Unilateral attentional neglect is the inability to attend to sensory stimuli on one side of the visual field. Most commonly, this occurs when people with strokes experience damage in the right hemisphere, impairing their ability to attend to stimuli in the left hemisphere. Neglect can be tested by asking the patients to perform visual search on an array of objects – patients with unilateral neglect will fail to identify targets on one side of the array. In contrast, patients who are blind on one side will simply redirect their vision so that the entire array lies within their intact visual field.

One of the puzzling things about unilateral neglect is that its symptoms can be quite varied across individuals. The presence of such diverse symptoms might indicate that attentional control – that is, the ability to modulate the focus of attention – is distributed across many brain regions. In 1981, M.-Marsel Mesulam proposed a four-part network of directed attention [1]. In his view, the parietal component encodes a sensory map of the world, the frontal component coordinates actions using that representation, the cingulate cortex regulates motivations for such actions, and the reticular component provides the baseline level of arousal needed for directed attention. Such a theory makes very particular predictions about the symptoms of unilateral neglect that should be experienced due to damage to or manipulation of the respective network components.

Later studies have since broadly supported Mesulam’s cortical network for directed attention. For example, a PET study by Nobre and colleagues showed that attentional control tasks activated the same regions postulated by Mesulam from lesion studies [2].

Surprisingly, unilateral neglect in humans is rarely caused by left hemisphere damage. Using this fact, Mesulam deduced that the right hemisphere is important for directed attention in both halves of the visual field. This theory, sometimes called the hemispatial theory, would explain why right hemisphere damage should have a disproportionate effect on directed attention. However, the hemispatial theory has recently been challenged by neuroimaging evidence. In 2010, Szczepanski and colleagues used topographical mapping techniques with fMRI to demonstrate that both the left and right parietal cortex are broadly biased in favor of the contralateral visual field [3]. That is, the brain areas corresponding to symmetric areas of the left and right visual fields systematically favored inputs from their assigned side, with the notable exception of the right hemisphere area SPL1 and the left hemisphere frontal eye field and intraparietal sulcus.

This evidence – that both the left and right frontoparietal areas are systematically biased towards their contralateral visual fields – challenges the hemispatial theory of unilateral neglect. As an alternative theory, the authors propose an interhemispheric competition theory, which argues that the left and right hemispheres act in constant competition for the focus of directed attention [4]. To explain the prevalence of left visual field neglect, this theory’s proponents argue that damage to the right area SPL1 is simply more common than damage to the left IPS. In further support of this theory, another study found that, by using transcranial magnetic stimulation to interfere with activity in either the IPS or right SPL1, the researchers could induce visual neglect in both fields of vision. Moreover, several patients experiencing lesions confined to the left IPS do experience visual neglect in the right visual field. So converging evidence from fMRI, TMS, and lesion studies support the idea of interhemispheric competition as a refinement upon Mesulam’s original attentional control network.

If the interhemispheric competition account of unilateral visual neglect is correct, then it hints at a deeper idea about competition between sensory inputs. Consider the common real-world situation in which two stimuli lie in the same neuron’s receptive field – for example, a banana and an apple. What should that neuron send as an outgoing signal to represent what it receives? If it simply sends an average or sum of the signals for “banana” and “apple,” how can downstream neurons disambiguate between the two signals? How can they discriminate between bottom-up neural inputs in general? According to the biased competition account of attention, this is exactly the purpose that attention serves in the brain – by biasing the reception of bottom-up sensory inputs in favor of the attended stimulus, they sway the competition between those inputs. For example, a study by Moran and Desimone in 1985 recorded neurons from the visual areas V2, V4, and IT in monkeys. At first, they showed two stimuli in the same neuron’s receptive field with equal relevance for the monkey. The firing rate of the neuron converged on an average of the signals for the two stimuli in isolation. Then the researchers modified the setup so that only one of the stimuli was relevant to the monkey. When the monkey focused on only one stimulus, the firing rate of the neuron matched the firing rate corresponding to the attended stimulus alone. As the biased competition theory explains, the top-down signal of attention picked a winner in the competition between the bottom-up stimuli, based on high-level goals from working memory [5].

