Discovered serendipitously whilst recording single motor neurons in macaques, mirror neurons are cells that fire both when an action is performed, and when that same action is observed (Di Pellegrino et al., 1992). Homologous neurons and mechanisms have been inferred in humans via a range of techniques, with the inferior parietal lobule (IPL) and the inferior frontal gyrus (IFG) typically implicated. These structures, along with the superior temporal sulcus (STS), are together referred to as the parietofrontal mirror neuron system (MNS; Iacoboni and Mazziotta, 2007). Some assert that the MNS is responsible for, or implicated in, social cognitive processes such as action/goal/intention understanding (Gallese et al., 2013), imitation (Williams et al., 2001), empathy (Iacoboni and Mazziotta, 2007), and theory of mind (Perkins et al., 2010). Others have challenged these claims (Dinstein et al., 2008; Hickok, 2009).

These potential links between the MNS and social cognition have led to hypotheses such as the broken mirror theory of autism spectrum disorder (ASD). This posits that if MNS function is impaired, this might lead to a decreased ability to understand and/or imitate what we observe, which would in turn contribute to the social and mentalising deficits characteristic of ASD (Ramachandran and Oberman, 2006). Relevant results in this regard are inconsistent, however. One possible explanation for such inconsistency is that MNS dysfunction may be secondary, influenced by factors such as atypical visual processing. The MNS necessarily receives input from occipital and temporal regions involved in visual processing. It has been suggested that abnormalities of the STS (a region inputting to the MNS and involved in integrating visual biological motion cues), for example, may provide a neural basis for aberrant social cognitive performance (Dakin and Frith, 2005; Hickok, 2013). More generally, MNS activity (both in the immediate and longer term) in the broader population is likely to be influenced by the quality of visual input, which is in turn shaped by factors such as gaze variables and early visual processing.

Two relevant gaze variables in this regard are gaze pattern (the number and duration of fixations in particular areas of a visual scene) and predictive gaze (PG; eye movements that proactively arrive at a visual target before a moving stimuli does). Given that mirror neurons are most active in response to observing transitive actions (Rizzolatti and Craighero, 2004), it could be hypothesized that gaze patterns that attend relatively strongly to the biological motion aspect of a transitive action (rather than the object being acted upon) would be more likely to promote greater MNS activity. Regarding PG, Flanagan and Johansson (2003) found similar PG profiles when participants observed and when they performed the same action. They inferred that observers implement eye motor programs directed by their own motor representations of the performance of that action. This inference has been supported by others and hypothesized to be related to MNS activity (Falck-Ytter et al., 2006; Elsner et al., 2013).

Whilst this proposed influence of such gaze variables on MNS activity may seem straightforward, the relationship has not been empirically examined in humans as far as the authors are aware. One study did explore this relationship in monkeys, however. Maranesi et al. (2013) recorded premotor cortex mirror neuron activity (via microelectrodes) and eye position in macaque monkeys during the observation of grasping actions. Most neurons recorded discharged more strongly when the monkey was looking directly at the action, and discharged earlier and more strongly when the gaze was proactive rather than reactive. The authors therefore concluded that mirror neuron activity was related to gaze behavior.

Direct electrode recordings of mirror neurons are generally not possible in humans. One paradigm used to putatively measure MNS activity employs transcranial magnetic stimulation (TMS) during the observation of transitive actions. TMS is a non-invasive means of stimulating the cortex via pulsed magnetic fields. When applied to the ‘hand region’ of the primary motor cortex, TMS can be used to elicit a motor evoked potential (MEP) in contralateral hand muscle which can be measured by electromyography (EMG). The observation of another’s hand movement during such TMS application results in a facilitated (i.e., increased) MEP (Fadiga et al., 1995; Theoret et al., 2005). In this context, MEPs are regarded as a measure of interpersonal motor resonance (IMR), a concept alluding to the activity of one’s motor/sensorimotor system whilst viewing another’s motor actions (Uithol et al., 2011). IMR is in turn an index of putative MNS activity.

The current study utilized this TMS paradigm to investigate the human mirror neuron response, whilst eye tracking recordings were also taken to investigate aspects of gaze and visual processing during action observation. PG times [(time of hand or object arrival at target) – (time of first fixation on target after the first fixation on arriving hand or object)] and fixations in areas of interests (AOIs) were measured, with the latter focusing on hand movement areas versus object areas. It was hypothesized that IMR would be positively associated with PG times and fixation counts in hand AOIs, and negatively associated with fixation counts in object AOIs.

