Review
Revisiting the role of persistent neural activity during working memory

https://doi.org/10.1016/j.tics.2013.12.001Get rights and content

Highlights

  • Recent work suggests that the standard neural model of WM is in need of revision.

  • Multivariate analyses find that sensory cortex stores WM information.

  • lPFC simultaneously encodes multiple task variables to bias sensory regions.

  • WM is mediated by dynamic population codes in addition to persistent neural activity.

What are the neural mechanisms underlying working memory (WM)? One influential theory posits that neurons in the lateral prefrontal cortex (lPFC) store WM information via persistent activity. In this review, we critically evaluate recent findings that together indicate that this model of WM needs revision. We argue that sensory cortex, not the lPFC, maintains high-fidelity representations of WM content. By contrast, the lPFC simultaneously maintains representations of multiple goal-related variables that serve to bias stimulus-specific activity in sensory regions. This work highlights multiple neural mechanisms supporting WM, including temporally dynamic population coding in addition to persistent activity. These new insights focus the question on understanding how the mechanisms that underlie WM are related, interact, and are coordinated in the lPFC and sensory cortex.

Introduction

WM comprises the set of operations that support the active retention of behaviorally relevant information over brief intervals. Given the central role of WM in goal-directed behavior, establishing the neural basis of WM has been a priority of neuroscience research. Early WM studies observed that selective increases in neural activity during the presentation of a to-be-maintained sample item persisted throughout the blank ‘delay’ interval of a WM delay task, bridging the temporal gap between the sample and the subsequent contingent response 1, 2. This work inspired the theoretical framework that has predominated in the field: neurons or neuronal populations that are selectively tuned to the to-be-remembered information hold this information in an active state through persistent activation [3]. We refer to this model, which emphasizes stable persistent neural activity (see Glossary) in selective neurons as the fixed-selectivity model. Motivated by this model, functional MRI (fMRI) studies in humans and electrophysiological studies in monkeys have consistently identified persistent neural activity in the lPFC, leading many to conclude that the lPFC stores representations of WM memoranda.

A decade ago, we provided a critique of the literature on persistent activity in the context of contemporary models of prefrontal cortical function [4]. We hypothesized that, in contrast to existing theories of WM, persistent lPFC activity signifies attention directed to internal representations maintained in sensory cortices. Viewed through the lens of the fixed-selectivity model, evidence for this proposal is limited. Studies of sensory and motor function, however, suggest that information is likely to be represented through the combined activity of neural populations with diverse tuning properties rather than individual highly-tuned neurons 5, 6. This notion offers a promising framework for understanding WM.

In recent years, analytic and methodological advances (Box 1) have expanded researchers’ ability to capture the multivariate nature of population coding and the causal relationships between neural activity and behavior. The findings generated using these approaches underscore the need for a revision of existing views of WM. In light of these results, we revisit the issue of how information remains active during WM. The studies we discuss here focus on visual WM, but the general principles discussed herein apply to WM in other modalities.

Section snippets

Evidence for persistent WM representations in visual cortex

Neurons in visual cortex are selectively tuned to visual stimulus features and are consequently well suited for maintaining high-fidelity representations of visual information in the service of WM [7]. Yet, from the perspective of the fixed-selectivity model, evidence for sustained WM representations in visual cortex has been equivocal. Although sustained responses have been observed in temporal cortex [8], studies typically describe transient neural responses to sample stimuli without any

The role of the lPFC in WM

The most pervasive observation in the WM literature is that lPFC activity persists throughout WM maintenance. This finding has been interpreted as evidence that lPFC delay activity encodes sensory features of WM items [3]. However, in addition to displaying coarse selectivity for WM items and features [33], lPFC activity exhibits selectivity for a broad range of task variables during the delay period of WM tasks. For example, lPFC neurons show differential preferences for task rules [34],

Persistent neural activity revisited

Persistent neural activity, particularly in the lPFC, has become synonymous with WM. However, this equivalence is misleading. First, the lPFC does not appear to be privileged in its ability to generate persistent activity. Particularly when analyses focus on neurons or voxels that are highly stimulus selective, persistent neural activity can be observed nearly everywhere in the brain 8, 64, 65, 66. Second, although persistent neural activity is a key mechanism for forming temporal links between

Concluding remarks

An understanding of the neural mechanisms underlying WM is critical for gaining insight into the wide range of goal-directed behaviors supported by WM. In this review, we present a perspective on WM that emphasizes the notion of distributed population activity in encoding WM information. Methodological advances in the past ten years, and in the past few years in particular, have highlighted the sensory nature of sustained WM information in sensory cortices and the high-dimensional nature of

Acknowledgments

This work was supported by grants from the National Institutes of Health (R01 MH63901 to M.D. and R01 EY016407 and R03 MH097206 to C.E.C.).

Glossary

Delay tasks
the experimental paradigm typically used to study the neural basis of working memory (WM). A trial in a delay task begins with a brief presentation of a sample item. The subject encodes this item into WM and maintains this item over a blank ‘delay’ period of a few to several seconds. At the end of the delay period, a probe stimulus appears and the subject initiates a behavioral response contingent on the WM representation of the sample item. A key feature of delay tasks is that they

References (96)

  • B.R. Postle

    Working memory as an emergent property of the mind and brain

    Neuroscience

    (2006)
  • A.J-W. Chen

    Goal-directed attention alters the tuning of object-based representations in extrastriate cortex

    Front. Hum. Neurosci.

