Preview

PDF, English Download (16MB) | Terms of use

Abstract

All three papers included in this thesis, rely on an exemplar trait-impression formation paradigm (Asch, 1946), but throughout each paper different theoretically important aspects of the sampling task are analyzed. Prager, Krueger and Fiedler (2018) elaborate a possible solution to the "less-is-more" debate, that is the question whether more information leads to more or less extreme judgments. The correlation between sample size and how amplified judgments are depends crucially on how the respective sample was stopped. When sample size was experimenter-determined (random), judgments tended to be more conservative for small than for large samples. In contrast, judgments on small samples were polarized for self-truncated sampling (i.e. when the judging participants can themselves decide on when to stop the sampling sequence). Prager and Fiedler (2021b) transferred the self-truncation principle to an inter-group context, where impression targets were groups rather than individuals. By assessing perceived within-group homogeneity in addition to the likeability judgment, we could demonstrate that perceived homogeneity is part of the self-truncation principle: Early truncated samples are not only more polarized but also more homogeneous than samples that are expanded further. Given that out-groups are associated with smaller information samples, these self-truncation effects might constitute a sufficient explanation of outgroup homogeneity and out-group polarization. In Prager et al. (2018) and Prager and Fiedler (2021a) we apply different versions of a yoked controls design: Whereas a primary participant engages in self-truncated sampling, a secondary yoked control receives exactly the same samples passively. This procedure results in regression: Small samples are perceived less polarized for the yoked control than for the primary self-truncating participant. This difference is an exclusive result of different cognitive processing of the sampling input, since the yoked controls design keeps the sampling input itself identical. All three papers analyze the impact of diagnosticity: Especially negative and extreme information is more diagnostic than positive and moderate input. Highly diagnostic input results in earlier truncation and more polarized judgment. The diagnosticity concept is further elaborated by considering multi-dimensional density in Prager and Fiedler (2021b).