Descriptive statistics are summarized in Figure 1. All statistical analyses were performed in SPSS (Version 20.0).

In the present study TMS and eyetracking technologies were used to examine the relationship between IMR (motor cortical facilitation while observing motor actions) and gaze pattern during the observation of grasping actions. No relationship was found between PG and Mean MEP. However, fixation counts in several AOIs were found to have strong associations with Mean MEP, our index of IMR. It should also be noted, however, that only linear relationships were examined here. Non-linear relationships not tested for may also exist.

The non-significant PG result found here is not necessarily inconsistent with prior research. Although the rationale for a relationship between PG and MNS activity is well established (Flanagan and Johansson, 2003; Falck-Ytter et al., 2006; Elsner et al., 2013), no previous studies have directly linked PG to a neurophysiological measure of IMR in humans. Furthermore, PG results in clinical samples with theorized MNS deficits (such as ASD) have been mixed (Falck-Ytter, 2010; Falck-Ytter et al., 2012). One study found a relationship between ‘proactive gaze’ and MNS activity in monkeys, but their operationalisation of PG was far less restrictive than that used here (Maranesi et al., 2013). PG for their study required only target fixation prior to hand-target contact, whereas PG here also required a prior fixation on the hand/object moving toward the target. Supporters of the PG-MNS link generally invoke the direct matching hypothesis, which holds that by simulating observed actions in our mirror systems, we gain some level of direct understanding of those actions. This has also been challenged, for example by Eshuis et al. (2009), who showed that PG – in the same paradigm utilized by Falck-Ytter (2010) – was more greatly influenced by agent intention and desirability of a goal state than simply whether a human agent is moving an object. These authors argue that MNS activity may reflect action understanding rather than cause it, in which case searching for a strong IMR-PG relationship is unlikely to be fruitful.

Other explanations less critical of PG-MNS theory are possible. For example, if the direct matching hypothesis is accurate, our eye movements should be approximately the same whether we are observing or performing an action. For example, in one recent study regional motor activity was shown to be highly similar irrespective of whether an arm or an eye was being utilized to control a virtual object (Modroño et al., 2015). It is unlikely, however, that when performing actions we employ PG in the way operationalised herein (i.e., looking at our own hand and then predictively at the object we are reaching to grasp). Flanagan and Johansson (2003) noted in their study that gaze was rarely directed to the hand in either performance or observation of hand actions. Furthermore, when viewing videos of simple grasping actions repeatedly, ‘predictive’ gazes might soon manifest as fixations directly to the target, in which case typical PG might bypass the agent’s movement toward the object. Approximately 50% of participants here had previous exposure to the videos during the TMS component, and may have ‘known’ the outcome without the need for prediction. A final possible contributing factor concerns movement availability. PG behavior has been reported to be impaired when participants are positioned so that the execution of an observed action is not possible (Ambrosini et al., 2012). Whilst participants here did not have tied hands, they were required to be still during both TMS and eyetracking components of the study, and may not have felt in a position to perform the viewed actions, or to produce subtle head movements that might normally occur within the motor programs associated with the action.

Overall gaze preferences align well with expectations based on knowledge regarding MNS network structure and function. Increased fixation counts in hand movement AOIs tended to be positively related to IMR, in contrast to the negative associations observed with object AOI fixation counts. The strongest distinction in this regard was between Mug Body (negatively related) and Total Hand and Spoon Handle (both positively related) in the Tea Stir video. In the present context, the Spoon Handle AOI is largely a site of biological motion, as the hand grasps the handle to stir for much of the video. The key structure that feeds visual input into MNS areas – the STS – is thought to be involved in the integration and ‘interpretation’ of biological motion cues (Saitovitch et al., 2012). Reduced biological motion perception has been associated with hypoactivity in the STS and parietal mirror regions (Freitag et al., 2008). In individuals who display a relative lack of preference for biological over non-biological visual stimuli, this may be associated with aberrant STS and subsequent MNS functioning.

