This paper concerns the differential effects of use of ‘thematic’, rather than abstract (symbolic), content on the ‘Wason selection task’. An effect of thematic content has been reported several times, originally by Wason & Shapiro (1971), but Manktelow & Evans (1979) report five experiments that failed to obtain an effect and argue against the validity of the content effect in general. However, it is pointed out that four of their experiments use a novel content and that there are previous reports of non-effective content in the literature. It is concluded that a particular type of thematic content cannot be treated as a ‘random sample’ of thematic content in general, and that their argument thus rests upon their failure to produce an effect of the specific (’towns and transport’) content used by Wason and Shapiro. This produces an essentially empirical controversy which is settled in two ways. First, Manktelow and Evans’s criticisms of two previous replications are shown to be misleading. Second, an experiment is reported that exactly copies the Wason and Shapiro design and produces a significant effect of content. It is concluded that the Manktelow and Evans result most probably was due to a Type 2 error. Finally, a hypothesis is developed to account for both the effect of certain types of content and the lack of effect of others, including that used in Manktelow and Evans’s first four experiments.