The representational–hierarchical view of pattern separation: Not just hippocampus, not just space, not just memory?

https://doi.org/10.1016/j.nlm.2016.01.006Get rights and content

Highlights

  • PS may be fundamental to many aspects of cognition including perception.

  • PS might be thought to happen in many regions, not just the hippocampus.

  • The DG is unlikely to be a truly domain-general pattern separator.

  • PS may have much wider explanatory power than previously suspected.

Abstract

Pattern separation (PS) has been defined as a process of reducing overlap between similar input patterns to minimize interference amongst stored representations. The present article describes this putative PS process from the “representational–hierarchical” perspective (R–H), which uses a hierarchical continuum instead of a cognitive modular processing framework to describe the organization of the ventral visual perirhinal–hippocampal processing stream. Instead of trying to map psychological constructs onto anatomical modules in the brain, the R-H model suggests that the function of brain regions depends upon what representations they contain. We begin by discussing a main principle of the R–H framework, the resolution of “ambiguity” of lower level representations via the formation of unique conjunctive representations in higher level areas, and how this process is remarkably similar to definitions of PS. Work from several species and experimental approaches suggest that this principle of resolution of ambiguity via conjunctive representations has considerable explanatory power, leads to wide possibilities for experimentation, and also supports some perhaps surprising conclusions.

Introduction

The importance of complex conjunctive representations for the resolution of ambiguity in lower-level representations has been a major focus of our research and that of others. Much work from several species and experimental approaches suggests that this principle of resolution of ambiguity via conjunctive representations has considerable explanatory power, particularly regarding how best to understand the effects of focal brain dysfunction (e.g., Barense et al., 2005, Barense et al., 2012, Bartko et al., 2007, Bussey and Saksida, 2002, Cowell et al., 2006, Cowell et al., 2010a, Graham et al., 2006, Lee, Buckley, et al., 2005, Lee, Bussey, et al., 2005, Lee and Rudebeck, 2010, McTighe et al., 2010). We have referred to this way of thinking about brain organization as the “representational–hierarchical” perspective (R–H) (e.g., Bussey and Saksida, 2002, Bussey and Saksida, 2005, Murray et al., 2007, Saksida and Bussey, 2010).

Recently there has been a great deal of interest in a process referred to as “Pattern Separation” (PS). PS has been defined as “… reducing interference among similar inputs by using non-overlapping representations…” (e.g., Reagh et al., 2014) and “…the ability of the network to reduce the overlap between similar input patterns before they are stored in order to reduce the probability of interference…” (Neunuebel & Knierim, 2014). Clearly the main principle of the R–H framework, the resolution of the “ambiguity” of lower level representations via the formation of unique conjunctive representations in higher level areas, is a strikingly similar idea to the above conceptions of PS. The present article is, therefore, aimed at researchers interested in PS, and explores the question: To the extent that researchers are interested in PS because it results in the formation of new, interference-reducing representations, what insights into PS might be offered by considering some of the conclusions resulting from R–H theory?

A quick word about the scope and aims of this review. R–H theory grew out of an interest in the functional relevance of processes and representational content in the brain. That is, what such processes and representational content do for us, in aspects of cognition such as memory and perception. The way to determine this is through empirical, causal behavioral experiments that measure cognition. Non-empirical approaches such as computational modeling have been hugely important in the area of PS (O’Reilly and McClelland, 1994, Rolls, 1987, Rolls, 1989, Rolls, 1990; Norman and O’Reilly, 2003, Rolls and Kesner, 2006, Treves and Rolls, 1994) and indeed, experiments testing the R–H view have been guided by computational modeling (e.g., Barense et al., 2012, Bartko et al., 2007, Cowell et al., 2006). But we must remember that these are models, and at some point experiments on real brains must be done. Correlational experimental approaches, particularly electrophysiology, have shed much light on PS at the cellular and network levels (e.g., Nakashiba et al., 2012, Knierim and Neunuebel, 2016), and focus on PS as a specific mechanism involving the transformation of an input representation to an output representation, in which the output is less correlated than the input (in line with computational models). However this process is of interest because of the representations thus formed and in particular their requirement for cognition (usually memory). Correlational approaches generally do not address this requirement. Additionally, correlational approaches have largely focussed on the hippocampus, whereas a main aim of what follows below is a suggestion that we need to consider other areas of the brain with respect to PS. Fully understanding any function of the brain cannot be accomplished by any single method alone, but requires converging, complementary approaches.

Furthermore, our focus in this article is the relevance to cognition of the result of processes such as PS, i.e., the representations that are formed from such processes. We will not, in this review, discuss how these representations are formed. For example, we will consider that such representations may be housed in the ventral visual stream, but not the way the formation of such representations might be related to the receptive fields of neurons. Comprehensive treatments of such issues, along with the properties of such representations (e.g., invariance), can be found in other sources such as Rolls (2016).

