30222
Predictive Processing in Changing Environments in Autism: Electrophysiological, Pupillometric and Behavioral Assays
The brain continually produces predictions of upcoming sensory inputs rather than passively waiting for input to act upon. Impairments in the ability to flexibly generate and update predictions about upcoming events could be stressful, and it could lead to resistance to even trivial changes in life, which is a major feature in autism spectrum conditions (ASC). In a stable and predictable environment, the neurotypical brain generates expectations with a high level of confidence; while in a volatile and unstable environment where contingencies fluctuate wildly, the predictions earn lower levels of confidence. Thus, the brain not only makes predictions, but also assigns them an expected error rate.
Objectives:
Here we test the hypothesis that individuals with ASC are impaired in the ability to flexibly adjust the confidence level of their predictions according to changes in level of volatility in the environment. This is done by investigating whether & how quickly a new prediction model is adapted upon new experience in environments with different levels of volatility (high, medium, low).
Methods:
Data from 15 high functioning adults with ASC (18 to 25 years, three females) and 15 IQ- and sex- matched neurotypical controls were collected. We first remotely trained subjects to perform a sequence pattern learning task, where three consecutive shapes appearing in a particular order can create a pattern. The subjects were instructed to respond to a target, which is the final item the pattern. On the laboratory day, we then presented conditions with varying levels of pattern violation (environmental volatility), while high-density electroencephalography (EEG), behavioral responses and pupillometry were recorded. To examine how the level of surprise was modulated upon pattern violations in different volatility conditions, we measured changes in pupil dilation. Evoked response potentials (ERPs: CNV, P3, and error related responses) and reaction times were analyzed to infer predictive processing mechanisms.
Results:
In neurotypical subjects, as expected, P3 was present for the target, and increased in amplitude for low volatility through high volatility conditions. Further, there was a graded P3 function for the predictive items leading to the target in all volatility conditions. In the ASC group, a clear target P3 was only seen for the high volatility condition. In both groups, response times (RT) varied according to the level of volatility, showing an increase in RT when the condition switched to a higher volatility, and a decrease as it switched to a lower volatility. Additional analyses will focus on how fast this RT change occurred to determine learning rates of subjects, as well as how the CNV response and pupil dynamics modulate across volatility conditions and differ between groups.
Conclusions:
The P3 data reveal that the level of volatility and predictability had a lower influence on confidence of predictions in ASC, compared to neurotypical controls. This implies that people with ASC are less capable of processing the differences in volatility in the environment, which could make the changes in daily life unexpected and discomforting. Further analyses will show whether the data fully supports these conclusions.