fMRI replicability during emotional scene viewing: Functional regions and sample size

Recent findings have questioned the replicability of functional magnetic resonance imaging (fMRI) in the study of affective processing, reporting low replicability of emotional enhancement during a face-matching task. However, poor replicability may instead reflect a lack of emotional engagement for face matching. In the current study, replicability of emotional enhancement was tested in a large (N = 160) sample when emotional engagement was assessed during pleasant, neutral, and unpleasant picture viewing, which reliably engages affective reactions in both the brain and the body. Replicability was computed using a subsampling technique, in which random sets of subjects of different sample sizes (N = 20, 40, 60, 80) were selected from the entire dataset, and replicability of emotional enhancement for peaks, clusters, and voxels were averaged across 500 permutations for each sample size. Consistent with previous findings, fMRI replicability increased with increasing sample size. On the other hand, even with relatively small samples, fMRI replicability for peaks, clusters, and voxels during emotional, compared to neutral, scene viewing was good to excellent. Importantly, replicability varied in different brain regions, with excellent replicability at both the cluster and peak level with an N of 40, at the most conservative threshold (p < .001), in the amygdala and the visual cortex. The data argue against general recommendations regarding sample size in fMRI studies of emotion, suggesting instead that degree of replicability depends on successful emotional engagement in task-related brain regions.