The Temporal Dynamics of Social Attention and Emergence of Autistic Group Differences

Background:

Many eye-tracking studies indicate that individuals with autism spectrum conditions (ASC) exhibit reduced attention to social stimuli relative to Neurotypical (NT) observers (Tegmark et al, Res Dev Disabil, 2015). However, despite the rich, time-varying gaze data produced by eye-tracking, almost no work has been conducted to provide models that quantitatively predict social gaze behavior over time in individuals with and without an ASC diagnosis. Instead, social gaze behavior is typically described by summary metrics, collapsed across time (e.g. total gaze duration) that neglect substantial information contained within the gaze time series. Recently, we demonstrated the explanatory power of exploring gaze time series: we found that individual differences in trait empathy predict the temporal evolution of social gaze behavior - empathy maintains social attention after prolonged viewing (Proc Roy Soc B, in press). Previous research indicates that ASC is associated with differences in some empathy-related processes (Baron Cohen et al, J Aut Dev Dis, 2004). We adopted a model-driven, time-series approach to characterise differences in the temporal structure of social attention between individuals with and without an ASC diagnosis.

Objectives:

i) Characterise the temporal structure of social gaze behavior across participants.

ii) Characterise how this temporal structure varies between ASC and NT observers.

iii) Characterise the influence of individual level social trait characteristics on social gaze bias over time.

Methods:

53 observers (31 neurotypical, 22 with an ASC diagnosis) observed 60 competing pairs of social and nonsocial reward stimuli for 5 seconds in an unconstrained free-viewing task. Gaze was recorded via a tobii T60 eye tracker. All observers completed social trait measures designed to measure empathic (Empathy Quotient, EQ) and autistic traits (Autism Quotient, AQ).

Results:

Consistent with previous literature, ASC observers exhibited a generalised decrease in social attention. Critically, the groups also differed with respect to the temporal structure of their social gaze behavior. NT observers demonstrated i) an initial increase in social attention, ii) a gradual decline and iii) partial recovery towards the end of the trial. However, this latter recovery component was not observed in ASC observers. Accordingly, a divergence analysis revealed that the difference in social attention between groups increased over time.

Moreover, this increased divergence between groups over time was mirrored by a corresponding increasing influence of EQ and AQ on gaze behavior over time. Replicating previous work, we found that higher EQ/ low AQ sustained gaze towards social images after prolonged viewing.

Conclusions:

Considering our data with previous literature, we speculate that empathy enhances the value of social rewards, potentiating a ‘gaze cascade’ type effect that maintains enhanced perceptual selection of social inputs over time. This provides a plausible mechanism for the social attention deficits observed in ASC. In general, our analyses demonstrate that considering the temporal structure of gaze signals provides more refined quantitative endophenotypes for conditions marked by deficits in one or more empathy related processes (e.g. ASC, Psychopathy).