What Strategies Do Autistic Children Use When Learning New Categories? an Eye-Tracking Study

Background: Autistics have shown a range of performance on category learning tasks (Gastgeb et al., 2012; Schipul et al., 2012; Soulières et al., 2011), possibly associated with differences in how they process learned material (e.g., reduced attention to social elements, attend to fewer elements). Recent studies have tried to understand how and when autistics learn efficiently (Foti et al., 2015), including by spontaneously extracting regularities from large arrays of information (Mottron et al., 2013). Such findings suggest that how information is presented may affect the way autistics explore their visual environment, ultimately affecting leaning.

Objectives: To study how autistic and typically-developing children process new information during a probabilistic categorization task, and specifically; 1) Do autistics attend to the same stimulus elements and show the same pattern of preferences compared to typical children?, and 2) Do the learning strategies used by the autistic children differ according to the way material is presented?

Methods: 38 autistic (M_age 10.1 years, SD 1.9; M_{WISC-IV PRI} 102.5, SD 22.7) and 43 typically-developing (M_age 9.4 years, SD 1.4, p>.05; M_PRI 111.6, SD 20.9, p>.05) children completed two sets of probabilistic categorization tasks (Nader et al., 2018) using a Tobii T120 eye tracker. Fourteen artificial stimuli varying across 4 features had to be classified into 2 categories based on 5 different probabilities. Each stimulus was probabilistically associated with an outcome. Tasks varied in the way information was presented; (1) one item at a time with feedback (sequentially), or (2) a non-feedback version of the task with the observation of the items-outcome association (simultaneously). A 200-trial learning phase preceded a test phase in which each stimulus was presented twice (total 28 trials), and eye tracking data collected. All stimuli were partitioned into 4 areas of interest (AOI) (4 main features of the stimuli), which could be either present or absent on a given stimulus. Analyses were conducted on the length and the frequency with which children attended to the different AOIs on items containing most of the AOI (i.e. presence of 3 features out of 4). Further analyses are ongoing.

Results: Consistent with Nader et al. (2018), performance of autistic children was better -and similar to that of typically-developing children- when information was presented simultaneously, and decreased when presented sequentially. Collapsed across conditions, both groups attended to the same features, but with autistics having shorter and reduced number of fixations on AOIs than typically-developing children. Interestingly, autistic displayed less (F (1, 37) 6.2, p=.02) and shorter (F (1, 37) 7.7 p=.009) fixations than typical children when information was presented simultaneously during learning, while the length and the frequency were similar for both groups when information was presented sequentially (p>.05).

Conclusions: In a probabilistic category learning task, autistic children based their learning using the same features as typically-developing children, but were more efficient when information was presented simultaneously, resulting in a reduction in information processing time. These findings suggest increasing access to information (i.e., simultaneous) while learning may be beneficial for autistic children.