Elsevier

Neural Networks

Volume 17, Issues 5–6, June–July 2004, Pages 695-705
Neural Networks

2004 Special Issue
Perceptual grouping and the interactions between visual cortical areas

https://doi.org/10.1016/j.neunet.2004.03.010Get rights and content

Abstract

Visual perception involves the grouping of individual elements into coherent patterns, such as object representations, that reduce the descriptive complexity of a visual scene. The computational and physiological bases of this perceptual remain poorly understood. We discuss recent fMRI evidence from our laboratory where we measured activity in a higher object processing area (LOC), and in primary visual cortex (V1) in response to visual elements that were either grouped into objects or randomly arranged. We observed significant activity increases in the LOC and concurrent reductions of activity in V1 when elements formed coherent shapes, suggesting that activity in early visual areas is reduced as a result of grouping processes performed in higher areas. In light of these results we review related empirical findings of context-dependent changes in activity, recent neurophysiology research related to cortical feedback, and computational models that incorporate feedback operations. We suggest that feedback from high-level visual areas reduces activity in lower areas in order to simplify the description of a visual image—consistent with both predictive coding models of perception and probabilistic notions of ‘explaining away.’

Introduction

One of the extraordinary capabilities of the human visual system is its ability to rapidly select and group related elements in a complex visual scene. This capability serves to bring together information likely to belong to a common cause, such as the same contour, surface or object. Grouping also reflects a general function of cognitive systems in that it greatly simplifies the description by exploiting redundancy in the input pattern (Barlow, 1959). For example, the image of a set of parallel lines can be succinctly described as a single texture pattern (‘N repetitions of feature X’) without needing to specify each element within the pattern.

These pattern-processing capabilities appear to be reflected in the activities of neurons at various stages of the visual system. For example, the response of a neuron in V1 to a single bar oriented along a receptive field's preferred axis can be suppressed by parallel bars on the two sides or enhanced if orientations differ and a collinear bar can enhance the response (Kapadia et al., 2000, Knierim and van Essen, 1992). Such pattern context effects in V1 appear to be mediated by both local connections (Das & Gilbert, 1999) and by interactions with higher areas (Hupe et al., 1998).

Grouping local features that belong to an object is particularly interesting from a physiological perspective because object shape is believed to be represented in higher stages of the visual system beyond V1, so any influence of perceived shape on lower areas would require feedback. Feedback is generally thought of as a process where activity in lower areas is positively correlated with the activity occurring in higher areas. However, recent work on predictive coding models has suggested that feedback may operate to reduce activity. In these models, higher-stages of a network compete by projecting their predictions about the stimulus to lower stages, where they are then removed from incoming data. In these models, the activity of neurons in lower stages will decrease when neurons in higher stages can ‘explain’ a visual stimulus, but will increase when the top-down explanation is poor (Mumford, 1992, Rao and Ballard, 1999). Other mechanisms for reducing activity via feedback are also possible and are discussed below. We present results from our own research and those of others suggesting that feedback from high-level visual areas reduces average activity in lower areas in order to simplify the description of a visual image.

Section snippets

Feedback

Feedback projections are a prominent anatomical feature of the primate visual system (Felleman & Van Essen, 1991) and recent evidence suggests they play a critical role in visual perception (Pascual-Leone & Walsh, 2001 and see Bullier, 2001, Lamme and Roelfsema, 2000 for reviews). Nevertheless, the functional significance of these connections has been open to interpretation. There are two basic possibilities: feedback may act by modifying input-driven activity in an existing active neural

Experimental results

To examine the possible role of feedback during object perception, we conducted a series of fMRI experiments (Murray et al., 2002) using stimuli with features that could either be perceived as ungrouped elements or perceived as being grouped into a single perceptual ‘explanation’—specifically, a single shape or object. Our first experiment used random-dot structure-from-motion stimuli. In one condition (‘SFM’), random dot patterns were projected onto the surfaces of simple geometric shapes

Information processing functions of feedback

In Section 3, we discussed predictive coding and sharpening as possible explanations for the observed decrease in V1 activity as a function of shape perception. However, these ideas tell us little about how these mechanisms might be involved in solving the computational tasks of vision. Although the empirical basis for feedback between cortical areas is becoming increasingly well-established, understanding its role in information processing poses a major theoretical challenge. We can get some

Remaining questions

Our empirical results demonstrate that neuronal activity, even in V1, does not simply represent the signaling of features in a visual scene but is strongly influenced by high-level perceptions of object shape. Though these results, in combinations with other studies, offer a compelling example of the potential role for feedback processes in vision, there are still many unanswered questions. For example, having timing information about the relative changes in V1 and LOC is crucial to

Acknowledgements

Portions of this work were reported earlier in Murray et al. (2002) and at Human Brain Mapping (Shen, Kersten, and Ugurbil, 1999), ARVO (Kersten, Shen, Ugurbil, and Schrater, 1999) and Soc. Neurosci. (Murray Olshausen, and Woods, 2001) conferences. Supported by NIH R01 EY015261, NIH P41 RR08079, pre-doctoral NRSA MH-12791 and post-doctoral NRSA EY015342-01 (S.O.M.), NSF SBR-9631682 (D.K.), NIH MH-57921 (B.A.O.), NIH MH-41544 and VA Research Service (D.L.W.). We thank Peter Battaglia for helpful

References (79)

  • P Lennie

    The cost of cortical computation

    Current Biology

    (2003)
  • C Moore et al.

