A systematic review and meta-analysis of atypical visual attention towards non-social stimuli in preschoolers with autism spectrum disorder.

More than half of all the children in the pooled sample came from just four large studies, so the headline result reflects those four far more than the other 25. The children averaged 31.62 months and spanned a narrow age range, so the finding that age made no difference says little about attention at other ages. Studies averaged 85.535% boys and several included only boys, so the absence of a sex effect may mean the analysis simply could not detect one. Only 4 of the stimulus-interest studies used still pictures against 15 using video, so the null result for static versus dynamic images is insensitive to smaller differences. Only 13 studies reported an ADOS score and 12 reported non-verbal developmental quotient, and social affect and repetitive behaviour subscales could not be analysed separately at all. Every included study was cross-sectional, so nothing here shows whether early attention to these images predicts later interests or development. The funnel plot for all the studies combined was asymmetric (Egger test z = -2.127, p = 0.033), which the authors put down to mixing stimulus types rather than to publication bias.

The authors declare no conflicts of interest. The review followed a protocol registered on PROSPERO on 03.05.2023 (CRD42023385175).

Do not read this as evidence that young children with autism simply prefer objects to people. Once the images were ordinary non-social objects, looking time did not differ from typically developing children; the difference appeared only for images tied to the intense interests reported in autism. Nothing here tests whether changing what you offer a child in a session changes anything.