Dysfunctional expectations and interpersonal problems are associated with depression, so we measured expected emotions towards interaction partners and compared them with actual emotions. We hypothesized that, between persons, individuals with higher subclinical depression would display stronger, more stable, and less accurate negative expectations. Within persons, we hypothesized that momentary negative expectations would predict subsequent negative affect. Fifty-three students completed 6 days of Experience Sampling, consisting of one morning expectation-assessment (9 am), three assessments on actual interpersonal emotions (1 pm, 5 pm, 9 pm), and six random affect-assessments. We regressed expected emotions, experienced emotions, expectation fluctuations, and expectation violations on subclinical depression. Using mixed model analyses, we further examined whether negative expectations preceded negative affect, and whether expectation violations preceded adjustments of expectations. Higher subclinical depression predicted more negative expectations. Within persons, worse-than-expected interpersonal interactions preceded negative affect whereas better-than-expected interactions preceded reductions of negative expectations. Despite problems with skewed data, our approach appears well-suited to examine interpersonal expectations in vivo.