Dance has long been recognized as a universal, non-verbal communicative medium through which emotional states, social narratives, and myths may be conveyed(1). Evidence from dance and emotion research suggests that humans are highly attuned to dancers’ affective states, communicated through body and limb movements; for overview(2). For many world cultures, music and dance are not separable cultural categories; for example, Merriam 1964 quotes Gbeho as stating that in the indigenous music of the Gold Coast “If we speak of a man being musical we mean that he understands all the dances, the drums and the songs”(3). From a global, cultural perspective, dance is frequently integrated into ritualistic and religious ceremonies, both in contemporary societies(4) and historically(5). Moreover, dance is considered an essential ingredient in the formation and maintenance of group identities(6, 7, 8). For example, researchers investigated the effects of an instructional program in creative dance on the social development of preschool children(9). Teachers and parents’ ratings of the children's social skills, both before and after the program, revealed significant gains in social competence, suggesting that creative dance instruction for (at-risk) preschoolers improves social interactions and behaviour(9). Therefore, dance might exist today due to an evolutionary past in which it promoted prosocial behaviour, increasing one’s ability to survive within ‘primitive’ communities(10, 11, 12, 13, 14).

Key to dance’s ability to engender social affiliation is, arguably, its capacity to direct the attentional foci and cognitive resources of individuals. Evidence for this, and its effect upon interpersonal memory, was investigated(15). In brief, the study required untrained dancers to recall various attributes of one another after having danced in groups, some synchronously, others asynchronously(15). Results showed that those who danced together in time were more likely to remember each other. It was hypothesized that this may facilitate social bonding, which would presumably be difficult to achieve in situations where interpersonal memory was absent(15). For recent related research regarding dance and social bonding, see(16).

In a related study, eye movements of participants observing pairs of dancers, one of whom danced in synchrony with a musical track, while the other danced asynchronously was investigated(17). Gaze dwell-times amongst participants were significantly greater for the music-synchronous dancer, indicating a possible mechanism through which attention may have been directed towards the in-tempo dancers in the group study(15). Moreover, they investigated fixations across different body regions, including head, torso, legs and feet(17). Perhaps, paradoxically, given the importance of legs and feet in most dancing, feet attracted significantly less dwell time than any other body region. In sum, dance, in combination with music, appears to have the ability to direct attention as detected in the eye movements of observers. This is on par with everyday motion, which also produces gaze toward the head(18).

Two types of eye movements are usually studied in scene-perception research: fixations, during which the eyes remain still and new information is acquired from the visual field, and saccades, movements between fixations during which vision is suppressed and no new information is gained(19). In reading research, regressions—reverse saccades in which the eyes backtrack to the previous fixation point—are frequently examined in relation to syntactic comprehension(20, 21). In scene perception, fixations are usually between 260–330 ms, interspersed with saccades lasting about 50 ms(19). Saccade lengths can differ significantly depending on the type of image being viewed, and are about 40% longer for complex natural scenes than abstract patterns(22).

While several individuals may observe the same dance, various studies indicate that how each individual completes this potentially cognitively demanding task depends upon context and experience. For instance, researchers found that emotional arousal increases upon repeated exposure to a musical piece, suggesting that, in general, familiarity influences cognition, possibly by allowing individuals to develop schematic expectations(23); see also(24). And while, since the early 1970’s, it has been known that expertise and familiarity of the static visual stimuli significantly influences eye movements(25), it was only in the 1990’s that researchers thoroughly investigated how expectations influence human motion perception. For example, individuals’ schema of biological motion guides them to fixate on the ends of limbs to track the movement of human extremities(18).

With respect to dance, in a study examining the influence of expertise on the observation of dance, it was found that people with advanced dance training had shorter fixation durations and faster saccades than novices(26). In their analysis of body-directed fixations, the experienced choreographer in the study attended mostly to the head of the dancers, while novices attended equally to the head, neck, torso and arms(26). It was suggested that the eye movements of the expert in their study was “likely guided by the expectancies and schemata in long-term memory”, and that this was due to them being “adept at abstracting and extracting key information from complex movement material” (p. 23)(26). Which is to say, fixations and saccades are influenced by the type of image being viewed and expertise that, in turn, are suggestive of viewers’ underlying cognitive processes, attentional foci, and schematic knowledge. The differences between experts and novices that were found(26) are consistent with earlier findings that found the performance of expert dancers on a dance sequence recall task depended in part on the amount of structure in the material, which implied that subject’s knowledge base impacted memorization(27). For related neurological research see(28).

Common to the work of (17, 26) and others(29) is the treatment of dance as a relatively abstract form of human movement, seemingly divorced from its richer cultural setting in which, for example, narrative meaning may be conveyed; although, see(30) for the effect of culture on eye movements during scene perception. The influence that narrative contexts can have on perception, and, in particular, eye movements, has recently received increased attention. Researchers studied the relationship between film viewers’ eye movements and their comprehension of film narrative by investigating whether eye movements differed based on understanding(31). Referred to as the mental model hypothesis, this notion is distinct from the alternative tyranny of film hypothesis, which stipulates that differences due to understanding are overwhelmed by viewers’ attentional synchrony(31).

