Effects of mnemonic load on cortical activity during visual working memory: Linking ongoing brain activity with evoked responses

https://doi.org/10.1016/j.ijpsycho.2013.04.001Get rights and content

Highlights

  • Event-related EEG activity was modulated by varying visual working memory load.

  • Multivariate analysis was used to examine co-variations between ERPs and ERS/ERD.

  • Alpha ERD co-varied with sustained event-related potentials over parietal cortex.

  • These co-variations are consistent with notion of asymmetric amplitude modulations.

Abstract

The mechanisms generating task-locked changes in cortical potentials remain poorly understood, despite a wealth of research. It has recently been proposed that ongoing brain oscillations are not symmetric, so that task-related amplitude modulations generate a baseline shift that does not average out, leading to slow event-related potentials. We test this hypothesis using multivariate methods to formally assess the co-variation between task-related evoked potentials and spectral changes in scalp EEG during a visual working memory task, which is known to elicit both evoked and sustained cortical activities across broadly distributed cortical regions. 64-channel EEG data were acquired from eight healthy human subjects who completed a visuo-spatial associative working memory task as memory load was parametrically increased from easy to hard. As anticipated, evoked activity showed a complex but robust spatio-temporal waveform maximally expressed bilaterally in the parieto-occipital and anterior midline regions, showing robust effects of memory load that were specific to the stage of the working memory trial. Similarly, memory load was associated with robust spectral changes in the theta and alpha range, throughout encoding in posterior regions and through maintenance and retrieval in anterior regions, consistent with the additional resources required for decision making in prefrontal cortex. Analysis of the relationship between event-related changes in slow potentials and cortical rhythms, using partial least squares, is indeed consistent with the notion that the former make a causal contribution to the latter.

Introduction

The relationship between ongoing activity and evoked responses has been a matter of debate for several decades. Two alternative mechanisms (additive activity and phase reset) have been widely debated as possible mechanisms for the generation of evoked responses (Boonstra et al., 2006, Fell et al., 2004, Makeig et al., 2004, Mäkinen et al., 2005, Sayers et al., 1974). In the classic additive model, afferent activity is linearly added to ongoing brain activity – which is treated as noise – and can be hence extracted by averaging across trials, suppressing the contribution from background activity that is not phase-locked to the stimulus. In contrast, the phase-reset model states that a stimulus induces a partial alignment of the phases of ongoing brain oscillations. Accordingly, averaging these phase-aligned oscillations over trials will result in an ERP. Although both models suggest opposing mechanisms for the generation of ERPs, they appear difficult to distinguish in experimental data (e.g., Ritter and Becker, 2009).

Recently, a third generative mechanism for event-related potentials has been proposed. Ongoing brain oscillations may not be symmetric: the peaks and troughs may have a different magnitude such that oscillations have a non-zero mean. Changes in the amplitude of ongoing oscillations will therefore result in a baseline shift that will not average out (Mazaheri and Jensen, 2008, Nikulin et al., 2007, Nikulin et al., 2010). This model implies that slow event-related responses are created as a direct consequence of amplitude modulations in brain oscillations. Indeed, a recent study showed that the cognitive modulations of alpha power and event-related responses were strongly correlated over subjects (van Dijk et al., 2010). Whereas the additive and phase-resetting models focus on early evoked components, the amplitude asymmetry model particularly focuses on late occurring components. Asymmetric alpha activity may result from an asymmetry of the intracellular currents that propagate forward and backward down the dendrites (Mazaheri and Jensen, 2010). Indeed, EEG and MEG signals are generally thought to result from these dendritic currents in pyramidal cells (Hamalainen et al., 1993). Mazaheri and Jensen (2010) point out that amplitude asymmetry can thus easily arise from physiological models of oscillatory activity and posit that amplitude asymmetry may in fact be the norm and amplitude symmetry the exception.

Multivariate analysis techniques are particularly suitable to disambiguate the contribution of these alternative mechanisms underlying event-related potentials (Turi et al., 2012). The causal link between amplitude modulations and event-related responses will manifest itself as consistent correlations across channels. The redundant information can be exploited by multivariate techniques to reliably quantify these relationships. In this paper we directly test the predicted correlations between amplitude modulations in ongoing brain oscillations and slow event-related responses using partial least squares (PLS). PLS is a regression technique that captures the covariance structure between two multidimensional data sets into orthogonal modes (McIntosh and Lobaugh, 2004). We use the event-related changes in the amplitude of oscillations at different frequencies as multivariate contrast, or latent variable, to predict the observed event-related potentials. In particular, we investigate the correlations between the experimental changes in both event-related activities to test the hypothesis of asymmetric amplitude modulations.

