a single-group, pre-/post-test design and received ethical approval by the La Trobe University Faculty of Health Sciences Human Ethics Committee (ref. FHEC13/026). The trial was registered with ANZCTR, number ACTRN12613001058763.

We recruited participants with idiopathic PD from the metropolitan Melbourne region in Australia. As the current body of literature does not provide an arbitrary method for determining sample sizes in pilot research (21), there were various considerations involved in determining sample size. Primarily, it was anticipated that a dance class with six to eight participants could be adequately monitored (18).

### Participants

The eligibility criteria tested what was planned for a future clinical trial. Criteria were also established to maximize safety and reduce risk of falls during the interventions. Eligibility criteria therefore included: (i) a diagnosis of idiopathic PD confirmed by a neurologist; (ii) mild-to-moderate disease severity (Hoehn and Yahr stages I–III) (22); (iii) community dwelling; (iv) aged 18–75 years; and (v) medically safe to participate. Volunteers were excluded if they were unable to provide informed consent (Mini Mental State Examination, MMSE, score <24) (23) and did not have sufficient English to be able to follow instructions. Volunteers were recruited through local Parkinson's support groups, movement disorder clinics, and the website and newsletter of Parkinson's Victoria, Australia.

At intake, participants signed informed consent and were screened for eligibility. The Unified Parkinson's Disease Rating Scale motor subsection (MDS-UPDRS-III) (24) and a health status screening form to identify any comorbidities that could require extra monitoring during intervention were also administered. They were then asked to provide written consent to contact their medical practitioner for written medical clearance.

### Intervention

Participants were invited to participate in 1-hour dance classes that ran twice weekly over 4 weeks at a dance studio in metropolitan Melbourne. They were required to make their own way to classes, and were provided with taxi vouchers if required. Classes of Argentine tango choreography were designed and led by a professional dance instructor, with experienced physiotherapists present to spot participants at risk of falls and serve as dance partners. People who were partners were all able-bodied participants. Classes included regular rest breaks, typically at 20-min intervals, to minimize fatigue (18). Modifications were made to the intervention by physiotherapists to reduce falls risk as needed. The participants essentially had the same class, although as is customary with tango dancing, there were some gender differences in the dance routines. Women are usually the followers in tango, and were required to frequently do steps backwards, in contrast to men. All of the participants were using PD medications and they were in the "on" phase of the medication cycle during the dancing classes. An outline of class structure and content is detailed in **Table 1**.

### Primary Outcomes


(v) Resource requirements: to document whether attaining the personnel required was feasible (19), the attendance of supervising physiotherapists was recorded.

To objectively establish if the proposed research protocol was successful, the following *a priori* criteria for feasibility success were developed (19): (i) recruitment strategy and screening process enable recruitment of eight participants within 1 month; (ii) adherence ≥70%, which has been deemed as high in older adults with a physical impairment (25); (iii) attrition ≤15%, an arbitrary figure for acceptable attrition established by the PEDro scale (26); and (iv) safety: the dance intervention was considered safe if there were no falls or injuries during the intervention and no "hands on" assistance was required for prolonged periods.

### Secondary Outcomes

The effects of the intervention on HRQOL were measured with the Euroqol-5D (27, 28), a validated measure of HRQOL in PD, using the rating questionnaire and the visual analog scale (VAS). The summary index was computed for the descriptive system using normative data sets from the United Kingdom, with a possible value between −0.59 and 1, where 1 represents "full health." The VAS scale is a self-perceived rating of health status scored from 0 to 100, with 100 indicating best imaginable health state. The Beck Depression Inventory (BDI) (29) was used to measure depression. This questionnaire has established validity and reliability for assessment of depressive symptoms in the PD population (30). The BDI scores are tallied to obtain a score out of 63, with a higher score indicating greater severity of symptoms. Both questionnaires were administered at baseline and at the end of the 4-week intervention phase.

### Statistical Analysis

Therapy outcomes at pre- and post-intervention were analyzed descriptively as medians and interquartile ranges (IQR) to


#### Table 1 | Class structure and content.

better describe the small data set and control for extremes in scores (31). Where participant data were missing at post-test, baseline data were carried over to complete the analysis. To establish statistical significance of within-group change between the two time points, a Wilcoxon matched pair signed-rank test was performed with *p*-value set at <0.05. A non-parametric technique was necessary given data were collected from a small sample and was not normally distributed (31). Statistical analysis of therapy outcomes was performed using SPSS statistics (IBM, Armonk, NY, USA).

### Results

### Participants

Participant demographic data are detailed in **Table 2**. Participants had mild PD as indicated by low mean scores on the UPDRS part III, motor examination, and with all participants scoring 2 on the modified Hoehn and Yahr scale.

### Feasibility

### Recruitment

The proposed recruitment strategy did not meet *a priori* criteria. Six individuals were recruited over 2 months. **Figure 1** indicates the flow of participants in the recruitment and intervention phases.

### Adherence

Attendance to the dance classes was 89%, exceeding the *a priori* criterion of 70%. One participant was unable to attend one class due to an unrelated medical matter, and another participant was unable to attend two classes secondary to a scheduled vacation.

### Attrition

One participant dropped out for the final week of the dance intervention (including the post-intervention assessment) citing a combination of medical and personal reasons. Attrition rate was thus 17% for the 4-week dance intervention, failing to meet the feasibility criteria of 15%.

### Personnel Requirements

Attendance of the supervising physiotherapists throughout the program was 86%, lower than the required 100% to provide dance partners for participants.


*Values are means (SD) unless stated otherwise.MMSE, mini-mental state examination; HY, modified Hoehn and Yahr scale; IQR, inter quartile range; MDS-UPDRS, Movement Disorders Society-Unified Parkinson's Disease Rating Scale; PD, Parkinson's disease.*

### Safety

There were no falls, injuries, or medical emergencies throughout the intervention period. Three participants withdrew from the class on separate occasions due to (i) one episode of transient vertigo following rehearsal of fast-paced turning; (ii) pain at an arthritic hip joint; and (iii) medication-related motor complications. There was one documented "near miss" without injury where a participant destabilized secondary to a dyskinetic tick. Incidents of mild pain at the back (*n* = 1), shoulder (*n* = 1), foot/ankle (*n* = 1), and hip (*n* = 1) were also documented. In health status screening forms, these participants cited arthritis at these joints. "Hands on" assistance was required in only one instance and distant (1 m) supervision was provided to all participants at all other times.

### Modifications to Dance Intervention

The intervention was modified on four occasions to ensure participant safety; no walking backwards in the first session, steps requiring single leg stance to be performed while seated, practice of turns to be performed over an arc rather than on the spot, cessation of dance to music with fast tempo. The tango hold position was also modified for one participant due to shoulder pain.

### Health-Related Quality of Life and Depression

The EuroQol median score increased from a baseline score of 0.796 (IQR: 0.73–1.00) to a post intervention score of 0.890 (IQR: 0.80–1.00), indicating improvement; however, this did not reach statistical significance (*p*= 0.317). There was a small but not significant improvement in the VAS score from a median score of 80 (IQR: 77.5–91.25) at baseline to 82.5 (IQR: 77.5–91.25) post intervention (*p* = 0.854).

Beck Depression Inventory scores at both time frames were distributed around the lower end of the scale, suggesting that overall depressive symptoms were low (32). The BDI scores decreased, indicating an improvement in depression, from a median score at baseline of 5.5 (IQR: 4.75–8.50) to a median of 0 (IQR: 0–5.50) post intervention, which reached significance (*p* = 0.042).

### Discussion

The Argentine tango dance intervention was found to be feasible for this sample of people with comparatively mild PD. It was shown to be safe with no serious adverse events, such as falls, injuries, or medical emergencies, occurring. The minor adverse events were mostly related to comorbidities, and were readily managed with modifications to dance routine and steps. Modifications were, however, made infrequently. All documented modifications reflected the recommendations for tango classes for adults with PD developed by Hackney and Earhart (18), and Rios Romenets, Anang (16). It is evident that the safety of the intervention was enabled by the presence of supervisors to monitor instability and to advise appropriate modifications. Formal education of dance instructors on screening and monitoring for adverse responses to interventions may result in earlier modifications to the interventions and a subsequent minimization of adverse outcomes (33). The selected dosage was shown to be safe and appropriate, and none of the participants reported pain, stiffness, or fatigue after classes.

Recruitment in the short time frame available was challenging. Researchers continued to receive expressions of interest in the month following commencement of the dance program, indicating a 3-month recruitment period may have successfully recruited eight participants. Recruitment difficulties are common in PD research (34), with similar issues reported by Batson (33), who rescheduled her dance program secondary to recruitment difficulties, and Duncan and Earhart (14), who reported only 50% of participants screened were eligible to participate in a tango dance intervention. It is possible that the potential benefits of dance are not widely understood, narrowing those who volunteer to participate (33). People with Parkinson's were more reluctant to participate if they had to commute long distances. This is supported by research identifying transportation as a barrier to exercise uptake for individuals with movement disorders (35).

Adherence to the Argentine tango dance program was high, satisfying feasibility criteria (25). Adherence has only been reported in a small number of studies evaluating dance therapy for PD. Duncan and Earhart (14) reported 78% adherence over 12 months of tango, Volpe et al. (13) showed 90% adherence with 6 months of Irish dance, and 90% adherence over 3 weeks of modern dance was reported by Batson (33). The high adherence to dance therapy supports the hypothesis that therapeutic dance may facilitate uptake and enjoyment (9). However, these findings may have been influenced by selection bias, where individuals that volunteer to participate are more likely to adhere to a program (36). Further research is warranted to explore adherence to a longer program, and to determine if systematic differences in adherence exist between tango and a second dance genre, or a comparison intervention. Attrition was relatively low over the short intervention phase. While this narrowly fell short of *a priori* criteria, a 100% retention rate was necessary to meet this criterion as the eventual sample included only six participants. Importantly, data showed that the participant lost to post-intervention assessment did not report dislike or disinterest in the intervention, citing personal reasons for discontinuing.

This study adds to the limited body of evidence evaluating the influence of dance therapy on outcomes related to QOL and wellbeing, which are important to a holistic management of individuals with PD (28). This pilot study identified significant improvements in depressive symptoms and a trend toward improved HRQOL. For BDI scores, this was not only statistically significant but also research indicates a change from 5.5 to 0 may also be clinically meaningful (32). To the authors' knowledge, this is the first study that has found positive effects of dance for depressive symptoms in adults with PD. The HRQOL findings support the work of Hackney and Earhart (37) and Volpe and colleagues (13), who found small, non-significant improvements in HRQOL compared to control and physiotherapy. Like this feasibility study, it is likely these preliminary trials were underpowered and may have failed to identify significant effects.

There are a number of factors that could have contributed to these observed improvements in depression and HRQOL. As a rehabilitative therapy, tango may improve specific functional impairments (18) that are known to influence perceived QOL for adults with PD (38). Research shows exercise to be an effective treatment for depression (39), and Ballroom dance has been shown to improve depression in the geriatric population (17). It is thought that dance may also improve well-being through enabling emotional expression and engagement (17), as well as building social support networks which may improve QOL for adults with PD (9).

There are a number of limitations of this pilot study. The small size of the sample decreases the external validity of findings (19, 40). Similarly, the short duration of the intervention phase means that few inferences can be made regarding the longer-term adherence and retention in adults with PD. Without a pilot control group, it cannot be determined if a comparison intervention (e.g., physiotherapy, exercise, or another dance genre) is also feasible pertaining to *a priori* criteria or if there are systematic differences between two intervention groups. Without a control group, this study could not counter for other potential factors that may have contributed to the observed improvements in therapy outcomes such as natural recovery or regression to the mean. Data were collected on a sample that was homogenous in disease severity; thus, conclusions regarding the safety of the dance intervention may not be applicable to individuals assessed as Hoehn and Yahr stages III and higher where postural instability is a marked impairment (22). This study did not evaluate participant acceptability of the intervention, which is suggested to be an important variable in pilot research (19). Furthermore, this study did not monitor if participants were concurrently undergoing active therapy for depression, which could have contributed to the observed improvement.

To facilitate the safety of the intervention delivery, it is proposed that dance instructors are educated regarding the disease processes, movement impairments, and likely comorbidities experienced by adults living with PD. Practical sessions would be useful to train instructors in pacing and movement selection (33). These sessions could educate instructors in therapeutic cuing, which

### References


could maximize movement and enhance motor learning for participants. Additional literature regarding community-based exercise programs for PD advocates basic training in first aid, emergency procedures, and exercise physiology (7).

It is recommended that the next stage of pilot research adopts a recruitment strategy based on the projection of recruiting 8–10 eligible participants over a 3-month period. Additionally, greater consideration should be given to location of the dance studio to limit barriers associated with long travel.

It is necessary for participant safety that the appropriate supervisors are present to monitor for fatigue, instability, and to provide appropriate suggestions regarding modifications. Hackney and Earhart (18) suggest that physiotherapy students may provide enthusiastic assistants, with the appropriate knowledge in risk management.

