Gamma oscillations in gerbil auditory cortex during a target-discrimination task reflect matches with short-term memory

Abstract

Tightly stimulus-locked “evoked” and not tightly stimulus-locked “induced” gamma-band oscillations have been implicated in a variety of cognitive functions as well as more basic stimulus-related aspects of neuronal activity. The present study aimed at dissociating both aspects using a preparation in rodents trained to perform in a target-discrimination task while recording the intracerebral electrocorticogram in parallel from left and right auditory cortex. While stimulus-related aspects of gamma-band activity were already evident in the naïve subjects, aspects related to successful target-discrimination only emerged with learning. Frequency-modulated tones were employed as stimuli and were varied with respect to two stimulus dimensions, viz. “spectral content” and “modulation direction”. A target stimulus had to be discriminated in a GO/(NO-GO) paradigm from three non-target stimuli matching the target in only one or none of these dimensions. Analysis of the event-related potentials indicated that the physical stimulus parameters explained most of the variance in amplitude of most event-related potential components. For the evoked gamma-band activity no learning related modulations were found. The induced gamma-band activity showed a relationship with the similarity of a stimulus with the target that was most prominent in the right hemisphere. The correspondence between task-specific behavior and the amplitude of induced gamma-band activity, both developing with learning, supports a functional relevance of this form of oscillation beyond merely processing physical stimulus attributes.

Introduction

Fast (∼ 30–100 Hz) oscillations of electrical potential recorded within or near sensory neocortical structures (or analogues thereof) have been reported in a wide variety of species (e.g. Adrian, 1942, Bressler and Freeman, 1980, Gray and Singer, 1989, Franowicz and Barth, 1995, Brosch et al., 2002) including man (e.g. Chatrian et al., 1960, Kaiser and Lutzenberger, 2003a, Jensen et al., 2007). Hypotheses about the potential functional implications of these so-called gamma-band oscillations range from the mere epiphenomenal reflection of neuronal processes at other frequencies (Jürgens et al., 1995) to specific roles in information processing and cognition (Bressler, 1990) like selective attention (e.g. Womelsdorf and Fries, 2007), short-term memory (e.g. Tallon-Baudry et al., 1998, Kaiser et al., 2003b) or long-term memory (Lenz et al., 2007).

In order to address the diversity of proposed functions for gamma-band activity in a more coherent fashion, a framework was recently proposed (Herrmann et al., 2004) in which two generic processes, (1) the comparison of stimulus-related information with memory contents and (2) the utilization of signals derived from this comparison, are considered elemental to various cognitive functions. This “match-and-utilization model” assumes early (∼ 100 ms) gamma-band activity to reflect neuronal interaction processes underlying the “match” operation and later (∼ 300 ms) processes to reflect “utilization”-related operations. It is a prediction of this model that gamma-band activity would be modulated by the number of features that have to be compared during successful match-and-utilization operations. In a previous EEG study on humans, Herrmann and Mecklinger (2001) used visual stimuli characterized by two feature dimensions (form and colinearity) in a target detection task and observed maximum gamma-band activity when a stimulus matched the target in both dimensions, intermediate gamma-band activity when the stimulus matched the target in only one of the two dimensions, and weakest gamma-band activity when the stimulus did not match the target in either dimension.

In the present paper and the companion paper (Lenz et al., 2008) we report about an analogous experiment employing auditory stimuli in a rodent preparation and in humans, respectively. This parallel approach allows exploiting the complementary advantages of rodent and human subjects, viz. easy intracerebral recording from primary sensory cortex and monitoring of learning-effects on gamma-band in the rodent experiment, and whole scalp accessibility of EEG signals and straightforward instruction of subjects in the human experiment. Both, the transfer of the described experiment to another modality and the parallel investigation in two separate species aid in judging the general validity of the match-and-utilization model. As stimuli, both studies employed linearly frequency-modulated tones, because animal lesion studies have demonstrated that the discrimination of rising and falling frequency-modulated tones, traversing the same frequency interval is dependent on intact primary auditory cortex while, for example, discrimination of pure tone pitches is not (Ohl et al., 1999). Furthermore, in the rodent preparation we aimed at recording and analyzing data from both left and right auditory cortex as lesion studies in rodents have implied a preference of right auditory cortex for the processing of frequency-modulated tones (Wetzel et al., 1998a, Rybalko et al., 2006).

Since stimulus properties are also known to modulate both early, stimulus-locked and late, not tightly stimulus-locked, gamma-band activity (Busch et al., 2004), supposedly independently of cognitive functions, it is necessary to dissociate the influence of such “bottom-up” processes on gamma-band activity from those “top-down” influences associated with task-specific cognitive operations. In the companion study on humans (Lenz et al., 2008) this issue is addressed using stimulus normalization techniques, i.e. controlling for potential confounds due to stimulus intensity (as shown by Schadow et al., 2007) by using a fixed intensity relative to individual hearing threshold. In the present report on the rodent preparation we use the fact that animal subjects have to be trained in order to perform in the task. The naïve state of a subject, before the task is learned, therefore provides a natural situation of unsuccessful match-and-utilization operations in which only stimulus-related bottom-up influences, but not task-related top-down processes can modulate gamma-band activity. Correspondingly, when the animal subject is well trained, gamma-band activity might also reflect physiological correlates of successful match-and-utilization operations.