The term motor imagery has been used in literature to indicate the visual (i.e., imagine to “see”) and kinaesthetic (i.e., imagine to “feel”) imagination of a movement without execution (Jeannerod, 1994, 2001; Decety, 1996; Mulder, 2007). Since the first seminal studies in the 1930s (Sackett, 1934, 1935), a large body of evidence shows that motor imagery and actual action execution share similar cognitive and neural processes. For example, the time taken to mentally imagine versus execute a movement has been shown to be similar (Decety and Michel, 1989; Sirigu et al., 1995) suggesting that imagery is based on overlapping action execution cognitive processes. Similarly, execution and motor imagery both adhere to Fitts’s law, whereby the time to execute or imagine a movement is moderated by accuracy demands (Decety and Jeannerod, 1995; Macuga et al., 2012; Macuga and Frey, 2012). According to the “functional equivalence hypothesis” (following motor simulation theory, cf. Jeannerod, 1994, 2001), similarities between actual execution and motor imagery emerge from shared motor-cognitive (neural) processes, allowing for the imagined rehearsal of movement using cognitive motor planning processes (Jeannerod, 2001). The functional equivalence hypothesis has received strong support, primarily from brain imaging studies which have demonstrated that motor imagery and actual motor execution involve similar activation of brain areas (Porro et al., 1996; Grezes and Decety, 2001; Sharma and Baron, 2013). These brain activations include a distributed premotor-parietal network, involving several subcortical structures, such as the putamen and cerebellum (Grezes and Decety, 2001; Hardwick et al., 2018). Further indirect evidence for equivalence comes from sport (Guillot and Collet, 2008; Williams et al., 2012) and clinical literature, which have shown that the mental practice of movement through motor imagery (i.e., the repetitive exposure to motor imagery of movements) can be effective for learning motor skills (Dijkerman et al., 2010; Barclay et al., 2020). The potential efficacy of mental practice in sport training and in motor rehabilitation is supported by the finding that motor imagery practice can induce plastic changes in motor networks (for a review see Ruffino et al., 2017).

Despite this common notion of the evidence for functional equivalence, technical constraints make measures of motor imagery extremely challenging for researchers, even in experimentally controlled settings. Motor imagery is a covert process. This means that, without comparison to actual execution, it is not always clear what is being measured by recording brain activations during motor imagery. Furthermore, neural measurement during movement execution is also limited, in particular for whole-body movements such as walking (Ladouce et al., 2017). Understanding the motor imagery of walking requires capturing the neural processes during whole-body movements, which cannot be done while lying in a brain scanner (nor while seated in a typical neurophysiology lab). Instead, researchers investigating functional equivalence between motor imagery and execution have constrained actions to simplistic hand movements that are feasible in an fMRI scanner. A further constraint is that imagined movements in experimental fMRI studies are performed lying down, whereas the imagined movements are more normally performed upright, or sitting (Jahn et al., 2004, 2008; Bakker et al., 2007; La Fougere et al., 2010; Malouin et al., 2013; Hamacher et al., 2015). It is possible that incongruence between body posture and imagined action causes an issue for ecological validity, also because the compatibility of the body position has been shown to affect motor imagery performance. For example, the time taken to perform motor imagery is most similar to actual execution when participants are positioned in a congruent compared to incongruent posture for the imagined relative to executed movement (Parsons, 1994; de Lange et al., 2006; Conson et al., 2011; Saimpont et al., 2012). These latter findings are important, as investigating the neural substrate and cognitive processes underlying motor imagery and action execution must be investigated using congruent and equivalent complex motor actions (Barton and Pretty, 2010; Menicucci et al., 2020).

