We reasoned that uneven performance could relate to differences
in the way context determines the expected levels of sensory
precision, which is an aspect of top–down modulation that has
only recently been described (Feldman & Friston, 2010) and linked
to ASD in the context of predictive coding models of perception
(Friston et al., 2013; Paton et al., 2012). Conceptually, expectations
regarding the precision of sensory input are of importance to the
relative weighting of bottom–up and top–down perceptual processes
in response to state-dependent (i.e., changing) levels of
uncertainty. This proposal therefore predicts that differences will
become apparent in contexts and experimental set-ups where
changing conditions suggest changing levels of uncertainty in the
sensory signal. In particular, individuals with ASD, as well as
nonclinical individuals with ASD-like traits, may be less sensitive
than individuals with few ASD-like traits to contexts that suggest
increased uncertainty. This would predict that in contexts that
suggest low uncertainty (i.e., high precision of sensory input) there
would be less difference between the groups, but that in contexts
that suggest higher uncertainty (i.e., suggests low precision of
sensory input) differences would begin to emerge. Sometimes
these differences would give rise to enhanced performance of the
ASD and ASD-like groups, namely when the expectation for high
precision input leads to increased sensory sampling and less
integration under prior expectations relevant to the context.
Sometimes this would lead to compromised performance for these
groups, namely when expectation for high precision leads to
blindness to underlying patterns of hidden, influencing factors.