Neuronal Mechanisms of Spatial Attention in Visual Cerebral Cortex
Abstract and Keywords
Attention is associated with improved performance on perceptual tasks and changes in the way that neurons in the visual system respond to sensory stimuli. While we now have a greater understanding of the way different behavioural and stimulus conditions modulate the responses of neurons in different cortical areas, it has proven difficult to identify the neuronal mechanisms responsible for these changes and establish a strong link between attention-related modulation of sensory responses and changes in perception. Recent conceptual and technological advances have enabled progress and hold promise for the future. This chapter focuses on newly established links between attention-related modulation of visual responses and bottom-up sensory processing, how attention relates to interactions between neurons, insights from simultaneous recordings from groups of cells, and how this knowledge might lead to greater understanding of the link between the effects of attention on sensory neurons and perception.
It is well established that the responses of individual visual neurons are modulated when subjects shift their attention from one location in visual space to another (reviewed by Olson 2001; Bisley and Goldberg 2010; Reynolds and Chelazzi 2004; Carrasco 2011; Gilbert and Sigman 2007; Maunsell and Cook 2002; Braun et al. 2001). Most neurons respond more strongly when the subject attends to a stimulus within their receptive fields relative to when attention is directed to a distant location. Shifts in spatial attention have been associated with changes in the responses of individual neurons in every visual cortical area that has been examined, including primary visual cortex (V1; Roelfsema et al. 1998; Thiele et al. 2009; McAdams and Maunsell 1999; Motter 1993).
More recently, electrophysiological studies have begun to elucidate the mechanisms underlying attention-related changes in the responses of sensory neurons and to establish a link between the modulation of sensory neurons and improvements in performance on psychophysical tasks. Progress in three areas has been particularly notable. First, while it has long been understood that a bottom-up sensory processing mechanism known as normalization can explain some aspects of modulation with attention (Lee et al. 1999; Reynolds et al. 1999; Carandini and Heeger 2012), recent results have underscored that normalization is critical for interpreting many aspects of the way that sensory responses change with attention. This insight has led to a refined understanding of the range of attention-related effects that are seen under different stimulus conditions. Second, studies that have recorded simultaneously from many neurons during attention tasks have found that attention has striking effects on correlations between (p. 319) the responses of nearby neurons. Such changes in the way that populations of neurons encode sensory signals cannot be seen when recording from one neuron at a time, but they might prove critical to the behavioural advantages associated with attention. Third, simultaneous recordings from a few dozen neurons have provided the statistical power needed to link attention-related modulation of sensory neurons with improvements in performance on a trial-by-trial basis. This discovery allows for temporally precise measures of attentional state (milliseconds), compared with the traditional approach of averaging over many trials (minutes).
These conceptual advances, combined with the advent of new recording technologies that have made identifying neurons and recording simultaneously from groups of neurons experimentally tractable, have improved our understanding of how spatial attention is linked to neuronal responses and have opened new avenues for future research. Here, we review these recent findings, focusing on experiments that have recorded the responses of individual neurons (recorded one at a time or simultaneously) in the visual cortex of trained, behaving monkeys. Although our subject is spatial attention, we will also refer to results from feature attention when they bear directly on the interpretation of the data. A comprehensive treatment of feature attention can be found elsewhere in this volume (see Wolfe, chapter 2; Scholari, Esther, and Serences, chapter 20; and Treue, chapter 21). We also discuss the potential of these ideas for extending our knowledge of how the responses of sensory neurons in different states of attention can affect perceptual performance.
Single Unit Measures: Modulations and Normalization
Sensory response modulation
Scaling of tuning curves
Attention to a visual stimulus is typically associated with increases in the responses of cortical neurons that represent that stimulus. Early attempts to understand the mechanism underlying this attentional modulation focused on the link between attention and the selectivity of neurons to different stimulus attributes. Neurons in visual cortex respond selectively to stimulus properties such as colour, orientation, and direction of motion. This selectivity is captured in plots of neuronal response as a function of the value of a given stimulus attribute (tuning curves; see Fig. 11.1). Although some early observations suggested that tuning curves for orientation in area V4 might be sharper when subjects attended to the stimuli (Haenny and Schiller 1988; Spitzer et al. 1988), it has subsequently been shown that in many situations responses to all stimuli are enhanced more or less proportionally, resulting in an overall multiplicative scaling of tuning curves.
A scaling of tuning functions with attention has been seen for responses to many stimulus dimensions by neurons in several visual areas. In area V4, responses to all orientations increase by approximately the same proportion when attention is directed toward the stimulus. Fig. 11.1a shows average orientation tuning curves for single neuron responses when attention was directed to or away from the visual stimulus in the neuron’s receptive field (McAdams and Maunsell 1999). In the middle temporal area (MT), attention to a stimulus is similarly associated with a scaling of average direction tuning curves (Fig. 11.1b, Treue and Martinez-Trujillo 1999b). Spatial attention can also be associated with scaling of contrast response functions for neurons in MT (Fig. 11.1c, Lee and Maunsell 2010b) and V4 (Williford and Maunsell 2006). Measurements of the temporal integration function used by MT neurons show that attention is associated with a scaling of the function with no shift in time or change in direction selectivity (Cook and Maunsell 2004). In area V4, receptive field profiles scale when attention is directed to different points around the receptive field (Connor et al. 1996, 1997), and in inferotemporal cortex (IT) behavioural differences between reward-contingent stimuli and other stimuli are coupled to scaling of orientation-tuning curves (Vogels and Orban 1994).