    Neural response to perception of volume in the lateral occipital complex

    Neuron

    (2001)
  • A Oliva et al.

    Diagnostic colors mediate scene recognition

    Cognitive Psychology

    (2000)
  • E Peterhans et al.

    Subjective contours—Bridging the gap between psychophysics and physiology

    Trends in Neuroscience

    (1991)
  • J.K Tsotsos

    Limited capacity of any realizable perceptual system is a sufficient reason for attentive behavior

    Consciousness and Cognition

    (1997)
  • Barlow, H. B (1959). Sensory mechanisms, the reduction of redundancy, and intelligence. Paper presented at the...
  • C Buchel et al.

    The predictive value of changes in effective connectivity for human learning

    Science

    (1999)
  • G.A Carpenter et al.

    ART2: Self-organization of stable category recognition codes for analog input patterns

    Applied Optics

    (1987)
  • P Cavanagh

    What's up in top-down processing?

  • Clark, J. J., & Yuille, A. L (1990). Shape from shading via the fusion of specular and Lambertian image components....
  • A Das et al.

    Topography of contextual modulations mediated by short-range interactions in primary visual cortex

    Nature

    (1999)
  • R Desimone et al.

    Neural mechanisms of selective visual attention

    Annual Review of Neuroscience

    (1995)
  • D.J Felleman et al.

    Distributed hierarchical processing in the primate cerebral cortex

    Cerebral Cortex

    (1991)
  • K.J Friston et al.

    Analysis of functional MRI time-series

    Human Brain Mapping

    (1994)
  • K Fukushima

    Neocognitron: A self organizing neural network model for a mechanism of pattern recognition unaffected by shift in position

    Biological Cybernetics

    (1980)
  • P Girard et al.

    Feedforward and feedback connections between areas V1 and V2 of the monkey have similar rapid conduction velocities

    Journal of Neurophysiology

    (2001)
  • U Grenander

    Elements of pattern theory

    (1996)
  • J.M Hupe et al.

    Cortical feedback improves discrimination between figure and background by V1, V2 and V3 neurons

    Nature

    (1998)
  • L Itti et al.

    Computational modelling of visual attention

    Nature Reviews. Neuroscience

    (2001)
  • J.S Johnson et al.

    Timecourse of neural signatures of object recognition

    Journal of Vision

    (2003)
  • M.K Kapadia et al.

    Spatial distribution of contextual interactions in primary visual cortex and in visual perception

    Journal of Neurophysiology

    (2000)
  • D Kersten et al.

    Object perception as Bayesian inference

    Annual Review of Psychology

    (2004)
  • D Kersten et al.

    Pattern inference theory: A probabilistic approach to vision

  • D Kersten et al.

    Bayesian models of object perception

    Current Opinion in Neurobiology

    (2003)
  • J.J Knierim et al.

    Neuronal responses to static texture patterns in area V1 of the alert macaque monkey

    Journal of Neurophysiology

    (1992)
  • C Koch et al.

    Predicting the visual world: Silence is golden [news; comment]

    Nature Neuroscience

    (1999)
  • V.A Lamme

    The neurophysiology of figure-ground segregation in primary visual cortex

    Journal of Neuroscience

    (1995)
  • D.K Lee et al.

    Attention activates winner-take-all competition among visual filters

    Nature Neuroscience

    (1999)
  • T.S Lee et al.

    Hierarchical Bayesian inference in the visual cortex

    Journal of the Optical Society of America. A. Optics, Image Science and Vision

    (2003)
  • Cited by (122)

    • The N400 in silico: A review of computational models

      2022, Psychology of Learning and Motivation - Advances in Research and Theory
    • Topographic signatures of global object perception in human visual cortex

      2020, NeuroImage
      Citation Excerpt :

      Building upon previous research involving the diamond stimulus (de-Wit et al., 2012), it is important to highlight that our results in lower visual cortex across all experiments seem to contradict suggestions of predictive coding theories that suppressive effects should be confined to cortical sites encoding the physical stimulus and accompanied by unchanged activity in the background region (e.g., Mumford, 1992; Murray et al., 2004; Rao and Ballard, 1999). They furthermore seem to conflict with alternative accounts, such as response sharpening (e.g., Kersten et al., 2004; Kersten and Yuille, 2003; Murray et al., 2004). Response sharpening accounts assume that predictive feedback from higher-tier areas sharpens diffuse responses in lower-tier areas (due to noise or ambiguity) by increasing activity matching the global interpretation of the bottom-up input and decreasing non-matching activity.

    View all citing articles on Scopus
    View full text