Reading Movement in Autism Gets Harder as Inference Grows

What the research found

Attach a small light to each of a walker’s major joints, film the person moving in a dark room, and show a viewer nothing but the lights. Roughly a dozen bright dots drift against black. Frozen on one frame they read as a meaningless scatter, but set them in motion and almost anyone sees a person walking, within a fraction of a second, and can often say whether the walker is heavy or light, hurried or tired. Gunnar Johansson described the effect in 1973, and the point-light display has served as the standard laboratory instrument for biological motion since.

Stripping away colour, outline and face isolates movement itself, which is why the display became the preferred tool for asking whether autistic people perceive human motion differently. Alessandra Federici, Valentina Parma and colleagues at the IRCCS E. Medea institute in Bosisio Parini gathered that literature into a meta-analysis published in Scientific Reports in 2020, searching PubMed and PsycInfo for every study comparing an autistic and a non-autistic group on a point-light task. Thirty of the 146 articles retrieved qualified.

Pooled across 29 of them, autistic participants performed less well, with a Cohen’s d of 0.60 and a confidence interval from 0.36 to 0.85. Cohen’s d states a group difference in standard deviations, and 0.2 counts as small by convention, 0.5 as medium, 0.8 and above as large. Correcting for studies estimated to be missing reduced the figure to 0.47, and neither age nor cognitive level moderated it.

The studies disagreed considerably, 76 percent of the variation exceeding what sampling error would produce, and the authors pursued that disagreement by sorting the tasks according to what each asked of the viewer. Tasks recording only whether a viewer looks longer at a walker than at a scrambled version yielded a d of 0.55 across seven studies. Tasks asking for an explicit judgement about the display, walker or scramble, leftward or rightward, yielded 0.40 across sixteen. Tasks using the walker to reach something further, the emotion in a gait or the intention behind an action, yielded 0.98 across eleven, and only that last gap survived correction for multiple comparisons.

A second analysis compared the foils. Where the control display kept each dot’s timing and moved it to a new position, the two groups did not reliably differ, at 0.38 with an interval crossing zero across seven studies. Where the dots stayed in position and their motion was pushed out of phase, the difference was reliable, at 0.67 across six. Six and seven studies are thin ground, though the direction agrees with wider work on temporal processing in autism. The displays that separated the groups were the ones disturbed in time.

What it means for you and your child

Nothing here overturns the standard account. A difference in point-light performance was present at every level of task tested, and the meta-analysis specifies where it concentrates: at the point where a moving figure is being used to work out something that the figure does not directly show.

That specification comes with a boundary the authors keep in view. These are laboratory judgements about dots on a screen, made under timed conditions against a foil designed to be difficult. Reading a friend’s posture across a playground lies some distance beyond anything the review measured, and any bearing of one on the other is a step the reader takes without the evidence.

Whether the difficulty belongs to human motion in particular is also open. Global motion perception, the ability to extract a common direction from a field of moving dots, is a separate paradigm with its own developmental history, and only two of the included studies used it as a control. Among the five studies that tested non-biological point-light stimuli, no reliable group difference appeared, though the comparison stimuli were mostly scrambled walkers. Specificity to the human figure remains untested.

For a parent, the practical value is in the distinction between two questions that look alike from the outside. Whether a child can see that a figure is walking, and whether a child can read what that walking means, are separated by an interval of processing, and the evidence suggests they can come apart. A child who is slower with the second may be entirely intact on the first, which matters when an adult is deciding whether a child failed to notice something or noticed it and had not yet worked out what it meant. Those two situations call for different responses and are easily confused at a distance.

The flat age curve in this analysis does not describe a static picture, since the authors suspect compensation accumulates over years and point to work in which autistic adults recognised actions accurately while the underlying perceptual adaptation remained weak. That would make the developmental horizon a long one. What a child makes of movement outside a laboratory has not been measured, and nothing in this analysis sets a schedule for it.