Like visual responses, spontaneous activity also changes with attention. Spontaneous activity in cortex typically amounts to only a few spikes per second, and modulation of such modest rates can be difficult to detect. Nevertheless, reliable changes in spontaneous activity have been reported in many visual areas (Luck et al. 1997; Reynolds et al. 2000; Williford and Maunsell 2006; Li and Basso 2008). Relatively indiscriminate scaling of both spontaneous activity and visual responses suggests that attention is associated with an increase in the overall sensitivity of neurons to all their inputs.
While an increase in sensitivity might be the primary signature of attention in sensory neurons, responses are affected in other ways. In particular, as we describe in the following section, shifting attention between preferred and non-preferred stimuli inside a neuron’s receptive field can be associated with more complex effects. Even with a single stimulus in the receptive field, changes that are inconsistent with a simple scaling of sensitivity can occur with shifts in attention. For example, although contrast response functions are often scaled upward during attention to a stimulus (Fig. 11.1c), in some cases contrast response functions can instead shift to the left (Reynolds et al. 2000; Martinez-Trujillo and Treue 2002). Additive offsets of tuning curves have also been described (Williford and Maunsell 2006; Thiele et al. 2009). Moreover, attention to stimulus motion has been associated with changes in the way that MT neurons sum motion in space and time that differ from what would be expected from a simple scaling of responses (Ghose and Bearl 2010). Attention to stimulus features has been linked to shifts in the orientation and spatial frequency tuning of neurons in V4 (David et al. 2008).
Results like these show that there is considerable diversity in attention-related changes in the selectivity of individual visual neurons. While a range of effects on the selectivity of sensory neurons has been observed, the relative potency of each effect remains to be established. Currently it seems that a simple scaling of responses is the most common and strongest effect associated with attention, while other effects might require careful measurement or specific circumstances.
(p. 322) Modulation with multiple stimuli in the receptive field
Typically, the neuronal correlates of attention are studied by shifting attention between a single stimulus inside a neuron’s receptive field and a distant stimulus. This manipulation usually produces a modest modulation, in the range of 5–20% of the rate of firing. However, it has long been known that shifting attention between two stimuli inside a neuron’s receptive field can yield much stronger changes in response. Moran and Desimone (1985) were the first to show this effect. They trained monkeys to shift attention between a preferred stimulus that strongly drove the neuron being recorded and a non-preferred stimulus that produced little or no response when presented alone, both of which were placed inside the neuron’s receptive field. When the animal attended to the preferred stimulus, responses of the neurons they recorded in V4 and IT were greatly enhanced compared to when the animal attended to the non-preferred stimulus. Strong response modulation from shifting attention between preferred and non-preferred stimuli has since been shown many times (Moran and Desimone 1985; Treue and Maunsell 1996; Luck et al. 1997; Reynolds et al. 1999; Ghose and Maunsell 2008; Ghose 2009).
A direct comparison of modulation with one versus two stimuli in a neuron’s receptive field has been difficult to obtain. Attending to one of two closely spaced stimuli in a receptive field is more demanding than attending to a single stimulus, and differences in the effort of the subject could account for much of the stronger modulation with two stimuli in the receptive field (Spitzer et al. 1988; Boudreau et al. 2006; Chen et al. 2008). This problem has been addressed by randomly interleaving brief presentations of either one or two receptive field stimuli that are too fleeting to allow the animal to alter its attentional effort. Such brief presentations facilitate comparisons of the modulations of the responses of MT neurons under equivalent attention conditions (Lee and Maunsell 2010a). Shifting attention from outside a cell’s receptive field to a single, preferred stimulus in the receptive field was associated with a 9% increase in response (Fig. 11.2a). However, shifting attention from outside the field to a preferred stimulus paired with a non-preferred stimulus inside the receptive field yielded a 28% enhancement (Fig. 11.2b; red vs black). Responses increased by 59% when attention was shifted from a non-preferred stimulus in the receptive field to a preferred stimulus in the receptive field (Fig. 11.2b; red vs blue). These measurements show that stimulus configurations greatly affect how much neuronal responses vary when spatial attention is shifted.
Normalization and attention
As described above, attention to a single stimulus inside a neuron’s receptive field is generally associated with an increase in that neuron’s response, as if attention was related to an overall increase in the sensitivity of the neuron. With two stimuli, however, attention can be associated with either an increase or a decrease in a neuron’s response, depending (p. 323) on whether a preferred or non-preferred stimulus lies at the focus of attention. This effect on cells’ responses cannot be accounted for by an overall increase in sensitivity. Recently, it has become widely recognized that the attention-related changes seen with either one or two stimuli in a neuron’s receptive field can be explained by a mechanism called response normalization.
Normalization describes the way that neurons respond when presented with more than one stimulus at the same time in their receptive fields. It was first introduced to explain failures of linear summation when sensory neurons are presented with multiple stimuli (Albrecht and Geisler 1991; Heeger 1992, 1993). In particular, normalization explains why adding a weakly excitatory stimulus to a receptive field containing a strongly excitatory stimulus will reduce, rather than increase, the response of a neuron (see Carandini and Heeger 2012). Normalization accurately explains responses to a broad range of stimulus conditions for neurons in V1 and extrastriate cortex, both for single units (Heeger et al. 1996; Carandini et al. 1997) and for populations of neurons (Busse et al. 2009). In the absence of attention, normalization causes a neuron’s response to reflect the similarity between the entire set of stimuli in its receptive field and its preferred stimulus. Therefore, the response to the combination of a preferred and non-preferred stimulus is more similar to the average than the sum of the responses to each stimulus alone.
