The Effect of Stress on Semantic Memory Retrieval: A Multiverse Analysis

Abstract

Stress is often associated with negative consequences, and this also applies in the context of memory retrieval. However, Smith, Hughes, et al. (2019) proposed that this relationship only holds for information stored in episodic memory, because it relies on the hippocampus. In contrast, conceptual knowledge is stored in semantic memory, which is associated with neocortical and striatal brain regions that are upregulated during stress. Indeed, Smith, Hughes, et al. (2019) found that people experiencing acute psychosocial stress performed better on a subsequent trivia questionnaire compared to a control group. Moreover, performance correlated positively with cortisol reactivity. These findings are important, novel, and perhaps somewhat surprising, hence it is important to establish their generalizability. The latter is accomplished in the present study through a multiverse analysis. The results showed that the effect is relatively robust to variations in the scoring rules for the trivia questionnaire, the type of statistical model being used, and the inclusion versus exclusion of gender as a covariate in the analyses. However, we obtained mostly null effects when using the change in psychological stress levels as a predictor variable, and/or when only considering questions that were actually answered. The latter finding in particular is worrisome as it might point to alternative explanations. That is, stress might improve performance, because participants are more engaged with the task or are more prone to guess, regardless of its (potential) impact on semantic memory retrieval. Hence, it is premature to conclude that stress enhances semantic memory.