Abstract
Although eye movements are voluntarily controlled, gaze behavior while viewing complex, dynamic stimuli is highly consistent between individuals—people tend to look in the same place at the same time when watching commercially produced movies (Shepard et al., 2010; Hasson et al., 2008). Here, we assess gaze reliability while viewing natural motor tasks to address two questions. First, can action reliably direct the eye movements of a passive observer? Second, how are the gaze patterns of actors and observers related in space and time?
We compared the gaze trajectories of "actors" engaged in a natural action task to those of "observers" who watched videos recorded from the actors’ perspective. The actors made a peanut butter and jelly sandwich while their eye movements were tracked with a 30 Hz head-mounted eye tracker. Videos from the eye tracker’s scene camera (capturing the actors’ head-centered field of view) were presented to the observers. Observers’ gaze was recorded with a 120 Hz remote eye tracker and downsampled to 30 Hz to allow comparison between actors and observers.
Consistent with previous studies (Land et al., 1999; Hayhoe et al, 2003), actors’ gaze was tightly linked to the task at hand: They fixated objects prior to reaching to them (e.g., knife prior to picking it up) and monitored ongoing movements (e.g., jelly jar while unscrewing lid). Videos from the actor’s perspective evoked reliable gaze in the observers: Average intersubject correlations between observers was r = .38, on par with eye movement reliability while watching commercially produced movies (Shepard et al., 2010; Hasson et al., 2008). Moreover, the average intersubject correlation between actors and observers was r = .41, suggesting that observers’ gaze was spatiotemporally linked with the actors’ gaze.
Meeting abstract presented at VSS 2012