Demand–withdraw: the conflict loop that drives couples apart

In the demand–withdraw pattern, one partner (the demander) raises a problem, asks for change, criticises or presses for discussion, while the other (the withdrawer) becomes defensive, goes quiet, changes the subject or physically leaves. The crucial feature is that the two moves feed each other: pressure makes withdrawal more likely, and withdrawal makes pressure more likely.

Andrew Christensen and Christopher Heavey described and measured the pattern in a 1990 study of married couples. Since then it has become one of the best-documented interaction patterns in relationship research.

A 2014 meta-analysis by Paul Schrodt, Paul Witt and Jenna Shimkowski pooled 74 studies with 14,255 participants. The headline findings:

• Demand–withdraw is moderately associated with poorer outcomes overall — lower relationship satisfaction, poorer communication, and to a smaller degree poorer individual wellbeing.
• The association is similar whichever partner demands: woman-demand/man-withdraw and man-demand/woman-withdraw showed almost identical effect sizes.
• The link is stronger in distressed couples than in community samples, which fits the idea that the pattern both reflects and deepens distress.

Christensen and Heavey's original work, replicated in later studies, showed that the pattern follows who wants change. When the topic was something the wife wanted changed, wife-demand/husband-withdraw was common; when it was something the husband wanted changed, the roles were much more even. The gendered version of the pattern is partly a reflection of who, in a given couple and culture, more often wants something to be different.

Several mechanisms keep the loop running.

Physiology. John Gottman and Robert Levenson's observational research found that during conflict some partners — more often men in their samples — became physiologically flooded: heart rate up, attention narrowed, capacity for listening gone. Withdrawal ("stonewalling") in that state is self-protection, not strategy. But to the partner it reads as a wall.

Attachment. The pattern maps closely onto the anxious–avoidant pairing. A partner higher in attachment anxiety experiences silence as abandonment and intensifies; a partner higher in attachment avoidance experiences intensity as engulfment and deactivates. See Attachment styles explained.

Reinforcement. Each move works in the short term. The withdrawer's exit ends the unpleasant conversation. The demander's pressure occasionally produces a response. Both strategies are intermittently rewarded, which is exactly the schedule that makes behaviour persistent.

Narrative. Over months, each partner builds a story — "they never want to talk" / "nothing is ever enough" — that filters new events. The story then runs ahead of the evidence.

The exit points are at steps 1 and 3 — how the issue is raised, and what each partner does with the first sign of the other's move.

Because the loop is maintained by both partners, either partner can change it by changing their own move. This is good news: you do not need the other person's cooperation to start.

For the demander

1. Soften the start. Gottman's newlywed research found that the first few minutes of a conversation predicted how it ended. "I'd like to talk about the budget — when's a good time today?" is a start the other person can accept. "We need to talk about how you spend money" is not.
2. Ask for a time, not an answer. Many withdrawers can engage when they are not ambushed. "Can we do twenty minutes after dinner?" respects that.
3. Notice the first sign of withdrawal and name it gently. "I think I'm coming on strong — I'm not trying to attack you, I just feel worried about this."

For the withdrawer

1. Say what is happening instead of just leaving. "I'm getting overwhelmed and I'm going to stop making sense. I need twenty minutes." The content is the same; the effect is completely different.
2. Set a return time and keep it. Withdrawal without a return reads as abandonment. "I'll come back at nine" is a pause, not an exit.
3. Use the break to calm down, not to build a case. Flooding takes at least twenty minutes to subside, and it does not subside while you are rehearsing the argument.

For both

• Agree, outside conflict, on a time-out signal and a maximum delay before resuming.
• Practise the other person's role for one minute: the demander says what they imagine the silence is protecting; the withdrawer says what they imagine the pressure is asking for.

Demand–withdraw is a dance between two people with roughly equal power. Some things that look like it are not:

• Withdrawal used as punishment. Days of silence intended to make the other person comply is control, not regulation. The tell is that it ends when you give in.
• Demand that is actually monitoring. Repeated interrogation about whereabouts, friends or money is not a bid for connection.
• Fear. If one partner withdraws because the other's "demand" has previously ended in intimidation or violence, this is not a communication pattern and should not be treated as one.

• The Conflict Avoider — the withdrawing pattern as a profile, with what helps.
• Attachment styles explained — the dimensions beneath the loop.
• The Four Horsemen — criticism, defensiveness, contempt and stonewalling, and what the research actually shows.
• Staying steady in a difficult conversation — the regulation skills that make the time-out work.