(p. 324) Neurophysiological support for a role of normalization in spatial attention
Several reports have shown how normalization models can account for many aspects of the way that the activity of sensory neurons varies with attention (Boynton 2009; Ghose 2009; Lee and Maunsell 2009; Reynolds and Heeger 2009). Normalization readily explains why shifting attention between a stimulus in a neuron’s receptive field and a distant stimulus produces relatively little modulation compared to shifting attention between two different stimuli that both lie within the receptive field (Fig. 11.2). This difference occurs because normalization mechanisms have relatively little influence when only one stimulus is present, but come into play strongly when multiple stimuli are present. Normalization can also explain how shifts in attention can scale the tuning response function and why neuronal modulations differ when shifting attention between stimuli that have either the same contrast or different contrast (Khayat et al. 2010). Some normalization models can produce leftward shifts of contrast response functions (contrast gain; Ghose 2009; Reynolds and Heeger 2009; Boynton 2009). More elaborate normalization models allow attention to be distributed narrowly or broadly in space (Reynolds and Heeger 2009; Ghose and Bearl 2010), thereby capturing a characteristic of attention that is well established in behavioural studies (LaBerge 1983; Eriksen and Yeh 1985) and seen in some single-unit experiments (Boudreau et al. 2006; Ghose 2009). When spatial attention is misaligned with the receptive field of a neuron, these models account for how attention can shift receptive fields (Connor et al. 1997; Womelsdorf et al. 2006). The ability of normalization to robustly accommodate a wide range of effects from attention suggests that it is an important mechanism for changing sensory responses when attention is directed to different stimuli.
While many neurophysiological results can be explained by normalization, few studies have provided evidence for a direct relationship between normalization and attention in the responses of neurons. Reynolds and colleagues (1999) measured the responses of neurons in V2 and V4 to two stimuli when they were presented alone or together. As in other studies, when the two stimuli were presented together, attention to the preferred stimulus was associated with an increase in the neuron’s response, while attention to the non-preferred stimulus was associated with a lower response. Because each neuron was tested with many stimulus pairs, some pairs included stimuli that were equally preferred by a cell. Normalization models of attention predict that there should be no modulation from shifting attention between two such stimuli, and that was what these investigators found. In MT, shifting attention between two identical stimuli is also associated with little response modulation compared to shifting attention between two differently preferred stimuli (Lee and Maunsell 2010a). Similarly, Chelazzi and colleagues (1998) showed that two stimuli placed within the receptive field of an inferotemporal neuron showed more modulation from shifting attention when both stimuli were placed on the same side of the vertical meridian, a configuration that associated with stronger stimulus interactions.
A cell-by-cell correlation between normalization and attention-related modulation was observed in an experiment that examined responses in MT (Lee and Maunsell 2009). Each cell was tested with a pair of stimuli within the receptive field that included (p. 325) a preferred stimulus and a non-preferred stimulus. With this pairing, cells are expected to show pronounced modulation when attention is shifted from one stimulus to the other. Normalization should also cause the addition of the non-preferred stimulus to reduce a cell’s response to a preferred stimulus. That was the case for many, but not all cells. MT neurons that showed little evidence of normalization also showed little modulation when attention was shifted between the stimuli. This strong correlation between the strength of normalization and the strength of modulation with attention was apparent in the population of MT neurons (Fig. 11.3). The modulation associated with attention approached zero when normalization approached zero. These results suggest that normalization and modulation by attention are tightly coupled in sensory neurons.
Implications of normalization for attentional modulation
Sensory normalization has long provided an important framework for understanding stimulus interactions in the generation of neuronal response. Recent work also shows that normalization provides insight into many aspects of the modulation of sensory responses associated with shifts in spatial attention. However, important questions remain unanswered. In particular, the precise relationship between normalization and attention remains to be established. It has been suggested that attention shifts are associated with changes in the potency of normalization (Lee and Maunsell 2009), which would imply that normalization mechanisms are critical for attention-related modulations in sensory cortex. However, the relationship might be less direct. Attention might primarily be associated with a scaling of the visual signals that provide input to a downstream neuron. Following this scaling, inputs might be combined according to normalization rules that are constant over different attention conditions, as in input gain models (Ghose 2009). In that case, normalization might amplify attention-related modulations but have no special relationship to attention. It will be important to determine whether any essential relationship exists between attention and normalization. (p. 326)
Another important question is whether normalization will have as prominent a role for feature attention as it does for spatial attention (Boynton 2009; Reynolds and Heeger 2009). Normalization models are generally conceptualized with each stimulus driving a separate population of neurons that spans a range of stimulus preferences, with shifts in attention differentially modulating these separate populations (Boynton 2009; Ghose 2009; Lee and Maunsell 2009; Reynolds and Heeger 2009). It is easy to envision two populations of neurons when two stimuli are spatially offset, as in experiments on spatial attention. However, the effects of feature attention span the visual field (Treue and Martinez-Trujillo 1999a, 1999b; Maunsell and Treue 2006). While it is possible that normalization can work with populations of neurons that are intermixed in cortex, it remains to be seen whether normalization is closely related to feature attention when it involves this more challenging configuration.
