Ohio State researchers found speakers exercise substantial control over whether they pick up an accent — and that expectations about how someone sounds can override what was actually said.

People drift toward the speech of those around them. Phonetic accommodation — adopting features of an interlocutor’s pronunciation — is one of the better-documented findings in sociolinguistics, and it usually gets described as automatic: a thing that happens to a speaker rather than something they do.

A study in Laboratory Phonology from Ohio State University tests how automatic it really is, by the simple expedient of asking people to stop. The answer is that they largely can.

The lead author is Kathryn Campbell-Kibler, professor in Ohio State’s Department of Linguistics, with Cynthia Clopper, professor of linguistics, as co-author.

The design

One hundred and thirty-six American English speakers listened to recordings of an Australian woman and were given one of three instructions: imitate her, avoid imitating her, or simply repeat the words they heard.

The three-way split is what makes the study informative. Comparing imitation against avoidance against a neutral baseline separates what a speaker does by default from what they are capable of doing on request — which a study that only asked people to imitate could not.

Participants showed a high degree of control.

“People are pretty good at controlling whether they sound like somebody, if you ask them to,” Campbell-Kibler said.

The baseline condition is the one that preserves the classic finding. “If we don’t tell you whether or not to imitate somebody, you’re going to sound like them,” she said — though without instructions, participants tended to pick up just one feature, such as dropping the r in “near”, rather than shifting their accent wholesale.

Which features, and why that matters

Control was not uniform across the accent. Awareness of particular features strongly shaped which sounds participants imitated, and the clearest case involved the vowels linguists label BATH and TRAP.

Those two words are the standard diagnostic for Australian and southern British English against American English: the vowel in “bath” is the one Americans notice, and it is a feature with enough cultural salience to be imitated deliberately. Features speakers can hear and name are features they can manipulate; features below the level of awareness are not available for deliberate control, which is why an instruction to imitate does not produce a uniform shift.

The finding that surprises

The most interesting result concerns what happened when the speaker’s actual pronunciation conflicted with what participants expected of her.

Participants initially copied her TRAP vowel, then stopped doing so once she used broad BATH vowels. The interpretation is that hearing a strongly Australian feature activated their stereotype of an Australian accent, and that stereotype then governed their output more than the specific recording in front of them did.

“What you think she usually sounds like ends up clobbering the token that you just heard,” Campbell-Kibler said.

That is a real challenge to the automatic-accommodation picture. If imitation were a low-level perceptual-motor process, it would track the acoustic signal. Instead it tracked a mental model of the talker — which means accommodation involves a representation of who someone is, not just a recording of what they said.

The prestige question

Participants showed no hesitation about imitating this speaker, and the authors attribute that to the status of her dialect. Australian English is, in Campbell-Kibler’s words, “a positively valued group for most Americans”, with the consequence that “you don’t feel like you’re making fun of somebody in the same way that you might” with other accents.

The paper makes the same point, noting that “the model talker spoke a dialect that was likely to be in high regard for the participants, allowing us to examine the imitation of different dialect features in a prestigious variety.”

This is a methodological choice with a visible cost, and it is worth being plain about. The study used one model talker: a college-educated white Australian woman in her twenties from Perth. Everything about the prestige interpretation rests on a comparison the study did not make — against a stigmatised accent, where participants might refuse to imitate, or imitate in a mocking register, or produce a caricature rather than an approximation.

With a single speaker, her dialect, her gender, her age, her class and her individual voice are all one variable. That she was easy to imitate without discomfort is reported rather than demonstrated, and the authors present it as the reason for choosing her rather than as a result.

What it adds

The applied reading Campbell-Kibler offers is modest and matches the data: “People who move to a new area often end up with an accent that’s somewhere in the middle.” That partial drift is consistent with the baseline condition here, where speakers picked up one feature rather than a whole accent.

Two further limits belong with the result. The release reports no statistical values — no effect sizes, no measure of how much imitation differed between conditions — so the strength of “a high degree of control” cannot be judged from it. And repeating recorded words after an instruction is not a conversation: the social pressures that drive real accommodation, and the stake a speaker has in being understood or accepted, are exactly what a laboratory task removes.

What survives those caveats is the clobbering result, which does not depend on the task being naturalistic. A speaker whose imitation follows their expectation of a talker rather than the talker’s actual speech is doing something cognitive, and that is a finding about how accents spread.

The paper was published on 28 August 2026. Several research assistants were supported by National Science Foundation Research Experiences for Undergraduates grants; the release gives no overall funding statement and states no conflicts of interest.

---

Notes

1. Campbell-Kibler, K., and Clopper, C. “Cognitive control in cross-dialect phonetic imitation.” Laboratory Phonology, 2026. DOI: 10.16995/labphon.24504.
2. Ohio State University. “How much control do we have over picking up an accent? A lot, study finds.” EurekAlert!, 22 October 2026. https://www.eurekalert.org/news-releases/1148794.

Photo by Andrea Piacquadio on Pexels.com