Oscillatory correlates of auditory working memory examined with human electrocorticography

Abstract

This work examines how sounds are held in auditory working memory (AWM) in humans by examining oscillatory local field potentials (LFPs) in candidate brain regions. Previous fMRI studies by our group demonstrated blood oxygenation level-dependent (BOLD) response increases during maintenance in auditory cortex, inferior frontal cortex and the hippocampus using a paradigm with a delay period greater than 10s. The relationship between such BOLD changes and ensemble activity in different frequency bands is complex, and the long delay period raised the possibility that long-term memory mechanisms were engaged. Here we assessed LFPs in different frequency bands in six subjects with recordings from all candidate brain regions using a paradigm with a short delay period of 3 s. Sustained delay activity was demonstrated in all areas, with different patterns in the different areas. Enhancement in low frequency (delta) power and suppression across higher frequencies (beta/gamma) were demonstrated in primary auditory cortex in medial Heschl's gyrus (HG) whilst non-primary cortex showed patterns of enhancement and suppression that altered at different levels of the auditory hierarchy from lateral HG to superior- and middle-temporal gyrus. Inferior frontal cortex showed increasing suppression with increasing frequency. The hippocampus and parahippocampal gyrus showed low frequency increases and high frequency decreases in oscillatory activity. This work demonstrates sustained activity patterns during AWM maintenance, with prominent low-frequency increases in medial temporal lobe regions.

Keywords: Auditory working memory; Electrocorticography; Hippocampus; Neurophysiology; Oscillatory activity.

Fig. 1

Paradigm for working memory task. Subjects were cued to maintain a particular tone for a period of 3 s following the presentation of two different tones. They were then asked to compare the maintained tone with a target tone, and given feedback as to whether they were correct, followed by a rest period.

Fig. 2

Behavioral performance of all subjects in the task.

Fig. 3

Example intracranial data from a single patient (L275). a. Location of implanted electrodes, with an 8-channel depth electrode along the axis of HG, a single 96-channel grid over STG, MTG and IFG, a strip electrode overlying the PHG and a depth electrode into HC. b. Corresponding event-related spectral perturbations (ERSPs) recorded from the ROIs during the working memory task.

Fig. 4

Change in power during the delay period, for each canonical frequency band and each ROI, plotted across subjects (n = 6). Thick coloured lines represent median across subjects. Grey markers indicate individual subject data. Axes are scaled to the maximum value for each ROI.

Fig. 5

Median change in power across subjects, relative to baseline, plotted across the entire delay period for HGpm. Time referenced to start of delay period. Green shading indicates median absolute deviation (MAD), light grey lines show individual subject data, whilst red areas highlight periods of significance, determined using cluster-based permutation tests with an alpha level of 0.05. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6

Median change (±MAD) in power relative to baseline, across subjects and over time, in IFG. Red areas highlight periods of significance. Single subject data are overlaid (light grey lines). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7

Median change (±MAD) in power across subjects, relative to baseline, plotted across the entire delay period for HC (A) and PHG (B). Red areas highlight periods of significance. Single subject data are overlaid (light grey lines). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)