Normalization might also explain differences in the strength of attention-related modulation observed both within (Fig. 11.3) and between cortical visual areas. One possibility is that this variance is optimized for enhancing behaviour. For example, it might be beneficial for neurons to be modulated differently depending on how well their responses distinguish targets from distractors (Navalpakkam and Itti 2007; Scolari and Serences 2009; Scolari et al. 2012). However, if attention-related modulation depends on the strength of normalization, its variance might have little to do with behavioural strategies. Instead, variance in attention-related modulation might be simply incidental to the variance in normalization mechanisms (due to inherent differences in normalization between cells or details of how experimental stimuli are configured relative to a cell’s receptive field). The data in Fig. 11.3 suggest that at least half of the variance in attention-related modulation across neurons in MT reflects variance in normalization. Similarly, the larger attention-related modulations seen in later stages of visual cortex (see Maunsell and Cook 2002; O’Connor et al. 2002) might depend more on changes related to sensory normalization, such as differences in receptive field size or the need to remove redundancy in sensory coding (Schwartz and Simoncelli 2001), than on differences in the strength of inputs from higher cortical centres.
Population Measures: Correlations and Fluctuations in Attention
Most of our knowledge of how the responses of visual neurons change with attention comes from single-neuron studies such as those described in the previous sections. Typically, the average response of a neuron across many trials is compared between attention conditions. In general, attention is associated with improvements in the sensitivity of single neurons that arise from increased mean responses and, sometimes, decreased trial-to-trial variability (Mitchell et al. 2007). The resulting improvement in (p. 327) neuronal sensitivity is consistent with the hypothesis that the attention-related modulations of visual responses are responsible for the associated improvement in perception, although the improvement in the sensitivity of any individual neuron is typically small (McAdams and Maunsell 1999).
Rather than averaging the responses of single neurons over many stimulus presentations, however, animals must act on the sensory information encoded in populations of neurons over a short period. Single-neuron studies cannot measure several critical aspects of the way changes in attention are linked to changes in cortical circuits, interactions between neurons, or the relationship between specific subpopulations of cells and behaviour. The average responses of single neurons provide only limited access to the relationship between attention and changes in the underlying circuit, in part because in most studies the cell type or laminar location of the recorded studies is unknown. Furthermore, because so many neurons respond to any visual stimulus, it seems unlikely that the small changes in the signal-to-noise ratio of individual neurons underlie the dramatic perceptual improvement associated with attention. Theoretical studies suggest that interactions between neurons profoundly affect the amount of information encoded by a population of neurons (Shadlen and Newsome 1996; Abbott and Dayan 1999; Averbeck et al. 2006), so it is critical to consider how such interactions might change between states of attention. Finally, the activity of single neurons is only weakly related to an animal’s perceptual decisions (Parker and Newsome 1998; Nienborg and Cumming 2010). Analysing the activity of large groups of neurons circumvents this problem and provides the statistical power necessary to determine subtleties in the relationship between attention-related modulation of particular groups of neurons and perception (Cohen and Maunsell 2010, 2011b; Nienborg et al. 2012).
Recent advances in multi-neuron recording technologies such as multielectrode arrays, laminar probes, and two-photon imaging have made recording from populations of neurons in behaving animals more tractable and more popular. Here, we review studies of attention-related changes in populations of visual neurons and discuss possibilities for using population metrics to uncover the mechanisms underlying attention.
Attention and correlations between the activity of pairs of neurons
Beyond properties of single neurons, the simplest higher-order measures of attention-related changes in visual neurons concern interactions between pairs of neurons. Cortical neurons respond variably to identical presentations of a sensory stimulus (Tolhurst et al. 1983). The extent to which this response variability is shared across a population of neurons can in principle affect two of the hypothesized functions of attention: the efficacy of spikes in driving downstream cells and the amount of information available in the population. (p. 328)
The most straightforward way of assessing the shared variability in a neuronal population is to measure correlations in the trial-to-trial fluctuations in the responses of a pair of neurons. The word correlation can refer to a relationship between responses over timescales that range from millisecond-level precision (synchrony) to the timescale of sensory stimuli or behavioural responses (typically hundreds of milliseconds). Correlations on these longer timescales (called spike count correlations, noise correlations, or rSC) are typically quantified as the Pearson’s correlation coefficient between spike count responses from two neurons measured over trials with identical stimulus and task conditions (for review, see Kohn et al. 2009; Cohen and Kohn 2011).
Neuronal correlations can vary depending on a variety of factors. Both synchrony and spike count correlations depend on the sensory stimulus (Espinosa and Gerstein 1988; Aertsen et al. 1989; Ahissar et al. 1992; Kohn and Smith 2005), learning (Ahissar et al. 1992; Gutnisky and Dragoi 2008; Komiyama et al. 2010; Gu et al. 2011), and behavioural context (Vaadia et al. 1995; Cohen and Newsome 2008; Poulet and Petersen 2008; Cohen and Maunsell 2009; Mitchell et al. 2009), as well as on cognitive factors including attention. Synchrony on short timescales is thought to improve the ability of cortical neurons to drive downstream cells, and increasing synchrony has been hypothesized as a mechanism by which signal efficacy might be improved when attention is directed to a stimulus (for review, see Womelsdorf and Fries 2007). Synchrony between cortical areas has also been suggested as a mechanism by which inter-area communication might be enhanced in some attention conditions (Buschman and Miller 2007; Saalmann et al. 2007; Fries 2009; Gregoriou et al. 2009). Because synchrony between pairs of cells in visual cortex is typically weak and requires large amounts of data to measure, most studies of attention and synchrony have focused on coherence measures based on spiking or local field potential responses (for review, see Fries 2009). The two studies that have measured neuronal correlates of attention on short-timescale synchrony between spiking responses have found either very weak (Roy et al. 2007) or no (Roelfsema et al. 2004) differences between attention conditions (Steinmetz et al. 2000).
