Momentary ratings of affective states with a pair of strict antonyms (“happy” vs. “sad”) were studied with an experience-sampling method in a group of 110 participants during 14 consecutive days at 7 randomly determined occasions per day. Before and after the experimental session participants also retrospectively rated how happy or sad they had been during the previous 2 weeks. Multilevel analysis showed that, at the level of single measurement trials, the momentary ratings of happiness and sadness were moderately negatively correlated (r = –.32, p < .001). A between-subject correlation of the two antonyms, however, was in a positive direction (r = .13, p = .123). Participants experienced mixed feelings during a considerable number of measurement trials, whereas the tendency to feel mixed emotions was predicted by all Big Five personality traits except Agreeableness. A configural frequency analysis (CFA) demonstrated that, although there was no strict bipolarity between momentary ratings of happiness and sadness, they were nevertheless used in an exclusive manner in many occasions. (PsycINFO Database Record (c) 2012 APA, all rights reserved)