They Learned the New Rule, Then Drifted Back

What the research found

For a long time, a familiar idea has followed autism around: that people on the spectrum struggle to shift gears, to move from one rule or routine to another. This study, published in the Journal of Autism and Developmental Disorders in 2015, looked closely at that idea and found it was not quite right. The problem, it turned out, was not switching. It was staying switched.

The researchers tested 115 people: 60 with autism spectrum disorder and 55 typically-developing controls, matched for age, sex, and nonverbal reasoning ability. Participants ranged in age from six to the mid-forties, with an average age of around fifteen in both groups. Everyone completed the Penn Conditional Exclusion Test, a touchscreen sorting task. You are shown objects and must work out the hidden rule for sorting them, first by shape, then by size, then by the thickness of the lines. The rule changes without warning. Feedback tells you when you are wrong. You have to notice, adjust, and carry on.

The task lets researchers separate two very different kinds of mistake. A perseverative error is failing to let go of an old rule, sticking with the previous way of sorting even after the feedback says it no longer works. A regressive error is different. It happens after you have already found the new rule and used it correctly, then slide back to the old one anyway. You knew the answer. You drifted.

On perseverative errors, the two groups were indistinguishable. The statistic is almost comically flat: F(1, 101) = 0.00, p = 0.99. People with autism let go of the old rule exactly as readily as everyone else. But on regressive errors, a gap opened. The autism group made an average of 11.29 such errors (SD = 12.24), against 6.25 (SD = 9.92) in the control group, a difference that reached statistical significance, F(1, 101) = 5.29, p = 0.02. In the authors’ own summary: individuals with autism were able to initially shift sets, but they had difficulty maintaining new response sets.

The study then asked whether this laboratory finding connected to anything in real life. It did, modestly. The researchers used the Autism Diagnostic Interview-Revised to measure insistence on sameness: the difficulty tolerating changes to routine and environment that many families recognise well. The tendency to regress to an old rule was correlated with more severe insistence on sameness, r = 0.33, p = 0.03. A correlation of 0.33 is real but small; it means the two things travel loosely together, not that one is a reliable predictor of the other.

What this means for you and your child

The most useful thing here is a shift in how you might picture a moment of difficulty. When a child seems stuck on an old routine, the intuitive reading is that they cannot see the new option or cannot bring themselves to abandon the old one. This study gently pushes against that. In the lab, at least, the children and adults tested could see the new rule and could adopt it. What they struggled with was holding it steady once the initial effort of switching had passed.

That distinction has a practical flavour. It suggests that transitions may not be the whole story, and that the harder work is sometimes sustaining a new pattern after the transition is over. A child who manages a change well in the first few minutes may still drift back toward the familiar without any obvious trigger. If you have seen that and found it puzzling, this finding offers a plain description of it: the initial shift was fine; maintaining it was the cost.

It is worth being careful about the correlation with insistence on sameness. The link was found, and it was statistically significant, but at r = 0.33 it is loose. It tells us that a laboratory measure of set maintenance and a parent-report measure of rigidity share some common ground. It does not tell us that the sorting task explains a child’s routines, nor that improving one would improve the other. The authors themselves were measured on this point. They wrote that a tendency to revert to old preferences may contribute to repetitive behaviours. May contribute is honest language, and it is worth keeping.

This is also a single, cross-sectional study. Everyone was tested once, so it captures a snapshot rather than a developing trajectory. The authors flag their own limits candidly: a small number of participants with autism could not learn the initial rules and warrant separate study; the youngest children may already have passed through relevant developmental change; and two different IQ tests were used across participants, adding a little noise. None of this undoes the central result, but it does mean the finding is a well-drawn observation, not a settled mechanism.

For a parent, the honest takeaway is narrow and genuinely useful. Difficulty may live less in the act of changing and more in the effort of not changing back. Support that anticipates the second problem, the quiet slide toward the familiar rather than the visible resistance at the moment of transition, is working with the grain of what this study actually found.