Recent studies have shown profound attention-related changes in spike count correlations (rSC) measured over hundreds of milliseconds. In two different tasks that modulated spatial attention, spike count correlations between pairs of neurons in V4 with overlapping receptive fields decreased by approximately 40% when the animals attended to locations inside the joint receptive fields of the neurons under study (Cohen and Maunsell 2009; Mitchell et al. 2009). This reduction in spike count correlation was substantially larger than the effects of attention on firing rates (<20%) and the variability of the individual neurons’ responses (quantified as a ratio of the variance of the spike counts to the mean, or Fano factor; <10%) in the same studies.
Implications for population coding
Pairwise correlations can profoundly affect the amount of information encoded by a population of neurons, so any process that modulates correlations will affect the sensory information available to guide behaviour. Attending to a location or feature improves perception of the attended stimulus, and recent evidence suggests that the (p. 329) associated large reduction in correlated activity might account for these perceptual improvements.
Correlations affect the information in neuronal populations because they affect the extent to which coding is improved by incorporating the responses of multiple neurons. If the fluctuations between neurons are independent, the amount of information available in a neuronal population grows with the square root of the population size (Zohary et al. 1994; Shadlen and Newsome 1996, 1998; Abbott and Dayan 1999). Intuitively, increasing the population size improves accuracy because independent noise can simply be averaged away.
The effect of correlations on population coding depends on how sensory information is read out by downstream cells, but it is almost always dramatic (Zohary et al. 1994; Shadlen and Newsome 1998; Abbott and Dayan 1999; Nirenberg and Latham 2003; Averbeck et al. 2006; Berens et al. 2011). For example, if a downstream cell were to average the responses of neurons with similar tuning properties, correlations would limit the benefit of having multiple cells. In this case, correlated fluctuations could never be averaged out, leading to a more variable (and less accurate) estimate of the mean rate. However, if a downstream cell responded to the difference between the responses of two neurons (for example, comparing the responses of cells with opposite tuning properties as in a discrimination task), a higher proportion of correlated noise would be beneficial as these shared fluctuations would be subtracted away.
Under most conditions, the amount and structure of correlated variability has a greater effect on the coding capacity of the population than do the signalling capabilities of single neurons. Many neurons will respond to any given sensory stimulus, so variability in the responses of single neurons can be compensated for by simply considering the responses of large groups of neurons. Correlations, however, can limit or enhance the benefit of reading out sensory information from multiple neurons.
Two recent studies in area V4 suggest that the reduction in correlation associated with spatial attention can account for the vast majority of the observed behavioural improvements (Cohen and Maunsell 2009; Mitchell et al. 2009). In these studies, pairs (Mitchell et al. 2009) or groups of neurons (Cohen and Maunsell 2009) were recorded simultaneously, the attention-related improvement in the stimulus sensitivity of these neuronal populations was quantified (Cohen and Maunsell 2009) or modelled, and the relative importance of different physiological changes for this improvement was measured. There was a strong link between the improvement in the sensitivity of the population and the monkey’s behavioural improvement due to attention (Fig. 11.4a), suggesting that the physiological factors that account for the improvement in sensitivity may also play a role in improving perception.
Increased attention was associated with at least three effects on populations of sensory neurons. Consistent with results from single-unit studies, firing rates increased and the variability of individual neurons decreased (quantified as the Fano factor; Mitchell et al. 2009). Additionally, spike count correlations decreased substantially. While each of these physiological changes could have contributed to the improvement in population sensitivity, the decrease in pairwise correlations was by far the most important factor in (p. 330) explaining the improvement in population sensitivity, accounting for over 80% of the observed improvement (Fig. 11.b). Importantly, the ideal way for a downstream brain region to combine the responses of V4 neurons to solve the detection task in the study that measured behavioural improvement was similar to averaging (Cohen and Maunsell 2009), so decreasing correlations is beneficial to coding efficiency. It remains to be seen whether increased attention can be associated with increased correlations or no change in correlations in situations in which that would be beneficial.
The effects of correlations on the representation of stimulus information in populations of neurons in general and the way this representation changes with attention are complicated and only beginning to be understood. These results suggest, however, that studies of the average responses of single neurons miss interactions between neurons that can have critical effects on behaviour. Consequently, the future of studying population coding must rely on multielectrode or imaging technologies that allow glimpses of population activity on the timescale of a single behavioural decision.
(p. 331) Implications for the neuronal mechanisms underlying attention
Studying correlations is also useful because it can provide insights about the neuronal mechanisms underlying attention that are inaccessible from the responses of single neurons. For example, correlations can shed light on the question of whether attention is associated with modulations of all aspects of neuronal responses or a particular communication channel in a specific frequency range (for review, see Womelsdorf and Fries 2007). A study by Mitchell and colleagues (2009) showed that attention is predominantly associated with changes in correlations on the timescale of tens to hundreds of milliseconds rather than precise synchrony that occurs over shorter periods. The authors trained monkeys to do a stimulus tracking task and recorded from pairs of V4 neurons during 1000 ms periods in which the attended object paused in the joint receptive field of the recorded neurons. They computed the spike-to-spike coherence as a function of frequency (which is monotonically related to correlation with sufficient amounts of data) and spike count correlation as a function of counting window. Consistent with other studies in visual cortex (Bair et al. 2001; Kohn and Smith 2005), the correlations they measured were dominated by co-fluctuations on the timescale of tens to hundreds of milliseconds (Fig. 11.5). Attention appears to be linked to substantial decreases in correlations at longer timescales, but only barely measurable effects on (p. 332) the already weak high-frequency synchrony. These results suggest that attention-related modulation of correlations is not frequency-specific, but instead affects the same mechanisms that cause these correlations in the first place.