We acquired scalp EEG from healthy participants performing a visual working memory (VWM) task across three levels of difficulty by parametrically increasing mnemonic load. Visual working memory is a key cognitive process that involves active encoding, retention and retrieval of information in distributed cortical networks. Stimulus encoding occurs principally in posterior cortical regions, whereas strategy, working memory and decision making derive from prefrontal cortex (Kochan et al., 2011). The variety of processes and regions underlying VWM hence appeals as an attractive paradigm in which to study traditional ERPs, event-related spectral changes, and their relationship (Freunberger et al., 2009, Freunberger et al., 2011, Mitchell and Cusack, 2011). We decomposed event-related data into raw potentials and their spectra and studied the effect of memory load in each domain separately. Finally, we used partial least squares to study patterns of co-variation across scalp regions and frequency content.

Section snippets

Participants

Eight healthy volunteers (four females, mean age 24.9, SD 4.0) with no history of neurological or psychiatric disorders, participated as paid volunteers in this study. Informed consent was given in accordance with the National Health and Medical Research Council guidelines and the Human Research Ethics Committee of The University of New South Wales.

Experimental design

Subjects were seated in a light and sound attenuated room and completed an event-related visual working memory (VWM) paradigm that manipulated

Behavioral results

The effect of memory load on task performance was assessed through accuracy (percentage correct) and speed (reaction times). With increased level of memory load there was a significant decrease in correct responses from 93 ± 2.3% (± SEM) in easy to 61 ± 6.6% in hard (F = 11.7, P = 0.001; Fig. 2, top panel). The reaction time increased from 1.5 ± 0.08 s in easy to 2.4 ± 0.20 s in hard (F = 12.8, P < 0.0005; Fig. 2, bottom panel).

Event-related potentials

Event-related potentials were dominated by a typical visually evoked response over

Discussion

The recently proposed amplitude asymmetry model (Mazaheri and Jensen, 2008, Nikulin et al., 2007) suggests that amplitude modulations in brain oscillations underlie slow event-related responses, and hence predicts that task-locked changes in frequency spectra and cortical potentials should be correlated. In the present study we test this hypothesis, employing partial least squares to assess the cognitively induced co-variations between event-related potentials and spectral changes in a visual

Acknowledgments

This research was supported by the ARC Thinking Systems grant TS0669860; the National Health and Medical Research Council; BrainNRG collaborative award JSMF22002082, and the Netherlands Organization for Scientific Research (NWO #45110-030).

References (38)

  • A.J. Langdon et al.

    Multi-frequency phase locking in human somatosensory cortex

    Progress in Biophysics and Molecular Biology

    (2011)
  • S. Makeig et al.

    Mining event-related brain dynamics

    Trends in Cognitive Sciences

    (2004)
  • V. Mäkinen et al.

    Auditory event-related responses are generated independently of ongoing brain activity

    NeuroImage

    (2005)
  • P. Manganotti et al.

    Task-related coherence and task-related spectral power changes during sequential finger movements

    Electroencephalography and Clinical Neurophysiology

    (1998)
  • A.R. McIntosh et al.

    Partial least squares analysis of neuroimaging data: applications and advances

    NeuroImage

    (2004)
  • C. Neuper et al.

    Event-related dynamics of cortical rhythms: frequency-specific features and functional correlates

    International Journal of Psychophysiology

    (2001)
  • V.V. Nikulin et al.

    Non-zero mean and asymmetry of neuronal oscillations have different implications for evoked responses

    Clinical Neurophysiology

    (2010)
  • M. Petrides

    The mid-ventrolateral prefrontal cortex and active mnemonic retrieval

    Neurobiology of Learning and Memory

    (2002)
  • G. Pfurtscheller et al.

    Event-related EEG/MEG synchronization and desynchronization: basic principles

    Clinical Neurophysiology

    (1999)
  • View full text