A less expected finding was the direction of order effects in terms of IMR. It might be predicted that participants completing the eyetracking tasks first would exhibit reduced IMR during the subsequent TMS component, due to habituation or fatigue effects after repeated earlier viewings of the images. The reverse pattern was found. Participants completing eyetracking tasks first displayed significantly greater IMR profiles on average, with facilitation generally greater during the early pulse versions of the videos. This is also unexpected as these are typically associated with reduced MEP facilitation compared to the late-pulse versions (Enticott et al., 2012a). Thus, our results may suggest a ‘priming’ or ‘learning’ effect of prior viewings. In terms of MNS theory, this does not imply a straightforward Hebbian or associative learning effect of the type advocated by Heyes (2010), which would require repeated sensory-motor pairings to instantiate a conditioned facilitation effect. Rather, it might support a model that emphasizes either top–down (Csibra, 2008) or bi-directional relationships with higher cognitive processes (Kilner et al., 2007). According to Csibra (2008), MNS function depends on prior intention ascription and conceptual processing. This process might be enhanced in the present context of repeated viewings, so that the ‘predictive validity’ of each intention ascription is gradually enhanced, leading to augmented MNS activity. Alternatively, repetition of the same grasping stimuli may lead to a short term enhancement in excitability – a ‘priming’ of the relevant neural networks. For example, Kilner et al. (2004) found that the knowledge of an upcoming observed hand movement enhanced the excitation of the observer’s own motor system, increasing their ‘readiness potential.’ The present videos were certainly more predictable for Group 2 (who did eyetracking then TMS) when their MEPs were measured, which may partially explain their strong/early response patterns.

There were several limitations to this study, the most notable being the separation of TMS and eyetracking components. Ideally, participants would undergo eyetracking analysis and TMS simultaneously. However, the Eyelink II headstrap prevented TMS coil access to the primary motor cortex, precluding in-time gaze pattern and IMR comparison and thereby reducing correlational accuracy. Whilst this issue was partially addressed through a counterbalanced design and a separate test–retest group for the eyetracking component, there is cause for caution. In addition to the order effects for IMR, there were weak test–retest correlations in some eyetracking variables. Separate Group 1 (TMS first) and Group 2 (eyetracking first) IMR-AOI correlations yielded no significant relationships at the corrected alpha level. Although this partly reflects reduced power, inferences based on overall IMR-AOI correlations here need to be carefully considered.

Second, the assumption that MEP facilitation reflects mirror neuron activity has been challenged (Dinstein et al., 2008; Hickok, 2009). The premotor cortex is understood to contain more canonical neurons than mirror neurons, and these are object-responsive (Nelissen et al., 2005). If canonical neurons activate in response to the mere presence of objects, this raises the question of whether the static hand alone is an adequate control, or whether both an object and static hand would be the minimum baseline stimuli required to isolate ‘pure’ mirror neuron activity. Whilst there was no difference between MEP responses to ‘static hand only’ and ‘static hand plus mug’ stimuli in one study (Enticott et al., 2010), there was a weak, non-significantly greater MEP response in the latter condition. The exact contribution of canonical neurons to the premotor ‘mirror response’ remains unclear. Furthermore, only the left cerebral hemisphere was examined, which is potentially significant given that left and right ventral premotor cortices may have divergent functional roles (Kilner, 2009).

There were also several eyetracking limitations. An exploratory analysis of AOI fixation percentages suggested a similar trend to that observed with the fixation count data, yet only one relationship (Mean MEP and Spoon Handle fixation percent) remained significant at the adjusted alpha level. It is possible that fixation counts reflect general attentional factors or saccadic activity rate rather than (or as well as) interest in particular AOIs. Whilst the reduced relationship strengths in fixation percentages are a concern, they are reassuring to the extent that they confirm the directionality of fixation count relationships. An alternative approach would be to measure dwell times (percent and total) in each AOI. However, ‘vacant’ or ‘blank’ staring can produce misleading dwell profiles (particularly in the context of numerous repetitive images), and fixation numbers were felt to better reflect engaged visual attention in AOIs. Finally, the AOIs were created with hand-movement versus object distinctions in mind. However, in the static/grasping videos in particular, hands move through Total Hand AOIs relatively quickly, and interact with the objects for longer periods, so that there is hand overlap in some object AOIs, potentially blurring the intended hand/object distinction. The rapid movement through hand AOIs precipitated minimal PG in the static/grasping hand videos.

In terms of more general limitations, a failure to meet parametric assumptions meant that non-parametric correlational analyses were utilized rather than regression models. Nor did this study control for IQ or concentration levels during testing. These are all shortcomings worth redressing in future studies, which would also benefit from measuring MEPs and eyetracking variables simultaneously (by using headstrap-free eyetracking equipment while applying TMS), utilizing a large sample with ASD and appropriate controls (including other developmental and/or language delay groups), and employing stimuli of a broad enough visual scope to allow greater scanning opportunity during observation of the transitive action. To the author’s knowledge, the present study is the first to explore and find a link between gaze patterns and inferred MNS activity in humans. The findings provide some (albeit qualified) support for the possibility that MNS atypicalities may be influenced by visual processes such as relative preferences for objects and biological motion.

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.