A group of 78 young employees in service and non-service professions reported 848 task related interactions at work over 1 week using a variant of the Rochester Interaction Record which measured emotion work requirements, emotional dissonance, and deviance. Multi-level analyses showed that dissonance was more likely in interactions with customers, whereas deviance, that is, the violation of display rules by acting out one's felt emotion, was more likely in co-worker interactions. Well-being in the interaction was lower (a) for interactions with emotion work requirements, (b) for dissonance, even after controlling for felt negative emotions, and (c) for deviance. Negative emotion displayed partially mediated the relationship between deviance and well-being. Regarding the relationship of more stable job related attitudes, psychosomatic complaints, and aggregated scores of social interactions, fewer effects were found than in questionnaire studies, which may be due to the fact that only interactions that lasted at least 10 minutes were assessed, as is customary in research with this instrument. Among the effects found, however, many involved proportions rather than frequency of interactions, which raises the possibility of balancing and legitimizing effects of non-stressful interactions.