Changes in correlations may also provide clues about attention-related changes in the underlying functional circuitry. Spike count correlations are thought to reflect the balance between the activity of common same-sign (excitatory or inhibitory) direct or indirect inputs and common opposite-sign inputs (e.g. mutual inhibition) to the two neurons (Zohary et al. 1994; Bair et al. 2001; Kohn and Smith 2005; Smith and Kohn 2008). Changes in correlations may therefore indicate changes in either the connection in the underlying circuit or in the relative activity levels of different inputs.
Recent evidence suggests that the modulations of rates and correlations associated with spatial attention are linked: the pairs of neurons that show large increases in rate show large decreases in correlations, and pairs with very little rate change show little correlation change (Cohen and Maunsell 2011a; Fig. 11.6, black line). A similar relationship was observed for a form of feature attention in which the same monkeys were trained to detect, in alternating blocks of trials, a change in either the orientation or spatial frequency of a stimulus. Like other forms of feature attention (for review, see Maunsell and Treue 2006), the stimulus feature that the animals expected to change either increased or decreased firing rates depending on the similarity of the neuron’s tuning to the attended feature (Cohen and Maunsell 2011a). Like spatial attention, the rate changes associated with feature attention were accompanied by changes in correlations, and large increases in firing rate were associated with large decreases in correlation (Fig. 11.6, right side of the grey line). The opposite relationship held as well: pairs of neurons whose rates decreased with feature attention showed increases in correlation (Fig. 11.6, left side of the grey line). The relationship between rate and correlation changes therefore appears to be similar for the two types of attention.
The results reviewed here show that studying correlations can provide new insights into both the mechanisms underlying attention and the impact of attention-related modulation of sensory neurons on the amount of information those neurons encode. These early results open several avenues for future research. Characterizing the conditions under which correlations change will be a first step towards understanding the way neural circuits reflect shifts in attention. For example, knowing whether shifts in attention are selectively associated with changes in correlations between neurons of different subtypes or in different cortical layers can provide clues about how the responses of populations of sensory neurons change with attention. Similarly, understanding whether there is always a fixed relationship between rate and correlation changes places limits on the mechanisms underlying modulation by attention or other cognitive, sensory, or motor factors. If the link between gain and correlation changes holds for simple processes such as normalization, the underlying mechanisms could be studied using genetic tools in rodents or other species without training complex attentional tasks. Finally, the role of correlations in population coding could be further investigated using tasks in which information coding would be improved by a different relationship between rate and correlation changes. (p. 333)
Using fluctuations in attention to study its neuronal basis
Changes in perception or sensory representations with attention are typically studied by comparing psychophysical performance or neural responses between sets of trials that differ in their instructions to the subject. Analyses that compare mean perceptual performance or mean neuronal responses implicitly assume that subjects follow task instructions reliably, making every trial within an attention condition identical. Despite the best efforts of experimenters and subjects, however, a subject’s attentional state will vary, even within the same task condition. Such uncontrolled fluctuations in attention can have important consequences for both perception and neuronal responses. The dynamics and cortical extent of these fluctuations and the extent to which fluctuations in different types of attention are coordinated can also provide new insight into the mechanisms underlying attention.
Measuring uncued attentional fluctuations requires an estimate of the subject’s attentional state on each trial, which has been impractical using either behavioural or neuronal responses. A single trial typically yields a single behavioural response, which is not easily decomposed into contributions from the subject’s behavioural state and perceptual capacity. The responses of individual sensory neurons are so variable that they have been similarly difficult to use to determine a subject’s attentional state. While many previous studies have found a relationship between single-neuron responses and behavioural choice (e.g., choice probability), the relationship between any one neuron and behaviour is typically weak (Parker and Newsome 1998; Nienborg and Cumming 2010). A neuron’s response on a single trial therefore cannot provide useful information about behavioural state.
The resolution lies in combining the responses of many neurons. Fluctuations in signals from scalp recordings (Thut et al. 2006; Bollimunta et al. 2008) and the BOLD signal from fMRI studies (Ress et al. 2000; Grill-Spector et al. 2004; Sapir et al. 2005; Fox (p. 334) et al. 2007; Leber 2010), which presumably reflect the activity of a large number of cells, are related to behavioural outcome. As with the responses of single neurons, however, fluctuations in these signals are only weakly correlated with behaviour and EEG and imaging methods do not allow flexibility in the way the responses of different neurons are combined for analysis.
Recent work shows that basing a single trial measure of attention on the responses of a few dozen simultaneously recorded neurons provides the power to accurately predict behaviour on a trial-by-trial basis (Cohen and Maunsell 2011a, 2011b). In the next section, we discuss this method for measuring attention at a single moment, some insights into attentional mechanisms, and avenues for future work.
Measuring attention on a single trial
A set of recent studies showed that a single-trial measure of attention can be based on the responses of a few dozen simultaneously recorded neurons in each hemisphere of visual area V4. These neurons can be used to identify fluctuations in attention in the context of a task involving either spatial attention alone (Cohen and Maunsell 2010) or spatial and feature attention simultaneously (Cohen and Maunsell 2011a). Just as traditional measures of attention compare mean responses in two attention conditions, the population analysis assessed the monkey’s attentional state on each trial in a similar way: by using a linear discriminator to quantify the similarity of the population response at a given moment to the average in each of the attention conditions.