### Conclusion

A 4-week Argentine tango program was safe and enjoyable for people with relatively mild Parkinson's and was associated with alleviation of depression in some. There remains a need to verify the safety of the intervention for people with PD, who are more disabled or with a greater fall risk. Arguably, people with more advanced disease and with a greater fall risk might show greater gains with this type of physical activity.

### Author Contributions

LB, MM, and WB participated in the design of the study, oversaw data collection, data analysis, preparation of the manuscript, gave final approval of the version to be published, and agreed to be accountable for all aspects of the work.

### Acknowledgments

The authors would like to thank Rina Ohashi, who designed and instructed the dance classes, to the physiotherapists that assisted with supervision, to Parkinson's Victoria who helped with recruitment, and finally to the participants with Parkinson's who graciously donated their time in participating in this trial. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Blandy, Beevers, Fitzmaurice and Morris. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The Embodied Self in Parkinson's Disease: Feasibility of a Single Tango Intervention for Assessing Changes in Psychological Health Outcomes and Aesthetic Experience

Sabine C. Koch1, 2 \*, Katja Mergheim<sup>1</sup> , Judith Raeke<sup>1</sup> , Clarissa B. Machado<sup>1</sup> , Eliane Riegner <sup>1</sup> , Joachim Nolden<sup>1</sup> , Gudrun Diermayr <sup>1</sup> , Dorothee von Moreau<sup>1</sup> and Thomas K. Hillecke<sup>1</sup>

*<sup>1</sup> Faculty of Therapeutic Sciences, SRH University Heidelberg, Heidelberg, Germany, <sup>2</sup> Department for Creative Arts Therapies and Therapy Sciences, Alanus University, Alfter, Germany*

Background: Dance is an embodied activity with benefits for mobility, balance, and quality of life (QoL) of persons affected by Parkinson's Disease (PD). It is enjoyable and likely to support adherence to movement prescriptions. The objective of this study was to assess the feasibility of measuring changes in psychological outcomes, specifically well-being, body self-efficacy, outcome expectations, and experienced beauty after a single Argentine Tango intervention in a workshop format. To anchor experienced beauty in a theory, the article introduces a *model of embodied aesthetics* featuring active art-making as a central aspect of healing in arts-based interventions.

#### Edited by:

*Marta Bienkiewicz, ´ Aix-Marseille University, France*

#### Reviewed by:

*Gunter Kreutz, Carl von Ossietzky Universität Oldenburg, Germany Andrea Schiavio, The Ohio State University, USA*

> \*Correspondence: *Sabine C. Koch Sabine.Koch@alanus.edu*

#### Specialty section:

*This article was submitted to Auditory Cognitive Neuroscience, a section of the journal Frontiers in Neuroscience*

> Received: *14 January 2016* Accepted: *08 June 2016* Published: *07 July 2016*

#### Citation:

*Koch SC, Mergheim K, Raeke J, Machado CB, Riegner E, Nolden J, Diermayr G, von Moreau D and Hillecke TK (2016) The Embodied Self in Parkinson's Disease: Feasibility of a Single Tango Intervention for Assessing Changes in Psychological Health Outcomes and Aesthetic Experience. Front. Neurosci. 10:287. doi: 10.3389/fnins.2016.00287* Methods: In a single-group pre–post design, we tested the feasibility of measuring psychological changes of 34 PD patients from Southern Germany after an introductory workshop in Argentine Tango. They participated in a 90 min *Tango for PD* intervention and completed the *Heidelberg State Inventory* (HSI-24; (Koch et al., 2007)), the *Body Self-Efficacy* Scale (BSE; (Fuchs and Koch, 2014)) with a sub-dimension on *aesthetic experience*, and the *Credibility-Expectancy Questionnaire* (CEQ; (Devilly and Borkovec, 2000)) before and after the intervention. A subgroup completed the *therapeutic factors of arts therapies*-scale, a new measure to elaborate on the aesthetic experience. We analyzed pre–post-differences with a *t*-test for paired samples.

Results and Discussion: The study supports the feasibility of measuring health-related psychological changes from a single Argentine Tango intervention for PD patients, as well as acceptance and appropriateness of the intervention for the patient group. After the tango intervention, well-being, body self-efficacy, and outcome expectancies increased. Participants also experienced an increase in beauty of their movements and other aesthetic aspects. We suspect that, in addition to the functional and psychological factors identified so far, the aesthetic experience in dance may be an important therapeutic factor mediating several outcomes of dance and other arts-based interventions. A controlled study for evidence-based testing of targeted variables can now follow to examine the new hypotheses.

Keywords: embodiment, Parkinson's disease, Argentine Tango, dance movement therapy, arts therapies, body self-efficacy, beauty, aesthetic experience

### INTRODUCTION

Parkinson's Disease (PD) is a progressive neurodegenerative disorder associated with movement disorders including impaired functional mobility and postural instability (Hackney and Earhart, 2010). The death of dopamine cells in the substantia nigra causes bradykinesia (slow movements), akinesia (absence of or impoverished voluntary muscle movements), resting tremor, rigidity ("stiff " movement, decrease in flexibility), freezing (no movement), and postural instability (Lindenbach and Bishop, 2013). Classical therapies for PD employing dopamine-receptor agonists or deep brain stimulation (DBS) improve PD symptoms but may cause side-effects such as drug-induced dyskinesia and surgical complications (Lindenbach and Bishop, 2013).

In recent years, there has been an increasing interest in non-pharmacological therapies and the effects of movement in general, and dance in particular, for people with PD. Integration of dance, and particularly tango, as an innovative approach into PD rehabilitation is supported by a number of recent reviews (e.g., Hackney and Bennett, 2014; Lötzke et al., 2015; Sumec et al., 2015; Abbruzzese et al., 2016) reporting dance-related effects on physical functions, as well as on cognitive and psychological outcomes such as depression level, enjoyment, and well-being. Here, we focus on psychological effects of Tango for PD from an embodiment perspective, and the aesthetic component that has not yet been taken into account as a mechanism. This component is therefore in this work first anchored in a theory model.

An embodiment perspective addresses the "living body" ("to be a body"—as opposed to the "body proper," i.e., "to have a body;" Merleau-Ponty, 1962), that is the body–mind unity as an entity, and provides a timely framework to understand and explain phenomena we meet in Parkinson's Disease (Fuchs and Koch, 2014; Schiavio and Altenmüller, 2015, this volume). In our study, we drew upon embodied enactive approaches from cognitive and neurosciences (e.g., Niedenthal et al., 2005; DeJaegher and DiPaolo, 2007), clinical psychiatry and clinical psychology (Fuchs and Schlimme, 2009; Koch, 2011; Ramsayer and Tschacher, 2011; Michalak et al., 2015), and the arts therapies (Koch and Fischman, 2011; Koch and Fuchs, 2011; Schiavio and Altenmüller, 2015); we focused on the effects of a body-based intervention on bodybased psychological outcomes. Embodied perspectives represent a novelty in understanding psychotherapeutic interventions and assume that our body is a therapeutic entry point that can operate relatively independently from the verbal level (Ramsayer and Tschacher, 2011; Michalak et al., 2015), but equally related to meaning. The body thus often provides an unexplored potential, a second option next to a verbal entry point to therapy. Embodied approaches assume that emotions are closely connected to our bodies (e.g., via embodied simulations; Gallese, 2001; Niedenthal et al., 2005), and that cognition and abstract concepts are based on sensorimotor processes (Lakoff and Johnson, 1999). The enactive view (Varela et al., 1991) draws from theoretical biology and looks at individuals as living systems. Principles of living systems are agency, plasticity (moment-to-moment adaptations), striving for balance, self-organization (Haken, 2010), autonomy, sense-making (DeJaegher and DiPaolo, 2007), embodiment, emergence, experience, and action-perception coupling as well as organism–environment coupling (Kelso, 1995). Enactive approaches have many of the same premises than embodied and dynamic systems approaches.

Embodiment approaches provide a framework from which PD and its effects are understood as a disorder of the embodied self, a concept focusing on the organismic and animated nature of human beings (Sheets-Johnstone, 1999; Fuchs and Schlimme, 2009), and the body–mind unity. The subjective bodily experience of the patient is taken seriously and the embodied interaction with the patient as experienced by the clinician is an important source of information. In Tango for Parkinson, movement and embodied therapies are employed in the service of increasing quality of life (QoL) and well-being. The body is the place where the illness happens, which endows it with an important role as a therapeutic entry point for increasing well-being and QoL. Bodily arousal and regulation mechanisms effect affect, attitudes, and behavior in coping with the illness. Our own previous work includes an embodied enactive theory framework for dance therapy (Koch and Fischman, 2011), a framework on embodied affectivity (Fuchs and Koch, 2014), and evidence-based studies on the effects of dance movement therapy on health-related psychological outcomes (e.g., Koch et al., 2014).

Dance has previously been investigated intensively in the older adult population showing that this relatively moderate form of physical activity yields improvements in balance and cognition (Kattenstroth et al., 2013; for reviews see Hwang and Braun, 2015; McNeely et al., 2015). While Lötzke et al. (2015) and McNeely et al. (2015) point out that there are still open questions in dance for PD, recent findings encourage its implementation into clinical practice. Studies report general positive effects of partnered dancing (Ashburn et al., 2014), music-based movement therapy (DeDreu et al., 2012), and social dance sessions (Lewis et al., 2016) on PD, with tango representing the most investigated form of dance for PD (Hackney et al., 2007; Sumec et al., 2015). Hackney and Earhart (2010) found that a 10-week program of dance classes improved balance, walking velocity and cadence among people with mild or moderate PD. This was true for non-partnered dance as well as partnered dance, whereas increased enjoyment and interest to continue the program was higher in the partnered dance group. In a similar vein, McKee and Hackney (2013) showed that community-based Tango lessons over 12 weeks improved spatial cognition, balance, and executive functions, while disease severity decreased, compared to a control-group receiving educational lessons. Duncan and Earhart (2012) found decreased PD severity and better physical functioning in a randomized trial comparing a tango intervention group with a control group. Safety issues of such tango interventions for PD patients have recently been addressed by Blandy et al. (2015; this volume).

In addition to effects on gait and balance (e.g., Fisher et al., 2008; Goodwin et al., 2008, 2011; Dibble et al., 2009; Morris et al., 2010; Duncan and Earhart, 2012, 2014), tango increased QoL (e.g., Hackney and Earhart, 2009), personal and social activities (Foster et al., 2013), and cognitive and psychological variables (Hashimoto et al., 2015). Support for the latter findings comes from a survey on the benefits of dancing among adults from Quiroga Murcia et al. (2009, 2010); participants reported that dancing affects emotional and physical aspects of health as well as social and spiritual domains and in particular selfesteem and coping strategies (Quiroga Murcia et al., 2010; see also Kreutz and Quiroga Murcia, 2015). In sum, however, metaanalyses and systematic reviews (e.g., Hackney and Bennett, 2014; Mandelbaum and Lo, 2014; Shanahan et al., 2014; Sharp and Hewitt, 2014; Lötzke et al., 2015) agree that the largest effects of Tango for PD exist for improving gait, balance, and QoL.

Why tango? Hackney and Earhart (2009) have found that Tango Argentino, with its improvisational nature, works better in improving gait, balance, and QoL in PD than ballroom dancing. One reason may be that patients are so focused on following and attuning to the partner in the present moment that they are not even realizing that they walk backward over extended periods of time, that they turn securely backward and forward, and that they repeatedly initiate movement without any problems. There have been recent findings on the health improving nature of nongoal-oriented improvisational dance movement that argue along the same lines (Wiedenhofer et al., 2016; Wiedenhofer and Koch, under review).

How short can the intervention be to still realistically lead to psychological effects? When Earhart (2009) or Duncan and Earhart (2012) talk about short-term interventions, they think of 6- to 10-week programs of Tango for PD (vs. 6- to 12 month programs that they have also been running). In our case, we assessed an ultra-short-term intervention in form of a 1.5 h workshop regarding feasibility, acceptance, and changes on psychological outcomes.

In conclusion, the Tango for Parkinson studies highlight the improvements in motor function for people with PD and high attrition rates in the dance groups compared to traditional exercise groups. While different motor symptoms were investigated by most of these studies, psychological variables are not well-understood. Duncan and Earhart (2012) stress the importance of additional work to explore the effects of exercises and dance on non-motor symptoms and activities of daily living. Therefore, the purpose of the exploratory part of the study was to assess the effects of a single intensive tango intervention on wellbeing, body-self efficacy, and patients' outcome expectancies.

In clinical practice, dance interventions can support adherence to keep high levels of daily movement and social activities, among other factors by causing pleasurable and aesthetic experiences from and with one's own body. The goal of this study was to employ tango for PD patients to explore its impact on health-related psychological outcomes in the course of assessing the feasibility of a workshop format, and to explore the aesthetic experience as a therapeutic factor, an aspect previously unaddressed. We anchor this aspect in the theory model of embodied aesthetics (Koch, 2016). This model can help us understand how dance therapy works from an arts therapies perspective.

Aesthetics has been defined as "a sensory experienced knowledge" by Baumgarten (1750/2007). Allesch (2006) pointed out that there is a lack of an aesthetic theory model suited for the arts therapies, and demands us to think big, that is to expand our thinking to include all forms of art, and not exclude nature, and everyday aesthetic phenomena. Given the lack of a theory model suited to explain the therapeutic factors of art-making, the model of embodied aesthetics was recently developed by Koch (2016) on the basis of the embodied affectivity model (Fuchs and Koch, 2014) and applies a circular causality and dynamic systems logic rather than a linear-causality logic (Thelen and Smith, 1994; Salvatore et al., 2015). The model extends the present state-of-the-art cognitive sciences model by Leder et al. (2004; **Figure 1**). As a classical input–output model, Leder's model focuses on the stages of cognitive and affective processing of an aesthetic stimulus with the result of a cognitive and an affective outcome (i.e., aesthetic judgment and aesthetic affect).

While the model of Leder et al. (2004) is a classical input– output model, the model of embodied aesthetics of (**Figure 2**) can be viewed both as an extended input–output model, and as a circular, systemic model, since input and output are in fact assumed to be parallel and not sequential processes. Such a "hybrid model" (of systemic and input–output model) at this stage allows us to investigate contexts with both experimental and dynamic systems methods. This is useful for our research context, as the evidence-based experimental studies in the health sciences are still regarded as the main pathway to advancing the general base of knowledge (whereas growing lines of science proceed to use systems models to be able to more adequately reflect the complexities of the "reality" of living organisms in living environments).

The two models are not exclusive of each other. The main point we want to make here is that the model of Leder et al. (2004) focuses solely on aesthetic perception (just as other cognitive science models of aesthetic experience; e.g., Martindale, 1984; Ramachandran and Hirstein, 1999; Reber et al., 2004; or updates of Leder's model by Chatterjee and Vartanian, 2014, or Leder and Nadal, 2014; even the model of Gifford, 1997, who bases his ecological aesthetic theory on Brunswick's lens model, is not helpful in terms of an embodied enactive perspective). Aesthetic action is missing. Yet, aesthetic action is central to the idea of healing and to the therapeutic process in the arts therapies.

The circular model of Koch (2016) complements aesthetic perception (impression side), addressed by the model of Leder et al. (2004; and other cognitive science models), with the previously unaddressed side of active art-making (expression) as practiced in creative arts therapies (Wallbott, 1982; Scherer and Wallbott, 1985; Wallbott, 1990; Fuchs and Koch, 2014; Koch, 2016). An aesthetic experience does not only result from an impression (perception) of art, but also from the expression of it (action). This experience may pass from mere playful expression and enjoyable experimenting (e.g., with music and dance), via a self-efficacy experience, for example, when moving or playing an instrument, to symbolic expression (e.g., how would your joy sound?), or the creation or formation of something beautiful/authentic in any arts modality.

The cycle is here described in movement, since movement provides the most immediate body feedback (Koch, 2014). The moment I move and experience my movement (e.g., as beautiful, authentic, or creative), I am affecting myself in a twofold way (cf.

Merleau-Ponty, 1962, 1964): I am moving and being moved both at once—there is an overlay in sensory and motor-mapping and it is a mere attention issue where consciousness is focused on in a given moment; there is no movement without perception and no perception without movement (vgl. von Weizsäcker, 1940; Merleau-Ponty, 1964; Gibson, 1966). Likewise, when I stand in a museum absorbing a piece of art, my body resonates with it (my breath may reflects a change of my organismic system' arousal, my emotions may get involved, cognitive conflicts may emerge and strive for solution); when I see a movie, a theater piece, a music performance, my body resonates with it; my body is going into synchrony or asynchrony with it. This resonance is providing body feedback (aesthetic impression) and is also a starting point for aesthetic expression<sup>1</sup> .

In tango dancing there are external stimuli that initiate an organism–environment coupling in relation to the music, the partner, and the group. The "art product" is just a transient movement, a fleeting moment of beauty (or heightened authenticity). Music and partner affect and move the person in particular yet unpredictable and emergent ways that resonate within the lived body. The person's bodily resonance (kinesthetic) and the dance (kinetic) change and influence one another on a moment-to-moment basis (Merleau-Ponty, 1962, 1964; Koch, 2011), and bring body and mind into an experienced unity (Mainka, 2015; this volume). This may be observed, for example, by the synchronization of body rhythms on an individual as well as an interpersonal level (Koch, 2014; Edelhäuser et al., 2015; Heusser, 2015).

### The Present Study

The aim of our study was to show the feasibility of measuring health-related psychological changes following a single tango intervention, to ensure patient acceptance and that the workshop format is appropriate for patients with severe neurodegenerative health problems such as PD. We explored the following question (with the rationale for the selected outcome variables provided below): Can a single tango-intervention improve well-being, outcome expectations, and body-self efficacy, including aesthetic experiences, in PD patients?


<sup>1</sup>Embodied simulation theory has been put forward to explain bodily resonance in light of the discovery of a mirror mechanisms in the brain (Rizzolatti et al., 1996; Freedberg and Gallese, 2007).

The idea that tango can increase patients' body self-efficacy has not yet been investigated in PD patients.


We assumed that well-being, body self-efficacy and outcome expectations increases after a single tango-intervention. Moreover, we expected an increase in aesthetic experience, particularly experienced beauty.

### METHODS

### Sample

Thirty-four participants with PD from Southern Germany participated in three groups. Twenty-six participants were women and eight were men, the age range was 40–82 years (M = 60.5; SD = 11.06), with the mode at 50 years (n = 10), and thus a skewness on the younger side. Participants were recruited through contacts to the PD-support groups of Heidelberg, Sulzbach (Taunus), Mannheim, and Ludwigshafen. Two of them had had previously danced tango, 13 more reported to have formerly danced as a hobby. The live music was a new element for them. All but one patient successfully participated in the classes: the one patient in the second Heidelberg group sat from beginning to end and was not included in the data analysis. Most participants were pensioners (the regular retirement age in Germany is 65); their degree of handicap was between 50 and 100% with a mean of 72.78 and an SD of 18.31, nine persons were classified with a 70% handicap (mode), 5 with 50%, 5 with 100%, the remaining in between with a range from 40 to 100%. All participants were Caucasian. The study was carried out with written informed consent from all subjects in accordance with the Declaration of Helsinki, and followed the data protection requirements of the outpatient center at SRH University Heidelberg, without being separately submitted to an ethics committee.

### Procedure

The study was conducted in three introductory workshops to tango therapy for PD patients. Workshops took place at two different sites with two different dance movement therapists as session leaders. For an overview of the sample and settings, see **Table 1**. The first workshop was organized in February 2014 via the Sulzbach PD-support group who traveled to SRH University Heidelberg in order to participate in the study (18 participants, 7 partners). The second workshop was announced via the Heidelberg PD-support group and took place in June 2014 at SRH University of Heidelberg (5 participants, 1 partner), and the third workshop took place in May 2015 in the Ludwigshafen PDsupport group setting (11 participants, 3 partners). In all three workshops participants first heard a presentation of ∼30 min by the first author, head of the dance movement therapy Master Program at the SRH University Heidelberg, introducing them to the effects of arts therapies, particularly dance, movement, and music, on PD. Then participants provided informed consent and filled in the pre-test questionnaires (15 min). Then the tango intervention took place. After a short warm-up and introduction of the therapist (Workshop 1: Clarissa Barcellos Machado, dance movement therapist and tango teacher from Argentina; Workshop 2 and 3: Eliane Riegner, dance movement therapy advanced student and tango teacher from Germany), the participants were invited to join different exercises, for example, walking next to each other in pairs, one person leading, while the other person was following (e.g., walking backwards). They practiced basic steps of Tango Argentino first on their own and then in pairs. During the partnered dance, the PD patients worked with their spouses, relatives/friends, or students of the

#### Koch et al. Embodied Self in Parkinson


#### TABLE 1 | Sample characteristics.

*Sample characteristics, therapists, and scales employed; PD, Parkinson's Disease; ATs, Arts Therapies.*

SRH Dance Movement Therapy Master Program as partners, or in rare cases with other patients. The entire intervention lasted ∼90 min (for description of the intervention see Appendix A in Supplementary Material). The Argentine Tango instructor and dance movement therapist (CM) instructed in English and a German dance movement therapy student translated after each few sentences; this first workshop used recorded music, but the last three songs were accompanied by live bandoneon music. The second and third workshop, led by a tango instructor, and advanced dance and movement therapy student from Germany (ER), employed only recorded music. Participants in all three workshops were told that they could take their shoes off, if that was more comfortable for them, and that they could sit down and take a rest during the session whenever needed, which they did selectively. They completed the post-test immediately after the intervention (∼15 min). Some needed help reading the items, which was provided by student helpers, however, all participants filled in the questionnaires by themselves.

### Materials and Instruments

### Material for the Intervention

To find the correct body posture for Argentine Tango, the instructors used yellow "post-its" (sticky paper) that each patient attached to his/her sternum. The "post-it" symbolized a light that shines from the chest to the outside just like a beam. This device helped participants to find the correct upright posture, position of the spinal column, muscular tension, and convexity of the upper body for the dance and to connect to the partners via this point of the body.

### Psychometric Instruments (See Appendix B in Supplementary Material)

### **Psychological well-being**

Well-being was measured by the 24-item "Heidelberg State Inventory" (HSI-24; Koch et al., 2007; unipolar version in the Appendix of Supplementary Material) with a range from "1" ("does not apply at all") to "6" ("applies exactly") assessing tension, anxiety, coping, positive affect, depressed affect, and vitality. The scale is based on a review of central outcomes of dance movement therapy by Goodill (2006). The internal consistency of the entire scale in previous studies was acceptable to excellent with Cronbach's alphas between 0.68 and 0.97. (e.g., Koch et al., 2007, 2015), and factor analyses mostly yielding a general factor of "positivity vs. negativity."

### **Body Self Efficacy**

Body Self Efficacy (BSE) was assessed with the BSE-scale (Fuchs and Koch, 2014). The validated German version of this questionnaire contains a 10 item scale measuring the selfperception of bodily abilities ("I can's;" Husserl, 1952) with items such as "My body is flexible," "My body feels like in pieces," etc., on rating scales from "0" ("does not apply at all") to "5" ("applies exactly"). In former studies, the internal consistency of the German version of the scale was Cronbach's alpha = 0.75 in a student sample, and Cronbach's alpha = 0.83 in a clinical sample (Kelbel, 2013; Fuchs and Koch, 2014).

### **The BSE-beauty subscale**

The BSE-beauty subscale consisted of the two items "My movements are beautiful" and "I can move elegantly/with grace." We looked at this subscale separately, because in working with PD patients, we became increasingly aware that art-based intervention have additional therapeutic factors other than functional and classical psychological ones.

### **Expectancies**

Patients' expectancies of the therapy outcome were measured using the Credibility Expectancy Questionnaire (CEQ; Devilly and Borkovec, 2000) with four items on cognitive expectancy (credibility), e.g., "How logical does the therapy offered to you seem?" on a rating-scale from "1" "not at all" to "10" "very much," and four items on affective expectancy, e.g., "How much do you really feel that the therapy will help you to reduce your symptoms?"). In previous studies (e.g., Devilly and Borkovec, 2000) high internal consistencies were found for both the cognitive (Cronbach's alpha = 0.86) as well as the affective expectancy factor (Cronbach's alpha = 0.90). The English version of the CEQ was translated into German for the purpose of this study by co-author Judith Raeke.

### **Therapeutic Factors of Arts Therapies in PD related to the Aesthetic Experience**

The scale on therapeutic factors of arts therapies in PD (Mergheim, 2015; see Appendix C in Supplementary Material) was composed on the basis of face validity of the symptoms and needs of the PD patients and the assumed aspects of the aesthetic experience such as beauty, flow, happiness, unity with self and unison with partner. It further contained items on pleasure/joy, expressiveness, and communication fluency.

Feasibility of appropriateness and acceptance of the workshop format was assessed by observations, conversations with participants, as well as short interviews with volunteering patients (collected by a research assistant; and for a radio report after the second workshop).

### Statistical Analysis

Outcome measures of the exploratory study were analyzed with a t-test for paired samples for pre–post differences with time as the factor using SPSS (version 23.0). The alpha-level was set to 0.05. After Bonferroni correction the new Alpha-level for the primary outcome was 0.01, and for the secondary outcome 0.008.

