Opinion
Domain generality versus modality specificity: the paradox of statistical learning

https://doi.org/10.1016/j.tics.2014.12.010Get rights and content

Highlights

  • Statistical learning (SL) theory is challenged by modality/stimulus-specific effects.

  • We argue that SL is shaped by both modality-specific constraints and domain-general principles.

  • SL relies on modality-specific neural networks and partially shared neural networks.

  • Studies of individual differences provide targeted insights into mechanisms of SL.

Statistical learning (SL) is typically considered to be a domain-general mechanism by which cognitive systems discover the underlying distributional properties of the input. However, recent studies examining whether there are commonalities in the learning of distributional information across different domains or modalities consistently reveal modality and stimulus specificity. Therefore, important questions are how and why a hypothesized domain-general learning mechanism systematically produces such effects. Here, we offer a theoretical framework according to which SL is not a unitary mechanism, but a set of domain-general computational principles that operate in different modalities and, therefore, are subject to the specific constraints characteristic of their respective brain regions. This framework offers testable predictions and we discuss its computational and neurobiological plausibility.

Section snippets

The promise of statistical learning

Humans and other animals are constantly bombarded by streams of sensory information. SL (the extraction of distributional properties from sensory input across time and space) provides a mechanism by which cognitive systems discover the underlying structure of such stimulation. Therefore, SL has a key role in the detection of regularities and quasi-regularities in the environment, results in discrimination, categorization, and segmentation of continuous information, allows prediction of upcoming

Domain generality versus domain specificity

Originally, domain generality was invoked to argue against language modularity; therefore, its definition implicitly implied ‘something that is not language specific’. Consequently, within this context, ‘domain’ implies a range of stimuli that share physical and structural properties (e.g., spoken words, musical tones, or tactile input), whereas ‘generality’ is taken to be ‘something that does not operate along principles restricted to language learning’. However, this approach says what domain

Towards a mechanistic model of SL

Our approach construes SL as involving a set of domain-general neurobiological mechanisms for learning, representation, and processing that detect and encode a range of distributional properties within different modalities or types of input (see [13], for a related approach). Crucially, however, in our account, these principles are not instantiated by a unitary learning system but, rather, by separate neural networks in different cortical areas (e.g., visual, auditory, and somatosensory

The neurobiological bases of SL

Recent neuroimaging studies have shown that statistical regularities of visual shapes result in activation in higher-level visual networks (e.g., lateral occipital cortex and inferior temporal gyrus 40, 41), whereas statistical regularities in auditory stimuli result in activation in analogous auditory networks (e.g., left temporal and inferior parietal cortices; frontotemporal networks including portions of the inferior frontal gyrus, motor areas involved in speech production [42]; and the

Individual and group differences in SL

The proposed framework leads us to argue that individual differences provide key evidence for understanding the mechanism of SL. In past work, it has often been assumed that individual variance in implicit learning tasks is significantly smaller than that of explicit learning (e.g., [58]). Consequently, the source of variability in performance in SL has been largely overlooked, and has led researchers to focus on average success rate (but see 19, 59, 60, 61).

However, in the context of SL,

Concluding remarks

Here, we offer a novel theoretical perspective on SL that considers computational and neurobiological constraints. Previous work on SL offered a possible cognitive mechanistic account of how distributional properties are computed, with explicit demonstrations being provided only within the domain of language 65, 67. Our perspective has the advantage of providing a unifying neurobiological account of SL across domains, modalities, neural, and cognitive investigations, and cross-species studies,

Acknowledgments

This paper was supported by The Israel Science Foundation (Grant 217/14 awarded to R.F.), by the NICHD (RO1 HD 067364 awarded to Ken Pugh and R.F., and PO1 HD 01994 awarded to Haskins Laboratories), and by a Marie Curie IIF award (PIIF-GA-2013-627784 awarded to B.C.A.).

References (89)

  • E. Fedorenko et al.

    Reworking the language network

    Trends Cogn. Sci.

    (2014)
  • E.A. Karuza

    The neural correlates of statistical learning in a word segmentation task: An fMRI study

    Brain Lang.

    (2013)
  • H. Eichenbaum

    Memory on time

    Trends Cogn. Sci.

