Conversation has been described1 as “the primordial site of human sociality”. We all have a lifetime’s experience to draw on if asked how it works, or when we reflect on the conversations we have participated in. But because conversation is something that we know tacitly how to do, scientific attempts to understand it are often relegated to the ‘soggy’ end of social psychology. Conversation certainly differs from other subjects of scientific scrutiny. For instance, black holes do not exist to be understood by people, whereas conversation exists only to be understood by people and to help us understand each other. Writing in Proceedings of the National Academy of Sciences, Mastroianni et al.2 report how they have taken up the challenge of researching conversation scientifically.
The authors focused on the question of whether conversations end when people want them to, and gathered data from two studies. In the first one, individuals (806 in total) taking part in an online survey were asked to recall the most recent conversation they had in person, report its duration and indicate whether it ended when they wanted it to. If they indicated that the conversation didn’t end when they wanted, they were asked to estimate how much longer or shorter they would have liked it to have been. Participants were also asked how they thought the person they were speaking to might have answered the same questions. These conversations were mostly between people who were familiar to each other; 88% were between those who had known each other for at least a year, and 84% of the participants spoke to the person in question at least a few times each week.
The authors explain that their second study was designed, in part, to deal with the limitations of the first, such as relying on participant recall of an event and accessing only one party’s view of the interaction. In the second study, the authors brought 366 previously unacquainted participants from university study pools into the laboratory for a one-hour experiment. Participants were paired up to have a conversation about whatever they wanted, for any duration, up to a maximum of 45 minutes. This was followed by another task until the 60 minutes were up. The participants answered the same questions as those used in the online survey.
The authors concluded that the conversations evaluated in both studies almost never ended when both individuals wanted them to, and rarely ended when even one person wanted them to. Interaction with intimates or strangers made no difference to this mismatch. Participants were unaware of when their partner wanted to end the interaction, or that their own perceptions were so different from their partner’s. For example, in laboratory conversations, there was a 46% discrepancy between the actual and desired duration. Interestingly, the authors excluded 57 of the pairs because they spoke for the full 45 minutes and did not end their conversations.
The study’s novelty is in its examination of how people feel about their conversations at this level of empirical scrutiny. It represents a clear advance in psychology, in getting closer to where the action of social life happens, especially in the second study. One future direction for research might be to record or analyse the laboratory conversations themselves, and to ask participants to use transcripts to inform their responses. Transcripts would help the participants to identify precisely the point at which they wanted the conversation to end, and help researchers to understand exactly what each party was doing at the time. Were they mid-story, repeating something or giving a minimal or an expansive response? People show how attuned they are to tiny nuances in social interaction even as it unfolds3, and transcripts might enable the authors to gain extra insights about their findings.
Some have commented that, despite psychology being a discipline associated with “professional people watchers”4, psychologists rarely investigate “where moment-to-moment behaviour naturally happens”4, or deploy “direct observation of actual behaviour”5. Indeed, Mastroianni et al. say that scientists know little about conversation: “how it starts, how it unfolds, or how it ends.” One possible direction for future research, therefore, is to combine laboratory studies of the kind conducted in the second study by Mastroianni and colleagues with investigations of naturally occurring talk.
For more than 50 years, the cumulative science of conversation analysis has examined audio and video recordings of anywhere from single cases to thousands of cases of conversation. One benefit of augmenting laboratory studies and surveys with such data and methods is to avoid the limitations of post-hoc survey methods as identified by Mastroianni et al., as well as the limitations of laboratory settings. All conversations have a reason for occurring, whether mundane or dramatic. In the authors’ second study, the reason was to be a research participant, making the experimental setting itself the ‘invariant occasion’ for the conversations that happened6. We know, however, that people interact differently when they are in a simulation or experiment compared with their behaviour in life ‘in the wild’, because the reason for the event and their stake in its outcome are different7,8.
The authors conclude that people cannot coordinate what each participant wants from a conversation, in terms of ending it at a mutually satisfactory time, according to their responses when asked later. This striking observation tells us something interesting about the difference between what happens inside a conversation and what people say about it afterwards. As Mastroianni et al. point out, analysis has shown that conversations have ‘closing rituals’, which are systematically coordinated. In other words, a typical conversation does not usually end abruptly; it must be brought to a close7. Endings take shape through highly routine practices, such as making arrangements (“So let’s sort out what time on Monday”), or re-stating the reason for the conversation (“Well, I just wanted to see how you were doing”), combined with a ‘terminal exchange’9 such as:
A: Bye bye
Endings are so systematic and recognizable that it can be easy to locate in transcripts the place at which someone wants the conversation to be over, whether by giving a delayed or minimal response or by saying something that indicates they are moving to draw the conversation to a close.
