Prediction Dynamically Modulates Gain and Width of Population Frequency Tuning in Human Auditory Cortex

During perception, the brain must veridically portray sensory representations while simultaneously highlighting unexpected events. According to two-process theory of predictive perception, this balance is achieved via complementary mechanisms: precision-weighted predictions sharpen sensory tuning around expected features, while momentary surprisal scales response gain, dampening expected inputs and amplifying unexpected violations. However, standard approaches to investigate these processes, such as categorical multivariate decoding, primarily capture shifts in overall signal-to-noise, making it difficult to disentangle changes in population tuning width from modulations in response gain. Here, we combine ultra-high-field functional magnetic resonance imaging (7T fMRI) with population receptive field (pRF) modelling and a continuous Bayesian ideal observer to test how expectations dynamically modulate frequency tuning across the human auditory hierarchy. We first established static pRF preferences and tuning widths independently of predictive inference. We then separately compared this static pRF tuning against two dynamically modulated alternatives. The first model evaluated how tone-by-tone precision-weighted predictions dynamically sharpen and broaden pRF tuning width, while the second model evaluated how gain (by amplification and dampening) was modulated by tone-by-tone information-theoretic surprisal. Across primary, belt, and associative auditory cortices, both a sharpening model driven by precision-weighted predictions and a gain model driven by surprisal explained significantly more variance compared to a static baseline model. Counterfactual models inverting these driving features yielded no improvement, ruling out non-specific model mimicry. Additionally, contrasting local and global tonotopic implementations revealed a feature-selective dissociation: whereas precision-weighted predictions locally sharpened pRFs tuned to expected frequencies, surprisal acted as a global broadcast signal that multiplicatively scaled pRF gain across all frequency channels of the tonotopic map. Together, these findings demonstrate that the auditory system continuously balances selectivity for expected structure and responsiveness to surprising input through local sharpening and global gain modulation.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-16
DOI
https://doi.org/10.64898/2026.11.16.766995
Primary Topic
Neuroscience and Music Perception
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Prediction Dynamically Modulates Gain and Width of Population Frequency Tuning in Human Auditory Cortex

Ryszard Auksztulewicz, Jorie van Haren, Sonja A. Kotz, Federico De Martino et al.
bioRxiv (Cold Spring Harbor Laboratory)
Neuroscience and Music Perception
preprint

Prediction Dynamically Modulates Gain and Width of Population Frequency Tuning in Human Auditory Cortex

Ryszard Auksztulewicz, Jorie van Haren, Sonja A. Kotz, Federico De Martino, Floris P. de Lange
preprint en

Abstract

During perception, the brain must veridically portray sensory representations while simultaneously highlighting unexpected events. According to two-process theory of predictive perception, this balance is achieved via complementary mechanisms: precision-weighted predictions sharpen sensory tuning around expected features, while momentary surprisal scales response gain, dampening expected inputs and amplifying unexpected violations. However, standard approaches to investigate these processes, such as categorical multivariate decoding, primarily capture shifts in overall signal-to-noise, making it difficult to disentangle changes in population tuning width from modulations in response gain. Here, we combine ultra-high-field functional magnetic resonance imaging (7T fMRI) with population receptive field (pRF) modelling and a continuous Bayesian ideal observer to test how expectations dynamically modulate frequency tuning across the human auditory hierarchy. We first established static pRF preferences and tuning widths independently of predictive inference. We then separately compared this static pRF tuning against two dynamically modulated alternatives. The first model evaluated how tone-by-tone precision-weighted predictions dynamically sharpen and broaden pRF tuning width, while the second model evaluated how gain (by amplification and dampening) was modulated by tone-by-tone information-theoretic surprisal. Across primary, belt, and associative auditory cortices, both a sharpening model driven by precision-weighted predictions and a gain model driven by surprisal explained significantly more variance compared to a static baseline model. Counterfactual models inverting these driving features yielded no improvement, ruling out non-specific model mimicry. Additionally, contrasting local and global tonotopic implementations revealed a feature-selective dissociation: whereas precision-weighted predictions locally sharpened pRFs tuned to expected frequencies, surprisal acted as a global broadcast signal that multiplicatively scaled pRF gain across all frequency channels of the tonotopic map. Together, these findings demonstrate that the auditory system continuously balances selectivity for expected structure and responsiveness to surprising input through local sharpening and global gain modulation.

bioRxiv (Cold Spring Harbor Laboratory)
Radboud University Nijmegen (NL), Maastricht University (NL)
Neuroscience and Music Perception
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon