Our brains focus on important events and filter out distracting ones. An investigation in monkeys reveals a surprising dissociation between the neuronal and behavioural manifestations of attention. See Letter p.434
Why are drivers more likely to have an accident if they are talking on a mobile phone? The obvious answer is that they are not paying attention to the road. But what is attention, and what goes on in our brains when we are 'paying' it? For decades, psychologists have proposed that we can direct something rather like a mental spotlight towards particular regions of our surroundings, and that this selectively enhances our perceptual sensitivity in that region. For example, if you were instructed to attend to an area to your left (without looking there), you would be able to detect a dimmer light in that region than elsewhere. Indeed, neurophysiologists have shown that attention amplifies the responses of neurons whose preferred spatial region (the neuron's 'receptive field') corresponds to the attended region1. However, on page 434 of this issue, Zénon and Krauzlis2 report that, in monkeys, inactivating a brain structure called the superior colliculus impairs visual attention but retains the enhanced responses of neurons in the brain's cerebral cortex.
In addition to amplifying — or, more precisely, increasing the gain of — neurons' responses, attention tends to make the relevant neurons slightly less 'noisy' and more independent (of their neighbours) in their responses; both changes allow them to collectively encode sensory information more reliably. All of these changes make sense, and seem to account for why an animal's perception is enhanced by attention3. This is why Zénon and Krauzlis's results2 will come as a surprise to many. By inactivating the superior colliculus, which has previously been shown to be important for attention4, they impaired the monkeys' ability to detect a visual target while ignoring an irrelevant, distracting stimulus in another part of the visual field.
Inactivation of the superior colliculus (by injecting a drug that inhibited neuronal activity) did not create a basic sensory deficit, like a blind spot, because the impairment was evident only when there were competing stimuli5. But, despite the severe impairment in the animals' ability to pay attention to the relevant stimulus, all of the known neural correlates of attention (including increased gain) in two sensory areas in the cerebral cortex — the middle temporal area (MT) and the medial superior temporal area (MST) — remained intact. Thus, the authors uncoupled the neuronal changes that are thought to underlie attention from the behavioural manifestation of attention.
What are we to make of this? For a start, we can conclude that the superior colliculus is not the only source of signals responsible for the changes in early sensory areas (those closer to the sensory receptors). We can also conclude that the improvements in the encoding of sensory information in the MT and MST are not sufficient to produce the perceptual effects of attention.
However, although Zénon and Krauzlis found no changes in any of the known neural correlates of attention, it is conceivable that they missed the right neurons — we know that, at any given location in a sensory area, only a subset of neurons contributes to any given task6. Moreover, it is possible that inactivation of the superior colliculus impairs the attentional system in other ways, and that the neuronal changes in the MT and MST are insufficient to overcome the deficit. For example, if selective attention emerges as the result of competition among visual representations in multiple brain regions7, the increases in gain in the MT and MST might simply be overbalanced by the loss of enhancements in other areas more directly connected to the superior colliculus.
Another possibility is that attention follows a two-stage mechanism: a first stage produces the gain changes in early sensory areas, whereas a later stage selects among these enhanced signals. In this model, the superior colliculus would act as part of the selection filter, the activity of which determines whether signals from a particular sensory region will be used to guide behaviour. Without the superior colliculus, the corresponding part of visual space is effectively filtered out, or ignored, as it is for patients with brain damage who have 'unilateral neglect'8 — they may fail to eat food from one side of their plate, for example, or to shave one side of their face.
Such a filtering stage would explain why humans often miss large changes in the visual scene. In one famous example, observers are asked to count the number of passes of a basketball among teammates, and they fail to notice a person in a gorilla suit who wanders through the scene9. This happens even though neurons in an early visual area accurately represent — and can therefore be used to detect — the gorilla and all such highly salient changes. At some stage, even these otherwise obvious events are filtered out, presumably to focus processing on the behaviourally relevant information. We speculate that inactivation of the superior colliculus, as described by Zénon and Krauzlis, may impair this latter stage.
It must be that a brain area other than the superior colliculus is responsible for the gain changes observed in MT and MST neurons. One possible candidate is a region of the cortex known as the frontal eye fields, which are involved in visual attention and eye movements. Indeed, there is evidence that electrical stimulation of the frontal eye fields can produce gain changes in early sensory areas similar to those produced by attention10. Future experiments will be necessary to determine how the activities of the superior colliculus and those of areas such as the frontal eye fields are coordinated to converge on an attended location. In particular, as the convergence of enhanced signals has been proposed to occur in a region of the parietal cortex called the lateral intraparietal area11, it will be important to determine whether inactivation of the superior colliculus leads to more-pronounced deficits in the effects of attention on neurons in this area than those observed in the MT and MST.
Zénon and Krauzlis's results suggest that there are at least two cooperating stages: attentional-gain modulation and subsequent selection. Their work therefore calls for further studies of how such systems interact to endow us with a mechanism that we depend on every day: the option to ignore our mobile phones and focus on the road ahead.
Reynolds, J. H. & Chelazzi, L. Annu. Rev. Neurosci. 27, 611–647 (2004).
Zénon, A. & Krauzlis, R. J. Nature 489, 434–437 (2012).
Cohen, M. R. & Maunsell, J. H. R. Nature Neurosci. 12, 1594–1600 (2009).
Kustov, A. A. & Robinson, D. L. Nature 384, 74–77 (1996).
Lovejoy, L. P. & Krauzlis, R. J. Nature Neurosci. 13, 261–266 (2010).
Britten, K. H., Newsome, W. T., Shadlen, M. N., Celebrini, S. & Movshon, J. A. Vis. Neurosci. 13, 87–100 (1996).
Desimone, R. & Duncan, J. Annu. Rev. Neurosci. 18, 193–222 (1995).
Brain, W. R. Brain 64, 244–272 (1941).
Moore, T. & Armstrong, K. M. Nature 421, 370–373 (2003).
Bisley, J. W. & Goldberg, M. E. Annu. Rev. Neurosci. 33, 1–21 (2010).
About this article
Cite this article
Smolyanskaya, A., Born, R. Attention is more than meets the eye. Nature 489, 371–372 (2012). https://doi.org/10.1038/489371a
Progress in Neurobiology (2015)