There is a particular moment in a conversation with someone new when you can feel a more interesting question arrive and decide not to ask it.
You ask what they do instead. Or how they know the host. Or whether they have been busy lately. The conversation remains perfectly pleasant, and both of you quietly cooperate in keeping it that way.
I have done this at more dinners and work events than I can count. The hesitation is not usually that I have nothing real to ask. It is that asking would make me responsible for whatever happens next. What if the question feels intrusive? What if the other person gives a polite answer and looks for an exit? What if I have mistaken a stranger’s friendliness for an invitation to know them?
A series of experiments published in the Journal of Personality and Social Psychology suggests that this forecast is often too pessimistic.
Michael Kardas, Amit Kumar and Nicholas Epley asked pairs of strangers to discuss questions that moved well beyond ordinary small talk. Before speaking, participants predicted how awkward the exchange would feel, how interested the other person would be, and how connected and happy they would feel afterwards. Then they had the conversation and answered the same questions again.
The conversations were less awkward, warmer and more connecting than people expected. When participants experienced both shallow and deep conversations, they preferred the deeper one more than they had predicted beforehand.
That last sentence is the accurate version of the more romantic claim that these were the conversations nobody wanted to end. This paper did not measure when participants wanted to stop talking. A separate study did examine conversational length, and I will come to it. Keeping the two findings separate actually makes the pattern more interesting.
The questions were genuinely personal
The paper, titled “Overly Shallow?”, was published online in 2021 and appeared in the journal’s 2022 volume. The American Psychological Association described the project as 12 experiments involving more than 1,800 participants in total.
The samples were broader than a single room of undergraduate psychology students. They included financial services executives, company employees, international MBA students, university students, community members and online participants. Most of the work was still conducted within the United States over a relatively narrow period, which becomes important when considering what it can establish.
Some participants received shallow questions. They discussed the weather, a recent television programme, Halloween or a memorable bad haircut.
Others received questions such as: What in your life makes you feel most grateful? If you could undo one mistake, what would it be? What would be important for a close friend to know about you? When did you last cry in front of another person?
Those are not clever conversation starters disguised as intimacy. They ask for actual disclosure.
In one experiment, the researchers removed the possibility that their own questions were creating the entire effect. Two hundred participants generated both the normal questions they would ask while getting to know someone and a set of questions deeper than they would ordinarily ask. Pairs were then randomly assigned to discuss one type. Even when people generated the questions themselves, they overestimated the awkwardness of the deeper exchange and underestimated how connected and happy they would feel.
The mistake was not expecting zero awkwardness
There is a temptation to turn this into a clean verdict: small talk is empty, deep talk is wonderful, so skip directly to asking strangers about their most painful memory.
The paper does not support that.
The deeper conversations were sometimes a little more awkward than the shallow ones in actual experience. The mistake was that participants expected the cost of depth to be considerably larger, and its rewards smaller, than they turned out to be.
Across the experiments, people expected conversations to produce weaker connections and less happiness than they actually experienced. They were especially likely to overpredict awkwardness when the questions were personal. In the experiments where each participant had both a shallow and a deep conversation, the deeper exchange created stronger feelings of connection and was preferred afterwards more than people had anticipated.
Results were not identical in every comparison. In some experiments the deep group felt significantly more connected than the shallow group; in others, the difference was not statistically clear. Shallow conversations also went better than people expected, and sometimes naturally became more intimate than their opening questions suggested.
That nuance matters. The robust finding was not that every deep conversation outperforms every shallow one. It was that people systematically misjudged how a conversation would feel, with the biggest forecasting error attached to depth.
We underestimate how much the other person will care
The most revealing part of the study, for me, is what participants appeared to be afraid of.
It was not only vulnerability. It was indifference.
People knew that they themselves would be interested in hearing something meaningful from another person. But they expected the stranger to care less about their own answer. We seem to grant ourselves curiosity and warmth while imagining that the person opposite us may be politely waiting for the personal part to finish.
