Summary: New research reveals that when people hear speech at different speeds, the auditory cortex does not change its internal timing to match the sped-up or slowed-down speech. Instead, it processes sound using a fixed temporal window. This challenges the long-standing belief that the brain simply speeds up or slows down its processing to follow speech rhythms.
Using high-precision intracranial electrode recordings from patients, the study shows that comprehension relies on higher-level brain areas interpreting a steady, time-locked stream of auditory information produced by the auditory cortex. These results refine our understanding of speech processing and could inform future approaches to diagnosing and treating speech and language disorders, as well as improving computational models of language.
Key Facts
- Fixed Processing: The auditory cortex integrates incoming sounds over a consistent temporal window regardless of whether speech is sped up or slowed.
- Direct Neural Recordings: Intracranial electrodes implanted for clinical monitoring in epilepsy patients provided precise timing data from regions close to active neurons.
- Clinical and Modeling Impact: These findings constrain neurocomputational models of speech processing and may help target therapies for people with comprehension difficulties.
Source: University of Rochester
You have 20 minutes to spare but your favorite podcast episode is a few minutes longer. Speeding up playback helps you fit it in. But what does that do to your brain?
Intuitively, one might expect the brain to speed up its internal processing when speech is faster and slow down when the speech is drawn out. The new study, published in Nature Neuroscience, finds that the auditory cortex does not do that. Instead, it responds according to a consistent, absolute time scale.
The research, led by Sam Norman-Haignere, PhD, at the University of Rochester’s Del Monte Institute for Neuroscience, was carried out in collaboration with teams at Columbia University, including principal investigator Nima Mesgarani, PhD, of the Zuckerman Institute, and Menoua Keshishian, who completed his PhD in Electrical Engineering in Mesgarani’s lab.
“We were surprised to find that slowing down a word doesn’t change the time window the auditory cortex uses,” said Norman-Haignere, the study’s first author. “The auditory cortex appears to integrate incoming sound over a fixed time scale.”
The researchers emphasize that understanding this fixed temporal processing is important for building better computational models of how the brain converts sound to linguistic meaning and for identifying what goes wrong in speech comprehension disorders.
The challenges of modeling speech and brain function
The auditory cortex contains multiple interconnected regions—including primary and secondary auditory areas—that work together to process sound. Beyond these, additional language regions interpret the auditory input and extract meaning. How each region contributes and how they are organized hierarchically, however, remains only partially understood.
Computational models have been essential tools for exploring these complexities. Such models use mathematical or algorithmic frameworks to predict neural responses and behavior from auditory input. The authors used modeling both to test their experimental approach and to evaluate competing hypotheses: does the auditory cortex integrate across fixed durations of time, or does it flexibly track speech structures such as syllables and words?
Some models trained by the team learned to integrate across speech structures, but the real cortical recordings did not show the same behavior. This discrepancy helped validate the experimental method and strengthened the conclusion that the auditory cortex is predominantly time-yoked rather than structure-driven.
Accessing precise neural signals in humans
Common noninvasive methods such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) have been indispensable in human neuroscience, but they lack the spatial and/or temporal precision to resolve activity at the scale of neuronal populations. To capture fine-grained timing, the researchers worked with patients undergoing clinical monitoring for epilepsy at NYU Langone Medical Center, Columbia University Irving Medical Center, and the University of Rochester Medical Center.
These patients had intracranial electrodes temporarily implanted to help localize seizure foci. These electrodes measure electrical activity close to active neurons, providing far greater spatial and temporal resolution than scalp EEG or fMRI. During recordings, participants listened to the same audiobook passage at normal speed and at slower speeds while researchers measured neural responses across auditory cortex regions.
The team expected to observe changes in the cortical integration window corresponding to the altered speech rate. Instead, they found little to no change: the auditory cortex’s integration window remained tied to physical time—on the order of tens to hundreds of milliseconds—rather than to the durations of speech units like phonemes or words.
“This challenges the intuitive notion that cortical processing is synchronized to speech units such as syllables,” said Mesgarani, a senior author. “Rather, the auditory cortex provides a consistently timed stream of information that downstream regions must interpret to extract linguistic content.”
Norman-Haignere added, “A clearer picture of how auditory and language systems transform sound into meaning will help us understand processing deficits and guide the development of better diagnostic tools and treatments.”
Funding and collaborators: Other contributing researchers included Guy McKhann and Catherine Schevon of Columbia University, and Orrin Devinsky, Werner Doyle, and Adeen Flinker of NYU Langone Medical Center. The work received support from the National Institutes of Health and a Marie-Josee and Henry R. Kravis grant.
About this auditory neuroscience research news
Author: Kelsie Smith Hayduk
Source: University of Rochester
Contact: Kelsie Smith Hayduk – University of Rochester
Image: The image is credited to Neuroscience News
Original Research: Open access.
“Temporal integration in human auditory cortex is predominantly yoked to absolute time” by Sam Norman-Haignere et al. Nature Neuroscience
Abstract
Temporal integration in human auditory cortex is predominantly yoked to absolute time
Sound elements such as phonemes and words vary widely in duration. This creates a distinction between integrating information across absolute time intervals (for example, ~100 ms) versus integrating across variable speech structures (for example, phonemes or words). Past auditory and cognitive models have emphasized either time-based or structure-based integration, but it has been unclear which principle predominates in human cortex.
To address this, the authors manipulated speech duration by stretching and compressing spoken passages and measured cortical integration windows using a combined experimental and computational approach applied to spatiotemporally precise intracranial recordings. The results showed a slight increase in integration windows for stretched speech, but the increase was very small (about 5%) compared with the large changes in structure durations. Even non-primary auditory regions implicated in speech processing showed only minimal lengthening.
These findings indicate that time-yoked computations dominate across human auditory cortex, placing important constraints on theories and models that aim to explain how the brain processes spoken structure.