Motor imagery is an established method of upper and lower-limb neuro-rehabilitation used with patients after stroke (Barclay et al., 2020). Considering that deficits of lower limb function are a common disability after stroke, it is surprising that there is just one single study to date that compares the neural activation of the execution and motor imagery of walking (La Fougere et al., 2010). They captured global neural activation during walking using PET imaging, and compared neural activation to motor imagery of walking using fMRI while lying flat. La Fougere et al. (2010) showed PET neural activation during walking execution involved primary motor cortex and somatosensory activation, while the fMRI neural activation measured during motor imagery of walking involved the activation of the supplementary motor area and basal ganglia. There are, however, some fundamental limitations to that study, in particular with regards to the design. Firstly, the measures of PET and fMRI are quite different, and it is not clear why the neural correlates of execution and motor imagery in that study were not measured using the same imaging methods allowing direct comparison. Secondly, activation of execution was captured through isotopic decay measured in the PET scanner after people had been walking. While ingenious as a solution and to be admired in its ambition as an early attempt of neuroimaging of real-world whole-body movements, the neural correlates are difficult to compare to motor imagery assessed through fMRI while the participant is performing the imagery. Thirdly, as stated above, differences are expected in brain imaging investigations of motor imagery where participants lay supine imagining the upright movement of walking. In conclusion, while groundbreaking, in many ways, the study of La Fougere et al. (2010) provided limited insight into the neural correlates of whole-body motor imagery functional equivalence.

Recent development of mobile electroencephalogram (EEG) technology solves many of the constraints discussed above, including the ability to record brain activity during naturalistic movement. EEG recordings allow a characterization of highly accurate temporal brain rhythm dynamics that reflect cognitive processing, overcoming many of the limitations of functional MRI scanning. Furthermore, the high temporal accuracy of EEG offers the possibility to disentangle cognitive processes occurring during motor imagery and action execution at different time points. It is crucial to investigate the neural correlates of these processes considering the growing applications of motor imagery in clinical settings. Indeed, motor imagery of movements represents one of the main cognitive tools being used within the brain-computer interface (BCI) approach, as it can be applied in absence of physical involvement and can be employed as a self-regulatory control signal of motor brain areas. The principle of the BCI approach is to detect neural signals to control external devices, such as exoskeletons or wheelchairs (Leeb et al., 2007; Choi and Cichocki, 2008; Lafleur et al., 2013) or prostheses (Pfurtscheller et al., 2003). It is therefore important to understand which are the neural correlates of walking imagery as they can be used as self-regulatory signals for motor learning and recovery in patients with gait impairments (Malouin and Richards, 2010; Kranczioch et al., 2014; Daeglau et al., 2020).

EEG investigations of motor imagery in isolated movements such as those of the foot or the hand have found decreased and increased spectral power, often termed event related desynchronization and synchronization, respectively. These changes in power are most often found in the alpha (8–12 Hz) and beta (13–35 Hz) frequency bands, occurring over sensorimotor (Pfurtscheller et al., 1997; Pfurtscheller and Neuper, 1997; Neuper and Pfurtscheller, 2001) and parietal-occipital (Salenius et al., 1995; Xie et al., 2020; Putzolu et al., 2022) brain areas. Generally, it is accepted that power decreases in the alpha and beta frequency bands are related to the activation of relevant brain areas during a given task, whereas power increases are associated with inhibition (Pfurtscheller, 1999; Pfurtscheller and Da Silva, 1999; Klimesch, 2012). Indeed, both alpha and beta decrease of power over sensorimotor areas are typically visible during movement preparation (Tzagarakis et al., 2010; Rhodes et al., 2018), the execution of movements (Pfurtscheller and Berghold, 1989; Stančák et al., 1997; Cassim et al., 2000; Kaiser et al., 2003; Kilavik et al., 2013) and have been consistently reported in previous EEG investigations of walking (Gwin et al., 2011; Bulea et al., 2015; Bradford et al., 2016). On the other hand, power increases in the alpha band have been associated with the inhibition of task-irrelevant information (Neuper et al., 2006; Klimesch, 2012), whereas beta power increases are associated with the recalibration of the motor system (Engel and Fries, 2010). Although previous EEG investigations have provided evidence of the engagement of sensorimotor brain areas during action execution and motor imagery, research has typically focused on upper limb movements, while participants were sitting, performing small hand or finger movements (e.g., finger-tapping). Studies of lower limb imagery were mostly limited to imagery of minimal movement, such as the dorsiflexion of the foot (Neuper et al., 2010; Pfurtscheller et al., 2006; Solis-Escalante et al., 2008, 2012; Müller-Putz et al., 2010; Hashimoto and Ushiba, 2013), stepping in place (Kline et al., 2021), or they did not contrast cortical activation with the actual execution of movement (Putzolu et al., 2022). Furthermore, existing EEG investigations on motor imagery have tended to focus on identifying reliable signals for brain-computer interface (BCI) control as mentioned above, allowing for connection between the oscillatory activity of the brain to control a computer (Wolpaw et al., 2002). Consequently, EEG recordings in these studies have typically involved a limited number of channels (leaving cognitive questions less explored). No existing studies report a systematic assessment of similarities and differences in brain activity and cognitive mechanisms underlying the actual execution and motor imagery of walking.

