Background and aims: This paper aims to have both a theoretical and a methodological contribution on conversational disfluency. Theoretically, reductions of self-esteem among people who stutter (PWS) are often explained by individual negative cognitions or emotions of the PWS or their conversation partners. We propose that the flow of a conversation can be seen as a representation of the relationship between speakers, and that by disrupting this flow, a stutter may directly threaten this relationship, and in turn, affect self-esteem. Methodologically, we present a new, event-contingent, method that assesses fluctuations in self-esteem over time and thereby allows one to assess the predictive value of specific conversational experiences. Method: Our focus is not on stable individual factors, but we expect that fluctuations within the flow of everyday conversations predict changes in self-esteem. We studied these fluctuations with an experience sampling method which prompted participants to answer a brief survey after each of ten conversations in a two-week period. Self-identified PWS (n = 58) and people who do not stutter (PWNS; n = 53) reported on their individual negative cognitions and emotions and experienced relational quality and state self-esteem in more than 1000 conversations. By assessing fluctuations in self-esteem over time, this method allows one to assess the predictive value of specific conversational experiences. Results: Correlational evidence demonstrates that flow disruptions are associated with temporary reductions in self-esteem. This association is mediated by increased individual negative cognitions and emotions, as well as threatened social relationships. This appeared to be true for both PWS and PWNS, although PWS experienced on average less conversational flow and lower state self-esteem. On average, PWS did not experience lower relational quality than PWNS. Conclusion: An event-contingent recording method is a useful way to assess momentary fluctuations in self-esteem. Findings are consistent with the notion that people monitor their relationship by attending to fluctuations in conversational flow; whereas a smooth conversational flow indicates a strong relationship, disruptions of flow (e.g., as caused by a stutter or other factors) signal a threat to the relationship.