Trends in Neurosciences
ReviewFeature ReviewMechanisms of top-down attention
Introduction
Language is infused with idiomatic expressions that make explicit the distinction between bottom-up (BU) and top-down (TD) processes of attention. We might ask someone to ‘pay attention to the road’ while driving, which implies a voluntary choice to allocate resources to a subset of the perceptual input. Alternatively, we might remark that the orange sports car really ‘caught our attention’. In this case, the resource has been involuntarily captured rather than voluntarily allocated. The distinction is not limited to idiomatic expressions, but rather stems from disparate modes of attentional processing [1]. BU attention is deployed very rapidly and depends exclusively on the properties of a sensory stimulus. By contrast, TD attention is slower and requires more effort to engage.
In the modality of vision, the two modes (BU and TD) give rise to the psychophysical phenomenon of pop-out and set-size effects. In a typical visual search experiment, a subject is presented with a number of items on a display and is asked to find a target item within this display, such as a bar with a particular orientation, or color, or a combination of the two. Pop-out occurs when the target item is significantly distinct from the surrounding items (distractors), such as a horizontal bar among several vertical bars. This different item automatically attracts BU attention (or pops-out) rapidly and independently of the number of distractors 2, 3. By contrast, when the target item is distinguished only by taking into account the conjunction of its features, such as color and orientation, BU cues alone cannot efficiently guide attention and TD attention must be recruited to scan the display. This gives rise to search times that increase with the number of distractors; in other words, a set-size effect is observed. In most real-life situations, the responses of the nervous system to a sensory input depend on both BU influences driven by the sensory stimulus and TD influences shaped by extra-retinal factors such as the current state and goal of the organism 4, 5.
A distinction is also made between two types of TD mechanisms. The first type is intuitively associated with TD and is called the volitional TD process, which can exert its influence through acts of will. The second type is known as a mandatory TD process and it is an automatic, percept-modifying TD mechanism that is pervasive and that volition cannot completely eliminate. The latter TD process can develop through experience-dependent plasticity or during development, and includes contextual modulation of perception 5, 6. A striking example of the dichotomy between these two mechanisms is presented in Figure 1.
Previous work has extensively studied the effects of TD attention on target brain regions, including modulatory effects in early sensory areas 5, 7, 8. Significant progress has been made in isolating the possible sources of TD signals [9], especially within the now well-studied frontoparietal attention network [10]. Much less understood at present are the exact pathways, contents, meaning and form of the signals that are sent from the top down. Here, we review recent findings from physiology, lesion and computational studies that have attempted to elucidate the mechanisms and signals involved in TD modulation of sensory processing. To focus this review, we mainly concern ourselves with visual perception and the volitional TD process, although similar principles can apply in other modalities.
Section snippets
Brain structures and circuits of visual attention
Visual processing begins in the retina, which sends parallel streams of information to the brain through its diverse set of retinal ganglion cells and their unique interactions within the retinal circuitry [11]. A majority of retinal projections reach the lateral geniculate nucleus (LGN) and a much smaller number (approx. 10%) connect to the superior colliculus (SC). The LGN sends projections to the primary visual cortex (V1), the initial site of processing in the cortical feed-forward visual
Pathways of TD attention
In this section, we focus on lesion and electrophysiological studies, particularly those using methods of microstimulation and simultaneous recordings in the brain areas identified in the previous section. These areas form an attentional network (Figure 2) and we consider how TD information is relayed in this network. Microstimulation, together with reversible inactivation [using either pharmacological agents such as muscimol or transcranial magnetic stimulation (TMS)] and permanent lesion
Subcortical influences on TD attention
Evidence suggesting that cortical areas have a strong influence on attention was discussed in the previous section. There are also several subcortical areas that play a crucial role in defining and communicating attentional signals (Figure 5a). It has been demonstrated that the phenomenon of change blindness, in which changes to a particular part of a visual scene go undetected, could be eliminated in monkeys by placing an attention-grabbing salient stimulus in the location where the blindness
The role of reward and emotion in TD attention
Until recently, studies of visual attention have traditionally tended to avoid non-visual aspects of cortical and subcortical neuronal responses to manipulations of attention. This has begun to change with a small number of psychophysical and electrophysiological studies that have explored the interplay between reward and attention.
To investigate the role of reward in modulating attention-related responses in the LIP, stimulus selection has been dissociated from motor selection in monkeys [71].
The role of oscillatory activity and neuromodulation in TD attention
It has recently been suggested that synchronous activity (in the gamma range, 50–80 Hz) between cortical regions might serve as the basis for attentional facilitation and cortical computations [84]. In this proposal, neuronal populations representing inputs and decision centers all consist of rhythmically active neural ensembles with distinct excitatory and inhibitory phases. Inhibitory interneurons in each ensemble rhythmically inhibit excitatory pyramidal neurons, thereby establishing a
Computational modeling
Physiological studies have guided several theoretical and computational models of attention. Building on the influential feature integration theory [2], guided search theory hypothesizes that massively parallel pre-attentive processes can be guided by TD biasing for features and locations [92]. This theory brings TD elements to a basic BU model of attention [93], which computes individual features at different scales and then combines these features to form a saliency map. A unifying
Conclusion
Attention modulates sensory signals early in the process, exerting its influence on the SC and the thalamus before further modulating signals in cortex. The cumulative effects of this modulation based on both TD and BU influences might be represented by a priority map over visual space. Although there is some debate about the exact locus of the priority map, it is clear that the LIP, FEF and SC exhibit properties that are compatible with the existence of a spatial map encoding behavioral
Acknowledgements
This work was supported by the Defense Advanced Research Projects Agency (government contract no. HR0011-10-C-0034), the National Science Foundation (CRCNS grant number BCS-0827764), General Motors Corporation, and the Army Research Office (grant no. W911NF-08-1-0360). The authors affirm that the views expressed herein are solely their own, and do not represent the views of the United States government or any agency thereof. We would also like to thank Robert Desimone, Jack Gallant, Jacqueline
Glossary
- BU influence
- influence on the nervous system due to extrinsic properties of the stimuli.