Perhaps the main theoretical alternative to the biased competition theory is the feature integration account of attention, notably proposed by Treisman and Gelade in 1980 [6]. In their review paper, they argue that attention acts like a spotlight to link features into a coherent object. They use this theory to explain experiments like visual search – for example, distinguishing a red T from a landscape of red I’s and green T’s is slower than distinguishing a red T from a landscape of green Ts. The amount of time it takes to solve this kind of visual search problem increases linearly with the number of confounding features, supporting the idea that attention is used serially to resolve object features. This view represents a kind of pessimism about the fidelity of human vision – if focal attention is required for most object recognition, then most of what we see outside of our focal attention is hallucinated and only the subject of our attention is determinate.

If the feature integration theory, as presented by Treisman and Gelade, were really true, then objects could seldom guide attention, because attention is needed to detect those objects in the first place. To the contrary, as Desimone and Duncan point out, objects often guide attention. For example, people are faster to identify two features of the same object than two equally collocated features – presumably because their attention to the features works upon already-binded object representations. Only by taking objects detected in parallel to be prior to the feature detection could the computational efficiency be improved. In contrast to feature integration theory, biased competition theory predicts that the brain greedily creates most object representations without requiring focal attention. Instead of being required to integrate features into objects, attention is better understood as modulating the competition among competing bottom-up stimuli.

As attention researchers began to converge upon the theory of biased competition, some began to investigate how it might be implemented. One promising – if unverified – theory comes from gamma-band synchronization. In a 2001 study of macaque monkeys, Pascal Fries and colleagues found that, when monkeys were presented with competing stimuli of which only one was relevant, they observed increased gamma-band frequency (35-90 Hz) synchronization among the neurons activated by the relevant stimulus compared to neurons activated by distractors [7].

What might explain this difference in synchronization? One theory is presented by Fries in a 2009 article [8]. In the 1990s, gamma-band synchronization had been presented by Wolf Singer, director of the Max Planck Institute for Brain Research during Fries’s time there, as a potential solution to the binding problem. The binding problem – which is also the problem that motivated feature integration theory – asks how it is that bits of information from sensory systems get merged into a coherent representation. In their electrophysiological research, Singer and colleagues observed gamma-band synchronization between related brain areas when cats responded to a sudden change in visual pattern [9]. Based on this and other evidence, they hypothesized that gamma-band synchronization helps the brain merge disparate brain areas into a single representational state. To provide a potential implementation for this hypothesis, Fries presents a mechanistic theory of gamma-band synchronization – that is, how gamma-band synchronization in a certain group of neurons mechanistically affects synchronization in adjacent groups.

Here’s how gamma-band synchronization might work: At the start, a certain sensory neuron starts firing. Nearby interneurons, called basket cells, start firing in response, and they are interlinked such that they tend to synchronize. Then, neurons downstream of the interneurons synchronize to the rhythm of the basket cells, such that the whole group is firing in tandem. Now imagine that two neuronal groups, A and B, both project onto a higher-level group C. Which group should group C sync its firing rate to? Probably whichever one is more in sync – because neurons in group C rely on temporal summation in short synaptic connections, high frequency synchronization (i.e., gamma-band synchronization) would be more likely to yield feedforward activation than the lack thereof. Let’s say group A has stronger gamma synchronization. Then A and C become coherent due to this feedforward coincidence detection. Finally, since incoming signals in group C are only significant when they arrive at times of maximum input gain, only signals in time with the A-C rhythm are received by C, effectively “locking in” the choice of A.

If the gamma-band synchronization binding hypothesis is true, then this mechanism provides one possible implementation. If neuronal groups for, say, the shape of a banana and the colour of a banana both fire in sync, then high-level cortical regions will synchronize to this rhythmic representation, which is experienced as the percept of a banana. Notably, this mechanism for an exclusive communication link also provides a good explanation for biased competition. First, attention biases higher cortical areas to select certain lower-level inputs. Then, input gain modulation via gamma-band synchronization forms exclusive communication links between neurons corresponding to the attended stimulus and higher cortical areas, cutting off the feedforward signals from distracting stimuli.

