Sex differences in atypical predictive processes from low spatial frequencies
Résumé
Background:
Fast visual recognition in non-autistic adults relies on coarse information conveyed by Low Spatial Frequencies (LSF), which possibly triggers predictions. Atypical predictive processes have been observed in autistic individuals. In visual perception, this might be related to a bias toward local information conveyed by High Spatial Frequencies (HSF). By reducing predictions in autism, this bias might impair facial processing. Additionally, sex differences have been reported in autism, including during attention to faces, but need further investigation.
Objectives
The objective of the present study was to investigate if autistic individuals have impaired predictive processes from LSF during face processing and if there were sex differences.
Method
The Mismatch Response (MMR) being an electrophysiological correlate of the prediction error, an oddball paradigm was employed in 40 autistic and 48 non-autistic individuals while recording their scalp activity with a 96-EEG channels cap. An unfiltered face was used as the standard stimulus, whereas HSF and LSF-filtered faces were deviants. Assuming that LSF were at the root of predictive processes, a smaller prediction error (i.e., smaller MMR) to deviants containing only LSF compared to deviants containing only HSF was expected in non-autistic individuals. In contrast, this difference was expected to be reduced in autism.
Results
Cluster statistics and ERP analyses on MMR over parieto-occipital areas both revealed that LSF-filtered face deviants lead to a smaller prediction error than HSF-filtered face deviants in two time-windows (130 -230 ms and 350-450 ms after stimulus onset). However, cluster statistics show that this difference was reduced in autism during the second time window (pcluster = .03). Planned comparisons on ERP showed that the difference in amplitude between HSF-MMR and LSF-MMR was significant in autistic males and females irrespective of the time window ( all p < .006 ) but not significant in autistic males, while in autistic females, it was significant in the early part (130-230 ms; p < .001) but not in the late part (350-450 ms).
Discussion
These findings confirm the predictive role of LSF in non-autistic individuals and highlight atypical predictive mechanisms from LSF during face processing in autistic individuals. Additionally, it emphasizes sex differences in autism, in line with recent research. These differences might contribute to better social skills in autistic females and confirm the need for sex stratification in autism research.