Insofar as PS can be considered to be a process of forming new representations that help resolve ambiguity, the results of experiments carried out under the auspices of R–H theory lead us to offer a number of possibly surprising hypotheses about PS, including:

  • 1.

    PS is fundamental to many aspects of cognition including perception; it is not just for memory.

  • 2.

    PS happens in many cortical regions, not just the hippocampus.

  • 3.

    PS happens for all stimulus material and not just ‘spatial’ or ‘episodic’ material. It happens for different types of representations, in the different regions and different levels throughout the “representational hierarchy”.

  • 4.

    The dentate gyrus (DG) is unlikely to maintain all levels of representation, and thus is not a truly domain-general pattern separator.

  • 5.

    PS – insofar as this term equates with the formation of interference-reducing conjunctive representations – may have much wider explanatory power for understanding the effects of brain dysfunction than previously suspected.

Below we provide a brief history and summary of R–H theory, with special emphasis on these particular conclusions regarding PS, and then return to these points.

Section snippets

The Representational–Hierarchical model: A brief history and summary

The R–H view was first presented in 1998 (Saksida & Bussey, 1998) and discussed in a review article in 1999 (Murray & Bussey, 1999). Eventually the first modeling and experimental studies testing the theory were published (Bussey and Saksida, 2002, Bussey et al., 2002, Bussey et al., 2003). Although R–H takes a broad view of cognition, initial studies focused on visual cognition, and in particular the issue of whether there are dedicated, anatomically segregated memory (Squire & Zola-Morgan,

What does any of this have to do with PS?

The relevance of R–H principles to PS is that all of the above and more can be explained by the formation of highly discriminable representations that resolve interference or ambiguity, which is thought to be the function of PS (see definitions above). This suggests that PS is a rather powerful process that happens in many parts of the brain, and may be a major player in many aspects of cognition.

It is worth taking a moment to focus on the last point above, the proposal that the hippocampus is

Concluding thoughts

To recap, the main points arising from this analysis are:

  • 1.

    PS is fundamental to many aspects of cognition including perception; it is not just for memory.

  • 2.

    PS happens in many cortical regions, not just the hippocampus.

  • 3.

    PS happens for all stimulus material and not just ‘spatial’ or ‘episodic’ material. It happens for different types of representations, in the different regions and different levels throughout the hierarchy.

  • 4.

    The DG is unlikely to maintain all levels of representation, so is not a

Acknowledgments

LMS and TJB were funded by Medical Research Council/Wellcome Trust grant 089703/Z/09/Z. BAK was funded by Gates Cambridge and the Natural Sciences and Engineering Research Council of Canada (NSERC).

References (95)

  • A.C.H. Lee et al.

    Perceptual deficits in amnesia: Challenging the medial temporal lobe ‘mnemonic’ view

    Neuropsychologia

    (2005)
  • A.C. Lee et al.

    Differing profiles of face and scene discrimination deficits in semantic dementia and Alzheimer’s disease

    Neuropsychologia

    (2007)
  • E.A. Murray et al.

    Perceptual–mnemonic functions of the perirhinal cortex

    Trends in Cognitive Sciences

    (1999)
  • T. Nakashiba et al.

    Young dentate granule cells mediate pattern separation, whereas old granule cells facilitate pattern completion

    Cell

    (2012)
  • J.P. Neunuebel et al.

    CA3 retrieves coherent representations from degraded input: Direct evidence for CA3 pattern completion and dentate gyrus pattern separation

    Neuron

    (2014)
  • R.Q. Quiroga et al.

    Sparse but not ‘grandmother-cell’ coding in the medial temporal lobe

    Trends in Cognitive Sciences

    (2008)
  • E.T. Rolls

    Limbic systems for emotion and for memory, but no single limbic system

    Cortex

    (2015)
  • E.T. Rolls et al.

    A computational theory of hippocampal function, and empirical tests of the theory

    Progress in Neurobiology

    (2006)
  • E.T. Rolls

    Pattern separation, completion, and categorisation in the hippocampus and neocortex

    Neurobiology of Learning and Memory

    (2016)
  • L.M. Saksida et al.

    The representational–hierarchical view of amnesia: Translation from animal to human

    Neuropsychologia

    (2010)
  • J.E. Warren et al.

    Sounds do-able: Auditory–motor transformations and the posterior temporal plane

    Trends in Neurosciences

    (2005)
  • M.A. Yassa et al.

    Pattern separation in the hippocampus

    Trends in Neurosciences

    (2011)
  • M. Azab et al.