Therefore, the aim of the present study was to investigate the similarities and differences of neural activations associated with action execution and motor imagery of walking movements using a mobile EEG approach. In a within-subject study design, we asked participants to walk, or to imagine walking (in a movement compatible standing position), at a natural pace, along a path. To enhance the sensorimotor experience, we asked participants to perform the imagery straight after walking, using a first-person perspective and focussing on the kinaesthetic experience (feeling the sensations as if actually walking), which has been shown to elicit greater motor activation compared to visual motor imagery from a third person perspective (Lorey et al., 2009; Mizuguchi et al., 2017). Furthermore, we used a mental counting task as a non-motor control condition to ensure that during imagery, participants were not just counting steps, but instead actually imagined themselves walking. We assessed similarities and differences between motor imagery and actual execution by comparing the temporal and spatial features of EEG patterns. Specifically, we examined the time course and the spatial distribution of power spectral changes in alpha and beta frequency bands, according to previous literature on motor imagery (Pfurtscheller et al., 1997; Pfurtscheller and Neuper, 1997; Neuper and Pfurtscheller, 2001).

The present study aimed to examine overlap and differences between the oscillatory brain activity related to motor imagery and actual execution of walking. The results showed a general match between the modulation of beta power regardless of whether participants were performing the actual execution of walking or the motor imagery of walking, whereas an equivalent correspondence did not emerge when participants were performing the control task of mental counting. Interestingly, the analysis also revealed clear evidence of differences between action execution and motor imagery. Specifically, a distinct pattern of beta activity was present during motor imagery, which was associated with a larger amplitude of beta power over central, parietal and occipital brain regions during motor imagery compared to actual walking. We also found that oscillatory activity in the alpha frequency band was characterized by a stronger power decrease at the beginning of the motor imagery task compared to during the actual execution and mental counting conditions. Therefore, the data suggests motor imagery and execution of naturalistic walking involve shared motor-cognitive activations, as shown by similar modulations in the alpha and beta frequency bands during motor imagery and action execution of walking, but that motor imagery requires additional cortical resources, indicated by the distinct pattern of more distributed brain activity involved during motor imagery compared to the other experimental conditions.

Consistent with the functional equivalence hypothesis (following motor simulation theory, cf. Jeannerod, 1994, 2001), our results demonstrated that motor imagery and actual execution of walking induced similar beta power decrease–increase dynamics, contrasting with the results observed during mental counting, suggesting that participants did not just count the steps mentally in the motor imagery condition. The decrease of beta oscillations during motor imagery signals the recruitment of neural circuits supporting the activation of motor-related information underlying the action representation of walking, which resembles typical cognitive mechanisms occurring during the preparation of overt motor responses (see Jeannerod, 2001; Glover and Baran, 2017). The role of decreased beta over the sensorimotor network has been defined as a mechanism that recruits neural processes to generate a motor output (Rhodes et al., 2018).

As such, we see neural overlap between the motor imagery of walking and actual walking in this motor-related beta decrease over the sensorimotor network. Beta power decrease has been previously reported during the kinaesthetic imagery of movement of body parts (i.e., hand, foot and tongue motor imagery) (Pfurtscheller and Neuper, 1997; Neuper et al., 2006; Pfurtscheller et al., 2006) and for motor imagery of skilled movements over sensorimotor areas (Pfurtscheller et al., 2002; Nakagawa et al., 2011; Di Nota et al., 2017). Furthermore, similar beta power decreases have been reported during motor imagery of simple and complex walking tasks (Putzolu et al., 2022), albeit with no comparison to execution of walking, and no inclusion of a control task, which lead to uncertainty about what was being measured. The neural overlap we see in the present study between the motor imagery of walking and actual walking evident in a motor-related beta decrease provides direct evidence for functional equivalence.

