Reduced task adaptation and contextual awareness in autistic adults during facial emotion recognition: evidence from mixed-effects modeling and automated facial analysis.

Everyone in the study was an autistic adult without intellectual disability and without acute severe depression. Nothing here tells you about autistic people with intellectual disability or about children. This is one specific computer task with three forced-choice answers and no time limit, done in a lab. The authors say the findings on how performance changes over the task and how much the stimulus matters could be different for other emotion recognition tests. The autistic participants were recruited from a psychotherapy trial, so they may not look like the autistic adults on your caseload. The automated salience measures were interpreted through theory rather than read directly off the model, and the authors say FaceReader's own errors do not match the participants' difficulties, so it is not an objective answer key. Depression and ADHD symptoms were higher in the autistic group. When the authors adjusted for these, some effects weakened, though the group-by-ambiguity interaction held up. Nobody's eye movements were recorded, so the idea that people were using the answer options to narrow things down is an interpretation the study cannot confirm. Only faces were tested. Emotion recognition also runs through voices, which this study did not look at. The sample size was chosen for the wider project's brain-imaging analyses, not for this emotion task.

The only funding statement supplied in the text names the University Medical Center Freiburg (Universitatsklinikum Freiburg). No author conflict-of-interest declaration appears in the text provided here, and none is described.

Do not read the autistic group's lower accuracy as proof of a fixed deficit in reading other people's emotions. Both groups got more accurate as the task went on and the autistic group improved less, so where a person is tested in the task changes the score they get. The explanation that non-autistic adults used the answer options to work by elimination rests on a single significant interaction; the follow-up test inside that group was not significant (P = 0.123), and no measure of stimulus salience had a direct effect on accuracy.