[PS-1.18] Contextual priors do not modulate action anticipation in children with autism

Amoruso, L. 1, 2 , Pinzino, M. 3 , Narzisi, A. 3 , Finisguerra, A. 2 , Fabbro, F. 2 , Muratori, F. 3 , Volzone, A. 4 & Urgesi, C. 2, 4

1 BCBL
2 University of Udine, Laboratory of Cognitive Neuroscience, Udine, Italy.
3 Scientific Institute (IRCCS) Fondazione Stella Maris, Pisa, Italy.
4 Scientific Institute (IRRCS) Eugenio Medea, Polo Friuli Venezia Giulia, I-33078, San Vito al Tagliamento, Pordenone, Italy.

Influential models on action comprehension postulate that optimal intention inference requires the combination of two types of information: the expectations about other?s likely behaviours based on previous experiences and the sensory evidence conveyed by perceptual movement kinematics. Recently, it has been suggested that individuals on the autism spectrum have difficulties in anticipating other people?s actions due to an attenuated influence of prior information. Here, we aimed to directly test this hypothesis in a group of 24 high-functioning children with autism and 24 ability-matched typically developing children. Based on the premise that the brain represents contextual information probabilistically, we developed a behavioural paradigm consisting of a probabilistic learning task (familiarization phase) followed by an action prediction task (testing phase).
During the familiarization phase, children were presented with videos depicting a child actor grasping different objects to perform actions in diverse contexts and asked them to recognize actor?s intention. Importantly, within this phase, we implicitly biased action-context associations in terms of their probability of co-occurrence. During the testing phase, children observed the same videos but, in this case, the second-half remained occluded from view and participants were asked to predict action unfolding. We reasoned that, during this phase where movement kinematics became ambiguous, children?s responses would be biased to contextual priors acquired during the familiarization, thus compensating perceptual uncertainty. We found that, unlike ASD children, TD children exhibited a probabilistically modulated behaviour. More specifically, TD children were better at predicting those actions embedded in contexts with higher probability as compared to those with lower probability and this effect was absent in the ASD group. Collectively, our results suggest that ASD children have an impairment in building contextual priors and do not benefit from them when predicting other people?s actions.