Trends in Cognitive Sciences
Volume 8, Issue 8, 1 August 2004, Pages 363-370
Journal home page for Trends in Cognitive Sciences

How parallel is visual processing in the ventral pathway?

https://doi.org/10.1016/j.tics.2004.06.003Get rights and content

Visual object perception is usually studied by presenting one object at a time at the fovea. However, the world around us is composed of multiple objects. The way our visual system deals with this complexity has remained controversial in the literature. Some models claim that the ventral pathway, a set of visual cortical areas responsible for object recognition, can process only one or very few objects at a time without ambiguity. Other models argue in favor of a massively parallel processing of objects in a scene. Recent experiments in monkeys have provided important data about this issue. The ventral pathway seems to be able to perform complex analyses on several objects simultaneously, but only during a short time period. Subsequently only one or very few objects are explicitly selected and consciously perceived. Here, we survey the implications of these new findings for our understanding of object processing.

Section snippets

Isolated objects and the binding problem

In laboratory conditions, when the visual system is stimulated with a single object, it seems that there is no binding problem, because all low-level elements in the visual field belong to the same object. However, one still needs a mechanism to organize correctly those elements in space. Indeed, a large number of TE neurons appear to be sensitive to specific element configurations such that a neuron might respond to element A above B (e.g. a circle and a square), but not to B above A 4, 5. It

Spatial information in TE

The ventral pathway is able to build very rapidly the representation of an isolated object, finally encoded in TE neurons. But because TE neurons have large RFs, ranging from about 10° to 30° and often more (e.g. 17, 18, 19), one might question whether the ventral pathway is able to deal with two or more objects simultaneously. Indeed, it is tempting to assume that spatial information is lost in the large RFs of TE neurons, making it impossible to combine the elements from different objects

Representation of two or more stimuli

Overall, the responses of TE neurons might incorporate enough spatial information to be able to encode the identity of several objects in parallel. Several studies have investigated the behavior of TE neurons in response to two stimuli (e.g. 28, 30, 31, 32). Two stimuli typically enter into competitive interactions following two important rules. First, at the level of a single neuron, competition is most likely to occur when both stimuli appear in the RF of a neuron and not when one stimulus is

Low-level and high-level constraints

The evidence reviewed so far suggest that the ventral stream might be able to encode high-level diagnostic information about objects in parallel, but competitive interactions would then strongly limit the number of detailed representations available. In this section, we consider two other important constraints that might limit parallel object processing.

Conclusions

Recent advances in monkey neurophysiology and computational neuroscience highlight the sophistication of the mechanisms implemented in the ventral pathway. Contrary to the claim that the large RFs in TE prevent object-recognition mechanisms from dealing with more than one object at a time, there is now clear evidence that the ventral pathway is well equipped for parallel processing in cluttered scenes. However, local competition, and low-level and high-level constraints considerably limit the

Acknowledgements

The authors would like to thank the Cognitique program, the Integrative and Computational Neuroscience Initiative and the CNRS for support. G.A.Rousselet is supported by a Tier I Canada Research Chair grant to Patrick J. Bennett.

References (80)

  • R. VanRullen et al.

    Surfing a spike wave down the ventral stream

    Vision Res.

    (2002)
  • P.R. Roelfsema

    The implementation of visual routines

    Vision Res.

    (2000)
  • A. Angelucci et al.

    Reaching beyond the classical receptive field of V1 neurons: horizontal or feedback axons?

    J. Physiol. (Paris)

    (2003)
  • G.A. Rousselet

    Processing of one, two or four natural scenes in humans: the limits of parallelism

    Vision Res.

    (2004)
  • J. Palmer

    The psychophysics of visual search

    Vision Res.

    (2000)
  • B.C. Motter et al.

    Cortical image density determines the probability of target discovery during active search

    Vision Res.

    (2000)
  • J. Gottlieb

    Parietal mechanisms of target representation

    Curr. Opin. Neurobiol.

    (2002)
  • J. Driver et al.

    Perceptual awareness and its loss in unilateral neglect and extinction

    Cognition

    (2001)
  • S. Thorpe

    Spike-based strategies for rapid processing

    Neural Netw.

    (2001)
  • N.K. Logothetis et al.

    Visual object recognition

    Annu. Rev. Neurosci.

    (1996)
  • C. von der Malsburg

    The what and why of binding: the modeler's perspective

    Neuron

    (1999)
  • C.I. Baker

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

    Nat. Neurosci.

    (2002)
  • K. Tanaka

    Columns for complex visual object features in the inferotemporal cortex: clustering of cells with similar but slightly different stimulus selectivities

    Cereb. Cortex

    (2003)
  • M.W. Oram et al.

    Time course of neural responses discriminating different views of the face and head

    J. Neurophysiol.

    (1992)
  • S.J. Thorpe et al.

    Seeking categories in the brain

    Science

    (2001)
  • H. Wersing et al.

    Learning optimized features for hierarchical models of invariant object recognition

    Neural Comput.

    (2003)
  • S. Edelman et al.

    (Coarse coding of shape fragments)+(retinotopy) approximately=representation of structure

    Spat. Vis.

    (2000)
  • M.C. Elliffe

    Invariant recognition of feature combinations in the visual system

    Biol. Cybern.

    (2002)
  • A.G. Leventhal

    Concomitant sensitivity to orientation, direction, and color of cells in layers 2, 3, and 4 of monkey striate cortex

    J. Neurosci.

    (1995)
  • P.H. Schiller

    Past and present ideas about how the visual scene is analyzed by the brain

  • D. Boussaoud

    Visual topography of area TEO in the macaque

    J. Comp. Neurol.

    (1991)
  • E. Kobatake et al.

    Neuronal selectivities to complex object features in the ventral visual pathway of the macaque cerebral cortex

    J. Neurophysiol.

    (1994)
  • H. Op De Beeck et al.

    Spatial sensitivity of macaque inferior temporal neurons

    J. Comp. Neurol.

    (2000)
  • A. Nowicka et al.

    Eye position-sensitive units in hippocampal formation and in inferotemporal cortex of the macaque monkey

    Eur. J. Neurosci.

    (2000)
  • M. Ito

    Size and position invariance of neuronal responses in monkey inferotemporal cortex

    J. Neurophysiol.

    (1995)
  • K. Tanaka

    Coding visual images of objects in the inferotemporal cortex of the macaque monkey

    J. Neurophysiol.

    (1991)
  • J.J. DiCarlo et al.

    Anterior inferotemporal neurons of monkeys engaged in object recognition can be highly sensitive to object retinal position

    J. Neurophysiol.

    (2003)
  • A. Lueschow

    Inferior temporal mechanisms for invariant object recognition

    Cereb. Cortex

    (1994)
  • Y. Wang

    Contribution of GABAergic inhibition to receptive field structures of monkey inferior temporal neurons

    Cereb. Cortex

    (2002)
  • E.T. Rolls

    The receptive fields of inferior temporal cortex neurons in natural scenes

    J. Neurosci.

    (2003)
  • Cited by (150)

    View all citing articles on Scopus
    View full text