- Conjunction search
- search task in which a subject is required to find a target item among several distractors, and the target is defined by a unique conjunction of features. In this type of search task, locating the target is more difficult because distractors share some of the features of the target and thus the target does not obviously stand or pop out.
- Covert attention
- attention paid to a subset of the
References (138)
- et al.
A feature-integration theory of attention
Cogn. Psychol.
(1980) - et al.
Brain states: top-down influences in sensory processing
Neuron
(2007) - et al.
Contextual cueing: implicit learning and memory of visual context guides spatial attention
Cogn. Psychol.
(1998) Top-down control of visual attention
Curr. Opin. Neurobiol.
(2010)Parallel processing in the mammalian retina: lateral and vertical interactions across stacked representations
Prog. Brain Res.
(2001)- et al.
Separate visual pathways for perception and action
Trends Neurosci.
(1992) - et al.
Salience, relevance, and firing: a priority map for target selection
Trends Cogn. Sci.
(2006) - et al.
Feature-based attention increases the selectivity of population responses in primate visual cortex
Curr. Biol.
(2004) - et al.
Attention as a decision in information space
Trends Cogn. Sci.
(2010) From thought to action: the parietal cortex as a bridge between perception, action, and cognition
Neuron
(2007)
The distinct modes of vision offered by feedforward and recurrent processing
Trends Neurosci.
Measurements of simultaneously recorded spiking activity and local field potentials suggest that spatial selection emerges in the frontal eye field
Neuron
Global feature-based attention for motion and color
Vis. Res.
The normalization model of attention
Neuron
Distinct mechanisms for top-down control of neural gain and sensitivity in the owl optic tectum
Neuron
The thalamus is more than just a relay
Curr. Opin. Neurobiol.
Pulvinar contributions to the dorsal and ventral streams of visual processing in primates
Brain Res. Rev.
The pulvinar and visual salience
Trends Neurosci.
The basal ganglia: learning new tricks and loving it
Curr. Opin. Neurobiol.
On the importance of the transient visual response in the superior colliculus
Curr. Opin. Neurobiol.
Shedding new light on the role of the basal ganglia-superior colliculus pathway in eye movements
Curr. Opin. Neurobiol.
Unraveling the attentional functions of cortical cholinergic inputs: interactions between signal-driven and cognitive modulation of signal detection
Brain Res. Rev.
Prefrontal cortical modulation of acetylcholine release in posterior parietal cortex
Neuroscience
Modeling the influence of task on attention
Vis. Res.
Search goal tunes visual features optimally
Neuron
Modeling visual-attention via selective tuning
Artif. Intell.
The Principles of Psychology
What attributes guide the deployment of visual attention and how do they do it?
Nat. Rev. Neurosci.
Computational modelling of visual attention
Nat. Rev. Neurosci.
Neural mechanisms of selective visual attention
Annu. Rev. Neurosci.
An integrative theory of prefrontal cortex function
Annu. Rev. Neurosci.
Control of goal-directed and stimulus-driven attention in the brain
Nat. Rev. Neurosci.
Enhancement of visual responses in monkey striate cortex and frontal eye fields
J. Neurophysiol.
Attention, intention, and priority in the parietal lobe
Annu. Rev. Neurosci.
A visual salience map in the primate frontal eye field
Prog. Brain Res.
The functional logic of cortico-pulvinar connections
Philos. Trans. R. Soc. Lond. B Biol. Sci.
Guarding the gateway to cortex with attention in visual thalamus
Nature
Attention modulates responses in the human lateral geniculate nucleus
Nat. Neurosci.
A backward progression of attentional effects in the ventral stream
Proc. Natl. Acad. Sci. U.S.A.
Effects of attention on orientation-tuning functions of single neurons in macaque cortical area V4
J. Neurosci.
The representation of visual salience in monkey parietal cortex
Nature
Neuronal activity in the lateral intraparietal area and spatial attention
Science
Cortical connections of area V4 in the macaque
Cereb. Cortex
Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices
Science
Representation of attended versus remembered locations in prefrontal cortex
PLoS Biol.
Neural mechanisms of spatial selective attention in areas V1, V2, and V4 of macaque visual cortex
J. Neurophysiol.
Parallel and serial neural mechanisms for visual search in macaque area V4
Science
Feature-based attention influences motion processing gain in macaque visual cortex
Nature
Neural correlates of attentive selection for color or luminance in extrastriate area V4
J. Neurosci.
Attentional modulation of visual processing
Annu. Rev. Neurosci.
Cited by (315)
Phase of neural oscillations as a reference frame for attention-based routing in visual cortex
2024, Progress in NeurobiologyFrom Driver to Supervisor: Comparing Cognitive Load and EEG-Based Attentional Resource Allocation Across Automation Levels
2024, International Journal of Human Computer StudiesVigilance end-spurt patterns in event-related potentials
2023, Brain ResearchThe secondary somatosensory cortex gates mechanical and heat sensitivity
2024, Nature Communications