crowded conversation

Vitaly Gariev on Unsplash

Study Finds a Split-Second Overlap When Attention Shifts

In A Nutshell

  • New brain-scan research finds that when attention shifts from one speaker to another in a noisy room, the brain briefly tracks both voices at once instead of switching instantly.
  • Locking onto a new voice starts and finishes earlier than fully letting go of the old one, showing these are two separate mental processes, not one smooth handoff.
  • A dip in a brain rhythm called alpha suggests the switch takes real mental effort, with the heaviest lifting lining up with locking onto the new speaker.
  • Word-by-word language predictions may reset after a switch, though the study is small and the researchers say it does not mean listeners forget the gist of what was said.

At a loud dinner party, one conversation holds full focus until something at the next table pulls attention away. In that instant, the brain doesn’t cut off the first voice like a switch flipping. New research finds that the brain briefly holds onto both voices at once, a split-second overlap scientists are only now mapping.

A study published in PLOS Biology found that when people shift focus from one speaker to another, the brain doesn’t fully let go of one voice before it grabs onto the next. For a brief moment, both get tracked at once.

That undercuts a widespread assumption: that tuning in and tuning out happen in lockstep. Picture a driver who starts turning the wheel before fully lifting a foot off the gas.

24 Adults Wore EEG Caps in a Simulated Cocktail Party

Researchers recruited 24 young adults with normal hearing and sat them in a circle of six loudspeakers to simulate a cocktail party. Two front speakers played separate TED Talk recordings, one male voice and one female voice, while four speakers behind played overlapping chatter as noise. Participants followed one front voice at a time, switching whenever an on-screen arrow flipped direction, with each stretch of listening lasting roughly 10 to 30 seconds. Electrode caps tracked brain activity throughout, a technique called EEG, letting researchers see which voice the brain was locked onto at any given instant. To confirm attention, researchers quizzed participants afterward on what they’d heard; they got it right about 86 percent of the time.

listening brain waves
The experimental setup, showing a subject with brain signal cap, listening to more than one conversation at once. (Credit: Prof. Alejandro López Valdés)

Engagement With a New Speaker Starts Before Disengagement Ends

By analyzing brain signals around each attention switch, researchers found something unexpected. The brain didn’t wait to finish letting go of one speaker before it started picking up the next. It began locking onto the new voice first, and finished doing so first, while its grip on the old voice was still fading in the background. The exact length of that overlap can’t be pinned down precisely, since it shifted depending on how the researchers sliced the data, but the order was consistent: engagement led, disengagement lagged.

Researchers also tracked alpha waves, brain rhythms linked to mental effort and focus. Activity in this range dipped noticeably after each switch cue, which the team reads as a sign the brain was working harder during the transition, though alpha waves aren’t a direct gauge of effort. That dip bottomed out slightly after the new voice had fully taken over, suggesting the heaviest mental lifting happens around the moment engagement finishes, not across the whole switch.

Brain May Reset Language Context After an Attention Switch

Beyond tracking which voice the brain followed, researchers asked a deeper question: when attention moves to a new speaker, does the brain carry forward what it built up about language, or start fresh?

To investigate, researchers used a large language model to calculate how predictable or surprising each word would be given accumulated context. They tested four theories. One assumed the brain carries all prior context forward, unaware a switch happened. Two others assumed the brain keeps context only from speakers it had actually attended to. A fourth, the Reset model, assumed context gets wiped after each switch.

Out of the four theories, Reset came closest to matching what the brain was actually doing, though the match was modest and inconsistent across every way the researchers checked it. A second measure of language surprise showed a weaker, shakier version of the same pattern. Taken together, the results offer tentative evidence that listeners rebuild word-by-word predictions after switching speakers, rather than picking up right where the old context left off. That doesn’t mean the brain wipes the conversation clean; listeners may still hold onto the gist even as moment-to-moment predictions start over. One caveat: the AI model used to test this, Mistral-7B-v0.1, predicts text, not speech, and may not handle a mid-conversation jump the way a human brain does.

Findings on Attention Switching Could Shape Future Hearing Aids

Most brain research on listening in noisy environments has focused on people holding attention steady while ignoring competing speakers. Real conversations shift constantly, and this study is among the first to closely examine those moments of transition.

These findings could carry practical relevance for hearing aids that try to follow a user’s shifting attention automatically. That said, the results shouldn’t be assumed to apply beyond the tested group: 24 young adults with normal hearing. How age, hearing loss, or cognitive differences might change these dynamics remains open, which future work should address.

Attention has long been treated as a spotlight that snaps from one thing to the next. This study suggests it’s closer to a dimmer switch with a lag built in, one that keeps a hand on the old conversation just long enough to make sure it’s really ready to let go.


Paper Notes

Limitations

This study’s participant group consisted of 24 young adults with normal hearing between the ages of 18 and 39 who were native English speakers, which limits how broadly the findings can be applied to older adults, people with hearing loss, or non-native speakers. Researchers excluded three participants from parts of the analysis because their brain responses during the attention switch did not meet a reliability threshold, leaving 21 participants in those portions of the study. Timing measurements for engagement and disengagement are also relative rather than absolute, since the exact neural timing of the switch cannot be determined with precision using sliding time windows, only the ordering and pattern. On the question of language context, the team notes that the AI tool used to model word predictability was not designed to mimic human neural processing, and differences between how the AI and the brain handle context may have influenced the results. Researchers also acknowledge that the instructed, task-based nature of the experiment makes it less naturalistic than real-world listening situations.

Funding and Disclosures

According to the paper, Sara Carta, Alejandro López Valdés, and Giovanni M. Di Liberto were supported by the William Demant Fonden under grants 21-0628 and 22-0552, and by Taighde Éireann – Research Ireland under grant No. 18/CRT/6223. Giovanni M. Di Liberto also received support from Research Ireland at ADAPT, the Research Ireland Centre for AI-Driven Digital Content Technology, at Trinity College Dublin under grant 13/RC/2106_P2. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors declared no competing interests. Emina Aličković and Johannes Zaar are affiliated with Eriksholm Research Centre, Oticon A/S.

Publication Details

Paper Title: Competing speech streams are simultaneously represented in the human cortex during attention switching | Authors: Sara Carta, Emina Aličković, Johannes Zaar, Alejandro López Valdés, Giovanni M. Di Liberto | Journal: PLOS Biology | Published: July 16, 2026 | DOI: https://doi.org/10.1371/journal.pbio.3003876


About StudyFinds Analysis

Called "brilliant," "fantastic," and "spot on" by scientists and researchers, our acclaimed StudyFinds Analysis articles are created using an exclusive AI-based model with complete human oversight by the StudyFinds Editorial Team. For these articles, we use an unparalleled LLM process across multiple systems to analyze entire journal papers, extract data, and create accurate, accessible content. Our writing and editing team proofreads and polishes each and every article before publishing. With recent studies showing that artificial intelligence can interpret scientific research as well as (or even better) than field experts and specialists, StudyFinds was among the earliest to adopt and test this technology before approving its widespread use on our site. We stand by our practice and continuously update our processes to ensure the very highest level of accuracy. Read our AI Policy (link below) for more information.

Our Editorial Process

StudyFinds publishes digestible, agenda-free, transparent research summaries that are intended to inform the reader as well as stir civil, educated debate. We do not agree nor disagree with any of the studies we post, rather, we encourage our readers to debate the veracity of the findings themselves. All articles published on StudyFinds are vetted by our editors prior to publication and include links back to the source or corresponding journal article, if possible.

Our Editorial Team

Steve Fink

Editor-in-Chief

John Anderer

Associate Editor

Leave a Comment