The researchers called this a miscalibration in perceived care. Participants underestimated how interested their partners would be in what they disclosed. When people were told to expect a caring partner, they chose deeper questions. When they were told that people tend to underestimate strangers’ care, they also became more willing to choose intimate topics than participants who received the opposite message.
This sits beside the well-known “liking gap” documented by Erica Boothby and colleagues: after meeting someone, people tend to underestimate how much the other person liked them. Both findings point to the same asymmetry. I have direct access to every hesitation and clumsy phrase on my side of a conversation. I see only the other person’s public response. It is easy to compare my internal doubt with their composed exterior and conclude that I am the less welcome participant.
I touched this from a more personal angle in an earlier article about people who make conversation feel effortless. Often their gift is not verbal brilliance. It is behaving as though the person in front of them is worth following beyond the first answer.
A separate study found that we also leave too early
The claim about not wanting a good conversation to end comes from a related but distinct line of work.
In five preregistered experiments involving 1,093 participants and 966 spoken conversations, Kardas, Juliana Schroeder and Ed O’Brien studied what they called the “hedonic trajectory” of talking with a new acquaintance.
Participants expected that once the first few minutes had passed, enjoyment would decline. They imagined running out of material or reaching the point where continuing would become work. In reality, enjoyment generally stayed stable or increased as the conversation continued.
In one experiment, participants chose to spend less time talking and more time alone than would have maximised their reported enjoyment. This did not mean every individual was desperate to continue, or that a longer conversation is always better. It was an average mismatch between how quickly people expected pleasure to fade and what they subsequently experienced.
Put the two papers together carefully and a consistent picture appears. Before a conversation, we may underestimate how much a stranger will care. During one, we may underestimate how long there will be something worth saying. Those forecasts can keep the exchange both shallower and shorter than the experience itself would justify.
This helped me put firmer research underneath something I had previously tried to describe in my article about the exhaustion of spending hours in polite small talk. The variable is not merely how long we talk. It is whether the conversation starts feeding both people enough to sustain itself.
Small talk is not the villain
Small talk performs a real social function. It lets two people test pace, attention and safety before either takes a risk. A question about the weekend may be boring as content, but useful as calibration. Does the other person answer generously? Do they ask something back? Do they seem hurried, guarded or open?
The researchers themselves note that asking someone in a supermarket queue about the last time they cried could land very differently from asking it in an experiment where both people have consented to discuss personal questions. Their studies produced short-term measurements in structured settings. They did not test the long-term effect of disclosing to someone you will never meet again, and they did not establish whether the pattern works identically across cultures.
Context matters. So does reciprocity. A conversation becomes deep because both people move towards it, not because one person forces the other to provide intimate material.
I wrote before about being told I was too intense for casual conversation. The lesson I take from this research is not that intensity was secretly correct all along. It is that my fear of making a conversation strange is probably less accurate than I assume, provided I am paying attention to whether the other person is coming with me.
The useful move is one level deeper, not all the way down
I do not think the practical response is to memorise the study’s most intimate questions and deploy them at the next dinner.
A better move is often one degree of honesty beyond the script already running.
“How is work?” can become “What part of work has been taking most of your attention lately?” A question about where someone lives can become what they like about living there. A conversation about children can move from ages and schools to what parenthood has changed in the way they see their own parents.
None of those questions demands a confession. Each gives the other person somewhere real to go if they want to.
I think this is what I have often misunderstood at social events. I assumed that keeping a conversation light was the considerate choice because depth would place a burden on the other person. The research suggests that the other person may be making exactly the same calculation about me.
So two people who would both welcome something more interesting protect each other from it.
The surprising part of these studies is not that strangers can disclose personal things when researchers tell them to. It is how poorly they anticipate the warmth waiting on the other side, even though each of them is helping to create it.
Perhaps the conversation we dread is not always the one we should avoid. Sometimes it is simply the one neither person wants to be brave enough to begin first.
