Changes of AI receptive fields with sound density

David Trumbull Blake, Michael M. Merzenich

Research output: Contribution to journalArticle

68 Citations (Scopus)

Abstract

Primates engage in auditory behaviors under a broad range of signal-to-noise conditions. In this study, optimal linear receptive fields were measured in alert primate primary auditory cortex (A1) in response to stimuli that vary in spectrotemporal density. As density increased, A1 excitatory receptive fields systematically changed. Receptive field sensitivity, expressed as the expected change in firing rate after a tone pip onset, decreased by an order of magnitude. Spectral selectivity more than doubled. Inhibitory subfields, which were rarely recorded at low sound densities, emerged at higher sound densities. The ratio of excitatory to inhibitory population strength changed from 14.4:1 to 1.4:1. At low sound densities, the sound associated with the evocation of an action potential from an A1 neuron was broad in spectrum and time. At high sound densities, a spike-evoking sound was more likely to be a spectral or temporal edge and was narrower in time and frequency range. Receptive fields were used to predict responses to a novel high-noise-density stimulus. The predictions were highly correlated with the actual responses to the 2-s complex sound excerpt. The structure of prediction failures revealed that neurons with prominent inhibitory fields had relatively poor linear predictions. Further, the finding that stochastic variance is limiting in prediction even after averaging 150 repetitions means that high-fidelity representations of simple sounds in A1 must be distributed over at least hundreds of neurons. Auditory context alters A1 responses across multiple parameter spaces; this presents a challenge for reconstructing neural codes.

Original languageEnglish (US)
Pages (from-to)3409-3420
Number of pages12
JournalJournal of Neurophysiology
Volume88
Issue number6
DOIs
StatePublished - Dec 1 2002
Externally publishedYes

Fingerprint

Neurons
Primates
Noise
Structure Collapse
Auditory Cortex
Action Potentials
Population

ASJC Scopus subject areas

  • Neuroscience(all)
  • Physiology

Cite this

Changes of AI receptive fields with sound density. / Blake, David Trumbull; Merzenich, Michael M.

In: Journal of Neurophysiology, Vol. 88, No. 6, 01.12.2002, p. 3409-3420.

Research output: Contribution to journalArticle

Blake, David Trumbull ; Merzenich, Michael M. / Changes of AI receptive fields with sound density. In: Journal of Neurophysiology. 2002 ; Vol. 88, No. 6. pp. 3409-3420.
@article{afeb9423c5c54393ad72dbb26c6eb89f,
title = "Changes of AI receptive fields with sound density",
abstract = "Primates engage in auditory behaviors under a broad range of signal-to-noise conditions. In this study, optimal linear receptive fields were measured in alert primate primary auditory cortex (A1) in response to stimuli that vary in spectrotemporal density. As density increased, A1 excitatory receptive fields systematically changed. Receptive field sensitivity, expressed as the expected change in firing rate after a tone pip onset, decreased by an order of magnitude. Spectral selectivity more than doubled. Inhibitory subfields, which were rarely recorded at low sound densities, emerged at higher sound densities. The ratio of excitatory to inhibitory population strength changed from 14.4:1 to 1.4:1. At low sound densities, the sound associated with the evocation of an action potential from an A1 neuron was broad in spectrum and time. At high sound densities, a spike-evoking sound was more likely to be a spectral or temporal edge and was narrower in time and frequency range. Receptive fields were used to predict responses to a novel high-noise-density stimulus. The predictions were highly correlated with the actual responses to the 2-s complex sound excerpt. The structure of prediction failures revealed that neurons with prominent inhibitory fields had relatively poor linear predictions. Further, the finding that stochastic variance is limiting in prediction even after averaging 150 repetitions means that high-fidelity representations of simple sounds in A1 must be distributed over at least hundreds of neurons. Auditory context alters A1 responses across multiple parameter spaces; this presents a challenge for reconstructing neural codes.",
author = "Blake, {David Trumbull} and Merzenich, {Michael M.}",
year = "2002",
month = "12",
day = "1",
doi = "10.1152/jn.00233.2002",
language = "English (US)",
volume = "88",
pages = "3409--3420",
journal = "Journal of Neurophysiology",
issn = "0022-3077",
publisher = "American Physiological Society",
number = "6",

}

TY - JOUR

T1 - Changes of AI receptive fields with sound density

AU - Blake, David Trumbull

AU - Merzenich, Michael M.

PY - 2002/12/1

Y1 - 2002/12/1

N2 - Primates engage in auditory behaviors under a broad range of signal-to-noise conditions. In this study, optimal linear receptive fields were measured in alert primate primary auditory cortex (A1) in response to stimuli that vary in spectrotemporal density. As density increased, A1 excitatory receptive fields systematically changed. Receptive field sensitivity, expressed as the expected change in firing rate after a tone pip onset, decreased by an order of magnitude. Spectral selectivity more than doubled. Inhibitory subfields, which were rarely recorded at low sound densities, emerged at higher sound densities. The ratio of excitatory to inhibitory population strength changed from 14.4:1 to 1.4:1. At low sound densities, the sound associated with the evocation of an action potential from an A1 neuron was broad in spectrum and time. At high sound densities, a spike-evoking sound was more likely to be a spectral or temporal edge and was narrower in time and frequency range. Receptive fields were used to predict responses to a novel high-noise-density stimulus. The predictions were highly correlated with the actual responses to the 2-s complex sound excerpt. The structure of prediction failures revealed that neurons with prominent inhibitory fields had relatively poor linear predictions. Further, the finding that stochastic variance is limiting in prediction even after averaging 150 repetitions means that high-fidelity representations of simple sounds in A1 must be distributed over at least hundreds of neurons. Auditory context alters A1 responses across multiple parameter spaces; this presents a challenge for reconstructing neural codes.

AB - Primates engage in auditory behaviors under a broad range of signal-to-noise conditions. In this study, optimal linear receptive fields were measured in alert primate primary auditory cortex (A1) in response to stimuli that vary in spectrotemporal density. As density increased, A1 excitatory receptive fields systematically changed. Receptive field sensitivity, expressed as the expected change in firing rate after a tone pip onset, decreased by an order of magnitude. Spectral selectivity more than doubled. Inhibitory subfields, which were rarely recorded at low sound densities, emerged at higher sound densities. The ratio of excitatory to inhibitory population strength changed from 14.4:1 to 1.4:1. At low sound densities, the sound associated with the evocation of an action potential from an A1 neuron was broad in spectrum and time. At high sound densities, a spike-evoking sound was more likely to be a spectral or temporal edge and was narrower in time and frequency range. Receptive fields were used to predict responses to a novel high-noise-density stimulus. The predictions were highly correlated with the actual responses to the 2-s complex sound excerpt. The structure of prediction failures revealed that neurons with prominent inhibitory fields had relatively poor linear predictions. Further, the finding that stochastic variance is limiting in prediction even after averaging 150 repetitions means that high-fidelity representations of simple sounds in A1 must be distributed over at least hundreds of neurons. Auditory context alters A1 responses across multiple parameter spaces; this presents a challenge for reconstructing neural codes.

UR - http://www.scopus.com/inward/record.url?scp=0036898602&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0036898602&partnerID=8YFLogxK

U2 - 10.1152/jn.00233.2002

DO - 10.1152/jn.00233.2002

M3 - Article

VL - 88

SP - 3409

EP - 3420

JO - Journal of Neurophysiology

JF - Journal of Neurophysiology

SN - 0022-3077

IS - 6

ER -