### RESULTS

### Feasibility and Acceptance

Feasibility of the intervention was evidenced (a) by the fact that merely one participant had to sit out from the intervention for physical reasons, (b) by observations of an increase in patients' positive affect, (c) by participant utterances in brief interviews (see below), and (d) by the fact that we received continued requests for more workshop offers from participants.

Participants profited regardless of their aims. A recently diagnosed woman had the aim "I want to fight the stiffness of the limbs and the difficulties with balance with movements that keep me as mobile as possible; my kids are 13 and 17 years old and still need me; that requires a certain speed in everyday life; I hope to keep up with them and be able to share their tempo" (age 52, for 3 years diagnosed with PD); and an older lady: "I am shaky and slow, and lately it has been getting worse; here I am fighting to keep what I have" (age 79, for 27 years diagnosed with PD).

Despite the challenge of the workshop for many participants, evaluations of the Tango for PD intervention were positive "I finally can breath again," "I feel happier, more free, and mobile" "When I arrived I was totally down... that has changed," or "There were more changes happening than I expected. I feel good and feel ready to continue," "The workshop has been fun and inspired me to continue, I want to do more," "This wonderful workshop has caused great joy, I want to continue in any case," or "Even though it was physically demanding, it worked for me." No negative voices were recorded in the interviews, yet since only volunteers were interviewed, self-selection bias needs to be accounted for.

The final statement of the mother with the two children from above was "When one realizes that it becomes increasingly more difficult to move, there is a high probability that one withdraws. The prescribed exercise is often an unloved duty. But here in the tango workshop, everybody is in a similar situation. This takes away the achievement pressure. Music and dance were completely relaxing, and the movement became increasingly easy; for a moment, I had totally forgotten that I am actually sick."

## Exploratory Study

### Primary Outcomes

In the post-test participants showed significantly improved scores on well-being, body self-efficacy and the cognitive aspect of outcome expectancy (see **Table 2**; **Figure 3**). Controlling for workshop group yielded no significant differences between groups regarding the outcome.

The changes of means from pre- to post-test scores increased significantly across the four measures. Well-being t(33) = −3.73; p = 0.001, d = 0.69, cognitive outcome expectancy t(33) = −4.02; p = 0.000, d = 0.55, affective outcome expectancy t(33) = −3.31; p = 0.002, d = 0.44, body self-efficacy t(33) = −3.59; p = 0.001, d = 0.65, and the beauty aspect of BSE t(28) = −2.81; p = 0.009. d = 0.66, all improved beyond the critical alpha of 0.01 (effects sizes are Cohen's ds).

### Secondary Outcomes

Given that the aesthetic experience is assumed to be an important active factor (mediating mechanism; Koch, 2016) in the arts therapies and the related input of our continuous groups' members, and the fact that the BSE subscale of beauty (with two items) improved with t(28) = −2.81; p = 0.009. d = 0.66, we constructed a measure to investigate the aesthetic experience in PD more closely. The therapeutic factors scale of eight items reflects central hypothesized active factors in arts therapies related to the aesthetic experience particularly geared toward PD patients, such as experienced beauty, pleasure/joy, happiness, expressiveness, fluency of movement and speech, body–mind unity, and unison with the partner. Participants reported an increase in the aesthetic experience after the tango workshop (see **Table 3**):

Happiness increased after the tango intervention t(10) = −3.60, p = 0.005, d = 0.99, and experienced beauty t(10) = −2.84, p = 0.018, d = 0.94; so did other active factors of arts therapies related to aesthetic experience, such emotional expression t(10) = −2.67, p = 0.024, d = 0.75, flow of movement t(10) = −4.08, p = 0.002, d = 1.65, body–mind unity t(10) = −2.81, p = 0.019, d = 1.06, and unison with partner t(10) = −2.73, p = 0.021, d = 0.76. Joy/pleasure t(10) = −2.19, p = 0.053, d = 0.67, and flow of speech t(10) = −2.14, p = 0.058, d = 0.67 suggest a tendency in the expected direction; the effect sizes provide arguments for further testing. The total score of the Therapeutic Factors of Arts Therapies for PD Scale showed a significant increase with t(10) = −3.82, p = 0.003, which was beyond the critical alpha of 0.008. The effect size of the entire scale was Cohen's d = 1.09.

Because the results of the Therapeutic Factors of Arts Therapies for PD Scale rest on a subsample of 11 patients only, they need to be interpreted with according caution, but can provide first ideas for follow-up studies.



*Descriptive results: Means (M), standard deviation (SD), and sample size (N); higher post-test scores represent improvement; Inferential statistics: t-value (t), degrees of freedom (df), probability-value (p), and effect size (Cohen's d); Design: pre–post within-group comparison; based on Bonferroni-correction values of p* < *0.01 reflect significant differences.*

### DISCUSSION

We tested the feasibility of a single Tango for Parkinson intervention for measuring effects on health-related psychological outcomes. After the intervention, we observed increased well-being, body self-efficacy and outcome expectancies; our exploratory results are in line with prior and present studies suggesting positive effects of dance on psychological outcomes (Koch et al., 2014). Results suggest feasibility of the single tango intervention regarding the appropriateness and acceptance of the workshop format for PD patients and the sensitivity of the chosen measures in such a


TABLE 3 | Secondary outcomes: therapeutic factors of arts therapies and effects of aesthetic experience in Tango for PD.

*First results of the scale on Therapeutic Factors in Arts Therapies for PD, created to capture the aesthetic experience and other active factors (Appendix D in Supplementary Material); Descriptive Results: Means (M), standard deviation (SD), and sample size (N); higher post-test scores represent improvement; Inferential statistics: t-value (t), degrees of freedom (df), probability-value (p), and effect size (Cohen's d); based on Bonferroni-correction values* <*0.008 reflect significant differences; items 1–3 showed significant correlations among each other.*

context. The intervention can now be included into a controlled design to compare, for example, Argentine Tango and other enjoyable non-dance movement interventions; a design that would allow to further identify/control for aesthetic experience.

The acceptance of patients may be particularly high, because Tango does not only address the physical and social aspects of the disease, but also the psychological co-morbidities of it and the beauty aspect that we find in cultural and artsbased interventions. It may thus be experienced as more holistic, affecting the entire person, as opposed to more functional techniques. One could rightfully ask, whether cultural techniques, in general, should thus be attributed a more central role in rehabilitation programs. The exploratory findings of the study were also encouraging, particularly considering the small yet appropriately powered—sample size and the short duration of the intervention. However, careful interpretation of the results is warranted, because of the non-controlled character of the study. The study suggests that the tango intervention may positively influence well-being, patient's cognitive expectancies, and body self-efficacy. The suggested increase in well-being with a medium effect size is in line with prior findings of effects of dance movement therapy on other patient groups such as depressed patients (Koch et al., 2007), or subclinical samples. Movement in a protected setting seems to generally stimulate positive affect, vitality, and coping, and to decrease tension, depressed affect, and anxiety. In this study, particularly positive affect, vitality and coping increased. We would like to encourage other groups to make use of measures such as the HSI-24 that cover mental health variables, since depressed and anxious affect—often poorly recognized—are in many cases influential comorbidities of PD.

This is the one of the first attempts to address the importance of patients' expectancies in dance movement therapy. Cognitive and affective expectancies increased after the intervention, leading to the assumption that the interest and motivation to continue the activity was present in the participating patients (which was also perceptible from repeated requests for more workshops of patients, who lived too far away to come traveling to the regular groups). Interestingly, the data suggests that patients with initially higher expectations showed a higher increase in well-being and body self-efficacy, and less negative affect than participants with lower initial expectancies of therapy outcome. Although this was not the focus of the study, it does highlight the importance of the expectancy construct for other studies exploring group differences, especially where the initial expectancies could explain some differences of the later therapy outcome and attrition (Ametrano, 2011).

Body self-efficacy increased in the PD patients after the intervention. Even though or maybe because of the fact that patients verbally reported the intervention to have been a challenge for them, the feeling to have mastered the challenge could have been one contributing factor to this increase of the belief in one's own bodily skills. An enactive approach supports this argument, putting agency at the base of identity and health. The embodied-enactive approach is more suited than a classic cognitivist account for explaining effect of Tango on Parkinson's disease, in that it takes into account bodily self-regulation, and action-perception coupling; those processes are characterized by multiple feedback relations, and cannot be captured by linear input–output schemata.

Clusters of therapeutic factors/active factors in arts therapies (Koch, 2016):

(a) Hedonism: Art for pleasure and play → probing and enacting (future and past), "as if" space


The intervention can be as short as 1.5 h and already show an effect on psychological outcomes. This extends the assumption of Earhart (2009), Hackney et al. (2007), or Duncan and Earhart (2012) that short-term interventions start at 6 weeks, possibly owed to the fact that these researchers are physical therapists focusing on functional changes; whereas psychological changes, emotional, as well as motivational, can sometimes come about much faster. For measuring psychological change, our intervention of only 1.5 h was feasible, acceptable and appropriate for the patient group.

### Experiencing Beauty

In arts therapies, we assume that the aesthetic experience is an important health predictor. In tango, the patients not only practice functional skills such as walking backward over an extended time, turning, initiating, and stopping, but also experience the pleasure of the music, of the company of their partner, and of the dance as a holistic experience per se. In a moment of aesthetic experience, I feel alive, I feel my body in unison with my mind (thoughts and feelings are in a heightened congruency), and I may experience my own body or movements as beautiful. These moments of aesthetic experience emerge in a complex pattern and are thus difficult to capture. Yet, they are crucial to understanding the workings of the arts therapies, such as dance movement therapy (see **Box 1**).

Well-being and positive affect are important preconditions of a more global cognitive processing of stimuli (Gasper and Clore, 2002; Bless and Fiedler, 2006). They may challenge the patient away from more detailed processing and rumination to a more holistic processing style where unity of body and mind possibly can be experienced more easily. Positive outcome expectations and the increased experience of self-efficacy/being in control may additionally lead to a more global processing style (Gasper and Clore, 2002), this could in turn support a more holistic perception and an increased experience of the body–mind unity (as parts of the aesthetic experience; see results for the body–mind-unity item).

The experienced unity of body and mind may be "recalibrated" in movement therapy sessions (Edelhäuser et al., 2015; Heusser, 2015), and the unison with the partner, or the group in tango therapy may play an additional role in that readjustment. Tango therapy can therefore be seen as keeping the brain-body-environment system adaptive. Although PD patients are often slower and more rigid in their movements, attunement and kinesthetic empathy are still possible. It seems that the plasticity of the brain-body-environment system can be increased by challenging it via movement (see results for the flow of movement item). It is the task of the therapist to find the right music, to synchronize to movements and rhythms of the patient, and to provide the amount of scaffolding that matches the needs of the patients.

### Limitations

Limitations of the feasibility study are the small sample size. Yet, for the exploratory test, assuming an effect size of d = 0.44, a p-value of 0.05, and a power of 0.80, the required sample size equals exactly N = 34 patients (computed with G∗Power, Faul et al., 2007). The biggest limitation of the exploratory test part of the study is its non-controlled character: all results reported here are merely suggestive and not conclusive. Moreover, the sample consisted of patients who were engaged in support groups and who were actively looking for alternative therapies, this could have contributed to higher levels of motivation and therapy expectancies compared to non-members of support groups. Moreover, we had no objective data other than selfreport concerning patients' PD diagnoses, its severity and their medication. Further studies with a bigger sample should include those variables and ensure generalizability. During the tango intervention the participants were free to choose their dance partners (spouses or relatives, students, or other patients), it is unclear whether and how the partners or the number of partner changes influenced the results. Because we had no control group, other factors such as the structured group activity, the degree of physical activation, etc., may have influenced the results.

Another limitation concerns the slightly different situational factors at the different workshops, such as live music at the end of the first workshop and a slightly shortened session in the last workshop due to external conditions. It would be important for further studies to keep these conditions standardized. Moreover, because of certain motor difficulties such as a hand tremor some patients needed help from partners and relatives to fill out the questionnaires, this could have influenced the result with respect to social desirability. In general, there may have been social desirability and demand effects, which were probably one of the biggest problems of the studies. Since the study was uncontrolled, we cannot rule out that the increase in values after the intervention was merely due to the fact that we created a pleasant atmosphere, transmitted a caring attitude, or a firm belief in the effectiveness of the intervention, or alike. Transmitted beliefs additionally may have caused expectancy effects, for example trough the speech of the first author at the beginning of the workshops. However, the expectancy questionnaire did not show a decrease but an increase in expectancies after the intervention, providing evidence against the latter assumption. In sum, future research is called to replicate this study with one or more control groups using a randomized allocation of patients.

