Why Is It So Hard to End a Conversation?

There is a point in some conversations when everyone appears to agree that the conversation is finished.

Someone says, “Well, I should get going.” The other person agrees. People stand. Coats go on. Keys appear. A body rotates toward the door. The social evidence is overwhelming.

Then somebody says, “Oh, before you go…”

Ten minutes later, nobody has gone anywhere.

This is one of the stranger things about ordinary conversation. Starting one usually takes almost nothing. Someone says hello, asks a question, makes an observation, or complains about parking. Ending one can require a multi-stage procedure involving verbal signals, body positioning, recap statements, future plans, well-wishes, explanations, repeated goodbyes, and sometimes a final conversation conducted entirely in the doorway.

The easy explanation is that people are awkward or overly polite.

The research points to something more interesting.

Humans appear to have sophisticated ways of ending conversations. We use recognizable closing signals. We coordinate turns. We move our bodies toward departure. We are not simply standing there because civilization forgot to invent goodbye.

And yet research also suggests that conversations often do not end when the people in them privately want them to.

That creates a better question.

If we are so good at performing goodbyes, why are we so bad at getting the timing right?

A conversation does not end just because someone wants it to

A conversation is shared territory.

One person can begin leaving mentally. That does not automatically end the interaction. If the other person is still talking, introduces a new topic, asks a question, remembers something important, or responds in a way that invites another turn, the conversation still has work to do.

Foundational conversation-analysis research by Emanuel Schegloff and Harvey Sacks treated conversational closing as an organized interactional accomplishment. The important idea is simple even if the underlying analysis is not: conversations do not usually stop the way a light turns off. Participants have to arrive together at a point where another turn is no longer expected.

That helps explain why the word “goodbye” is not always the end.

A closing often begins before the final goodbye. People may summarize what was discussed, make plans to talk again, offer well-wishes, explain why they need to leave, or use familiar phrases that signal the interaction is moving toward completion.

The participants are not merely exchanging information. They are changing the status of the conversation from still happening to about to stop happening.

That transition takes coordination.

Dunce Dynamics interpretation: a goodbye is less like pressing an off switch and more like landing an airplane together. You can see the runway. Everyone understands the objective. Somehow there are still several procedures between here and actually being on the ground.

Your body can begin leaving before the conversation does

The closing process is not only verbal.

Research on embodied interaction has examined how walking away, changing orientation, and other physical movements can help make an ending publicly recognizable. Mathias Broth and Lorenza Mondada studied mobile interactions in which movement became part of the closing sequence. Sylvaine Tuncer found similar embodied practices in informal office encounters.

The precise settings matter. These studies do not prove that putting on your coat universally means “please stop talking now,” or that every step toward a door reveals somebody’s private desire to escape.

They do establish something narrower and useful: physical movement can participate in the organization of closing.

That matches a familiar piece of everyday theater.

First, you are sitting.

Then you are standing.

Then you are standing while holding your coat.

Then you are wearing the coat.

Then you have your keys.

Then you are at the door.

Then the door is open.

At this point the body has submitted six separate resignation letters, but the conversation remains employed.

The reason is not necessarily that one person failed to notice the signals. A physical departure cue can make the closing more visible without mechanically terminating whatever is being said. If a new conversational reason appears—a question, memory, clarification, joke, plan, or “one more thing”—there is still something for the other person to respond to.

The body may be halfway down the runway while the talk has requested another lap.

Here is the surprising part: people may want different ending times

If conversational closings are this organized, it would be reasonable to assume people are fairly good at ending when they want to.

That assumption runs into an interesting result.

In 2021, Adam Mastroianni, Daniel Gilbert, Gus Cooney, and Timothy Wilson published research asking a remarkably direct question: Do conversations end when people want them to?

Across two studies involving 932 conversations, participants reported when they had wanted the conversation to end and estimated when they thought their partners had wanted it to end.

The answer was not especially reassuring for anyone who believed humans had solved this problem.

Participants’ preferred ending times often differed from the actual ending. They were also imperfect at estimating when the other person wanted to stop.

In the laboratory portion of the research, involving conversations between strangers, only 29.37 percent of conversations ended when one participant wanted them to. In 46.83 percent, the conversation continued beyond the point when both participants later reported wanting it to end. Another 9.52 percent ended while both participants reported wanting to continue.

Those percentages belong to the studied sample. They are not population estimates for every conversation, culture, relationship, or setting on Earth.

But the broader finding is difficult to ignore.

People can be in the same conversation and want different things from its ending.

More importantly, the relevant information is mostly private.

You know when you are ready to stop.

You do not have direct access to the other person’s preferred stopping point.

So two people can perform an impressively coordinated social closing while working with incomplete information about the very thing they are trying to coordinate.

This is where the problem becomes more interesting than awkwardness.

We may be solving the wrong problem extremely well

Imagine that ending a conversation had only one objective:

Stop at the exact moment both people privately prefer.

You could make those private preferences explicit. Every few minutes, somebody could ask, “On a scale from one to ten, how much longer do you want to continue this interaction?”

The timing information would certainly be less private. The conversation would also begin to resemble a customer-satisfaction survey administered by someone standing three feet away from you.