A putative ‘attention axis’ was defined as the line connecting mean responses before correct detections in each of two cued attention conditions (Cohen and Maunsell 2010, 2011a, 2011b). The population responses for each trial, which form a high dimensional space (where each neuron represents one dimension) are projected onto this axis (Fig. 11.7a). The attention axis was constructed based only on data from correct trials, so missed trials provided an important test of whether position on the attention axis correlates with behavioural performance. On average, projections for trials in which the animal missed the orientation change had smaller values than for correct detections, meaning that attention was shifted toward the mean of the opposite attention condition (Cohen and Maunsell 2010).
The population projection provides a metric of attention that strongly correlates with the animal’s performance (Fig. 11.7b). On trials in which the animal’s attention was putatively directed to the stimulus on the left, the monkey did well detecting stimulus changes on the left (Fig. 11.7b, grey line) and poorly detecting changes on the right (Fig. 11.7b, black line). Conversely, on trials in which the population response was more similar to the mean for correct attend-right trials, the monkey correctly detected most changes on the right but few on the left. This metric, which was based on the responses of a few dozen visual neurons over just 200 ms, therefore provides a glimpse into the animal’s attentional state at a given moment, and the results in Fig. 11.7 show that fluctuations in attention are associated with changes in perception. (p. 335) (p. 336)
Insights from fluctuations in attention
Beyond establishing a relationship between attentional fluctuations and psychophysical performance, examining these fluctuations can provide insights about the mechanisms underlying attention that are inaccessible from firing rates and pairwise correlations. For example, comparing fluctuations in feature and spatial attention provides a new way to address long-standing questions about whether these two forms of attention share underlying mechanisms (Cohen and Maunsell 2011a). Such insights illustrate the importance of studying fluctuations in attention and other cognitive processes in the future.
A recent study compared feature and spatial attention by recording from populations of V4 neurons in both hemispheres of monkeys that performed a mixed spatial and feature attention task (Cohen and Maunsell 2011a). On each trial, four estimates of attention were obtained: one for spatial and one for feature attention calculated from the responses of the subset of neurons recorded in each hemisphere. With these measures, it was possible to ask whether spatial and feature attention are co-modulated by calculating the correlation coefficient between projections onto the feature and spatial attention axes both within (Fig. 11.8b) and across hemispheres (Fig. 11.8a). In both cases, the mean correlation coefficient was statistically indistinguishable from zero. Therefore, although feature and spatial attention are associated with similar changes in the rates and correlation of nearby neurons (Fig. 11.7), and likely mediated by similar mechanisms, these mechanisms nevertheless appear to function independently.
Calculating the correlation coefficient between projections onto attention axes defined separately for neurons recorded from the two hemispheres can also provide insight into the cortical extent of modulation by spatial and feature attention. The correlation between projections on the spatial attention axes for the two cerebral hemispheres was indistinguishable from zero (Fig. 11.8a). This lack of correlation was not a result of insufficient statistical power: when neurons recorded within each hemisphere were randomly divided into two equal-sized groups, a positive correlation between projections onto spatial attention axes calculated from each subgroup was easily detected. These data indicate that fluctuations in the amount of spatial attention allocated to the two stimuli arise from fluctuations in groups of neurons within a hemisphere, rather than because the animal attends to the wrong stimulus.
The cortical extent of neuronal changes related to feature attention is qualitatively different. The statistical power for detecting correlations along the feature attention axes was similar for feature and spatial attention (Fig. 11.8b). However, in contrast to spatial attention, projections on the two feature attention axes were positively correlated across hemispheres (Fig. 11.8a, grey bar).
These results are consistent with the idea that spatial attention involves coordination among local groups of neurons, and that the amount of attention allocated to locations in opposite hemifields is independent. In contrast, attention to features appears to involve coordination across neurons representing the entire visual field, selectively (p. 337) co-modulating neurons located far apart, even across hemispheres. The idea that spatial and feature attention operate on different spatial scales is supported by psychophysical evidence. A subject’s ability to attend to an object in one hemifield is unaffected by attention to objects in the other hemifield (Alvarez and Cavanagh 2005). Conversely, feature attention can affect visual processing independent of stimulus location (Motter 1994; Saenz et al. 2002, 2003; Liu and Mance 2011). Together, these results are consistent with the idea that feature and spatial attention are mediated by a unified attentional mechanism that can modulate the responses of arbitrary subgroups of neurons. More generally, they suggest that studying populations of neurons can provide ways of distinguishing between mechanisms underlying different cognitive processes.
Attention and the dynamics of neuronal modulations
As the studies of attentional fluctuations show, attention varies over short intervals. These fluctuations can occur even within a trial. When subjects scan a visual scene, they shift their eyes every few hundred milliseconds (DiCarlo and Maunsell 2000), (p. 338) and each shift is associated with a change in spatial attention (Zhou and Desimone 2011). Studies that have examined shifts in attention have found that changes in neuronal responses can be correspondingly rapid. When distracting stimuli appear unexpectedly, the responses of neurons in the lateral parietal area (LIP) to distractors rise and fall quickly and closely match the dynamics of behavioural distraction (Bisley and Goldberg 2003, 2006). When monkeys are given a signal to shift their attention from one target to another, neurons in V4 reflect the change in attention within 100–200 ms (Motter 1994). Corresponding experiments that compared latencies for attentional modulation in LIP and MT found similarly rapid shifts, but with modulation in LIP responses ~60 ms earlier than those in MT, consistent with a top-down flow of attention-related signals in visual cortex (Herrington and Assad 2009).