### CONCLUSIONS

This exploratory study led to some interesting starting points for future research. Feasibility of measuring health-related psychological variables from a single tango intervention was given. Yet, long-term interventions and randomized controlled trials (RCTs) using Tango Argentino should investigate both health-related psychological symptoms as well as motor symptoms and their interactions in order to improve therapies for persons with PD. In fact, PD-symptoms should be directly included on scales, which was not possible here, because of the limited time frame of our study. Research could then test, whether short-term interventions have stronger effects on psychological, and long-term interventions have stronger effects on physiological variables. Specificity of the Tango intervention needs to be further addressed: can other interventions do just the same, and if so, which ones? Equally important is the separation of contributing factors, such as the role of music, rhythm, psychological factors, preferences, etc. For example, Nombela et al. (2013) suggest that music facilitates activation of motor networks that bypass the disease-affected networks via cerebellum–thalamic–cortical circuitry. Therefore, it is important to objectify the musical, personal, and contextual variables that influence motor behavior in PD and other neurological diseases.

Future studies can investigate patients' expectancies as possible mediators to explain group differences, and can help to derive indications and contraindications for dance therapy with this specific population. Future studies with artsbased interventions should look at the additional outcomes of depression, anxiety, and body image changes, as well as the mediating factors of rhythmic activity (Sandel et al., 1993; Hackney et al., 2015, this volume), and resonance with self and other (Koch, 2011; Fuchs and Koch, 2014), in terms of body feedback (Koch, 2014), and embodied intersubjectivity (Fuchs and Koch, 2014).

To summarize, the study supports that a single dance movement therapy Tango for PD intervention is feasible for measuring changes on health-related psychological outcomes. It finds a positive relation between the tango intervention and the health-and adherence-related psychological outcomes

### REFERENCES


of well-being, body self-efficacy, and outcome expectancies in PD patients, and identifies the potentially influential mediator of aesthetic experience. Due to the lack of a control group, results of this study are only suggestive, not conclusive. The usefulness of embodiment approaches and the role of the aesthetic experience, as a therapeutic factor of the arts therapies, is an aspect of Tango for PD that calls for further attention and investigation.

### AUTHOR CONTRIBUTIONS

All authors contributed substantially to this work. SK analyzed the data, is responsible for the model, and wrote the first article draft, SK, KM, and JR collected and analyzed the data of the first workshop, KM, NH, and JR collected the data of the second workshop, and KM, ER, and SK collected the data of the third workshop and wrote up the corresponding parts. KM connected the Tango for PD workshops with the model of embodied aesthetics in her Master's Thesis, CM and ER conducted the intervention, consulted on the manuscript from the dance therapy side, and wrote the intervention description, GD consulted and co-authored from the physiotherapy side, DM, JN, and TH consulted and co-authored from the music therapy side. DM and TH directed the institution under whose supervision the study was conducted. Authors discussed the results and commented on the manuscript at all stages.

### ACKNOWLEDGMENTS

We thank the patients of the workshops for their participation and their confidence in the researchers and therapists. Thanks to Heike Brunner-Wild, Claire Majerus, and Susanne Hofinger for help with data collection and entry, Sabrina Junker for music therapy consulting, and Nicole Hartmann for dance therapy consulting. Thanks to Jörg Zumbach and Lily Martin for proof reading, formatting, and helpful comments on the manuscript. Thanks to Catherine Stevenson for English corrections. The work was not supported by any funding institution.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fnins. 2016.00287

ballroom dancing on people with Parkinson disease: randomized controlled trial protocol. JMIR Res. Protoc. 3, e34. doi: 10.2196/resprot.3184


disease: a systematic review and meta-analysis. Mov. Disord. 23, 631–640. doi: 10.1002/mds.21922


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Koch, Mergheim, Raeke, Machado, Riegner, Nolden, Diermayr, von Moreau and Hillecke. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Verbal Auditory Cueing of Improvisational Dance: A Proposed Method for Training Agency in Parkinson's Disease

#### *Glenna Batson1 \*, Christina E. Hugenschmidt2 and Christina T. Soriano3*

*1Department of Physical Therapy, Winston-Salem State University, Winston-Salem, NC, USA, 2Section on Gerontology and Geriatric Medicine, Department of Internal Medicine, Wake Forest School of Medicine, Winston-Salem, NC, USA, 3Department of Theatre and Dance, Wake Forest University, Winston-Salem, NC, USA*

*Edited by:* 

*Marta Bien´kiewicz, Aix-Marseille University, France*

#### *Reviewed by:*

*Madeleine Hackney, Atlanta VA and Emory University, USA Daniele Volpe, Casa di Cura "Villa Margherita", Italy*

> *\*Correspondence: Glenna Batson glenna.batson@gmail.com*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 14 September 2015 Accepted: 02 February 2016 Published: 17 February 2016*

#### *Citation:*

*Batson G, Hugenschmidt CE and Soriano CT (2016) Verbal Auditory Cueing of Improvisational Dance: A Proposed Method for Training Agency in Parkinson's Disease. Front. Neurol. 7:15. doi: 10.3389/fneur.2016.00015*

Dance is a non-pharmacological intervention that helps maintain functional independence and quality of life in people with Parkinson's disease (PPD). Results from controlled studies on group-delivered dance for people with mild-to-moderate stage Parkinson's have shown statistically and clinically significant improvements in gait, balance, and psychosocial factors. Tested interventions include non-partnered dance forms (ballet and modern dance) and partnered (tango). In all of these dance forms, specific movement patterns initially are learned through repetition and performed in time-to-music. Once the basic steps are mastered, students may be encouraged to improvise on the learned steps as they perform them in rhythm with the music. Here, we summarize a method of teaching improvisational dance that advances previous reported benefits of dance for people with Parkinson's disease (PD). The method relies primarily on improvisational verbal auditory cueing with less emphasis on directed movement instruction. This method builds on the idea that daily living requires flexible, adaptive responses to real-life challenges. In PD, movement disorders not only limit mobility but also impair spontaneity of thought and action. Dance improvisation demands open and immediate interpretation of verbally delivered movement cues, potentially fostering the formation of spontaneous movement strategies. Here, we present an introduction to a proposed method, detailing its methodological specifics, and pointing to future directions. The viewpoint advances an embodied cognitive approach that has eco-validity in helping PPD meet the changing demands of daily living.

Keywords: Parkinson's, dance, improvisation, cognition, spontaneity, balance, function, agency

Batson et al. Improvisational Dance Cues for Parkinson's

*When we begin to appreciate that movement is what matters … an idea emerges with the potential to reinterpret existing theories of communal dancing in such a way that we can appreciate the ongoing value of dance as vital art here and now. If we privilege bodily movement rather than matter as the currency of evolution, if we insist upon seeing any and all bodily forms as potentials for movement making, then we can begin to perceive dancing, then and now, as a practice in which humans exercise a distinctive human potential to participate in the ongoing evolution of the universe in human-enabling ways* (1).

## INTRODUCTION

A hallmark of human movement is volitional control – the freedom to move easily, automatically, and safely within the changing demands of daily living (2). Clinical signs characteristic of Parkinson's disease (PD) include rigidity and tremor, hypokinesia, stooped posture, vocal alterations, facial masking, and progressive loss of peri- and extrapersonal use of space (3). These movement aberrations are one of many factors that increase the risk of falling and point to the need for improving fall prevention strategies (4). These disease-related limitations experienced by people with Parkinson's disease (PPD) increasingly rob them of movement freedom. Quality of life declines as conscious attention and effort in everyday living replace spontaneity of communication and safety in navigation (5).

### The Benefits of Dance for PPD

Among the non-pharmacological treatments available to PPD, researchers have shown group-delivered dance classes to be effective in improving functional gains and quality of life measures (6–10). The number of randomized, controlled trials is limited at present due to the newness of the field of the study. Controlled studies in ballet, modern dance, Argentinian tango, other forms of social and ballroom dance, however, have demonstrated meaningful physical and psychosocial benefits (6, 11). A recent review of controlled trials by Sharp and Hewitt (10) found that compared to no intervention, dance results in statistically significant improvements in Unified Parkinson's Disease Rating Scale (UPDRS) motor scores, balance, and velocity. While exercise, in general, is beneficial for PPD (12), dance may provide benefit beyond those substantiated in common forms of aerobic exercise (e.g., walking, cycling). Relative to aerobic exercise, for example, dance improves balance and quality of life, as measured with the Parkinson's Disease Questionnaire 39 (PDQ-39) (10).

Exercise can be defined as "physical activity that is planned, structured, and repetitive for the purpose of conditioning any part of the body" (13). Dance can be defined as a "choreographed routine of movements usually performed to music" [Ref. (8) as cited in Ref. (10)]. Using these definitions, dance is a form of exercise where a series of planned movements are choreographed and practiced to achieve mastery. The addition of music may have specific benefits of its own, including being an auditory cue that helps maintain rhythm and unity among the dancers.

The exact neurobiological mechanism through which exercise benefits the brain is not known, but current evidence suggests that exercise-induced increases in blood flow, trophic factors, and anti-inflammatory cytokines may help protect dopaminergic neurons and synapses, or other neural circuits that interact with them (2, 13). In fact, physical exercise may be important for brain health by creating an environment that facilitates plasticity (14). Thus, pairing physical activity with a cognitively demanding task may facilitate learning or performance of that task (14). Increasing numbers of scientific studies indicate the importance of multiple system interaction and its consequent impact on motor and cognitive impairments in PD (15). Researchers who are investigating processes underlying dance making (choreography) have generated the term "creative cognition" (16). These investigations increasingly reveal the importance of mind–body integration and an understanding of cognitive processes through creative movement training. In this way, the cognitive challenges that dance poses may enhance the benefits of physical activity alone.

Dance is viewed as a unique form of movement training because of its potential to create and express new movement patterns. This helps build a psychophysical sense of "embodied and extended agency" (17). Agency is the capacity of individuals to act in the world autonomously and independently (18). From a phenomenological viewpoint, moving freely and easily relies in part on having a sense of agency (19). Agency implies more than merely experiencing one's body; rather, it suggests a more "embodied" sense of "I can," that is, that persons own a body and can act on their own behalf and meet their own needs (16, 17). A sense of agency is basic to quality of life, allowing individuals to synchronize their actions with their intentions (19, 20). In the case of learning dance, movement cues draw the mover's attention to explore agency through expressive whole body gesture. Meaning arises out of participants' discovery of new physical expressiveness that impacts positively on their ability to act spontaneously and immediately in relationship to change contexts (21, 22).

### Improvisation Defined

To date, protocols for group-delivered dance classes for PPD largely have employed choreographed dance with rhythmic, musical accompaniment (e.g., modern dance/ballet), or prescribed partnering moves within a codified and widely used social dance system (e.g., tango and other familiar social dance forms, such as ballroom or folk dance). These approaches have shown good success in promoting functional gains in balance, gait, and UPDRS score.

Improvisation is the ability to create new gestures and movements on the spur of the moment (23). Improvisation is not limited to dance but is also part of other performance arts, such as music or drama. Within the performing arts, the definition of improvisation varies (24–28). However, a common conception is that improvisation evokes acting on the unexpected and unknown. The objective is that preplanned or prescriptive movement, copying or mimicking are replaced by the possibility for novel physical responses. Improvisation does not imply that the event lacks structure or that people are free to do whatever they want. While material may be unplanned or unexpected, it is not random (26). Instead, improvisation allows movers to make empowering choices within a structured environment of select constraints (28, 29). In this way, improvisation is directly linked with the idea of agency, of a person's ability to act autonomously.

### Improvisation as a Form of Dance

Although many dance forms may appear stylized or systematized, they can include elements of improvisation. For example, Argentinian tango is a form of social (predominantly partnered) "street" dance that allows for improvisation within its codified form. However, improvisational movement alone can be taught as dance, one that offers a distinct form of movement variability, in and of itself. This form of improvisation may have unique benefits for the PD community, benefits that await substantiation through comparative research.

Two key differences distinguish teaching improvisational dance as the primary dance form from encouraging improvisation around another form of dance. These are (1) the use of auditory cueing and music and (2) repetition of choreographed or codified dance movements. In improvisational dance, verbal auditory cueing (VAC) is used to convey improvisational ideas that elicit novel movement from students in the class, even at the earliest stages of learning. Because the VAC does not instruct explicit movements, participants self-select motor strategies in response to the prompt that vary in terms of shape of movement, spatial usage, and timing. Unlike many other forms of dance or exercise, improvisation is not learned by repetition and reinforcement of specific steps.

In addition, in improvisational dance, movement may or may not be rhythmically synchronized with the music. This is an important point because auditory cueing in the form of rhythmic tones has been shown effective in normalizing gait impairment in PPD (30). The presentation of tones can be as simple as rhythmic beeps, and benefit is often limited to when the tones are played. Regardless of delivery (teacher, caregiver, or affected person), cues help initiate, sustain, and terminate movement as tasks demand. In choreographed solo or partnered dance, the strong rhythmic musical beat arguably is essential in cueing timing of the basic steps needed for accuracy and safety in the initial stages of learning. Rhythmic music has been shown to be an effective form of auditory cueing in dance instruction in partnered Argentine tango (7). It is clear that the benefits of choreographed and partnered dance are not limited to moments when music is playing; yet, there may be a specific benefit to cue movement initiation without a rhythmic cue.