In brief, two groups were presented with a short clip from a James Bond movie in which a villain (“Jaws”) was about to fall from the sky onto a circus tent(31). Critically, one group saw only the clip while the other saw the preceding two-and-a-half minutes of the movie(31). The researchers hypothesized that the second group, who viewed the clip with its narrative context, would be better able to draw critical inferences and have more coherent perceptions than the group who viewed only the short clip(31). However, despite the difference in the stimuli, both groups showed strong attentional synchrony, and only small between-group variance(31). Overall, then, the results were more consistent with the tyranny of film hypothesis than the mental model hypothesis, suggesting that narrative context may contribute less to eye movements than visual features such as flicker and motion (i.e. temporal contrast) during free-viewing of videos(31); see also(32, 33).

Our intention in this eye-tracking study was to develop some of the research discussed above in an experiment that examined the effects of expertise and dance narrative on eye movements—that is, dances which, through various gesture sequences, attempt to convey specific, real-world and/or religious meanings. To preempt our hypotheses somewhat, we envisaged that the difference between experts and novices would lead to differences in eye movements. One such dance that lends itself particularly well to this is Indian Bharatanatyam dance. As discussed above, while prior eye-tracking studies have identified several factors that influence the processing of dance—such as expertise—these factors have yet to be explored within broader cultural contexts. Our study sought to provide an expanded cultural understanding of the effects of expertise on observing dance using videos of Bharatanatyam.

This study explored the extent to which expertise and narrative content influenced the eye movements of people observing Bharatanatyam dance. Two dependent variables were measured, fixation duration and dwell time. A further analysis calculated the dispersion of fixations within 1,000 ms time windows, advancing in increments of 500 ms. Despite the relatively small eta squared effect values, the analyses determined that expert viewers possessed greater attentional similarity than novices: lower fixation-position SDs within a given time period are indicative of greater fixation alignment, and thus increased attentional similarity.

The above was conducted in order to test four related hypotheses. Hypothesis (1)—that novices will have longer fixations than experts—was supported by the data: experts in our study did have shorter fixation durations than novices, suggesting that the inexperienced participants observed the dance videos less efficiently (Figure 2). This result is consistent with previous findings(26) who reported a similar effect in their novice-expert dance study. However, while those findings were concentrated on western contemporary dance(26), our study sought to expand and/or generalize this effect to a very different cultural tradition, i.e. classical dance of southern India.

Hypothesis (2)—that narrative dance versus non-narrative dance will produce differences in eye movements—was not supported by the fixation-duration data, which showed no statistical differences within factor Dance-type. This hypothesis was based on previous research that even in the absence of any visual stimuli, contextual information influences oculomotor responses(40). Prior to conducting the experiment, we had speculated that longer fixation durations might occur for the narrative videos as these dances contain richer semantic information, and thus, arguably, require greater attention; however, this was not the case in this instance. That said, a significant effect of Dance-type was observed when the dependent variable was fixation position SD (Figure 7). Whether this effect was genuinely cognitive (e.g. involving different attentional resources or processes), or simply due to the dancer moving more in the non-narrative videos is, however, uncertain—dynamic video-contrast analysis(32) would be required in order to answer this point definitively. Be that as it may, given that we are uncertain as to the cause of the effect of Dance-type on fixation position SD, the findings above cannot be said to support the conclusions discussed in the Introduction(31); namely, that narrative contexts may contribute less to eye movements than visual features such as motion (i.e. temporal contrast) during free-viewing of videos.

Hypothesis (3)—that greater gaze dwell times will occur in relation to the upper body—was conclusively found to be the case (Figure 4). This result strongly aligns with the findings that, perhaps counter-intuitively, found that the feet of the dancer in their experiment attracted the least fixations(17). Feet are arguably a dancer’s greatest asset(41); it could be considered paradoxically, then, that when observing dance, we seem to spend the least amount of time fixating on this part of the body. That said, an important confound should be mentioned. In scene viewing, gaze direction can be predicted using a combination of saliency and face detection(42), both of which do not depend on whether the scene involves dance. The effects of saliency and face detection significantly bias the upper screen area, in which the face of our dancer was almost invariably located. As a result, no significant conclusions can be made regarding whether dance per se specifically directs attention towards the head.

An interesting finding concerning factor ROI was that fixation durations were significantly shorter for the lower versus upper part of the dancer’s body (Figure 3). We confess to being somewhat puzzled by this, although it could be related to the fact that participants spent significantly less time observing the lower part of the dancer’s body (Figure 4). Maybe, observers genuinely find dancers’ legs and feet less interesting (or, at least, informationally impoverished relative to the upper body and head), in which case, when their gaze is drawn downwards, a vertical saccade abruptly directs foveal vision upwards. Recent research supports this conjecture; researchers found that dance communicates group coordination through combined movement dynamics within groups of performers(43). They showed that movement synchrony among a group of performers predicts the aesthetic appreciation of live dance performances(43). They concluded that their findings were in accordance with the evolutionary function of dance in transmitting social signals between people through human movement(43).

