Tracing Co-Rumination in Couples’ Conversations Through Language-Based Models: Understanding the Costs and Benefits of Co-Rumination at Scale

Discussing difficult experiences with a romantic partner can provide emotional support, strengthen intimacy, and help individuals make sense of stressful situations. Yet conversations intended to provide reassurance may sometimes sustain distress rather than alleviate it. When partners repeatedly revisit problems, dwell on negative emotions, or encourage continued discussion without moving toward greater understanding or reslution, supportive communication may develop into a pattern known as co-rumination.

In the MRI-funded project Tracing Co-Rumination in Couples’ Conversations Through Language-Based Models: Understanding the Costs and Benefits of Co-Rumination at Scale, Yunzhi Zheng (Florida State University) examines how co-rumination develops during conversations between romantic partners and how these interactions relate to emotional well-being and relationship quality. By combining behavioral observations with computational language analysis, the study seeks to develop a method for identifying co-ruminative communication as it unfolds between partners.

Co-rumination involves repeated discussion of problems, sustained attention to negative emotions, and continued speculation about the causes and consequences of distressing experiences. Unlike conversations focused primarily on problem-solving, co-rumination is characterized by reciprocal engagement in problem-focused discussion, often with both individuals encouraging the interaction to continue.

Previous research has suggested that co-rumination may involve competing emotional and relational consequences. Studies of friendships have associated these patterns with greater closeness and perceived support, but also with elevated symptoms of anxiety and depression. These findings raise an important question: how can conversations that strengthen interpersonal connection also be associated with increased psychological distress?

Within romantic relationships, this question may be particularly complex. Partners often serve as important sources of emotional support while sharing responsibilities, commitments, and ongoing experiences. The consequences of repeatedly discussing problems may therefore depend on the nature of the stressor, the responses of each partner, and the interactional patterns that develop during the conversation.

An important limitation of existing research concerns how co-rumination is measured. Studies have frequently relied on self-report questionnaires assessing individuals' general tendencies to engage in repetitive problem discussions. Although informative, these measures may not capture the reciprocal processes through which co-rumination develops during actual interactions. Partners may also interpret or remember the same conversation differently, making it difficult to distinguish individual perceptions from observable communication patterns.

To address this limitation, Zheng's project will develop and evaluate the Co-Rumination Dyadic Language Index (CORDLI), an automated approach designed to identify co-ruminative patterns in recorded couples' conversations. Using natural language processing, the researchers will examine linguistic features associated with repetitive problem discussion and the ways partners respond to and reinforce one another's contributions.

The study will recruit 40 romantic couples, comprising 80 individuals, who have been in their relationships for at least six months. During a laboratory session, each couple will participate in two recorded conversations about stressful experiences outside their relationship. In one discussion, the couple will address a concern identified by one partner; in the other, they will discuss a concern identified by the second partner. Each conversation will last approximately eight minutes.

Before the discussions, participants will complete questionnaires assessing relationship quality, emotional well-being, and their general tendencies toward co-rumination. Following each conversation, both partners will independently report their emotional experiences and perceptions of the interaction, including stress, depressed mood, perceived understanding, progress toward problem resolution, relationship satisfaction, and feelings of being understood, validated, and cared for.

Participants will also review recordings of their conversations in 30-second segments, evaluating the extent to which they and their partners engaged in specific co-ruminative behaviors. These assessments will provide a more detailed account of participants' perceptions as the interactions unfolded.

The researchers will then analyze conversation transcripts using computational language models. CORDLI will focus on four dimensions of co-rumination: mutual encouragement of problem-focused discussion, repetitive conversational patterns, continued speculation about problems, and sustained negative emotional expression.

By examining these features across both partners' contributions, the model aims to capture co-rumination as a dyadic process rather than a behavior attributable to one individual. This distinction is central to the study's systemic perspective: a conversational pattern may be maintained not only by what one partner says, but also by how the other responds.

A principal objective is to determine whether the automated measurements correspond with participants' assessments of their own conversations. The researchers will also examine whether CORDLI predicts emotional and relational outcomes following each discussion and whether it provides information beyond conventional self-report measures.

This approach may help distinguish conversations that support emotional understanding and connection from those associated with prolonged distress. Two couples may discuss similarly difficult experiences, yet their interactions may differ in the extent to which partners repeatedly revisit negative emotions, encourage further speculation, or move toward a clearer understanding of the problem.

The project's methodological contribution also has implications for relationship research more broadly. Traditional observational studies often require trained researchers to review and code recorded interactions manually, a process that can be time-intensive and difficult to extend to larger samples. If successfully validated, CORDLI could provide a more scalable approach to examining couples' communication while retaining attention to the reciprocal nature of their interactions.

Although the study focuses on measurement development rather than clinical intervention, its findings may inform future research on emotional regulation, couple communication, and relationship functioning. A more precise understanding of co-rumination could help clarify when repeated discussion of stressful experiences is associated with relational closeness, psychological distress, or both.

By integrating computational methods with the study of interpersonal processes, this MRI-funded project contributes to an interactional understanding of how couples manage distress together. Rather than treating co-rumination as an individual communication style, the research examines how patterns of emotional expression and reciprocal responding emerge within relationships—and how these shared processes may be associated with both the benefits and costs of seeking support from a romantic partner.