    (2012)
  • J.M. Fuster

    The prefrontal cortex – an update: time is of the essence

    Neuron

    (2001)
  • J. Driver

    Concurrent brain-stimulation and neuroimaging for studies of cognition

    Trends Cogn. Sci.

    (2009)
  • C.E. Curtis et al.

    Beyond working memory: the role of persistent activity in decision making

    Trends Cogn. Sci.

    (2010)
  • G. Deco

    Brain mechanisms for perceptual and reward-related decision-making

    Prog. Neurobiol.

    (2013)
  • A. Afshar

    Single-trial neural correlates of arm movement preparation

    Neuron

    (2011)
  • S. Druckmann et al.

    Neuronal circuits underlying persistent representations despite time varying activity

    Curr. Biol.

    (2012)
  • K. Jimura et al.

    Analyses of regional-average activation and multivoxel pattern information tell complementary stories

    Neuropsychologia

    (2012)
  • M.T. Todd

    Confounds in multivariate pattern analysis: theory and rule representation case study

    Neuroimage

    (2013)
  • J.T. Serences et al.

    Computational advances towards linking BOLD and behavior

    Neuropsychologia

    (2012)
  • S. Rossi

    Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research

    Clin. Neurophysiol.

    (2009)
  • R.T. Canolty et al.

    The functional role of cross-frequency coupling

    Trends Cogn. Sci.

    (2010)
  • J. Jacobs et al.

    Direct brain recordings fuel advances in cognitive electrophysiology

    Trends Cogn. Sci.

    (2010)
  • S.L. Bressler et al.

    Large-scale brain networks in cognition: emerging methods and principles

    Trends Cogn. Sci.

    (2010)
  • J.M. Fuster et al.

    Neuron activity related to short-term memory

    Science

    (1971)
  • S. Funahashi

    Mnemonic coding of visual space in the monkey's dorsolateral prefrontal cortex

    J. Neurophysiol.

    (1989)
  • A. Georgopoulos

    Neuronal population coding of movement direction

    Science

    (1986)
  • A.B.A. Graf

    Decoding the activity of neuronal populations in macaque primary visual cortex

    Nat. Neurosci.

    (2011)
  • T. Pasternak et al.

    Working memory in primate sensory systems

    Nat. Rev. Neurosci.

    (2005)
  • R.O. Duda

    Pattern Classification

    (2001)
  • C.P. Hung

    Fast readout of object identity from macaque inferior temporal cortex

    Science

    (2005)
  • F. Tong et al.

    Decoding patterns of human brain activity

    Annu. Rev. Psychol.

    (2012)
  • G.J. Brouwer et al.

    Decoding and reconstructing color from responses in human visual cortex

    J. Neurosci.

    (2009)
  • S.A. Harrison et al.

    Decoding reveals the contents of visual working memory in early visual areas

    Nature

    (2009)
  • J.T. Serences

    Stimulus-specific delay activity in human primary visual cortex

    Psychol. Sci.

    (2009)
  • E.F. Ester

    A neural measure of precision in visual working memory

    J. Cogn. Neurosci.

    (2013)
  • Y. Xing

    Decoding working memory of stimulus contrast in early visual cortex

    J. Neurosci.

    (2013)
  • D.E.J. Linden

    Mapping brain activation and information during category-specific visual working memory

    J. Neurophysiol.

    (2011)
  • T.B. Christophel

    Decoding the contents of visual short-term memory from human visual and parietal cortex

    J. Neurosci.

    (2012)
  • L. Woloszyn et al.

    Neural dynamics in inferior temporal cortex during a visual working memory task

    J. Neurosci.

    (2009)
  • K.K. Sreenivasan

    Distributed and dynamic storage of working memory stimulus information in extrastriate cortex

    J. Cogn. Neurosci.

    (2014)
  • A.C. Riggall et al.

    The relationship between working memory storage and elevated activity as measured with functional magnetic resonance imaging

    J. Neurosci.

    (2012)
  • E.K. Vogel

    Neural measures reveal individual differences in controlling access to working memory

    Nature

    (2005)
  • M. Scolari et al.

    Basing perceptual decisions on the most informative sensory neurons

    J. Neurophysiol.

    (2010)
  • S.M. Emrich

    Distributed patterns of activity in sensory cortex reflect the precision of multiple items maintained in visual short-term memory

    J. Neurosci.

    (2013)
  • V. van de Ven

    Topographic contribution of early visual cortex to short-term memory consolidation: a transcranial magnetic stimulation study

    J. Neurosci.

    (2012)
  • J. Silvanto et al.

    Causal evidence for subliminal percept-to-memory interference in early visual cortex

    Neuroimage

    (2011)
  • Cited by (314)

    • A construct-first approach to consciousness science

      2024, Neuroscience and Biobehavioral Reviews
    • Fast Hebbian plasticity and working memory

      2023, Current Opinion in Neurobiology
    View all citing articles on Scopus
    View full text