Hypothesis (4)—that experts will have greater attentional similarity—was supported by the data: the fixations of the Experts in our study were, in general, more tightly clustered than Novices’ fixations, both spatially and temporally (Figures 5). This is particularly noticeable in the dwell time heat maps in Figure 6: the lower image (novices) clearly has a larger and more diffuse gaze-pattern relative to the upper (experts). This result lends support to the conjecture, referred to in the Introduction, that experienced observers of dance have choreographic schemata stored in long-term memory, which enables them to target their attention towards (shared) salient elements within the dance(26). In turn, this results in a higher degree of attentional similarity between participants, as expressed in fixation-dispersion data. Conversely, inexperienced viewers produce more scattered fixations as they (variously) attempt to make sense of and predict the dancer’s movements. Interestingly, within Bharatanatyam, hand movements are generally considered more important than leg movements, a fact which would have been known to our expert participants, but not novices. Therefore, although the above finding may be due, in part, to dance schemata, the possibility that explicit Bharatanatyam knowledge may have also influenced our results cannot be ruled out.

The fact that our video stimuli were based on examples of Bharatanatyam has both benefits and drawbacks from an experimental perspective. A particular benefit of using excerpts from an authentic Bharatanatyam performance was that they provided the study with a degree of ecological validity—the stimuli were not unduly controlled or contrived, nor stripped of their cultural richness as expressed in the dancer’s costume and the accompanying Carnatic music. It is legitimate to claim, therefore, that our data, however imperfect, were at least produced in response to a real-world phenomenon, and are thus potentially applicable or relatable to other similar global dance practices. For example, future research could similarly examine Kathakali, which, like Bharatanatyam, is a major form of classical Indian dance involving story telling, but which in contrast to Bharatanatyam is predominantly performed by male actor-dancers.

With respect to our study’s limitations, by including only one dancer, factors specific to this individual not controlled for in the experiment may have skewed our results. That said, given the highly codified nature of Bharatanatyam, achieved through multiple years of training, it would seem likely that our findings would be replicated using other dancers. A further potential drawback, previously mentioned, is that our participants were all female. Given that Bharatanatyam is most commonly danced by females, experts with this discipline tend also to be female, if not exclusively so for all practical purposes. Our desire to avoid a gender imbalance between the experts and novices naturally led to the exclusion of male (novice) participants, which may, in itself, impose restrictions on the degree to which our results are generalizable. One further limitation concerns the interdependence of the music and dance—these two crucial factors were not separately manipulated and thus it is possible that there was an undetected interaction between the two, i.e. there may have been factors within the music that interacted with particular gestures, giving rise to specific eye movements. That said, as mentioned previously, all the music was of a similar emotional character and mood, and thus we believe it is unlikely that there were any significant interactions.

Eye-tracking cameras invariably produce a wealth of data that can be analyzed using systems’ proprietary software and/or exported to other analytical packages. In this regard, our decision to concentrate on fixation durations and dispersions, and dwell time may seem unduly restricted: saccade and pupillometric information could, in theory, have also been included in the analysis. Our reason for not doing so was due to the hypotheses we wished to test and their relationship to previous research(26, 17). Certainly, multiple additional statistics could, no doubt, have been included; the extent to which these would have enriched or detracted from the study is, however, open to debate.

Our aim was to extend to a broader cultural context a series of findings derived largely from western dance(17, 29), and in this regard the study achieved its main goal. Data consistent with three of our four hypotheses were produced by stimuli consisting of videos of an actual Bharatanatyam dance performance: experts had shorter fixations and greater attentional similarity; greater gaze dwell times occurred predominantly in relation to the upper body. Only one hypothesis achieved limited support: fixation durations showed no difference between narrative and non-narrative videos, whereas fixation dispersion patterns did differ. In sum, the study assists in building a nuanced picture of some of the eye movements associated with Bharatanatyam, and, in so doing, helps pave the way for research investigating other non-western dance forms.

The author(s) declare(s) that the contents of the article are in agreement with the ethics described in http://biblio.unibe.ch/portale/elibrary/BOP/jemr/ethics.html and that there is no conflict of interest regarding the publication of this paper.

This research was supported by funding from the School of the Arts, McMaster University, and the Arts Research Board—Social Sciences and Humanities Research Council (SSHRC), Institutional Grants (SIG) program.

We especially wish to thank Ms. Colleen Tang Poy for her detailed reading of the manuscript, and for her helpful edits and suggestions. Thanks are also due to Mr. James Anthony for his assistance in processing the timewindow information. In addition to the university students who took part in the experiment, special mention should be made of our Bharatanatyam experts without whom this study would have been impossible. We are deeply indebted to them for their input, made possible through years of dedication to dance.