Adaptation in 2-Year-Olds with High and Low Likelihood of Autism

I want to share my first first-author publication:

No Evidence for Neural Markers of Gaze Direction Adaptation in 2-Year-Olds with High or Low Likelihood of Autism

DOI of the official typeset version: 10.1037/abn0000518

Freely accessible version: https://osf.io/afnzy/

This paper will appear in the special issue "Predictive Coding and Psychopathology", in the Journal of Abnormal Psychology in August. We used adaptation as a specific case to test Predictive Processing accounts of autism in our 2yos with older siblings with and without autism.

Adaptation is a really cool phenomenon (I might be biased) that shows how our recent experience influences our perception. For example, check out this continuum, with Ben Affleck on the left, Heath Ledger on the right and 3 morphed images in between (from Walther et al., 2013).

The central face is morphed such that it's 50% Ben, 50% Heath. If you look at the leftmost picture (our friend Ben) for a few seconds, and then at the central one, the central one looks more like Heath. If you look at Heath for a few seconds, the central one looks more like Ben! The central face hasn't changed at all, so the input we're getting is identical, but our perception of it changed. The only difference is our recent experience. This happens with every new piece of info we get, our perception is *always* influenced by our previous experience.

Predictive Processing theories of autism claim that autistic people use their previous experience in a different way than non-autistic people when interpreting new information, and that this is also always true, for every new piece of info they get. This is important for our understanding of autism because at the moment we really don't know much: why do some people have certain social differences and also have special interests? How can we best support those who are struggling? To know this we need to find mechanisms.

Back to adaptation: autistic people sometimes experience less adaptation. The perceived identity of the central face changes less, based on previous experience, than for non-autistic people. This makes sense within the theory but we don't know exactly how or why it happens. If the theories are correct, we should be able to see this difference in much younger children. That's key to understanding autism's roots because older autistic people have learnt strategies to get on in the world. We look for group differences that pre-date these strategies. So, if the theory is right and all info processing in autism is different as a result of combining previous experience with current input, we can assume the high-likelihood group will show signs of reduced adaptation.

We used gaze direction instead of identity because in older children, these stimuli have been used in a behavioural task. After exposure to extreme left gaze, participants experience slight right gaze as actually being straight ahead, but the autistic kids experience this less.

Because 2-year-olds can't press buttons to answer (at least, not very well), we used EEG to see if their processing of the faces changed after adaptation. We looked to see if their N290 or P400 amplitudes differed to small gaze angles before and after exposure to an extreme gaze angle. We expected that high-likelihood kids would habituate to faces slower than low-likelihood kids, because if their perception is less influenced by experience, each face is more distinct. We also expected less adaptation in high-likelihood kids, as shown before in autistic kids.

What we actually found:

High-likelihood kids habituated more than low-likelihood kids. Unexpected based on this theory and based on prev habituation work. Maybe high-likelihood kids didn't notice the different gaze directions and low-likelihood kids did, but this is also counter to the theory.

When it came to adaptation, groups didn't differ and the participants as a whole didn't adapt, regardless of group. Also weird! This was a complicated interaction so we could've been underpowered, but Bayesian stats showed moderate evidence for a null effect.

We're not totally sure why we don't see any adaptation, but we had relatively few trials per cell (average of 8) and the gaze directions are quite subtle so if the kids were moving they would've potentially not differentiated between slightly left, straight on or slightly right. It's a shame we couldn't observe adaptation because it limits what we can say about Predictive Processing theories of autism, but the habituation result challenges their claims. We suggest that the theories may need to be refined based on this and some other null results.

I still like Predictive Processing theories of autism because they're the only ones so far that I think address all aspects of autism, but there's a pile of conflicting results now so I do think we need to revise them. A lot of cool work's ongoing so we'll likely know more soon!