Attention and visual perception

https://doi.org/10.1016/j.conb.2005.06.009Get rights and content

Somewhere between the retina and our conscious visual experience, the majority of the information impinging on the eye is lost. We are typically aware of only either the most salient parts of a visual scene or the parts that we are actively paying attention to. Recent research on visual neurons in monkeys is beginning to show how the brain both selects and discards incoming visual information. For example, what happens to the responses of visual neurons when attention is directed to one element, such as an oriented colored bar, embedded among an array of other oriented bars? Some of this research shows that attention to the oriented bar restricts the receptive field of visual neurons down to this single element. However, other research shows that attention to this single element affects the responses of neurons with receptive fields throughout the visual field. In this review, these two seemingly contradictory results are shown to actually be mutually consistent. A simple computational model is described that explains these results, and also provides a framework for predicting a variety of additional neurophysiological, neuroimaging and behavioral studies of attention.

Introduction

Much has been learnt in recent years about how attention influences the neuronal representation of visual stimuli. The majority of recent studies of attention can be categorized as studies of either spatial attention or feature-based attention. Cueing a monkey to shift spatial attention from outside to inside the receptive field (RF) of a neuron increases the responses of visual neurons, possibly by increasing the effective contrast of the stimulus [1, 2]. Cueing a monkey's feature-based attention to one of two stimuli presented together within the RF of a neuron biases neuronal responses as though attended stimulus was presented alone [3, 4]. This is conceptually similar to restricting the RF to the attended stimulus. This can be called ‘local feature-based attention’ because it describes feature-based attentional effects locally within the focus of spatial attention. Monkey electrophysiological [2, 5•] and functional magnetic resonance imaging (fMRI) studies [6] show evidence of ‘global feature-based attention’, in which attention affects the response of neurons with RFs well outside the focus of spatial attention. This might seem contradictory because local feature-based attention restricts the RF and global feature-based attention affects neurons regardless of RF location. These effects of spatial in addition to local and global feature-based attention might seem complex. However, a simple model of attention presented in this review predicts a wide variety of experimental results.

Section snippets

The neuronal contrast–response function

A variety of stimulus-driven responses in the macaque visual cortex [7, 8, 9] can be described by a divisive contrast normalization process. Neurons have an inherent peak sensitivity for a specific feature of a stimulus, such as direction of motion, color or orientation. If stimuli are defined as consisting of multiple components presented within the RF of a neuron, each of which has its own feature, xi, and contrast, ci, then the response of a neuron to a stimulus can be described by equation

Contrast-gain model of spatial attention

Recent electrophysiological studies of attention in monkeys suggest that spatial attention increases the effective contrast of a stimulus in color-selective area V4 [1] and motion-selective area MT [2]. This can be described either by multiplying all contrast components by a constant or, equivalently, by dividing the semisaturation parameter, σ, by the same constant. Dividing the parameter σ by a constant k>1 increases the effective contrast of the stimulus by the same factor, shifting the

The feature similarity gain model of global feature-based attention

Treue and Martinez-Trujillo [5•] have shown that, for example, if a neuron in area MT of the macaque is selective to upward motion, then attention to upward motion will enhance the response of this neuron, whereas attention to downward motion will suppress the response. Interestingly, this modulation is found even when the spatial focus of attention is outside the RF of the cell. These authors suggest a ‘feature-similarity gain model’ of attention. In this model, attention directed to a

Biased competition model of local feature-based attention

In a classic study, Moran and Desimone [3] measured electrophysiological responses in the monkey from area V4 and from neurons in the inferotemporal cortex (IT) neurons when two stimuli were presented within the RF of a neuron. One stimulus was called the ‘preferred’ stimulus because, when presented alone, it produced a larger response than the ‘non-preferred’ stimulus. When attention was directed away towards a fixation task, the response to the pair of stimuli fell between that to the

The feature similarity gain model predicts biased competition

The biased competition model is analogous to shrinking the effective size of the RF [4] but the feature-similarity gain model seems to achieve almost the opposite effect because it is a spread of attention to all neurons in the visual field [17]. However, the simple model described above can explain both effects by assuming that the same feature-based attention mechanism operates both within and outside the RF of a neuron.

Consider the case when attention is directed towards fixation when a pair

Spatial attention

The model described above provides a framework for understanding a variety of studies on spatial and feature-based attention, and helps reconcile apparent discrepancies in their results. For example, McAdams and Maunsell [18] measured the effects of attention on the orientation-selectivity of V1 and V4 neurons. Orientation tuning curves in both V1 and V4 were found to increase in a multiplicative fashion with spatial attention. This, at first, seems like a contradiction of the contrast-gain

Conclusions

The model in this review proposes a simple relationship among the effects of spatial, local and global feature-based attention on visual neuronal responses. Spatial attention acts as a contrast-gain mechanism. Local and global feature-based attention are mediated by the same mechanism, and operate separably with spatial attention. Although this model explains biased competition through global feature-based attention, a related model explains global feature-based attention through biased

References and recommended reading

Papers of particular interest, published within the annual period of review, have been highlighted as:

  • • of special interest

  • •• of outstanding interest

References (36)

  • M. Saenz et al.

    Global effects of feature-based attention in human visual cortex

    Nat Neurosci

    (2002)
  • D.J. Heeger

    Modeling simple-cell direction selectivity with normalized, half-squared, linear operators

    J Neurophysiol

    (1993)
  • D.J. Heeger et al.

    Computational models of cortical visual processing

    Proc Natl Acad Sci USA

    (1996)
  • M. Carandini et al.

    Linearity and normalization in simple cells of the macaque primary visual cortex

    J Neurosci

    (1997)
  • D. Tolhurst

    Separate channels for the analysis of the shape and the movement of a moving visual stimulus

    J Physiol

    (1973)
  • A.F. Dean

    The variability of discharge of simple cells in the cat striate cortex

    Exp Brain Res

    (1981)
  • D.G. Albrecht et al.

    Striate cortex of monkey and cat: contrast response function

    J Neurophysiol

    (1982)
  • G. Sclar et al.

    Orientation selectivity in the cat's striate cortex is invariant with stimulus contrast

    Exp Brain Res

    (1982)

    • Cited by (0)

    View full text
    Copyright © 2005 Elsevier Ltd. All rights reserved.