The Annex: 48 autistic adults, 90 others, one migraine treatment, and a sample size that depends on which page you read

The Annex reads the method behind the headline. Today’s: a brief report by Hidehito Honda, Hirokazu Kumazaki and Kazuhiro Ueda, published on 12 November in Psychonomic Bulletin & Review, on how people with and without autism spectrum disorder (ASD) turn probability words such as “likely” or “doubtful” into numbers, and into decisions¹.

The design. Participants read a hypothetical scenario: a friend is considering an expensive migraine treatment, and a doctor describes its chance of success either as a number (“a 30% chance”) or in words that lean positive (“it is possible”) or negative (“it is unlikely to succeed”)². They said how strongly they would recommend it, and separately translated each verbal phrase into a number^1,2. A hierarchical Bayesian model, followed by cluster analysis, sorted people by how much the wording’s direction and their own numerical translation drove their choice¹.

The headcount, read closely. The university release says 48 people with ASD and 90 without². The abstract is more precise: Experiment 1 had 48 participants with ASD and 41 without “in the final analyses”, and Experiment 2 an independent sample of 49 without ASD¹. Desk arithmetic: 41 + 49 = 90, so the totals agree, but the direct comparison inside one experiment is 48 against 41. That is a small study, and the journal labels it a brief report¹.

The result. Three profiles^1,2. One, mostly people with ASD, was swayed more by positive or negative framing and less by the numbers they had inferred. A second, mostly people without ASD, leaned on the inferred numbers. A third, mixed, weighed both². Participants with ASD also gave more similar numbers to different phrases¹. The authors are careful: participants varied in age and intellectual ability, so the differences cannot be attributed to ASD alone, and what places someone in a cluster is still unknown².

For you. The finding that travels is not about autism: “even subtle differences in wording can influence people’s decisions,” says Ueda². When a clinician tells you something is “unlikely”, it is fair, and not rude, to ask what that means in numbers, and to say back what you understood.