Figure 2. Mental state transition probabilities predict similarity ratings.

In Studies 1a-g, participants observed a sequence of mental states determined by a predefined transition matrix (heatmaps), and then rated the conceptual similarity between each pair of states. Mental state concepts were significantly predicted by the observed transition probabilities (line graphs) in all studies. Multilevel bootstrapping was used to estimate a meta-analytic effect size and confidence interval (circle radius reflects sample size). The transition probability matrices represent the ideal from which each participant’s sequence was independently sampled; the transitions probabilities each participant observed differed due to random sampling.