One hundred and fifty‐five participants had to solve a set of 2–4–6 like reasoning problems (Wason, 1960), in which they were told which hypothesis a majority (or a minority) proposed, as well as which example was used for the test. In a 2 × 2 design, participants were also told that the problems allowed either one single correct answer or several possible answers. Results show that, when the source is a majority and the problem allows one single answer, most participants adopt the source's hypothesis and use confirmatory testing. On the contrary, it is when the source is a minority and the problem allows several answers that most participants give alternative hypotheses and use disconfir