### Previous Studies of Improvisational Dance in PPD

To the best of our knowledge, two small clinical studies have been conducted specifically evaluating improvisational dance as an intervention for PPD and both have shown select improvements in balance over relatively short time spans (31, 32). Pilot work by Marchant and colleagues (32) showed statistically significant improvements in the UPDRS motor score, Berg balance scale, and features of gait (increased swing/decreased stance) in a 2-week intensive workshop in a specific form of improvisation known as "contact improvisation." In 2014, our group added to this nascent field with a pilot study of a 7-week improvisational dance class (31). Statistically significant improvements were observed in balance on the Fullerton Advanced Balance Scale, and improvements of a clinically significant magnitude were obtained on the Timed Up and Go. In addition, in a subsequent case study (31), resting-state functional magnetic resonance imaging of one participant was performed and a graph–theory analysis of community structure was performed. Briefly, community structure analysis shows what brain regions are more functionally connected with each other than with the rest of the brain, revealing brain "neighborhoods." As an example, the occipital lobe is often identified as a community or neighborhood in this analysis, as the visual cortex is more connected with itself than with the brain as a whole. For more detail on this method, please see Ref. (33). Before the intervention, it was observed that the basal ganglia were in a community that only consisted of the basal ganglia. That is, the basal ganglia were more connected with themselves than the brain as a whole. After the intervention, the community structure of the basal ganglia had changed, and now included premotor cortex. Now, the basal ganglia and premotor cortex were more functionally connected to each other than shown at baseline. While these studies are small, with neuroimaging data only available for one person, these results suggest that improvisational dance can result in meaningful changes in movement and perhaps also brain networks.

### The Potential Relevance of Improvisation for Automaticity in PPD

Daily life demands a high degree of automatic functioning. This means employing flexible, adaptive strategies to alter conditions, or switch tasks between automatic (unconscious, habitual) and goal-directed (conscious, volitional) movement (13). Effective switching between these modes of movement allows for flexibility and automatically to another task, but rapidly to recruit conscious control if the environmental demands suddenly change. Unanticipated events call upon the person's ability to create motor strategies in the moment in response to balance perturbations and challenges to planning, problem solving, and memory.

Automaticity of gait and balance are both affected in PPD (30). The hallmark of PD is the deterioration of dopaminergic neurons in the basal ganglia, a group of subcortical brain structures that are central to movement, learning, and motivation. Recent research suggests that the basal ganglia are topologically organized into functional divisions (34). The sensorimotor portion of the basal ganglia, the region first affected by PD, is characterized by high connectivity with sensory and motor areas of cortex and is thought to be key in maintaining automaticity of movement (13, 34, 35). The enhanced connectivity between the basal ganglia and premotor cortex seen in response to improvisational movement supports the idea that improvisational dance may be possibly be changing brain networks involved in automaticity in meaningful ways.

Because the loss of automaticity requires movement to be guided by conscious control, cognition is also taxed when automaticity deteriorates. PD has been shown to negatively alter automaticity, altering speed variability and quality of gait (locomotion), particularly in dual- and multi-tasking conditions (36). The ability to restore automaticity could therefore benefit both movement and cognition in PPD. Preliminary evidence is emerging that the use of contemporary dance improvisation may improve cognitive flexibility in normal aging populations (37). Here, measurable improvements have been shown in executive functions such as attentional control and goal directedness in the face of distracting activities and ability to switch tasks on cue (38). Improvisational dance builds on perceptual awareness of the moving body in a moving context (that is, a context in which other people also simultaneously are moving in the same space). This constantly changing context challenges balance, agility, attention, choice and decision-making, and other physical and cognitive skills.

Following this reasoning, one might expect that for someone with good automaticity (automatic movement responsiveness), the prefrontal cortex, which is highly linked with conscious cognitive control, would be in a different network neighborhood than the motor cortex. That is, not to say that those brain networks would not interact, but rather that they would not be preferentially connected. In someone with poor automaticity, who often relied on prefrontal cortex to compensate for losses in automatic movement control, prefrontal cortical activity might be expected to positively correlate with motor cortical activity, at least during certain tasks. Interestingly, neuroimaging literature on improvisation suggests that frontal cortex may be deactivated during improvisation, while motor and premotor regions are more active (39–41), suggesting that improvisation might train the motor and premotor regions to operate independently of prefrontal regions. In other words, improvisation may help to train brain networks involved in automaticity.

### METHODS

Reported here is one dance teacher's (Ms. Soriano) method in teaching dance to mild-to-moderate stage PPD over a 3-year period (2013–2015). Ms. Soriano is a tenured university dance professor also trained in the Mark Morris Dance for PD® program. Ms. Soriano developed her method by engaging in several pilot studies at her home university and teaching a community class series over this time frame. She has taught the method to medical professionals and artists and will continue to do so. She is trademarking the name IMPROVment for this method. In addition to the information provided here, introductory details regarding the method are posted at www.improvment. us. Detailing the method will serve as a model class structure that other dance and movement instructors can consider in designing movement classes for communities of PPD. As well, the model can be used by medical professionals, such as physical and occupational therapists, to evaluate the feasibility of IMPROVment exercises for specific groups of patients or types of care.

The primary aim of the improvisational movement approach described here is to propose a method of physical problem solving that enables students to develop more positive responses to psychophysical challenges faced in everyday living. The improvisational class structure explained here aims to promote the sense of agency needed to navigate independently, efficiently, and safely in daily living by promoting physical responsiveness. The method does not seek primarily to improve select functional outcomes (e.g., balance) or to employ compensatory strategies as a means of coping with dysfunction. Rather, the focus is on movement potential for PPD, the ability to generate new movement strategies. This affords a new vision of physical independence, one who assumes greater relevance as aging accompanies neurodegenerative illness.

### Description of Students/Participants

Participants in the original pilot study (31) were assessed as having mild-to-moderate PD with Hoehn and Yahr scores between 1.5 and 3 and their carepartners. For the ongoing community classes, participants are referred by their neurologist or through the Winston-Salem community support group and represent a range of both physical and cognitive function. In addition, one class has been taught to non-PD older adults who signed up for the class as a continuing education opportunity. Prospective students are invited to watch the class the first time and are asked to sign a waiver before active participation begins. The class includes both PPD and their carepartners, who may be spouses, family members, or paid caregivers. Some class members use assistive devices to walk.

### Description of Class Setting

Classes take place at a private dance studio in a central location of Winston-Salem, NC, USA. There is ample parking within 25 ft of the studio entrance. The studio affords stair-less access, with handicap bathrooms <20 ft from the center of the studio space. There is plenty of ambient light and all environmental obstacles are removed or otherwise controlled. Chairs are available and are arranged in a circle in the middle of the studio at the beginning of class. Two walls contain a ballet barre and mirrors, and a moveable barre is located along a third wall. The class lasts approximately an hour and is free of charge.

### Theory and Philosophy

From a theoretical perspective, the IMPROVment method finds its complement in embodied cognitive science. Embodied cognition states that bodily movement plays a constitutive role in agency and thought (42–44). The ability to act on one's own behalf efficiently and effectively does not come from mental computations alone but from direct experience of body movement within varied exposure to environments and tasks.

IMPROVment classes are grounded in a series of four core principles that shape the overall tone of the class and result in a sense of community and social belonging: non-judgment, non-competitiveness, curiosity and playfulness, and risk taking. Importantly, for implementing this method, the teaching vocabulary (the selection of verbal auditory cues) issues from these core beliefs, many of which are central to the spirit of dance improvisation. These are presented in **Table 1** and described in detail below.

### Non-Judgment

The ability to move without judgment of self or others may challenge group members. Many perceived fears (real and imagined) present barriers to social participation for PPD (45). A space is created where PPD enter into significant camaraderie, look after one another, and form deep friendships. Class participants might regularly see one

#### Table 1 | Principles of improvisational dance and methods for their inclusion in class.


another in various community support group settings. However, dance class goes further in encouraging a sense of inclusion as participants are brought into physical play without judgment.

Generating an atmosphere of non-judgment is an active process. Concern for both individual and the group is consciously established from the first introduction to a participant and carried through the class. The class is advertised as a movement class instead of a "dance" class to encourage participation of those who think they cannot dance. Classes are free of charge to encourage participation across economic sectors. Care partners or spouses are strongly encouraged to join. Students are explicitly instructed that there are no mistakes or "better" ways to move – only new movement options.

Peer witnessing plays a part in generating group enthusiasm, inclusion, and acceptance. The class enables participants to see each other "perform" – that is, letting go of self-consciousness and becoming spontaneously expressive through their whole bodies. Some of the exercises are done with eyes closed to help participants "convene with their own bodies." This cue helps support trust in the rightness of each person's own movement choices. Participants appear to overcome their anxiety to perform well and simply become themselves, each of who is perceived as a unique character in the colorful mix of group expression. Thus, participants are seen and received "differently," as "expressive" or "creative." Dance improvisation engenders a different layer of sociability as persons are witnessed as being "funny," "endearing," or "dramatic."

### Non-Competitiveness

A related core value in IMPROVment is that everyone can dance in some way regardless of age, stage, and condition. The overriding tone of the class is that all movement is an honest effort. The degree to which one moves and the type of movements expressed are of equal value. Movement cues focus on action, on generating movement itself, without value placed on quality of movement. Rather than attempting to move like someone else or move according to a preconceived ideal of rightness or normalcy, the focus is on learning to simply move for oneself, whatever the outcome.

The phenomenological experience of agency, the possibility and potential of "can do," underscores confidence, self-efficacy, self-acceptance, and empowerment. Improvisation is introduced as a palette of options or choices that are not final solutions, but possibilities for more choices. An auditory prompt can result in an infinite number of movement ideas and variations. The prereflective, non-planned awareness of the body moving (improvising) in space is each participant's unique and empowering form of agency. As participants come up with one movement choice, they are encouraged to choose another and another. "Yes, and …" is the cue, as opposed to saying "No, not that …" or "rather … try this."

### Curiosity and Playfulness

Exerting agency requires "active" curiosity. This means that participants become more aware of the various shapes and movements the body makes in space and in relation to others. Curiosity is emphasized over competitiveness. Participants simultaneously invest in what they themselves are doing while staying curious about what might happen next. Attention is drawn to keep pace with VAC within the changing movement conditions. Prompts enable self-generated movement that continues in real time without extra time to reflect, self-edit or otherwise change the original choice made. These movement cues themselves induce curiosity. The prompt to "keep going" (with one's movement choice) is synonymous with the prompt to "stay curious and interested in what you are doing." Playing the game often supersedes selfconsciousness. Participants may not be able to keep up with each and every cue. They can, however, stay intensely focused on and motivated to the present task, rather than becoming inhibited. Every person is invited to share (try on) gestures that created by others in the class with a playful spirit, as opposed to feeling pressed to perform. "Nothing is precious" is a repeated VAC, reminding participants that each spontaneous movement made is part of the ongoing group field of play.

### Risk Taking

Physical risk taking is encouraged within reasonable margins of safety. Increasing one's ability to take risks when moving is integral to the class structure in order to maximize the effects of improvisation. Feeling safe helps participants challenge themselves and promotes autonomy and a sense of personal agency. As participants become more comfortable and confident in responding on cue, they usually react to new instructions more quickly, moving with increased speed and demonstrating greater diversity and flexibility of physical responses.

As described in Section "Description of Class Setting," the class location and environment are designed to maximize safety. In addition, strategies are used in class to help maintain safety while facing various balance challenges. First, the IMPROVment method encourages each member of the class to self-select his or her participation level. As an example of self-selection, persons may not be able to carry out complex locomotor movements and choose to adapt these sitting or at the ballet barre. Participants receive frequent VAC reinforcing their self-selected choices. This allows for support of all class members, regardless of level of ability, physical limitation, or apprehension. Self-selection helps reinforce not only autonomy and personal agency but also safety.

Second, transitions from one phase of the class (e.g., seated) to another (e.g., barre) are monitored carefully for safety. Transitions are guided, as opposed to fully improvised. Care partners or registered health care assistants who take the class with their spouse or family member living with Parkinson's, assist at these times as needed. Ms. Soriano and a trained undergraduate assistant may "shadow" class members to and from their chairs or the barre. Velocity changes are common in these transitions. The ability to control velocity changes can depend on the ease of recruiting smaller joint sub-movements to execute task changes. This has been demonstrated in reaching tasks for this population (46). It is important to watch a participant's ease of movement from one movement from sitting to standing, or from one phase of the class to the next (from chairs to the stationary ballet barre, or from the barre into the center of the room).

Third, VACs train group attention to the constraints of the task and the environment, rather than evoking fear reactions. Rather than warning participants to "be careful" so they would not fall, the statement "Be aware of your surroundings and others moving simultaneously in the room" helps keeps everyone consciously aware of the task context.

