Search

Almaatouq et al. argue that an "integrative experiment design" approach can help generating cumulative empirical and theoretical knowledge. Here, we discuss the novelty of their approach and scrutinize its promises and pitfalls. We argue that setting up a "design space" may turn out to be theoretically uninformative, inefficient, and even impossible. Designing truly diagnostic experiments provides a better alternative.

Consistent with sampling theories in judgment and decision research, impression judgments depend on the number of traits drawn randomly from a population of target person traits in distinct ways. When sample size is determined externally by the experimenter, the sensitivity of resulting impression judgments to the prevailing (positive or negative) valence increases with the number of traits. In contrast, sensitivity is negatively related to sample size (more extreme judgments for smaller samples) when sampling is self-truncated. Building on previous findings by Prager et al. (2018), two new experiments corroborate the judgment pattern for self-truncated sampling and elaborate on the distinction of Brunswikian sampling (of stimuli in the environment) and Thurstonian sampling (of states within the judge's mind). Thurstonian sampling effects were evident in depolarized (regressive) judgments by yoked control participants provided with exactly the same trait samples as original judges, who could truncate sampling when they felt ready for a judgment. Experiment 1 included two kinds of yoked controls, receiving trait samples truncated in a previous stage either by themselves or by other judges, distinguishing between temporal and interpersonal sources of Thurstonian sampling variance. As expected, self-yoking yielded less regressive shrinkage than other-yoking. Experiment 2 provided convergent results with yoked controls manipulated within participants, dealing with higher dispersion of impressions on self-truncated samples (Thurstone, 1927). Across both experiments, individual impression judgments were highly predictable from theoretically meaningful parameters: expected valence in the population, sampling error, sample size, and different indices of trait diagnosticity. (PsycInfo Database Record (c) 2021 APA, all rights reserved).