Abstract Emotional events are remembered better than neutral ones. While many human neuroimaging studies have identified brain regions involved, relatively few—and typically small—studies have disentangled how arousal and valence shape the neural substrates of this enhancement. We leveraged a large single-centre fMRI sample (n=1006) in which healthy young adults viewed negative, neutral, and positive pictures during scanning followed by an unannounced free-recall test. Using whole-brain subsequent-memory analyses (PFWE 0.05), we contrasted successful encoding of emotional (negative + positive) vs neutral items, then tested valence-specific effects (successful encoding: negative neutral; positive neutral), and finally controlled for subjective arousal via serial parametric modulation. Behaviourally, recall was higher for emotional than neutral pictures. Consistent with prior meta-analytic evidence, emotional neutral successful encoding engaged occipito-temporal visual cortex, anterior cingulate, insula, and amygdala. Additionally, we observed an extensive temporoparietal network, while hippocampal/parahippocampal activations were absent. After controlling for arousal, amygdala and insula effects were no longer significant, indicating these regions were sensitive to arousal rather than valence. Overlap of negative- and positive-valence enhancement localised primarily to the occipito-temporal cortex. Negative-specific enhancement recruited the lateral occipital/fusiform and bilateral supramarginal regions; positive-specific enhancement involved the rostral/caudal anterior cingulate, superior frontal, and parietal cortex, as well as the precuneus. Successful neutral encoding preferentially engaged frontoparietal control regions and bilateral lingual/parahippocampal cortex. Together, these findings dissociate valence-dependent from arousal-dependent mechanisms and reveal both partially overlapping and distinct networks for negative and positive memory enhancement, refining neurocognitive models of emotional memory encoding.
Amini et al. (Thu,) studied this question.