Two-year-olds with autism exhibit no preferential attention to biological motion, while control children show significant preferences. (A) Example still images from point-light biological motion stimuli, with centering cue at start. Each animation showed an upright (UP) and inverted (INV) figure with accompanying soundtrack matching the actions of the upright figure. The upright figure enacted childhood games. Figures were identical except that the inverted figure was rotated 180 and its movements played in reverse order. In B, C, and D, visual scanning data of individual children are plotted as horizontal location by time. Breaks in data occur for blinks or offscreen fixations. (B) Visual scanning data from one toddler with autism (ASD), for one animation. (C) Data from one typically-developing toddler (TD). (D) Data from one developmentally-delayed but non-autistic toddler (DD). (E) For the ASD group, fixation to upright and inverted biological motion occurs at chance levels. (F) TD toddlers give preferential attention to upright animations. (G) DD toddlers also give preferential attention to upright animations. Horizontal guidelines denote percentages not significantly different from chance. Error bars are SEM.

When the animation contains a physical contingency, two-year-olds with autism do show significant viewing preferences. (A) During other biological motion animations, ASD toddlers show no preference; but when a physical contingency is present on the upright side, these toddlers show significant preference for the upright figure: different from chance (p < .01), and different from their viewing behavior to other animations (p = .044). While other animations presented only moving point-lights and human voice, one type of animation contained an additional cue: as two point-lights, representing the actor’s hands, collided, the sound of clapping could be heard (playing “pat-a-cake”). The collision of point-light “hands” actually caused a noise (the clap) to occur, localized to the upright figure and absent from the inverted (the inverted figure’s movements were not synchronous with the claps). (B) TD toddlers show no significant change in preferential viewing. (C) DD toddlers also show no significant change in preferential viewing. Horizontal guidelines denote percentages not significantly different from chance. Error bars are SEM. * = p < .05. Please see Supplementary Materials for movie data.

Quantification of audiovisual synchrony. (A) We measured spatial trajectories (X-Y location over time) of all point-lights throughout each biological motion animation. Example trajectories are for Inverted left hand and for Upright left hand. (B) Magnitude of change in velocity of Inverted left hand, | V|. (C) Magnitude of change in velocity of Upright left hand. (D) Magnitude of change in shortterm amplitude envelope of audio soundtrack, | A|. (E) Audiovisual synchrony, AVS, of Inverted left hand, obtained as pointwise product of parts B and D. (F) Audiovisual synchrony, AVS, of Upright left hand, obtained as pointwise product of parts C and D. (G) Two still frames from pat-a-cake animation. Color scale values range from low or no synchrony (dark blue) to maximum synchrony (red). Note that some point-lights are very synchronous (the hands, shown here during claps), while others are hardly synchronous (e.g., the feet). (H) Summation of audiovisual synchrony over the duration of an entire animation. Oblique view shows that while there is more audio-visual synchrony on the upright side, the inverted side also contains synchrony: by chance alignment (reverse motion signal aligned with forward audio signal), some change in movement of inverted point-lights can occur synchronously with the change in audio. If preferential viewing in our stimuli were related to level of audiovisual synchrony, then the relative levels of synchrony on the upright versus inverted side will provide predictions of expected viewing behavior.

Level of audiovisual synchrony is highly correlated with preferential viewing in two-year-olds with autism; is uncorrelated with viewing in control children; and can predict ASD viewing patterns in novel animations. (A) Preferential viewing is significantly correlated with audiovisual synchrony in ASD toddlers. Plots pair preferential viewing and audiovisual synchrony. When the animation with greatest upright audiovisual synchrony (pat-a-cake) is withheld from analysis, audiovisual synchrony is still significantly correlated with viewing behavior in ASD toddlers: r² = .95, p = .018 (plotted as thin regression line through remaining 4 data points). (B) Preferential viewing by TD toddlers is uncorrelated with audiovisual synchrony, across either 4 or 5 animations. (C) Preferential viewing by DD toddlers is also uncorrelated with audiovisual synchrony. (D) To test whether audiovisual synchrony could predict looking behavior in new animations, we created two additional animation types. The regression from the original data, with weighted binomial prediction intervals, provided a model for expected behavior. P₁ and P₂ denote prediction intervals for the new animations. Probability of obtaining results in these intervals is noted to the right of the regression plot. For an independent cohort of toddlers with autism, matched to the original cohort, preferential viewing was predicted on the basis of audiovisual synchrony (p =.0004). In all plots, axis Y shows preferential viewing as a difference score: percentage of fixation time to upright (UP) minus percentage of fixation time to inverted (INV). Positive values indicate increased looking at the upright. Likewise, axis X shows audiovisual synchrony as synchrony of the upright (as percentage of total synchrony) minus synchrony of the inverted (also as percentage of total). Positive values indicate greater synchrony in the upright figure.