Psychologist John Gottman spent decades studying how couples fight, tracking their conversations in a lab, sometimes for years afterward, to see which patterns predicted separation and which didn’t. Out of that research came what Gottman called the Four Horsemen, four communication habits that, when they became a couple’s default response to conflict, reliably predicted divorce: criticism, contempt, defensiveness, and stonewalling. Stonewalling is the one that involves walking away. It shows up as one partner shutting down entirely during an argument, going quiet, physically leaving the room, or simply refusing to engage, typically without explaining why or when they’ll be back. In Gottman’s research, this pattern showed up disproportionately in men, and its presence in a relationship’s conflict style was one of the more reliable warning signs his team identified.
What makes stonewalling so damaging isn’t simply that someone stopped talking. It’s the combination of withdrawal and ambiguity. The remaining partner is left in an argument with no partner in it, unsure whether the other person is coming back, still angry, or has simply checked out of the relationship altogether. A 2026 systematic review published in Frontiers in Psychology, pulling together 15 peer-reviewed studies on the closely related pattern of the silent treatment, found this same dynamic at work more broadly: the harm to the relationship was consistently worse when the person receiving the silence had no idea why it was happening or when it would end. Both the person giving the silent treatment and the person receiving it showed worse psychological health where the pattern became chronic, and the review’s authors describe silence used this way, unexplained and often as leverage or punishment, as functioning more like a slow erosion of trust than a pause.
Here’s where the research gets more interesting than a simple “never walk away” rule would suggest. Gottman’s own work on stonewalling grew out of an earlier finding, published with his longtime collaborator Robert Levenson, that during heated arguments, a person’s heart rate can climb past roughly 100 beats per minute, a state researchers call flooding. Once someone is flooded, the physiological systems needed for empathetic listening and flexible problem-solving are, functionally, offline. Trying to keep talking through flooding tends to make an argument worse, not better, because at least one person in it has temporarily lost the capacity to communicate well. Gottman’s own recommendation, based on this research, isn’t to power through. It’s to take a break, but a specific kind: one agreed on in advance, signaled clearly in the moment, lasting at least twenty minutes to allow the body to physiologically settle, and followed by a return to the conversation once both people have calmed down.
The distinction between that kind of break and stonewalling comes down almost entirely to communication and intent, not to whether someone physically leaves. A person who says “I’m getting overwhelmed, I need twenty minutes, I’ll come back and we’ll keep talking” and then does exactly that is doing something structurally very different from a person who goes silent mid-argument and gives no indication of whether or when they’ll re-engage, even if both scenarios look similar from the outside for the first few minutes. The first is a documented, evidence-based way to keep a nervous system from derailing a hard conversation. The second is the pattern Gottman’s research found among the strongest predictors that a relationship was heading toward divorce.
It’s worth being honest about where the certainty in this research runs out. Gottman’s studies, however influential, describe patterns and predictive associations built mostly from heterosexual married couples observed over specific windows of time, and predicting divorce statistically isn’t the same as proving that stonewalling itself, on its own, causes a relationship to fail rather than reflecting deeper problems that produce both the stonewalling and the eventual breakup. The silent treatment review similarly draws on a modest, though growing, body of research rather than one single definitive study. Neither literature claims that every instance of a partner going quiet is destructive, or that a well-executed pause guarantees a healthy relationship. What both bodies of research converge on is a narrower, more useful claim: walking away without a word functions very differently depending on whether the other person is left wondering what just happened, or already knows exactly what’s happening and when the conversation will pick back up.
The practical version of that distinction is less about the walking away itself and more about what happens right before it. A sentence, even a short one, seems to be doing most of the real work: not necessarily an explanation of the whole argument, but enough information that the person left behind isn’t also left guessing.