Summary: Researchers report that familiar voices are easier to understand than those of strangers, and this intelligibility advantage persists even when the familiar voice has been acoustically altered so that listeners no longer explicitly recognize who is speaking.
Source: APS.
Familiar voices are easier to understand, and this advantage remains even when listeners cannot consciously identify the speaker, according to research published in Psychological Science.
Researchers associated with Western University’s BrainsCAN initiative examined how familiarity with a speaker influences both voice recognition and speech intelligibility. Their results show that people extract different kinds of information from a voice depending on whether the task is to identify the speaker or to comprehend the words they say. In practical terms, a familiar voice can be more intelligible than an unfamiliar one, even when acoustic manipulations make the voice unrecognizable.
Emma Holmes (University College London), the study’s first author, explains that listeners emphasize different acoustic cues when trying to recognize a known person versus when trying to understand speech. The team—Holmes, Ingrid S. Johnsrude, and Ysabel Domingo—focused on two stable acoustic properties that vary between individuals: pitch (related to glottal pulse rate) and resonance (related to vocal tract length). They asked whether these features contribute differently to the ability to recognize a voice and to the ability to understand speech from that voice under competing conditions.
The study recruited 11 pairs of participants who were friends or partners, had known each other for at least six months, and spoke regularly. Each participant read a set of standardized sentences aloud that followed a fixed structure (name, verb, number, adjective, noun; e.g., “Bob bought five green bags”). The researchers digitally manipulated the recordings to systematically alter pitch and resonance for each speaker, creating versions that were acoustically modified but preserved the original linguistic content.
In testing, listeners heard sentences produced by their partner and by unfamiliar speakers matched for gender. Two tasks were used. In the intelligibility task, listeners heard two sentences simultaneously—one from the target talker and one from a competing speaker—and were asked to report words from the target sentence. In the recognition task, listeners heard individual sentences and indicated whether the voice belonged to their partner.
Results aligned with expectations in some respects: participants most reliably recognized their partner’s voice when they heard unaltered recordings, and recognition suffered when the recordings were manipulated. Pitch changes reduced recognition less than changes in resonance, indicating that resonance is a particularly salient cue for identifying who is speaking.
Crucially, even when resonance was altered enough to eliminate explicit recognition of the partner’s voice, listeners still found the familiar speaker easier to understand in the presence of competing speech. In other words, the intelligibility advantage for familiar voices can persist independently of conscious voice recognition. This dissociation suggests that different perceptual systems or representations support recognition versus comprehension: some voice features may boost intelligibility without contributing to explicit identification.
The study highlights resonance as a key acoustic cue for voice identity, while both pitch and resonance can influence speech understanding. The findings deepen our understanding of how listeners process voice information and how experience with a talker shapes perception. They also have practical implications: people with hearing loss, who face particular challenges when speech competes with background noise, may benefit more from familiar voices. The results could also inform the design of voice-enabled technologies—such as virtual assistants and robots—by showing which acoustic properties support intelligibility and which support identity recognition.
Funding: The study was supported by a Natural Sciences and Engineering Research Council of Canada Discovery Grant, a Canadian Institutes of Health Research Operating Grant, and BrainsCAN, Western University’s Canada First Research Excellence Fund program in cognitive neuroscience.
Source: Anna Mikulak – APS
Publisher: NeuroscienceNews.com (organized coverage)
Image Source: Image adapted from the APS news release.
Original Research: Holmes, E., Domingo, Y., & Johnsrude, I. S., “Familiar Voices Are More Intelligible, Even if They Are Not Recognized as Familiar,” Psychological Science. Published August 10, 2018. DOI: 10.1177/0956797618779083.
Abstract
Familiar Voices Are More Intelligible, Even if They Are Not Recognized as Familiar
We can recognize familiar people by their voices, and familiar talkers are more intelligible than unfamiliar talkers when competing talkers are present. However, whether the acoustic voice characteristics that permit recognition and those that benefit intelligibility are the same or different is unknown. Here, pairs of participants who had known each other for six months or longer were recorded while speaking, and the researchers manipulated two acoustic correlates—vocal tract length and glottal pulse rate. These manipulations had different effects on explicit recognition and on the intelligibility benefit from familiar voices. Even when explicit recognition was eliminated by altering resonance, familiar voices remained more intelligible than unfamiliar ones, demonstrating that intelligibility benefits do not require explicit recognition. Processing familiar-voice information therefore appears to depend on multiple, at least partly independent, mechanisms that are engaged depending on the listener’s perceptual goal.