    (2013)
  • J. Poppenk

    Long-axis specialization of the human hippocampus

    Trends Cogn. Sci.

    (2013)
  • F. Roux et al.

    Working memory and neural oscillations: alpha-gamma versus theta-gamma codes for distinct WM information?

    Trends Cogn. Sci.

    (2014)
  • A. Flöel

    White matter integrity in the vicinity of Broca's area predicts grammar learning success

    Neuroimage

    (2009)
  • P.A. Reeder

    From shared contexts to syntactic categories: the role of distributional information in learning linguistic form-classes

    Cogn. Psychol.

    (2013)
  • J.L. Elman

    Finding structure in time

    Cogn. Sci.

    (1990)
  • M. Carreiras

    The what, when, where, and how of visual word recognition

    Trends Cogn. Sci.

    (2014)
  • S. Laszlo et al.

    PSPs and ERPs: applying the dynamics of post-synaptic potentials to individual units in simulation of temporally extended Event-Related Potential reading data

    Brain Lang.

    (2014)
  • S. Laszlo et al.

    A neurally plausible parallel distributed processing model of event-related potential word reading data

    Brain Lang.

    (2012)
  • S.B. Kaufman

    Implicit learning as an ability

    Cognition

    (2010)
  • K.M. Petersson

    What artificial grammar learning reveals about the neurobiology of syntax

    Brain Lang.

    (2012)
  • N. Chomsky

    A review of B.F. Skinner's verbal behavior

    Language

    (1959)
  • P.D. Eimas

    Speech perception in infants

    Science

    (1971)
  • J.R. Saffran

    Statistical learning by 8-month-old infants

    Science

    (1996)
  • C.I. Baker

    Role of attention and perceptual grouping in visual statistical learning

    Psychol. Sci.

    (2004)
  • R.H. Hunt et al.

    Statistical learning in a serial reaction time task: access to separable statistical cues by individual learners

    J. Exp. Psychol. Gen.

    (2001)
  • K. Friston

    The free-energy principle: a unified brain theory?

    Nat. Rev. Neurosci.

    (2010)
  • N. Tishby et al.

    Information theory of decisions and actions

  • J.R. Saffran et al.

    Domain-general learning capacities

  • J.M. Toro et al.

    Statistical computations over a speech stream in a rodent

    Percept. Psychophys.

    (2005)
  • C.M. Conway et al.

    Modality-constrained statistical learning of tactile, visual, and auditory sequences

    J. Exp. Psychol. Learn. Mem. Cogn.

    (2005)
  • C.M. Conway et al.

    Statistical learning within and between modalities: pitting abstract against stimulus-specific representations

    Psychol. Sci.

    (2006)
  • A.D. Mitchel et al.

    Learning across senses: cross-modal effects in multisensory statistical learning

    J. Exp. Psychol. Learn. Mem. Cogn.

    (2011)
  • J.B. Misyak et al.

    Statistical learning and language: an individual differences study

    Lang. Learn.

    (2012)
  • M. Redington et al.

    Transfer in artificial grammar learning: a reevaluation

    J. Exp. Psychol. Gen.

    (1996)
  • R.J. Tunney et al.

    The transfer effect in artificial grammar learning: reappraising the evidence on the transfer of sequential dependencies

    J. Exp. Psychol. Learn. Mem. Cogn.

    (1999)
  • L.L. Emberson

    Timing is everything: changes in presentation rate have opposite effects on auditory and visual implicit statistical learning

    Q. J. Exp. Psychol. (Colchester)

    (2011)
  • T. Johansson

    Strengthening the case for stimulus-specificity in artificial grammar learning: no evidence for abstract representations with extended exposure

    Exp. Psychol.

    (2009)
  • M. Sigman et al.

    Learning to find a shape

    Nat. Neurosci.

    (2000)
  • P.J. Sjöström

    Dendritic excitability and synaptic plasticity

    Physiol. Rev.

    (2008)
  • R.D. Samson

    Computational models of reinforcement learning: the role of dopamine as a reward signal

    Cogn. Neurodyn.

    (2010)
  • L. Chen et al.

    Intersensory binding across space and time: a tutorial review

    Atten. Percept. Psychophys.

    (2013)
  • Cited by (346)

    View all citing articles on Scopus
    View full text