BackgroundPrevious research has shown evidence on the role of affective states for physical activity behavior. However, there is a lack of research investigating the interplay between affective states, intentions, and exercise behavior, especially with respect to maintaining regular exercise over time. The study aimed to investigate whether post-exercise affective states and changes in affect during exercise (i) are related to exercise intentions; (ii) moderate the relationship between intention and subsequent exercise behavior, and (iii) directly predict future exercise.MethodsParticipants from weekly voluntary sports and gym classes at two universities were recruited. For 13 weeks, 268 individuals’ (Mage = 24.5 years, SD = 5.6, 90% students, 67.4% female) class attendance was documented on a weekly basis. Before and immediately after training, participants self-reported affective states, including affective valence (Feeling Scale) and perceived arousal (Felt Arousal Scale). Participants also reported their intention to re-attend the class the following week. Mixed-effect linear models and Cox proportional hazard models were used to examine the relationships between affective states, change in affective states, re-attendance intentions, and class re-attention.ResultsAffective valence at the end of training was significantly positively associated with the intention to re-attend the class on the within-person level (β = 0.880, p < 0.001) as well as the between-person level (β = 0.831, p < 0.001), while higher increases of valence during class were related to smaller intention. For class re-attendance, significant effects of affective states were only found on the within-person level. A one-point increase on the valence scale increased the hazard ratio to re-attend by 8.4% (p < 0.05), but this effect was no longer meaningful after adjusting for intention. No moderation of the relationship between intention and subsequent class re-attendance was found.ConclusionThe results suggest that positive affective state immediately after exercise does not facilitate translation of intentions into subsequent exercise behavior (i.e., do not close the intention-behavior gap). Rather, affective valence was found to be an important predictor of exercise intentions but seemed indirectly related to behavior via intentions. Practitioners should plan exercise programs that allow for positive affective states especially at the end of a training.