Pet Talk Varies Wildly, and That May Explain the Pet Effect

A PHONE tucked into someone’s back pocket clicks on without warning, the way it does at random moments throughout the day, and catches its owner mid-conversation with a dog who has just knocked over the water bowl. A University of Arizona team built a coding system, tested for reliability on 240 people’s recordings, that suggests how differently people talk to their pets, from full narration to near total silence, may help explain why living with an animal seems to help some owners far more than others. The tool needed years of listening, arguing, and re-listening before anyone trusted it enough to run the numbers. Now that it exists, the numbers say something the pet effect literature never quite accounted for: talking to an animal is a wildly idiosyncratic habit.

The pet effect, the theory that living with a cat or dog measurably helps human health, has been notoriously inconsistent in the research: in 2022, a systematic review of two dozen studies on pets, loneliness, and social isolation found only a portion turned up a significant link, with the rest split between no effect and, in one case, a worse one. One explanation researchers keep circling back to is that having a pet is too blunt a category. Writing in 2021, a methods paper on the field argued that limited quantification of what owners actually do with their animals, rather than whether they own one, is a major source of the field’s mixed results.

What an Unwitting Recorder Caught

Enter the electronically activated recorder (EAR), a smartphone app that has quietly recorded ordinary life for psychology studies since 2001, switching on at random moments, roughly five times an hour, so nobody can perform for it. A review of the method credits its power to exactly that: unobtrusively collecting authentic real-life data rather than the version of a day someone would choose to report. The four archival studies mined for this project, covering older adults, meditators, couples coping with breast cancer, and people going through divorce, had no idea their pet chatter would ever be analyzed; participants consented to recording for entirely different research questions. That is precisely what let the team sidestep a problem that has dogged human-animal interaction research for years, namely that people who sign up for a pet study already know it is about their pet, and tend to answer accordingly. Sift the EAR archive for stray barks, meows, and the phrase who’s a good boy, though, and there was no coding system built to make sense of what turned up.

So the team built one, calling it AURAL-Pet, short for Audio-based Understanding of Relational Animal-directed Language, and spent multiple rounds of listening and arguing to whittle an unwieldy 55-category checklist down to 30 workable variables. The hard part was never deciding what to listen for; it was getting two independent coders to agree on what they had just heard.

Applied to 5,072 audio clips from 240 people the coders determined were pet-adjacent, meaning a dog or cat was audibly present or verbally mentioned, the system held up. 22 of 23 original yes-or-no categories cleared the reliability bar researchers use for this kind of data, once two coders’ scores were averaged together. Only one variable, whether someone addressed the animal as a fellow creature rather than as a near-human confidant, fell short, a reminder that some emotional shading is genuinely hard to hear from audio alone.

“Very often in human-animal interaction studies, people sign up knowing the topic of the study is their relationship with their pets,” says Emily Bray, the study’s senior author and an assistant professor in the College of Veterinary Medicine. “So, right off the bat, you have a bit of a biased sample.” An archive assembled for entirely different research, she adds, “is as unbiased as it can get.”

The Owners Who Never Sound Alike

Across every clip in which someone was speaking at all, whether to a partner, a group, or themselves, pet-directed talk claimed an average of 9.2 percent of that time, matching the one-tenth figure in the release, though individual owners ranged from almost never to nearly half of all their speaking clips. The average hides more than it shows. Across 11 of the speech behaviors the coding system tracked, praising, scolding, asking the pet direct questions, narrating an entire afternoon to it, some participants logged the behavior in every single clip while others never did it once. Most of that talk happened while the person was entirely alone: 44 percent of pet-adjacent clips involved no one else in earshot, a further 36 percent had one other person present, and the rest involved a group, a solitary pattern that echoes recent time-use research on how much of daily life now happens without anyone else in the room. None of these numbers say whether treating a pet like a roommate rather than an animal changes how much good the pet actually does its owner. They are, for the first time, numbers a future study could test against exactly that question.

One caveat sits underneath all of it. Because the recordings came from archives never built to study pets, coders could confirm only that an animal was audible or mentioned, not that it belonged to the person speaking. The paper is careful to call its subjects pet-adjacent rather than pet owners, a distinction the surrounding coverage does not always keep, since a neighbor’s dog can wander into a recording as easily as one’s own.

The demographic patterns are worth reading loosely for the same reason. With no hypothesis fixed in advance, and a sample built for studies about aging, meditation, illness, and divorce rather than pets, differences between older and younger participants, or between men and women, read as suggestive rather than settled. The team frames its own demographic breakdowns as groundwork for a future hypothesis, not a finding to lean on by itself.

What the paper does establish is a working instrument, and the team is already pointing it at new terrain. A companion project, led by colleagues Matthew Grilli and Jessica Andrews-Hanna, will use AURAL-Pet on adults over 60 to see whether the way older people talk to their pets tracks with measures of memory and well-being, the same population in which expectations about social connection have been shown to shape how socially engaged people actually become. Dara Jonkoski, the study’s lead author and a doctoral student in the Department of Psychology, says the point was never just to count words. “We hope AURAL-Pet can be used in future studies to understand how differences in these interaction styles map onto mental health and well-being outcomes in humans,” she says. Paired eventually with video-based coding of gestures and touch, the tool could give researchers a fuller picture of an interaction that, until now, mostly lived in owners’ self-reports.

“We can get straight to human behavior, which I love as an animal behavior researcher,” Bray says. “Humans are animals, too, right?”

FAQ