People imitate accent features they expect to hear, even without hearing them

Research from postdoc Lacey Wade confirmed this idea, what she calls expectation-driven convergence, in a controlled experiment for the first time. The work reveals just how much the subconscious factors into the way people speak.

An illustration of many different-colored heads with many different-colored talking bubbles.

Certain accents in the United States are unmistakable: the twang of someone from the South, the dropped “r” by a Bostonian. Without realizing it, people often copy this kind of accent when talking with someone who uses it, a concept in linguistics known as convergence. 

“There’s a lot of work looking at convergence toward observed features like sentence structure or imitation of speech sounds,” says Lacey Wade, a postdoctoral fellow in Penn’s Language Variation & Cognition Lab. Previous research had confirmed the existence of these temporary shifts in interpersonal communication and the context in which they most often show up. But Wade wanted to study another type of convergence, the kind focused on how people expect their conversation partners will sound before they’ve said a word. 

Prior theories on what Wade calls expectation-driven convergence had been mostly anecdotal. Now, in the first controlled experiment of its kind, Wade confirmed that people do imitate accent features they believe they’ll hear, even before they’ve heard them. She shared these results in the journal Language

Black-and-white photo of Lacey Wade sitting at a desk, with her head propped up on one arm.
Lacey Wade is a postdoctoral researcher in the Language Variation & Cognition Lab in the School of Arts & Sciences at the University of Pennsylvania. (Image: Courtesy of Lacey Wade)

“We didn’t realize how strongly expectations were influencing speech,” Wade says. “It’s never been established in this controlled way before. But we’ve now shown it’s something people do, with findings that are robust and replicable, too. People can build off this and ask a whole new set of questions about other ways expectations drive speech.” 

For the past several years, Wade has been trying to answer such questions herself, first as a doctoral student and then as a postdoc in the lab of Penn’s Meredith Tamminga. Broadly, Wade’s research looks at how social influences affect language. For this work on convergence, she hoped to delve into how memory and social factors play into speech patterns and variation.

She and Tamminga created an experiment around a particular aspect of a Southern accent, what happens with the vowel “i” in certain words. “This vowel has two parts,” Wade says. That is, when broken down, it sounds like ah and ee squished together. “In Southern-shifted speech, people often take the ee sound and reduce it. Instead of a strong movement over the course of the vowel, it’s weaker. It sounds more like the ah.” The words “dime” and “ride,” for example, often sound more like “domm” or “rod.” 

This particular feature, called glide-weakening, “is a very stereotypical feature associated with Southern speech,” Wade says. “It’s something people almost always use when putting on a Southern accent.” Regardless of how the researchers felt about that fact, they knew that made it a sociolinguistic feature people would likely expect, meaning it was a strong starting place to study this phenomenon. 

We didn’t realize how strongly expectations were influencing speech. It’s never been established in this controlled way before. Penn postdoctoral researcher Lacey Wade

To do so, Wade built a word-naming game akin to popular word games like Taboo or Catch Phrase. In each of three rounds, participants received clues meant to prompt them to say specific words out loud. In round 1, they read these hints from a computer screen, to provide the researchers a baseline. In round 2, someone with either a Southern or Midwestern accent—representing the experimental and control groups, respectively—read the clues, intending to elicit specific words. In the final round, participants again read them from a screen. 

“We wanted them to feel like they were doing something fun, playing this game, taking their mind off their own speech,” Wade says. “It’s been found that feelings of interaction, feeling like you want to align with who you’re talking to, promotes convergence.” 

As the researchers had hypothesized, simply hearing someone with a Southern accent say any words led participants to change how they spoke the vowel “i,” to do this glide-weakening. The same wasn’t true of hearing someone with a Midwestern accent. Wade replicated the experiment a second time, to similar results. She also replicated this with different model talkers of different genders and with different words. 

The findings revealed to her just how much the subconscious factors into how people speak. “The person we’re talking to, the situation, how we’re feeling, they all exert influence on our language,” she says. “At the very least, this work can make us aware of the powerful role stereotyped associations can play in our lives, even in our speech, and help us understand that sometimes our language is shaped by our expectations instead of reality. From a scientific perspective it’s useful to understand how we might draw from stereotyped beliefs if we want to combat them.” 

Funding for this research came from the National Science Foundation (Grant BCS-1917900). 

Lacey Wade is a postdoctoral researcher in the Language Variation & Cognition Lab in the School of Arts & Sciences at the University of Pennsylvania.