Further overlap in the beta oscillations in this study signaling functional equivalence is evident in a motor-related increase of power in the beta frequency band toward the end of the movement, both when imagining and executing walking. These power increases in the beta frequency band, also known as beta “rebound,” are typically observed over premotor and sensorimotor areas in the post movement phase (Pfurtscheller et al., 2005; Jurkiewicz et al., 2006; Pfurtscheller and Solis-Escalante, 2009). The beta rebound has been related to a resetting mechanism, reflecting processing that occurs at the end of a movement (Kilavik et al., 2013), and indicates the recruitment of neural circuits related to the reset (Engel and Fries, 2010), or similarly to inhibition (Salmelin et al., 1995), which is thought to be automatically recruited in motor imagery (Angelini et al., 2015). As a finding, it is remarkable that imagery of whole-body movement, that does not involve actual movement, generates this motor-related beta-rebound. That said, we also discuss differences related to the beta-rebound below.

The methodology and design of the current study, which allows for direct comparison between imagery and execution of whole-body movements, as well as having strong temporal resolution not only revealed overlap as presented above, but also motor-cognitive differences. As shown by our analysis, the beta power decrease seen in motor imagery task was shorter compared to actual execution of walking. However, as shown by the analysis, motor imagery required a larger recruitment of neural activation compared to the execution of walking, reflecting a greater effort when beginning to imagine the action without an actual overt motor output. Furthermore, the greater power decrease in the beta band observed in the initial phases of motor imagery compared to actual walking was evident over a broad area, including central, parietal and occipital cortical recording sites, suggesting the recruitment of brain regions included in the sensorimotor network (Solodkin et al., 2004).

Recent meta-analyses have revealed that motor imagery involves a greater range of brain areas which are not activated during actual execution (Hétu et al., 2013; Hardwick et al., 2018). This broader neural network, that includes motor brain areas, encompasses parietal and occipital regions, which are known to play a key role in motor planning and in the processing of visuo-spatial information during spatial navigation (Andersen, 1997; Ino et al., 2002; Cui and Andersen, 2007; Julian et al., 2016). According to recent accounts, a function of motor imagery is to promote action predictions supporting the processing of motor and sensory information in order to guide behavior (Bubic et al., 2010; Ridderinkhof and Brass, 2015; O’Shea and Moran, 2017). Consistent with these findings, our results showed a larger parietal and occipital activation when participants imagined themselves walking, compared to when they actually walked and when they simply mentally counted. These results indicate that the imagery condition required the active maintenance of action simulation as well as a dynamic integration of visual and kinaesthetic sensory information associated with the motor imagery of one’s body moving in the environment (Krüger et al., 2020, 2022). Taken together, therefore, current findings suggest that motor imagery involves cognitive mechanisms that go beyond the mere encoding of motor information and point toward the multisensory nature of motor imagery (Dahm, 2020; Kraeutner et al., 2020; Krüger et al., 2022; Frank et al., 2023).

Differences in the temporal dynamics were also evident at the end of the movement. The results showed an increase of power in the beta frequency band toward the end of the walking movement in all three tasks, but with an overall stronger effect in the motor imagery condition. The beta power increase was apparent at different times across the three experimental conditions, with two peaks occurring during the last two time intervals in the motor imagery condition. From a temporal perspective, it was notable that the first of these peaks coincided with the beta peak in the mental counting condition, whereas the second peak coincided with the beta peak in the actual execution condition. Therefore, the beta-rebound appears to occur twice in the motor imagery condition, but for different reasons. The first “early” increase of beta power might indicate the end of the mental counting of the steps as part of imagery, whereas the second “delayed” increase of beta power appears to be related to the reset of the motor representation, occurring at the same time as the end of walking execution. Notably, it is the neural overlap in the motor-related beta-rebound between motor imagery and actual execution of whole-body walking movements here that is most remarkable.

