Sensory Attenuation in Autism Spectrum Conditions

Background:

The last years have seen a growing interest in using approaches within the predictive processing framework (Dayan et al., 1995; Rao & Ballard, 1999) to investigate differences in the cognitive, perceptual and neural processes in autism spectrum conditions (Pellicano & Burr, 2012; Sinha et al., 2014; Lawson et al., 2014; Van de Cruys, 2014).
At the core of models of prediction lies the notion of the brain as an 'optimal inference machine' which tries to model the external world with the purpose of minimising prediction errors (= the difference between the expectation of the experience and the experience itself).

Self-generated actions are thought to be accompanied by an efference copy of the motor command and as such generally result in smaller prediction errors than externally generated sensory input. There is evidence that predicted sensory input (such as from self-generated actions) is attenuated in order to make unpredicted sensory information more salient (Bays & Wolpert, 2007).

Objectives:

We were interested in investigating sensory predictive processes during self-generated movements in autism spectrum conditions.

Methods:

A well-established ‘force matching’ task (Shergill et al., 2003) was performed by 26 control and 23 ASC participants. During the experiment a force (varying between 0.5N and 2.75N) was applied to the participants’ index finger via a lever and participants were are asked to match the force by pressing directly on the lever with their other index finger (‘finger condition’) or by adjusting a slider which controlled the torque motor and lever (‘slider condition’).
As the ‘finger condition’ models the sensory attenuation occurring during self-generated movements, it is expected that participants will overcompensate and apply a larger force than the reference force whereas limited experience with the ‘slider condition’ should produce more accurate matching results.
Participants also filled in self-report measures of autistic (AQ, Baron-Cohen et al., 2001) and schizotypal traits (PDI, Peters, 1999).

Results:

Regression lines were calculated for both the ‘finger’ and ‘slider’ condition.
Both group experienced sensory attenuation (p<0.000) as measured by a difference in slope and intercept between the ‘finger’ vs. ‘slider’ condition but there was no difference between the groups in the magnitude of overestimation (p=0.805).

Higher scores on the PDI were linked to smaller intercepts in the ‘finger condition’ in the control subjects only (p=0.012) whereas there was a trend in the ASC group relating higher scores on the AQ to larger intercepts in the ‘finger condition’ (p=0.067).

Conclusions:

Our experiment does not support the suggestion of autism as a generalised ‘disorder or prediction’ (Sinha et al., 2014) as attenuation of predicted sensory input seems to be preserved.
However while we replicated earlier findings of reduced sensory prediction in individuals with higher schizotypal traits in the control group (Teufel et al., 2010) the same did not hold for the ASC participants. This could be either due to the fact that the PDI is not measuring the same latent trait in the ASC group or because variance in sensory prediction is not mediated by the same factors in individuals with autism.