As a further refinement upon the mechanism of attentional selection via gamma-band synchronization, Fries suggests that lower-frequency theta-band and alpha-band synchronization may help to make and break gamma-wave synchronization patterns. Early evidence of this mechanism comes from the fact that gamma-band synchronization patterns appear to be built up and broken down over the course of a single theta-band oscillation (about 200 milliseconds). In a 2005 study by Rollenhagen and Olson, monkeys presented with a preferred stimulus and then a conflicting non-preferred stimulus – where the preferred stimulus robustly induces a higher firing rate in a single inferotemporal neuron – demonstrated an oscillating firing rate in the theta-band frequency, suggesting their attention alternated between selection of the preferred and non-preferred stimulus every 200 milliseconds [10]. However, this explanation of attentional oscillation is only a hypothesis, and needs to be verified by further research.

By providing a testable implementation of attention as a biased competition between bottom-up inputs, gamma-band synchronization lends further credence to the biased competition theory. At the same time, many questions remain. One question involves the function of topographical maps in the parietal cortex, of which there are apparently many. Another question involves the role of attention in enhancing visual acuity even in the absence of distracting stimuli. And the outstanding question, for which attentional neglect offers a compelling case study, involves the relationship between attention and awareness: attention seems to be necessary for some kinds of awareness, as in the awareness of the left visual field in patients with left visual neglect. On the other hand, attention also seems to be a fundamental property of information processing – the selection and prioritization of relevant information in a resource-constrained environment – applied by systems ranging from beetles to artificial neural networks. These two facts suggest that the further study of attention might yield universal insights into the nature of intelligence and awareness.

References

Mesulam, M.-.-M. (1981), A cortical network for directed attention and unilateral neglect. Ann Neurol., 10: 309-325. https://doi.org/10.1002/ana.410100402

Nobre, A. C., Sebestyen, G. N., Gitelman, D. R., Mesulam, M. M., Frackowiak, R. S., & Frith, C. D. (1997). Functional localization of the system for visuospatial attention using positron emission tomography. Brain : a journal of neurology, 120 ( Pt 3), 515–533. https://doi.org/10.1093/brain/120.3.515

Szczepanski, S. M., Konen, C. S., & Kastner, S. (2010). Mechanisms of spatial attention control in frontal and parietal cortex. Journal of Neuroscience, 30(1), 148-160. https://doi.org/10.1523/JNEUROSCI.3862-09.2010

Scolari, M., Seidl-Rathkopf, K. N., & Kastner, S. (2015). Functions of the human frontoparietal attention network: Evidence from neuroimaging. Current Opinion in Behavioral Sciences, 1, 32-39. https://doi.org/10.1016/j.cobeha.2014.08.003

Nobre, A. C., & Kastner, S. (Eds.). (2014). The Oxford Handbook of Attention (Oxford Library of Psychology). Oxford University Press.

Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97-136. https://doi.org/10.1016/0010-0285(80)90005-5

Fries, P., et al. (2001). Modulation of oscillatory neuronal synchronization by selective visual attention. Science, 291, 1560-1563. https://doi.org/10.1126/science.105546

Fries, P. (2009). Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annual Review of Neuroscience, 32, 209-224. https://doi.org/10.1146/annurev.neuro.051508.135603

Roelfsema, P., Engel, A. K., König, P., & et al. (1997). Visuomotor integration is associated with zero time-lag synchronization among cortical areas. Nature, 385, 157-161. https://doi.org/10.1038/385157a0

Rollenhagen, J. E., & Olson, C. R. (2005). Low-frequency oscillations arising from competitive interactions between visual stimuli in macaque inferotemporal cortex. Journal of Neurophysiology, 94, 53368-53387.