According to an influential theory in cultural evolution, within-group similarity of culture is explained by a human conformist-bias, which is a hypothesized evolved predisposition to preferentially follow a member of the majority when acquiring ideas and behaviours. However, this notion has little support from social psychological research. In fact, a major theory in social psychology (Latan and Wolf (1981) argues for what is in effect a nonconformist-bias: by analogy to standard psychophysics they predict minority sources of influence to have relatively greater impact than majority sources. Here we present a new mathematical model and an experiment on social influence, both specifically designed to test these competing predictions. The results are in line with nonconformism. Finally, we discuss within-group similarity and suggest that it is not a general phenomenon but must be studied trait by trait.