Fourth, increasing challenge is introduced by gradually increasing the variety and complexity of movement tasks. This is important in learning to be safe while increasing the embodied sense of confidence.

### Training Strategies

Every moment presents multiple challenges in terms of balance, awareness of space and environmental constraints, awareness of self and others within the space, requiring a response that integrates the brain and body, and automatic and intentional movement. Ms. Soriano uses the following main training strategies to maintain a level of challenge for cognition and physical activity: active imagination, variability, and pacing.

### Active Imagination

Working with imagery is crucial in an improvisatory practice. VACs are used to create movement scenarios that cue or activate the motor imagination. VAC takes primacy over rhythmic entrainment to music, although the music itself may be used as an improvisatory cue. The teacher calls the cue, demonstrating an optional response, and asks participants join in immediately with their own gestural inventions. As an example, students might be prompted during the seated warm-up to recreate a beach scene. VACs direct motor imagination by using rich language to act out a beach scenario – laying down their beach blankets, putting on sunscreen, opening their picnic baskets and setting out lunch on a blanket, running into the ocean, avoiding the shark swimming toward them, and so forth. Often, participants will do this seated, with eyes closed. This activates the imagination more strongly and adds a balance challenge.

### Variability

The IMPROVment method does not aim to learn a specific movement pattern and habituate to it. Rather, the aim is to intend to stimulate new pathways for motor learning by meeting unexpected environmental conditions arising in the moment and devising new physical solutions as a result. Daily life is fraught with the unexpected, with variable environmental encounters that call for ongoing problem solving. The variability inherent in improvisational movement training helps minimize the tendency for people to default to their habitual (familiar) ways of moving and consider new movement strategies. Preliminary research suggests that that variability acts as a novel stimulus to the motor cortex, facilitating new motor pathways (38).

Cues often are delivered quickly, one after another, to increase excitability of the motor cortical as well as enable participants to go beyond their habitual ("self-perceived") capability (13). Within an average of 2 min, tasks requiring quicker decision-making are implemented. Physical challenges are advanced by dual- and multi-tasking, such as being asked to direct traffic with the right side of the body while picking apples with the left.

Maintaining variability throughout the class can be a challenge. Variability is accomplished by adding complexity, presenting new prompts, or assigning a qualitative change to a specific exercise. For example, students might be given a simple prompt to make any shape with their upper body. Variability can be added by increasing complexity, e.g., a VAC to create a series of new shapes, each one different from the previous, and then to remember and reproduce the first shape. Variability can be added by presenting a new VAC, e.g., "Now make a second shape with your lower body." Variability can also be added by assigning a qualitative change, e.g., a VAC to swim with the upper body and lunge with the lower body. Variable movement themes (e.g., changes in body shape and movement direction) are intended to stimulate movement inventiveness and to avoid defaulting to habitual responses.

Improvisational prompts challenge the scope and speed of movement by introducing multiple body parts engaging in multiple tasks. For PPD, maintaining variability of movement becomes a particular challenge when cueing dual- and multitasking activities (36).

### Pacing

Pacing refers to the rate at which new movement prompts are presented. As with variability, quick changes in pace also avoid defaulting to habitual responses, thereby facilitating new movement options. Participants cannot rely on copying another, or necessarily their own memory or anticipation for the answer to the motor problem, Verbal cues might be delivered in a rapidfire manner, for example, in order to keep participants from having the time to think or reflect on movement choices. There is little time to change one's mind, become embarrassed, or be dissatisfied with the choice made. Participants simply need to move through the chosen sequence, even if it might not seem logical or conform to their ideal conception of the movement. Even if the outcome does not match their intention, participants often surprise themselves in seeing that they are capable of new movement choices.

### Music Selection

Participants are invited to bring music in for class. Music is used in different ways throughout the class. The playlist is random and variable, not the source of rhythmic entrainment. Sometimes, it is unrelated to movement instruction and is relegated to being ambient, such as something that might be heard at a cocktail party. Other times the music may serve as an improvisational cue. As examples, students may be asked to dance what comes to their minds listening to Otis Redding's "(Sittin' on) The Dock of the Bay" and then variations are cued from that initial movement, or "play" an instrument they can hear in a complex piece of classical or jazz music. Sometimes, participants themselves create movement through vocalizing or body-based percussive actions. Portions of the class also happen in silence. Video analysis of previous studies shows that participants are more likely to improvise rather than follow (entrain to) the beat of any music played.

### Class Structure

The general class structure supports the core philosophies of IMPROVment and has four phases. These include (1) group warm-up in chairs positioned in a circle, (2) standing barre with solo- and partnering exercises, (3) moving as a group through free space (with and without a partner), and (4) recuperation and rest.

The exercises in the IMPROVment method are designed to help PPD engage with the challenges of motor control and coordination. The exercises present a movement theme for each class and increase in complexity as the class progresses. For example, an exercise practiced in chairs might also reappear at the barre, challenging participants to interpret in standing a prompt previously given in sitting. An example of an exercise that might thread throughout class is what Ms. Soriano refers to as "Out Out, In In" (OOII). OOII asks for inventive and varied examples of distal extension and contraction of limbs (including the head) either toward the center of the body (proximally) or away from the body center (into space). This general objective invites a diversity of expression, speed, and options and is incrementally more challenging from seated to ambulating. Examples of OOII in each of these setting are detailed within the four phases in **Table 2**. Regardless of physical ability, participants are able to participate in OOII in multiple expressive and meaningful ways, even if never moving from the chair.

Similar exercises may appear within each phase creating a progression of motor skills, but the class series does not build incrementally. Throughout the class, there are recuperative phases that are essential not only to rest and recover metabolically but also cognitively. These movements are slower, simpler, and often more familiar. For example, simple familiar exercises are introduced such as a seated hamstring stretch. Here is the opportunity to let go of attending to task and enjoy more automatic movement patterns. Material varies between phases, helping to consolidate memory (40) without using repetition as a means of reinforcement. Concepts and tasks are heterarchical and interdigitated. VACs used in class are open-ended. That is, they do not provide a movement solution, but rather a cue that has multiple movement solutions. Below, each phase is described in more detail and an example exercise is presented.

### Group Warm-up

The initial phase of the class takes place with participants seated in chairs where their sitting balance is tested by a variety of ways of shifting weight. Prompts demand while moving expansively to the limits of their self-selected range while reinforcing curiosity, imagination, and playfulness. Even while sitting, rich imagery is used, pacing remains quick, and variability is rapid. Seated exercises are still demanding.

A commonly performed seated exercise in this phase is called "Pass the Energy." "Pass the energy" is designed to invite every participant-to-participate and be empowered by his/her movement choices. Just as it sounds, "Pass the Energy" asks participants to create a gesture with their whole body and "pass"

Table 2 | Demonstrating the tier system through one exercise: Out Out In In (OOII).


that gesture to the next seated person. Each person's "energy" has unique qualities and calls for eye contact as the recipient takes the "energy" and transforms it into his or her own version. After participants make their first movement choices, increases in pacing and variability add complexity to the exercise constraints (as described in Sections "Variability" and "Pacing").

### Standing Barre

The class usually progresses from seated movement to moving at a ballet barre. The barre is not treated as a "prop" to rely on for securing balance. OOII practiced here, for instance, encourages participants to imagine the ballet barre as a dance partner. With a VAC to reach away from one's "partner" and then come back to it, participants find novel ways to step out or away from the barre and then come back toward it. Balance is challenged and tested in safe ways while reaching high and low in the space. The VAC "Out Out, In In" encourages the possibility of someone reaching away (or outwardly) from the barre with eventually no support and then returning toward the barre (inwardly) with one or two hands. This not only can be repeated multiple times but eventually, participants also become curious to try turning around themselves as they long for new and inventive ways to practice OOII.

### Moving through Free Space

Walking variations follow, first as freely moving on one's own in the space and then interacting with others in the class in partnered and non-partnered interactions.

For example, Ms. Soriano sets up various obstacle courses in the studio by using chairs, requiring gradual and incremental changes in speed and direction at first. The rationale behind this decision is not to create a hazard, but to reflect the improvisational nature of everyday life where obstacles will inevitably happen. Similarly, like the ballet barre, the chair can become another dancing partner. Chairs can become sculpture pieces that participants lift and move around the room with the VAC to consider the design of the chair in the space, with relation to one's body, other bodies, and other chairs in the room.

Another example of a VAC for moving around chairs in the space is to walk with a certain "effort percentage." This encourages participants to consider that they are in control of when they start and stop, as well as how fast or slow they move. Whatever the configuration of chairs in the space, participants might be given the VAC, "if the pace with which you are walking around the room now is 40%, transition to what 60% looks like (e.g., 20 or 75%)." Ms. Soriano determines the range of percentage changes in walking speed, based on the average pace the class presents in the moment. Finally, another walking exercise could be the simple VAC to walk and pause in the room over and over again, and each time during a pause period, create a different shape with your upper or lower body. In keeping with the improvisational emphasis on encouraging agency, each participant determines for him- or herself how long one pauses, or how long to walk before pausing.

### Recuperation

Class closure always involves the ritual of circling up, sharing a common gesture of celebration or applauding each other. Often, this circle opens out to become a "free dance." Here, participants are encouraged to dance in free form about the room with or without music playing, making eye contact, and gesturing to each person in the class to bring full closure.

### DISCUSSION

This article describes the philosophy, methods, and class structure of one dance teacher's method in using VAC of improvisational dance movement to enhance agency in PPD. The method is designed first as a tool to expand the range of self-perceived movement potential in persons with early to middle stage PD, and second, to build a caring and supportive environment in which to develop a fuller range of physical options and strategies for movement to meet the challenges of everyday navigation and communication.

Improvisation is not a prescribed set of exercises. Exercises are provided here as examples and starting points with the hope that teachers who wish to use this method will improvise their own exercises and in the future, share them *via* the website (www. improvment.us). One challenge in both teaching and researching improvisation is the fact that exact repetition is explicitly avoided. What is repeated and can be taught, what unifies the method presented here, are the philosophies of non-judgment, non-competitiveness, curiosity, and risk taking; and the teaching strategies of active imagination, rapid pacing, and variability.

### Improvisation and Emotional Well-being

In addition to the physical benefits for PPD documented in the limited data collected thus far, improvisational dance also may benefit emotional well-being. Subjective reporting by participants in IMPROVment classes, indicate improvement in social and emotional well-being. A survey was developed for IMPROVment classes which queried the participants' ability to act for and by themselves. This survey asked for anonymous feedback at a randomly selected point in the ongoing community class. Questions were rated on a 5-point Likert scale, the scoring of which ranged from 1 "Not at All" to 5 "Very Much." Previously unpublished responses from the eleven respondents are reported here. The surveys were collected with ethical oversight and approval from the Wake Forest School of Medicine IRB in accordance with the Declaration of Helsinki. Overall responses were positive. Seven of the 11 respondents stated increased life satisfaction "Very Much." Nine respondents rated the class as 4 or 5 for increased empathy for fellow classmates, creativity and physical benefits, and movement skills. Six respondents noted improvements in balance (4 or 5 rating). Open-ended comments were positive and many focused on social or emotional well-being. Some examples are "increased mobility, more confidence, more energy, helps all the way around"; increased "camaraderie and community" and "social interaction"; "laughing, fun, getting to connect with new friends, body warming up."

To date, direct comparisons with other dance methods have not been made in this particular sample, however. A 2009 study comparing Argentinian tango with Waltz/Foxtrot and Tai Chi (47) sets a precedent for this, wherein improvements in quality of life (including emotional well-being) allude to the specificity of well-being afforded in dance training. The researchers hypothesized that significant findings were due to addressing balance and gait in the context of closely connected improvisational partnering, characteristic of tango.

### Future Directions

Clinical guidelines of current protocols for teaching any style of dance to PPD are still evolving, and therefore, this article serves to contribute to a growing understanding of safe and effective community-based, group-delivered dance programs for this population.

There is a clear need for additional research using larger sample sizes to examine the potential long-term effects of dance for those with PD (6). Most studies of dance for PD, to date, have used rather small sample sizes and have only examined the short-term effects of dance programs, as have most studies of creativity and improvisation. Future work should include larger samples and assessment of the long-term effectiveness of dance for PD, and optimal dosing of dance interventions with respect to frequency, duration, and intensity. Of particular relevance here would be to assess the effectiveness of VAC both in structured and improvisational contexts, as well as to compare behavioral results with brain mapping data.

While admittedly variable in its breadth and depth, current evidence suggests that aerobic exercise, mind–body exercise such as Tai Chi, choreographed dance, and social dance can all benefit people with PPD. This is good, because the more valid exercise alternatives that exist, the more likely people are to find a movement class that in enjoyable and accessible to them, increasing the likelihood that they will maintain a level of physical activity that benefits their health. One purpose of this article is to advance the idea that improvisational movement should be further investigated as another method of movement instruction that may contribute unique advantages to people with PPD.