These data also indicated a different alpha power modulation between the three experimental conditions. It is well known that during movement preparation and execution, a power decrease in the alpha and beta frequency bands occurs over sensorimotor areas (Pfurtscheller and Berghold, 1989; Leocani et al., 2001; Kaiser et al., 2003). In the present study, we found a sustained alpha decrease during actual execution, consistent with previous findings of walking (Gwin et al., 2011; Bulea et al., 2015; Bradford et al., 2016). Similar to our findings in the beta band, the results showed a prominent alpha power decrease in the motor imagery condition, but only at the beginning of the imagined walking, whereas there was no evidence of a modulation during the mental counting condition. Previous literature has suggested that such decreases in alpha power reflect the allocation of attention toward relevant task-related information, such as visual information (Foxe and Snyder, 2011; Brinkman et al., 2014) and furthermore, providing an index of cortical activation, in contrast to alpha increase of power which reflects cortical inhibition (the “idling” state, Pfurtscheller et al., 1996). From this perspective, the strong alpha decrease seen at the beginning of motor imagery could indicate either the initial allocation of attention toward relevant features of the action being simulated or to the active processing of motor and sensory information (Foxe and Snyder, 2011; Brinkman et al., 2014). Alpha suppression has been observed during motor imagery and movement observation (Pfurtscheller and Da Silva, 1999; Jurkiewicz et al., 2006; de Lange et al., 2008), and also specifically during motor imagery of simple and complex gait (Putzolu et al., 2022). The temporal specificity of the neural activity during motor imagery, and the contrast with actual execution of whole-body movements in this study, signals that motor imagery requires the allocation of attention, enabling active processing of motor and sensory information in order to simulate action.

Overall, the present results indicate that motor imagery and actual execution of whole-body walking movement share common neural features, as predicted by the functional equivalence hypothesis. Shared features occur during the early stages of walking action execution and motor imagery, likely reflecting the initiation of an action plan (Jeannerod, 1994, 2001; Decety, 1996). Overlap was also evident in the beta-rebound at the end of the action, suggesting motor imagery and execution of whole-body movements involve similar resetting or inhibition of motor processes (Engel and Fries, 2010). These similar mechanisms are reflected in beta and alpha power modulation during motor imagery, which resemble action execution patterns. The motor-cognitive processes of whole-body motor imagery was furthermore revealed through the differences between motor imagery and the actual execution of walking in this study. The data indicate that motor imagery elicits a wider range of brain activity, especially at the beginning of the task, compared to execution. The broader pattern of neural activation indicates that the cognitive mechanisms of motor imagery go beyond the mere encoding of motor-related information, and that motor imagery places greater cognitive demands than execution presumably due to the requirement to initiate and sustain a deliberate mental simulation of action. This is relevant information, for example when considering that stroke patients do not necessarily benefit from motor imagery as part of their neuro-rehabilitation (Ietswaart et al., 2011).

This is the first report of brain activity recorded during naturalistic full-body movement contrasted with the imagery of walking. Using the same imaging methods in both conditions, direct comparison between the neural correlates of execution and motor imagery of walking could be made. Furthermore, an upright body position was maintained during imagery to match the posture of actual walking. These design features were not present in the one other neuroimaging study (La Fougere et al., 2010) comparing imagery and execution of walking. EEG recordings furthermore allowed for the characterization of the temporal aspects of brain dynamics, to disentangle the cognitive processes occurring during motor imagery and action execution at different time points. As such, this is the first study that showed directly that motor imagery and execution of naturalistic walking involves shared motor-cognitive activations, and that motor imagery requires additional cortical resources.

The data supporting the findings of this study are available at https://osf.io/rth4c/?view_only=bea46044373d43dd95fde4bb6fa78b08.

The studies involving humans were approved by the Ethics Committee of University of Stirling. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

MM: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Writing – original draft, Writing – review & editing. DK: Conceptualization, Data curation, Formal analysis, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. ME: Conceptualization, Data curation, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. DD: Conceptualization, Data curation, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. MI: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Resources, Supervision, Validation, Writing – original draft, Writing – review & editing.