Negative emotions influence EEG correlates of inference formation during analogical reasoning

An experimental and simulation study of the impact of emotional information on analogical reasoning

We investigated whether and how emotional information would affect analogical reasoning. We hypothesized that task-irrelevant emotional information would impair performance whereas task-relevant emotional information would enhance it. In Study 1, 233 undergraduates completed a novel version of the People Pieces Task (Emotional Faces People Task), an analogical reasoning task in which the task characters displayed emotional or neutral facial expressions (within-participants). The emotional faces were relevant or irrelevant to the task (between-participants). We simulated the behavioral results using the Learning and Inference with Schemas and Analogies (LISA) model of relational reasoning. LISA is a neurally plausible, symbolic-connectionist computational model of analogical reasoning. In comparison to neutral trials, participants were slower but more accurate on emotion-relevant trials, and were faster but less accurate on emotion-irrelevant trials. Simulations using the LISA model demonstrated that it is possible to account for the effects of emotional information on reasoning in terms of how emotional stimuli attract attention during a reasoning task. In Study 2, 255 undergraduates completed the Emotional Faces People Task at either a high- or low-working memory load. The high working memory load condition of Study 2 replicated the findings of Study 1, showing that participants were more accurate on emotion-relevant trials than on emotion-irrelevant trials; in Study 2, this increased accuracy could not be accounted for by a speed-accuracy tradeoff. The working memory manipulation influenced the manner in which the congruence (with the correct answer) of emotion-irrelevant emotion influenced performance. Simulations using the LISA model showed that manipulating the salience of emotion, the error penalty, as well as vigilance (which determines the likelihood that LISA will notice it has attended to an irrelevant relation), could reasonably reproduce the behavioral results of both low and high working memory load conditions of Study 2.