For example, in a study of individuals calling their doctors10, receptionists often initiated the end of a call before the caller was ready. In the following example from that study, the caller ‘wants’ the call to continue after the receptionist has started to end it with “Okay then”, followed by “Thank you”. The square brackets indicate when both spoke at the same time. The final four lines are a classic ‘terminal exchange’.
Receptionist: Okay then,
(pause of 0.5 seconds)
Caller: [ So it’s th- ]
Receptionist: [Thank you, ]
(pause of 0.5 seconds)
Caller: That’s the sixteenth?
Receptionist: The sixteenth, [ at ten pa]st eleven.
Caller: [Okay then.]
(pause of 0.3 seconds)
Caller: Ten past eleven, thank you.
Receptionist: Thank you,
(pause of 0.2 seconds)
Caller: T[hank you,]
Receptionist: [ Bye ]
Furthermore, there was a correlation between surgeries in which the receptionists, rather than the callers, moved to initiate the ending of the call and lower patient-satisfaction scores with the surgery in general. Scrutiny of conversation transcripts reveals why this type of scenario creates dissatisfaction: in this case, the individual had to push past the receptionist’s move to end the call to get their appointment confirmed.
When one party walks out or puts the phone down on another, we have the exact definition of mismatched desires regarding a conversation’s ending. In the following call11, a salesperson ‘cold calls’ a company with the goal of selling printing systems (transcript simplified). Even when the call is nearing its end, the client still uses components of the terminal exchange before hanging up.
Client: Well we’re happy with, uhm, the people that we’re currently using.
(pause of 0.3 seconds)
Sales: I’m sure you are, but I wanted to find out when the contract’s up for review so then I can c[all maybe nearer the time]
Client: [Yeah no we’re happy with wh]o we’re currently using.
(pause of 0.5 seconds)
Sales: You don’t know when the [contract’s up for re]view?
Client: [Okay. Thank you.]
Asking ‘when did you want the call to end?’ is the wrong question here, because the conversation is likely to be unwanted by the client in the first place, and, once in it, the parties involved want different outcomes.
Both the medical and the sales calls show that, and indicate how, individuals ‘want’ their conversations to end at different points. We can identify this in real settings in which we understand the authentic purpose of the conversation. It would therefore be interesting to apply Mastroianni and colleagues’ methods to the analysis of such transcripts and recordings, to ask individuals later, on reflection, to identify at what point they wanted to continue or end the conversations.
What about conversations between loved ones — such as those recollected in the online survey? In the following conversation12 (transcript simplified) between Sue (not her real name), a young person with learning disabilities and in residential care, and her dad, Sue asks her dad to bring her extra pocket money when he visits. This is followed by the first turn that moves to close the conversation:
Dad: Right, well, I’m gonna get on now, I’ll be there for about half past nine tomorrow morning.
But the conversation continues for a further 45 seconds before another pre-closing event occurs:
Dad: Right, well I’m going to go now, darlin’.
Sue: Yeah I’ve got to finish my cards off.
Only after three more pre-closings, including those expressing love (Dad: “Okay, lovey?” Sue: “yeah”; Dad: “I love you”; Sue: “love you”), do they bring the call to its end.
How do you show that you care about someone? Mastroianni et al. rightly point out that conversation is the “bread and butter” of our psychological and physical health, and this is clear to see in Dad and Sue’s conversation. Staying longer in the conversation than external constraints allow (such as in a film scene in which people in a lift miss their floor to keep talking) is one way to do it. Closing rituals are so systematic that the conversational machinery allows us to see how the reopening of closings happen.
Mastroianni and colleagues’ findings are compelling. Some media headlines about their study (see go.nature.com/3sgIkup), such as “only 2% of conversations end when we want them to”, focused on the disconnect between the desired point for a conversation to end and its actual end. Although the headline news might be the scale of the disconnect, reducing conversations such as this chat between Dad and Sue to ‘who wanted what’ damages the empirical reality of their conversation and misses its purpose.
There are tremendous real-world benefits to analysing conversation with close scrutiny and rigour. For example, returning to the doctor’s surgery, the same research10 showed that when receptionists proactively confirmed an individual’s appointment time and date, rather than doing so only in response to a request for confirmation, the conversation ended collaboratively. Moreover, proactive confirmation was associated with higher patient satisfaction, and the finding was used to train receptionists.
Do conversations end when people want them to? Mastroianni et al. conclude that the answer is almost certainly no. Asking people to report on their conversations has shown this clearly. Apart from situations such as in an argument, people generally do not say, “I want this conversation to end.” They might tell other individuals, “I was trapped in that conversation for hours”, or “I don’t want to talk to her”, but, in real conversation, people usually convey such things tacitly. This is why examining conversations, including using transcripts, is informative. It is clear, as Mastroianni et al. state, that “The more we learn about conversation — about how it begins and ends, runs and stalls, delights and disappoints — the better positioned we will be to maximize its benefits.”