Experiential avoidance (EA) is an emotion-regulation strategy used to control or avoid unpleasant internal experiences. Experimental studies, however, have shown that EA is associated with an ironic increase in unpleasant experiences. While single manipulation laboratory experiments can demonstrate the immediate ironic detrimental effects of EA, a different methodology is needed to establish how such ironic processes unfold over time in the natural environment. The current study uses a longitudinal design and daily-diary methodology to examine daily associations between EA and negative affect (NA) over a three-week period among college-students who initially reported high levels of psychological distress. A daily measure of state EA based on several avoidant behaviors (thought suppression, emotion suppression, distraction, reflective pondering, and lack of experiential acceptance) was developed for this study. Before and after making daily web-based reports of EA and NA for 21 consecutive days, participants completed a standardized checklist of psychological symptoms, with pre-post change scores on this measure serving as an index of symptomatic improvement. Multilevel modeling analyses showed that, as predicted, symptomatic improvement was associated with decreasing trajectories of EA and NA during the 21-day study period. More symptomatic improvement was associated with weakening (decoupling) of same-day EA-NA links over time. Contrary to predictions, same-day and one-day lagged associations between NA and EA were not associated with symptomatic change. Additional multilevel analyses showed that symptomatic worsening was associated with more daily EA, over and above what was accounted for by daily NA. Likewise, traditional between-person regression analyses showed that overall mean levels of daily EA (aggregated across days) predicted symptomatic worsening, even after statistically accounting for mean levels of daily NA. The results of this study provide partial support for the hypothesis that EA and NA are related to each other in an ironic positive feedback loop that unfolds over time and that symptomatic improvement may involve a process by which EA and NA both decrease and decouple from each other over time. These findings emphasize the importance of using methodologies that track the relationship between EA and its consequences over time using within-person analyses, rather than solely relying on between-person designs.