    Contributions of human hippocampal subfields to spatial and temporal pattern separation

    Hippocampus

    (2014)
  • C.I. Baker et al.

    Impact of learning on representation of parts and wholes in monkey inferotemporal cortex

    Nature Neuroscience

    (2002)
  • A. Bakker et al.

    Pattern separation in the human hippocampal CA3 and dentate gyrus

    Science

    (2008)
  • M.D. Barense et al.

    Functional specialization in the human medial temporal lobe

    The Journal of Neuroscience

    (2005)
  • S.J. Bartko et al.

    Perirhinal cortex resolves feature ambiguity in configural object recognition and perceptual oddity tasks

    Learning & Memory

    (2007)
  • P. Bekinschtein et al.

    Brain-derived neurotrophic factor interacts with adult-born immature cells in the dentate gyrus during consolidation of overlapping memories

    Hippocampus

    (2014)
  • M.J. Buckley et al.

    Perirhinal cortex ablation impairs visual object identification

    The Journal of Neuroscience

    (1998)
  • S.N. Burke et al.

    Pattern separation deficits may contribute to age-associated recognition impairments

    Behavioral Neuroscience

    (2010)
  • T.J. Bussey et al.

    The organization of visual object representations: A connectionist model of effects of lesions in perirhinal cortex

    European Journal of Neuroscience

    (2002)
  • T.J. Bussey et al.

    Perirhinal cortex resolves feature ambiguity in complex visual discriminations

    European Journal of Neuroscience

    (2002)
  • T.J. Bussey et al.

    Impairments in visual discrimination after perirhinal cortex lesions: Testing ‘declarative’ vs. ‘perceptual-mnemonic’ views of perirhinal cortex function

    European Journal of Neuroscience

    (2003)
  • N. Butters et al.

    Alcoholic Korsakoff’s syndrome: An information processing approach

    (1980)
  • M.M. Chun et al.

    Memory deficits for implicit contextual information in amnesic subjects with hippocampal damage

    Nature Neuroscience

    (1999)
  • C.D. Clelland et al.

    A functional role for adult hippocampal neurogenesis in spatial pattern separation

    Science

    (2009)
  • N.J. Cohen et al.

    Memory, amnesia and the hippocampal system

    (1993)
  • R.A. Cowell et al.

    Why does brain damage impair memory? A connectionist model of object recognition memory in perirhinal cortex

    Journal of Neuroscience

    (2006)
  • R.A. Cowell et al.

    Functional dissociations within the ventral object processing pathway: Cognitive modules or a hierarchical continuum?

    Journal of Cognitive Neuroscience

    (2010)
  • R.A. Cowell et al.

    Components of recognition memory: Dissociable cognitive processes or just differences in representational complexity?

    Hippocampus

    (2010)
  • D.J. Creer et al.

    Running enhances spatial pattern separation in mice

    Proceedings of the National Academy of Sciences

    (2010)
  • R. Desimone et al.

    Stimulus-selective properties of inferior temporal neurons in the macaque

    The Journal of Neuroscience

    (1984)
  • R. Desimone et al.

    Neural mechanisms of visual processing in monkeys

    Handbook of Neuropsychology

    (1989)
  • J. Erez et al.

    Conjunctive coding of complex object features

    Cerebral Cortex (New York, NY: 1991)

    (2015)
  • S.E. Forwood et al.

    Multiple cognitive abilities from a single cortical algorithm

    Journal of Cognitive Neuroscience

    (2012)
  • D. Gaffan

    Against memory systems

    Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences

    (2002)
  • P.E. Gilbert et al.

    Recognition memory for complex visual discriminations is influenced by stimulus interference in rodents with perirhinal cortex damage

    Learning & Memory

    (2003)
  • Cited by (53)

    • Medial temporal lobe structure, mnemonic and perceptual discrimination in healthy older adults and those at risk for mild cognitive impairment

      2023, Neurobiology of Aging
      Citation Excerpt :

      By contrasting performance in Forced Choice and Yes/No tests, we test whether the association between MCI risk and mnemonic discrimination is dependent on the degree to which a task taxes hippocampal memory processes. We view mnemonic discrimination deficits in older adults through the lens of a representational account, in which target-lure similarity is one key determinant of impairment, consistent with the representational-hierarchical framework of MTL function (Bussey and Saksida, 2002; Cowell et al., 2010; Kent et al., 2016). This framework proposes that formation of complex, conjunctive visual representations necessary to differentiate stimuli with overlapping features is dependent on MTL, and lesions to as well as normal age-related and early pathological changes in these regions result in impoverished visual representations that compromise any cognitive process reliant on them, whether that be mnemonic or perceptual (Barense et al., 2005; Burke et al., 2018; Lee et al., 2007).

    View all citing articles on Scopus
    View full text