(Photo by Jarritos Mexican Soda on Unsplash)
Pink Noise in Conversation Is Linked to More Enjoyable Chats
In A Nutshell
- Conversations that balance staying on the same page with drifting apart, a pattern called pink noise, were rated as more enjoyable.
- Speed dates where both people were open to a second date showed more of this rhythm, though only 31 dates were studied.
- Pink noise grew stronger across podcast episodes, and an AI’s scripts of enjoyable chats showed more of it.
- The link was modest and does not prove the rhythm makes a conversation better.
Some conversations flow so easily that neither person wants to hang up. Others stall in small talk. New research in Science Advances finds that chats with a particular rhythm tended to be rated as more enjoyable, and that same rhythm was stronger in speed dates where both people were open to seeing each other again.
Dartmouth College researchers call the pattern pink noise, a kind of signal where change plays out at fast and slow speeds at once, and one also found in how people walk. Instead of tracking only how alike two people’s words become, the team tracked how that similarity rises and falls throughout a conversation.
Earlier research focused on how much partners start to mirror each other’s words and style. That idea falters at the extreme, since two people who echo each other nonstop make for a dull chat. Some drifting apart may be part of what keeps a conversation alive.
Most Stranger Conversations Show Some Pink Noise
Researchers sort conversation rhythms into three types. In white noise, how closely two people’s words line up jumps around randomly, like a chat that bounces between unrelated topics. In red noise, partners get stuck restating the same ideas for long stretches. Pink noise lands in between: partners stay on the same page for a while, one introduces something new, they drift apart briefly, and then they come back together as the other picks it up. The researchers describe it as a “goldilocks zone” between chaos and rigidity.
To look for it, the team analyzed 1,656 recorded video chats between strangers, who were asked to talk for at least 25 minutes as if they had just met at a social event. Software turned each turn of speaking into a set of numbers capturing its meaning, which let the researchers track how close each comment was to the one before it. Nearly every chat, 94 percent, showed a pattern more structured than pure randomness.
Stronger Pink Noise Goes With More Enjoyable Conversations
Participants also filled out surveys about how the chat went. Conversations with more pink noise were rated as more enjoyable and more connecting, and partners in them said they would have kept talking longer and had opened up more. Personality ratings showed no link, and the pattern held after accounting for how long people talked and how many words they used. The link was modest, though, a mild tendency rather than a strong one.
A second dataset of 97 conversations between friends recruited on college campuses, who spoke for about 20 minutes about a formative experience one of them had shared, showed the same link to enjoyment. Its link to feeling connected was too weak to count as reliable.
Matched Speed Dates Showed Stronger Conversation Rhythm
Survey answers can be skewed, since people may guess what researchers want to hear. A real-world outcome offered a check: whether two people on a speed date were open to meeting again. The team analyzed 31 ten-minute phone dates from a podcast called Blink Date, in which strangers chatted after a lighthearted prompt such as “What is your most controversial food opinion?”
Afterward, each person privately said whether a second date appealed to them. A pair counted as a match if both answered “yes” or “maybe,” and 19 pairs did. The study did not confirm whether any of those pairs actually went out again. Matched dates showed a pink noise pattern unlikely to be a fluke, while the 12 unmatched dates showed no reliable pattern. Matches also scored higher overall, even after accounting for conversation length. With so few dates, the authors call this “preliminary evidence.”
Podcast Hosts and an AI Model Showed the Pattern Too
Researchers also tested whether practice matters. Across 25 podcasts with two regular hosts chatting freely, pink noise grew stronger over each show’s first 50 episodes. That fits with partners coordinating better as they gain experience, though the data cannot confirm it.
In a final test, Claude Sonnet 3.7, an AI model, wrote 50 scripts of enjoyable conversations and 50 of unpleasant ones between strangers meeting at a coffee shop. Online readers rated the enjoyable scripts 4.8 out of 5 on average and the unpleasant ones 2.4. The enjoyable scripts showed stronger pink noise than the unpleasant ones, and only they rose above chance, even though the instructions never mentioned the pattern. The researchers suspect the AI absorbed the rhythm from the enormous amount of human writing it was trained on.
Pink noise looks like a clue to why some chats click, not a verdict on any date. The study doesn’t prove that creating this rhythm makes a conversation better, only that the two tend to go together. Experiments that deliberately change how partners coordinate would have to show whether the rhythm itself matters. If it does, the authors point to possible tools that could make conversation easier for people who find it hard, though that remains untested.
Paper Notes
Limitations
Observational by design, this work measured existing conversations and looked for correlations with outcomes like enjoyment, but did not manipulate conversations, so cause and effect cannot be established. The authors say experimental tasks that manipulate timing and coordination between speakers would be needed for a causal account. They also note that an individual speaker’s contribution is hard to separate from the pair’s shared pattern, since people shape and are shaped by the conversation they are in. In the large CANDOR dataset of stranger chats, the associations with enjoyment and connection were statistically significant but small (correlations of 0.11 and 0.09). Pink noise did not outperform every alternative: when tested against a combination of other similarity measures together, it no longer added anything extra to predictions of enjoyment or connection. In the friends dataset, the enjoyment association was no longer statistically significant once total conversation length and turn count were added, which the authors say is likely due to pairs who enjoyed talking continuing well past the required 10 minutes. The speed-dating analysis covered only 31 pairs (19 matches, 12 nonmatches), and a match meant both people answered “yes” or “maybe,” not that a second date happened, though the match difference held when conversation length was factored in. Finally, the explanation for the AI result is a hypothesis, since the authors note the model had no explicit training on this idea “to our knowledge.”
Funding and Disclosures
According to the paper, the work was supported by a JustX Grant from the Wright Center for the Study of Computation and Just Communities at Dartmouth College, the Stanford Center for Comparative Studies in Race & Ethnicity at Stanford University, and the Stanford Institute for Research in the Social Sciences at Stanford University. The authors declare no competing interests.
Publication Details
Titled “Pink noise in dyads’ semantic similarity trajectories predicts conversation enjoyment,” the study was authored by Kathryn C. O’Nell, Kiara L. Sanchez, and Emily S. Finn of the Department of Psychological and Brain Sciences at Dartmouth College in Hanover, New Hampshire. It was published open access in Science Advances, a journal of the American Association for the Advancement of Science, on November 6, 2026 (volume 12, issue 45, article eaeb2684), after submission on August 6, 2025, and acceptance on September 25, 2026. The DOI is 10.1126/sciadv.aeb2684. Corresponding authors are Kathryn C. O’Nell and Emily S. Finn.
