Testing domain-specific advantages after adversity in four open datasets: bias, not sensitivity, in anger recognition

The hidden talents framework proposes that people from harsh environments may perform relatively better on tasks that match the problems of those environments. Much of the evidence comes from emotion recognition, where accuracy alone cannot separate detecting a signal from reporting it more readily. I preregistered a synthesis of four open datasets (N = 1,014) covering facial emotion recognition, memory for dominance relations, and executive functions with ecological stimuli. In each, I computed a within-person specificity index: the association between adversity and performance in the environment-relevant condition relative to a matched control. The pooled index was positive but uncertain (β = 0.13, 95% CI −0.02 to 0.29), so the preregistered test was inconclusive. One small study drove the estimate; without it, the other datasets were equivalent to no meaningful effect (β = 0.05, equivalence test p = .037). In that study, children exposed to adversity did not detect anger better but labelled other expressions as anger more often, a pattern an independent sample also showed. In these four test cases, adversity was not linked to the predicted relative advantage, and the apparent anger advantage reflected a response bias. These results do not test the broader idea in other domains. Preregistration: https://osf.io/fd8n7/ · Materials, code and results: https://osf.io/p9yrq/