### REFERENCES


### AUTHOR CONTRIBUTIONS

All authors made equal contributions intellectually and practically to the conception, preparation, and execution of this manuscript with mutual mentoring and support.

### ACKNOWLEDGMENTS

The authors wish to thank our collaborators, Sukhada Gokhale, independent filmmaker; Nickolay Hristov, Ph.D., The Center for Design Innovation; Paul Laurienti, M.D., Ph.D., Jonathan Burdette, M.D., and Ashley Morgan, Wake Forest School of Medicine Laboratory for Complex Brain Networks; David Marchant, Washington University; and Eduardo Urena, University of North Carolina School of the Arts. We would like to thank our outstanding student collaborators, Rachel Bradley, Rachel Caskey, Quinn Cutchin, Sam Dunaway, Desaly Gonzalez, and Bryan Neth. And most importantly, thanks to the amazing members of the PD Forward movement group in Winston-Salem, NC, USA.

### FUNDING

The time and resources to develop the IMPROVment method outlined in this manuscript was supported by the Translational Science Center of Wake Forest University, The National Parkinson Foundation, Wake Forest University's Interdisciplinary Performance and the Liberal Arts Center, and Wake Forest University's Center for Undergraduate Research and Creative Activities and the Wake Forest School of Medicine Laboratory for Complex Brain Networks (LCBN). Support for this work was provided from the Wake Forest University Translational Science Center and Wake Forest University Wellness Initiative in partnership with Blue Cross Blue Shield North Carolina.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Batson, Hugenschmidt and Soriano. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Dancing with Disease: A Dancer's Reflections on Moving with People with Parkinson's and Memory Loss

*David William Marchant\**

*Performing Arts Department, Washington University in St. Louis, St. Louis, MO, USA*

Keywords: dance, dance therapy, therapy, Parkinson's disease, Alzheimer's disease, falling, improvisation, exercise

### INTRODUCTION: FIELD NOTES FROM AN EXPERIENTIAL MOVEMENT RESEARCHER

I am a professional dancer, choreographer, and Alexander Technique instructor. I joined the Performing Arts faculty at Washington University in St. Louis in 1994, where I teach contemporary concert art dance and somatic practices. Dancing is fundamentally a study of coordination, balance, and movement control. Because these essential goals are shared in movement therapies, I have become interested in contributing knowledge gained by dancing and collaboration with scientists working on therapeutic interventions. I am particularly interested in evidence that, for people attempting to mitigate symptoms of Parkinson's disease (PD), dancing in a variety of forms is emerging in scientific literature as an effective approach (1, 2). Furthermore, dancing is valuable to people who do it, not only for cognitive and motor issues but for social and personal fulfillment, while living with PD (3). I am also interested in correlative evidence suggesting that dancing, as a lifestyle activity, may confer a protective effect against risk of dementia (4). In this article, I would like to share several personal insights regarding dancing and why I believe it is so effective.

In 2008, I met Madeleine Hackney and Gammon Earhart who were engaged in a series of studies demonstrating that Argentine Tango improved balance and functional mobility in people with PD (5–7). I was invited to create and administer an untested intervention for PD using "contact improvisation" (8). We showed similar improvements as the Tango pilot study. However, participants reported preferring improvising (not having to learn step patterns), and the increased human touch that the practice fostered became valuable to them. This report in the participant survey is among the most meaningful to me; I firmly believe that live human touch is an essential ingredient in physical, mental, and emotional health.

During this past year, my students and I conducted a dozen non-scientific workshops for people with Alzheimer's disease (AD), applying creative dance practices I learned from dance artist Liz Lerman (9).

Memories are not just "in our heads," they are whole-person, embodied experiences – what I call "corps memories." I am intrigued by evidence that integrating a motoric component such as pantomimic gesture or sign language during learning improves memory retention (10, 11). In workshops, people shared memories, and we created dance from the spontaneous gestural movements that typically accompany speech when they become immersed in reverie. This organic combination of narrative and kinesthetic sensation seems to enhance the details of memory. Artistically, there is something poignant, "dancing" memories that will be lost.

### LOOKING BACKWARDS, WITH IDEAS FOR MOVING FORWARDS

When I began these workshops, I never knew what to expect. Perhaps, it was a virtue, I had so few assumptions about what people with PD and AD "can't" do. We simply started where we were

#### *Edited by:*

*Marta Bien´kiewicz, Aix Marseille Univ, CNRS, ISM, France*

#### *Reviewed by:*

*Christina E. Hugenschmidt, Wake Forest School of Medicine, USA*

## *\*Correspondence:*

*David William Marchant marchant@wustl.edu*

#### *Specialty section:*

*This article was submitted to Movement Disorders, a section of the journal Frontiers in Neurology*

*Received: 27 June 2016 Accepted: 09 August 2016 Published: 25 August 2016*

#### *Citation:*

*Marchant DW (2016) Dancing with Disease: A Dancer's Reflections on Moving with People with Parkinson's and Memory Loss. Front. Neurol. 7:137. doi: 10.3389/fneur.2016.00137*

and pushed the edge of our ability gradually, improvising and adapting along the way. Participants did movements that I would not have thought older adults would do, much less with disease. In the process, I think we all surprised ourselves, and that felt empowering to me and I believe to them as well.

Dancing with elders has changed the way I understand movement. Studying movement by feeling it from the inside has generated numerous ideas and questions. From this point of view, I will discuss several themes for scientists to consider with possibilities for further investigation: (1) how subjective, experiential research can contribute knowledge about movement complementary to objective science; (2) how improvised movement activity fosters adaptive capacity useful for recovering falls during unexpected loss of balance; (3) that dancing is more complex than "exercise," engaging cognitive, neuromuscular, esthetic, and social aspects of a person, integrally. Finally, I propose more collaboration between disciplines with a broad lens toward the goals of therapy, including how dancing offers meaningful ways to live with disease.

### OBJECTIVE AND SUBJECTIVE EVIDENCE ARE COMPLEMENTARY

"More evidence is needed;" I like this understatement deployed by scientists to estimate our insufficient understanding (and to nudge for continued funding). I suggest that more *kinds* of evidence are needed to fully explore human functioning. Objective measurement does not always capture the meaning of the lived experience. Subjective, experiential study can add different information and help point our investigations in new directions. Dancing is highly complex. It will be difficult to ever fully analyze why it is so widely beneficial, but the evidence that we feel good doing it is well documented.

Subjective experience is not as highly trusted as objective evidence. I understand why we feel less confident of data "corrupted" by non-quantifiable things like feelings. I argue subjectivity is not inherently unreliable; we only need more advanced education of sensation to train skills for reliable subjective self-assessment. Somatic and artistic practices offer such training and experiential methods of critical thinking and research in human movement (12).

Furthermore, subjectivity is where we live, and the way we feel is often what counts most to a person. For people with PD and AD, in my workshops, evidence that dance is beneficial is encouraging, but I find they mainly do it because they feel better and enjoy moving together. That is evidence enough it is worth doing.

### IMPROVISATION IS AN "ACTIVE INGREDIENT"

Studying movement from inside the sensory experience, I find that improvising develops superior motor control. In two studies of dance interventions for PD, I see a hint that improvising may be an "active ingredient" in their effectiveness (8, 13). Like real life, improvised movement generates dynamic, unpredictable interactions with the environment and/or people; this fosters capacity for adaptive response critical to successful living. Even rote activities feel "new," requiring adaptive response during the learning phase. But, in time, this virtue may fade, as the activity requires less attention to repeat. I am curious to see further investigation of "interactivities" that remain variable and challenge people to cope with the unpredictability of living, rather than avoid such risks with routines.

I also wish to correlate improvisation with Ellen Langer's research reporting wide ranging cognitive, emotional, and physical benefits of being "mindful" in novel ways (14–16). Her instructions to experimental, "mindful" groups often take the form of "thought experiments," teaching them to attend and improvise tasks in new ways (17). Improvising induces, if not requires, the quality of presence that Langer is recommending, suggesting that improvised interventions may be more beneficial than routine ones.

### FALLING IS NOT AN ERROR, IT IS A SKILL

One test of adaptive capacity appears when a person falls. Falling un-injuriously requires improvised adaptation to correct unexpected loss of balance. When I asked study participants, in 2010, about the risk of falling, they reported that their physicians said, "don't fall." Evidence shows that fear of falling may actually increase fall occurrence (18). Based on personal experience, trying not to fall is a terrible strategy for balancing, causing high muscle tension and overcorrection errors that ironically lead to falling. I find the secret to balance is becoming at ease with falling, fluidly oscillating between falling and recovering (a principle I learned in techniques by mid-century modern dance artist Doris Humphrey).

In Marchant et al. (8), we reported, "All together, strategies were intended to teach participants that stability in balance is a skill of continuous adaptive movement, rather than of fixity, holding or prevention of movement. Rather than suggesting participants try to 'avoid falling,' this workshop taught them how to fall more safely" (8). This is also a fundamental principle and technique in contact improvisation dancing.

Since publishing that statement, I have come to the opinion that falling is not an "error" we should avoid – it is *the* skill we should develop and maintain. Like Humphrey, my definition of "falling" includes weight transfers involved in walking; in this model, safely "losing" balance becomes equally fundamental to movement. I find that dancers' balance (suspension) becomes calmer and more natural when they are taught to allow their weight to "fall outward, in all directions equally." Corrections are achieved by falling in the opposite direction of an imbalance. When locomotion is desired, releasing counterbalance allows the body to move in the intended direction. In this approach, both suspension and motion are achieved with the same strategy, easing muscle tension instead of tightening, producing smoother transfers with fewer overcorrection errors.

Participants also said that it was not a matter of *if* they fall, but *when*. If the default prescription is "don't fall," there may be insufficient consideration or training of *how* a person should respond when they do fall. In our intervention, people were challenged to shift weight off center intentionally. The more familiar they became with instability, the more confidently they improvised fluid recoveries to perturbations.

I suggest we look closely at how we define and measure balance with care to not unintentionally send a message that good balance is something held still or that our balance is worse if we are moving to adjust.

### WHY DANCE IS SO BENEFICIAL: MORE THAN MERE "EXERCISE"

When we think of exercise as primarily muscle strengthening, cardiovascular conditioning, or even neuromuscular activity, we potentially miss something whole about the experience of moving. I suspect the complexity of dynamically interactive dependent variables that dancing introduces is not only why it is difficult to tease out causal relationships but also why it is so effective. It is tricky trying to isolate variables for "scienceability" without denaturing activities to the point that they lose their inherent value.

Dancing works on the whole person. Complex esthetic movement fosters relationships between brain areas, stimulating "mindbody" to work in novel, "more-than-necessary," neurogenerative ways. I also believe that esthetics – intangibles like "beauty" and "how it feels" – are not merely superficial embellishments. Appealing sensory experience comingled with conscious attention in movement integrates brain functioning. Exercise can become rote, reductive, and repetitive, performed in ways that diminish esthetic appeal, and, with it, the need/desire for cognitive engagement. I see people walking on treadmills listening to their iPod, reading a book on the easel, with a TV on in the background. If we need three distractions from how bored we feel, benefits of such exercise may be analogous to taking a multivitamin, but no substitute for healthy eating. I argue that cognitive presence is an essential nutrient of "whole" movement.

### MORE COLLABORATION BETWEEN DISCIPLINES IS NEEDED

What I know about human movement I know by moving with people in my classroom "laboratory." Collaborating with dancers,

### REFERENCES


I have learned ways to improve coordination, balance, and motor control. Collaborating with scientists, I learned that my knowledge about human movement might have valuable applications. And with scientific methods, I learned I could rationally seek more evidence to support my subjective hunches.

I predict that dancing will continue to emerge as a preferred therapeutic intervention and preventative activity for people with a wide variety of illness conditions. In support of this trend, I recommend scientists seek more collaboration with dancers; they may not yet fully realize that what they know is of value to you. Dancers interested in movement therapy may consider studying movement science. And I encourage scientists to dance. Dancing is not just something one watches, it is an experience. The sensation of moving cannot be replaced with objective observation, and if scientists are having these experiences, I am confident that they will generate new ideas for investigation.

### PEOPLE WITH DISEASE ARE NOT JUST DEGENERATING, THEY ARE ALIVE

Moving with elders has changed my definition of "dance" and what is "beautiful." The people I have worked with are some of the most alive people I have ever met – fully engaged and present. Brené Brown writes, "It didn't take long for me to learn that … for many of us, there is no form of self-expression that makes us feel more vulnerable than dancing. Its literally full-bodied vulnerability" (19). But in my experience, whatever vulnerability people with PD feel about dancing pales compared with the decline they are already facing. They are not as nearly as self-conscious or inhibited as many of my college students. Witnessing a person moving at the edge of their ability, whatever that may be, is what is most beautiful to me.

### AUTHOR CONTRIBUTIONS

DM is the sole author of this opinion article and has contributed all of its content.


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Marchant. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*