The impact of semantic distance and induced stress on analogical reasoning: a neurocomputational account

In a study of reasoning with four-term verbal analogy problems, we explored the relationship between the effects of an acute, mild stressor and the complexity of the reasoning process. Participants judged whether analogy problems in the form A:B :: C:D were valid or invalid, on the basis of whether the relation in the A:B term matched that in the C:D term. Half of the problems contained a C:D pair semantically near the A:B pair (e.g., NOSE:SCENT :: TONGUE:TASTE), and the other half contained ones semantically far from A:B (e.g., NOSE:SCENT :: ANTENNA:SIGNAL). After an initial block without stress, participants were randomly assigned to count backward by 13 s from 1,000 while being told to go faster, or to count forward by 1 s from 0. The stress-induced participants reported a significant increase in state anxiety as compared to controls immediately after the mental arithmetic task. Stressed participants performed less accurately (as measured by d') on both near and far analogy problems, mainly due to an increase in false alarms. We were able to model the influence of semantic distance using the "learning and inference with schemas and analogies" (LISA) model. Our findings indicated that even mild increases in stress impair analogical reasoning. However, the decrement does not seem to directly involve the integration of relations, but rather is due to a shift in decision strategy: Under stress, people show an increased tendency to endorse analogies as valid when the terms in the individual pairs are semantically related to each other, even if the overall analogical relationship is not valid.