Representation of CV-sounds in cat primary auditory cortex: intensity dependence

Abstract

The level-dependent representation of simple speech sounds in cat primary auditory cortex (AI) is explored in naive cats and in animals that have been exposed to these sounds in behavioral detection and discrimination tasks. Population analyses of multiple unit responses in the form of post-stimulus time histograms (PSTHs), neurograms, and spatial distribution were made for synthetic consonant–vowel sounds across AI. The temporal profile of cortical responses was robust across neurons, characterized by brief phasic responses at the onset of consonantal burst and voicing. The spectral profile of the sounds, i.e., the formant structure, was only weakly expressed in the response magnitude across characteristic frequency. The spatial response distribution across AI was discontinuous, and consisted of several patches of activation. Intensity-dependence in the spatial activity distribution was more strongly expressed than in population PSTHs and neurograms. Differences attributable to behavioral training were observed for rate-encoding and temporal encoding of speech sounds.

Introduction

In the analysis of neural representations of speech-like sounds and animal vocalizations, the spectral and temporal domains of the acoustic signal have emerged as the basic stimulus features represented at various stages of the auditory nervous system. Spectral and temporal characteristics of vowel sounds are also manifest in the firing rate and temporal response pattern of cat auditory nerve fibers (e.g., Delgutte and Kiang, 1984; Sachs and Young, 1979; Sinex and Geisler, 1983), cochlear nucleus neurons of cats (e.g., Blackburn and Sachs, 1990; Wang and Sachs, 1994) and in the inferior colliculus (Chen et al., 1996). In the primary auditory cortex (AI) in marmoset monkeys, spectro-temporal characteristics of species-specific vocalizations are integrated temporally and spectrally and represented by the synchronization of neural activity from spatially dispersed cortical cell assemblies (Wang et al., 1995; Nagarajan et al., 2002).

In the study of communication calls, the issue of behavioral relevance of the vocalization has been proposed as a factor in shaping neural representations of stimulus features at the cortex due to significant influences from auditory environment and experience (Merzenich et al., 1988, Merzenich et al., 1990; Wang, 2000; Wang and Kadia, 2001). A perceptual feature of speech syllables, voice-onset time (VOT), can be represented by time-locked activity of neurons in monkeys AI (Steinschneider et al., 1982, Steinschneider et al., 1990, Steinschneider et al., 1994, Steinschneider et al., 1992, 2000) and cats (Eggermont, 1995). In these experimental studies, however, animals were not trained or exposed to the speech stimuli to give these physiological responses behavioral salience. Thus, the findings from such studies reflect the ability of subsets of AI neurons to respond to changes in the acoustics of the speech stimuli without specific adaptation of the neural circuitry to the sounds. Addition of a behavioral component to physiological experiments examining the neural representations of speech stimuli provides behavioral salience of the sounds and may reveal specific adaptations to trained stimuli.

Therefore, the first objective in this study was to determine the distributed representation of speech syllables across many cells in AI; and second, to explore how behavioral training modifies the distributed and cumulative cortical representation of speech sounds.