Learning during noisy vision in 3-year-olds at high and low risk for autism

Abstract

Recent accounts of autism claim that due to altered predictive processing, autistic individuals cannot always identify signal from noise in sensory input, which leads to models of the world that are overfitted to specific experiences (Manning et al., 2015; van der Cruys et al., 2017). According to these accounts, information processing differences should be present at a very young age, even before autism can be reliably diagnosed. Due to the hereditary nature of autism, we can currently study the development of the condition by following younger siblings of autistic children. Here, we examined whether these younger siblings have difficulty in distinguishing signal from noise, by manipulating sensory noise in a touchscreen Serial Reaction Time Task (SRTT). In the task, a picture of a frog appeared in a sequence (see Figure 1), and the children were asked to press the frog. In block 1, the sequence was identical on every repetition, while in block 3, the frog's exact location included random jitter; it was possible to learn which lilypad the frog would appear on next, but not precisely where on the lilypad it would appear. According to Predictive Processing accounts of autism, high-risk siblings should show difficulty separating signal from noise, and should thus learn the pattern slower in the block with added jitter. This should be especially pronounced in the children who receive an autism diagnosis themselves. Data collection is ongoing, but so far the results suggest that high-risk children (N=12) do in fact learn the pattern when there is additional noise added to the signal. This is contrary to previous findings (Manning et al., 2015; van der Cruys et al., 2017) and raises questions around when autistic children’s performance is and is not influenced by noise in the signal.