Ordinary interaction has another complication: endings happen between people, and the way an interaction stops can carry social meaning.

That does not mean every goodbye is a delicate emotional negotiation. It means that stopping talk is also something one person does in relation to another person.

A recent qualitative study by Emmi Koskinen helps illuminate this tension, although its scope is specific. Koskinen analyzed 19 dyadic conversations in a Finnish research setting where participants were explicitly asked whether they wanted to continue or end after 45 minutes. The decisions were shaped by the local conversational sequence, and the participants used practices consistent with managing face concerns.

That does not establish that face-saving causes every prolonged goodbye.

It does suggest a useful possibility: the socially smooth way to close an interaction may not always be the same as the fastest way to express a private preference.

Dunce Dynamics interpretation: humans may be better at ending nicely than ending on time.

Those are different achievements.

A conversation can end warmly, respectfully, and without obvious rupture while still continuing longer than someone privately preferred. It can also end efficiently and leave someone feeling that the interaction was cut off abruptly.

The social system may be optimizing for something more complicated than the clock.

The problem is that your preferred ending time is invisible

Consider how much information is publicly available during a goodbye.

You can hear the words.

You can see someone stand.

You can watch them collect a bag.

You can notice the body orientation change.

You can hear, “I should probably head out.”

What you cannot directly see is the exact private preference underneath those actions.

Does the person want to leave immediately?

Would they happily talk another ten minutes?

Are they genuinely in a hurry?

Are they enjoying the conversation but also tired?

Did they stand because their leg fell asleep?

Did they say “I should let you go” because they think you want to leave?

The same visible closing behavior can sit on top of different internal states.

That matters because people are tempted to treat the outward ritual as a mind-reading device.

“He keeps talking because he is lonely.”

“She is trying to escape.”

“He does not know how to take a hint.”

“She is just being polite.”

Any of those could be true in a particular case. The research reviewed here does not allow us to infer them from one goodbye sequence.

The visible action is evidence that a closing is being negotiated.

It is not a psychological scan.

Then someone introduces one more thing

Even a closing that is progressing smoothly can be interrupted by new conversational business.

“Oh, I almost forgot…”

“Before you go…”

“Did I tell you what happened with…?”

“By the way…”

These phrases are dangerous because they introduce material that may legitimately require a response.

Once that happens, the participants are no longer merely completing the old conversation. They have created a new local reason for another turn.

This is one reason the doorway becomes such an important human habitat.

It is physically designed for transition between spaces. Socially, it sometimes becomes a second venue where the conversation continues after the first venue has already been decommissioned.

The living room conversation ends.

The doorway conversation begins.

Then there may be a driveway appendix.

If cars are involved, a final exchange can occur through an open window despite the existence of several previous goodbyes and a functioning brake pedal.

None of this proves that closings always reopen through a formal universal sequence. The conversation-analysis literature is more context-sensitive than that.

The narrower point is enough: a closing is locally revisable. If fresh talk becomes relevant, the interaction can move back toward conversation even after both people have begun doing recognizable leaving behavior.

The goodbye did not fail because the participants forgot how doors work.

The social system received new input.

Goodbyes are a coordination problem with incomplete information

Put the pieces together and the familiar long goodbye starts to look different.

People have organized ways of closing conversations.

They use words and bodies to make the ending increasingly recognizable.

They can still want different ending times.

They can misunderstand the other person’s preferred endpoint.

They may also be managing what the ending means socially while reacting to whatever is happening in the current conversational moment.

Then somebody remembers one more story.

This does not give us one universal explanation for prolonged goodbyes. The research on leave-taking varies across cultures, settings, relationships, and communication modes, and Lucy Baehren’s cross-disciplinary review emphasizes that the empirical literature still has important gaps.

So the safest conclusion is also the more interesting one.

A goodbye conversation is not simply two people failing to stop talking.

It is two people trying to coordinate an ending without direct access to each other’s private preferred endpoint, while preserving the ordinary social relationship that existed during the conversation and responding to whatever new material appears before the exit is complete.

That is a lot of work for a word as short as “bye.”

What this seems to tell us

Humans do not appear to lack closing skills.

If anything, we may have too many of them.

We summarize. We signal. We stand. We orient toward the exit. We put on coats. We collect keys. We make future plans. We wish each other well. We explain why we are leaving. We exchange terminal goodbyes.

And sometimes, after completing nearly the entire procedure, we begin discussing something else.

Research gives us a useful distinction between a socially coordinated ending and a perfectly timed ending. Conversation-analysis work shows that closings can be organized and collaborative. Research on desired ending times shows that private preferences can still diverge substantially from what actually happens. More recent qualitative work suggests that local conversational context and face-management can matter in how people negotiate whether to continue or stop.

That does not mean people are secretly trapped in every conversation.

It means that “when should this end?” is not always a piece of information both participants share clearly.

So perhaps the long goodbye is not evidence that humans are terrible at leaving.

It may be evidence that leaving is doing more than one job at once.

We are trying to stop the conversation.

We are also trying to stop it without making the other person wonder why we suddenly stopped the conversation.

That distinction may explain why the final thirty seconds occasionally requires twelve minutes.

The Institute therefore recognizes the doorway not as an exit, but as a transitional research environment where goodbyes go to receive additional funding.