Conversation Partners Literally Get on the Same Wavelength

“Getting on the same wavelength” with someone used to be a cute radio-frequency metaphor to describe how we develop a deep understanding and empathy for someone when a conversation is going well. When people are in a good conversation, they tend to mimic each other’s hand gestures, body postures, word choices, and even speech style – at least a little bit. The convergence of behaviors that happens in a conversation must have some degree of flexibility because there’s still something of a turn-taking aspect to a conversation. We can’t completely synchronize with each other, to the point of doing the same thing at the same time, or we would just wind up humming the same pitch together. (If we’re meditating together or chanting together, instead of having a conversation, perhaps humming the same pitch together is perfect. See work by Fred Cummins.)

The research on coordination and synchrony in conversation has been around for several decades (Clark, 1996) but only recently has it taken on some serious methodological and computational sophistication in the laboratory. For example, a number of researchers are using neuroimaging techniques to find correlations between the brain activity of a speaker and the brain activity of a listener (e.g., Kuhlen, Allefeld, & Haynes, 2012; Nastase, Gazzola, Hasson, & Keysers, 2019). Correlations between conversation partners have also been found in their eye movements (Richardson & Dale, 2005), their facial expressions (Louwerse, Dale, Bard, Jeuniaux, 2012), and even their postural sway (Shockley, Santana & Fowler, 2003).

But it’s not only simple correlations that can provide insight into how two conversation partners get coordinated. A conversation is so much more than a sine wave with some particular frequency. It doesn’t just go up and down and up again at some fixed pace. It is more like a weighted combination of many different sinewaves, with those weights changing over time. In fact, a statistical method called Fourier Analysis is often used to decompose speech into those differentially-weighted sine waves. Over the time course of a conversation, those waves have a hierarchical statistical complexity that carries in it some additional insights into how people get coordinated, over and above simply doing the same thing at the same time.

In all kinds of joint behaviors – such as ballroom dancing, carrying a table together, or having a conversation – doing exactly the same thing at the same time would be terrible. When one person does something in the joint action, the other person often must accommodate that action, not mimic it. Thus, while correlations can sometimes be informative for understanding how two conversation partners start “behaving like one system,” they are not the only tool.

Recent work in Chris Kello’s lab has been showing that the statistical techniques of Multiscale Matching are tools that can reveal how two joint actors converge onto the same “wavelength” even when they aren’t doing the same thing at the same time (Abney, Paxton, Dale, & Kello, 2015). For example, when one person was tapping targets on a touchscreen with their left hand and alternating with a partner using their right hand, the two people were using different limbs at different times but nonetheless gradually converged on a similar hierarchical structure in the time series of their movement times (Schloesser, Kello, Marmelat, 2019). Thus, even when the two partners are clearly not doing the same thing at the same time (and thus may not be directly correlated), they are nonetheless converging onto a shared statistical signature in the hierarchical temporal patterns of their behaviors. The two systems are becoming a little bit like one system, possibly optimizing information flow between its two subsystems.

A similar thing happens in conversation. Every person’s speech behavior has its own particular complex rhythm at the long timescale of minutes, a different complex rhythm at the medium timescale of seconds, and another different complex rhythm at the short timescale of milliseconds. But when two people are in a good conversation, their unique individual distributions of rhythms across those timescales will change over time and converge onto one shared distribution of rhythms. This kind of coordination even happens in a bilingual conversation where two bilinguals have one person speaking one language and the other speaking the other language (Schneider, Ramirez-Aristizabal, Gavilan, & Kello, 2020). In fact, some of this convergence even happens over videoconference platforms such as Zoom. Alviar, Dale, and Kello (2023) found that while in-person conversations show this convergence between people in both their speech rhythms and body movement rhythms, conversations over Zoom show only the convergence in speech rhythms. With less of the body visible over Zoom, less body movement convergence takes place. Perhaps when we are videoconferencing over platforms like Zoom, we should all back up a little from our cameras and include more of the torso and arms in frame to facilitate conversational convergence.

The overarching lesson from this work is the following: When one person (or biological system) has a good conversation with another person (or biological system), they don’t have to converge on doing the same thing at the same time in order to become like one conjoined system. Instead of simply synchronizing with each other, they can complexly syncopate with each other as a way of becoming one efficient system producing intelligently coordinated activity. Coordination does not require synchronization — just like a productive discussion does not require that all participants agree on the same opinion. For example, when liberal talk show host Bill Maher has conservative Ralph Reed on his TV show and sparks a lively political discussion, the two of them find a way to “get on the same wavelength” in a respectful debate. They certainly don’t copy each other, but they accommodate each other, and they syncopate with each other. May we all be so lucky.