Increased Access to Information, but Not Increased Feedback, Enhances Category Learning in Autism

Background: Popular autism interventions often feature high levels of feedback as well as the breaking down of tasks or information into numerous small increments presented sequentially. However, these practices are not necessarily consistent with a small but growing literature on how autistics learn well (Foti et al., 2015), including by spontaneously extracting regularities from large arrays of information (Mottron et al., 2009, 2013). Autistics have shown a range of performance in category learning tasks (Gastgeb et al., 2012; Schipul et al., 2012; Soulières et al., 2010), suggesting that changes in how and how much information is presented may affect how well autistics learn.

Objectives: Assess how access to information (sequential vs. simultaneous presentation) and level of feedback (high vs. low intensity) during learning affects the probabilistic categorization performance of autistic and typical children.

Methods: 108 children were tested on one of two probabilistic categorization tasks (Shohamy et al., 2004; Brown et al. 2010). To date, data have been analysed for 11 autistic (age=9.0 years, SD=1.5; WISC-IV PRI=108.0, SD=16.09) and 16 typical (age=9.0 years, SD=1.4, p=1.0; PRI=115.4, SD=12.0, p=0.17) children tested on the feedback task; and for different groups of 16 autistic (age=10.3 years, SD=2.0; PRI=105.9, SD=15.8) and 18 typical (age=9.2 years, SD=1.2, p=0.08; PRI=115.0, SD=11.0, p=0.06) children tested on the presentation task. 14 artificial stimuli varying across 4 dimensions had to be classified into 2 categories based on 5 different probabilities. Each stimulus was probabilistically associated with an outcome. Tasks varied either in information presentation (sequentially, one stimulus at a time vs. simultaneously, all stimuli together) or feedback level (low vs. high intensity nonsocial informative feedback) in the 200-trial learning phase, which was followed by two 70-trial test phases. Test1 used learning-phase stimuli, while Test2 used equivalent but new stimuli requiring generalization of learning.

Results:  Preliminary analyses were conducted of test-phase accuracy, reported here as mean number of correctly classified stimuli out of 70. Intensity of feedback (low vs. high) did not affect Test1 accuracy in either group (typical: low=42.2, SD=9.3, high=42.4, SD=7.2; autistic: low=40.0, SD=12.4, high=46.3, SD=9.3), p’s>.05. The same was true for Test2 (typical: low=42.0, SD=12.6, high=44.5, SD=11.4; autistic: low=45.8, SD=10.3, high=46.3, SD=12.0), p’s>.05. However, autistic children’s Test1 accuracy was significantly better when information was presented simultaneously during learning (46.9, SD 5.6) versus sequentially (36.8, SD 8.3, p=.01), while this made no difference in typical children (43.9, SD=8.2 vs. 49.6, SD=7.07, p=.176). In Test2, autistics again benefited from simultaneously (51.7, SD=6.21) versus sequentially (42.4, SD=4.7, p=.01) presented information during learning, with no difference in typical children (40.5, SD=13.2 vs. 48.2, SD=12.6, p=.125). Further analyses are ongoing.

Conclusions: In a probabilistic category learning task, increased access to information enhanced autistic children’s category learning and generalization, while increased feedback had no effect. Our preliminary results suggest the relative non-importance of feedback intensity to autistic learning, and that limiting autistic children to small increments of information, presented one at a time, may impede or undermine their learning. Increasing autistic children’s access to information they can process well should be a priority.