Subjects can also shift their attention during trials without any external cue or instruction. Monkeys that are familiar with the dynamics of a task will shift their attention to different spatial locations at appropriate times to maximize their chance of getting a reward. When the probability of receiving a reward varies predictably within trials, neurons in V4 (Ghose and Maunsell 2002) and LIP (Janssen and Shadlen 2005) are more strongly modulated with attention during periods of greater reward probability, with the strength of modulation varying over periods of no more than a few hundred milliseconds.
The dynamic aspects of attention pose a challenge for studies that aim to compare differences in neuronal activity across different tasks and conditions. When direct quantitative comparisons must be made, not only the stimulus conditions but also task difficulty and reward expectations must be carefully balanced. When considering the neuronal correlates of spatial attention, quantitative comparisons from different studies are problematic because neuronal modulations are greatly affected by the difficulty of the behavioural task. Subjects adjust the amount of attention that is allocated to different stimuli in response to task demands (Lavie and Tsal 1994; Urbach and Spitzer 1995; Lee et al. 1999), and this is reflected in neuronal activity. When subjects are faced with more difficult tasks, neuronal responses to given stimuli are stronger (Spitzer et al. 1988; Spitzer and Richmond 1991) and the modulation of neuronal responses associated with shifting attention between two stimuli is larger (Spitzer et al. 1988; Spitzer and Richmond 1991; Boudreau et al. 2006). Because task difficulty and attentional effort will vary between studies, precise quantitative comparisons of neurophysiological results from different laboratories are not practical. Studies that wish to quantitatively compare neuronal modulation linked to attention and other cognitive factors across stimuli and brain areas must therefore carefully control for task difficulty and reward expectation.
The ability to detect fluctuations in attention using populations of neurons provides a new avenue for studying the dynamics of switches in attention and other cognitive factors. The ability to estimate attention at a given moment will allow much more detailed studies of the dynamics of attention switching and the extent to which they depend on the demands of the behavioural task.
(p. 339) Attention and cortical cell classes
There is growing interest in whether attention is preferentially associated with modulation of the responses of specific cell classes in cortex. Extracellular recordings can distinguish two classes of neurons on the basis of spike duration. ‘Narrow’ spikes are thought to arise from inhibitory interneurons, while ‘broad’ spikes are thought to arise from pyramidal neurons, whose axons carry signals to other brain regions (McCormick et al. 1985; Connors and Gutnick 1990; Contreras and Palmer 2003). In prefrontal cortex, attention has been shown to be associated with changes in the selectivity of visually responsive neurons. When monkeys switch from a direction discrimination task to a speed discrimination task, the direction selectivity of narrow-spiking neurons in prefrontal cortex is reduced more than that of broad-spiking neurons (Hussar and Pasternak 2009). In V4, however, the responses of narrow-spiking and broad-spiking neurons scale by the same factor when spatial attention varies (Mitchell et al. 2007). Equivalent effects on both classes of neurons are consistent with the organization of the anatomical projection from prefrontal cortex to V4, which does not preferentially target inhibitory interneurons (Anderson et al. 2011a).
While both narrow- and broad-spiking V4 neurons change their mean rates proportionally when attention varies, they show differential effects on the variance of their responses across repeated presentations of the same stimulus. Mitchell and colleagues (2007) found that the variance of the broad-spiking neurons did not change with attention, and was the same as the variance of the narrow-spiking neurons when attention was directed toward their receptive fields. However, when attention was directed away from the receptive field, the narrow-spiking neurons increased their variance. Another difference in the effect of attention on narrow- and broad-spiking V4 neurons is the relationship between attention-related modulation and burstiness of firing. Among broad-spiking neurons, attention-related modulation is strongest among those broad-spiking neurons that have bursty patterns of firing (Anderson et al. 2011b), a relationship that is not found among the narrow-spiking neurons. Future studies of the circuits underlying attention might help explain why attention is associated with different effects for different cell classes in cortex.
Although one could imagine designing a nervous system in which cognitive factors do not affect basic sensory representations, it is clear that attention is associated with modulation of the responses of neurons throughout visual cortex. The studies described here have made promising strides toward understanding the neuronal mechanisms underlying this modulation and establishing a link between the modulation of sensory responses and changes in perception. Recent improvements in technology for recording from, identifying, and manipulating populations of neurons suggest that the field is well positioned to build on these insights to improve our understanding of how the changes observed in visual cortex improve observers’ perceptual abilities.
Abbott, L. F. and Dayan, P. (1999). The effect of correlated variability on the accuracy of a population code. Neural Computation 11: 91–101.Find this resource:
Aertsen, A. M. H. J., Gerstein, G. L., Habib, M. K., and Palm, G. (1989). Dynamics of neurons firing correlation: Modulation of ‘effective connectivity’. Journal of Neurophysiology 61: 900–917.Find this resource:
Ahissar, E., Vaadia, E., Ahissar, M., Bergman, H., Arieli, A., and Abeles, M. (1992). Dependence of cortical plasticity on correlated activity of single neurons and on behavioral context. Science 257: 1412–1415.Find this resource:
Albrecht, D. G. and Geisler, W. S. (1991). Motion selectivity and the contrast-response function of simple cells in the visual cortex. Visual Neuroscience 7: 531–546.Find this resource:
Alvarez, G. and Cavanagh, P. (2005). Independent resources for attentional tracking in the left and right visual hemifields. Psychological Science 16: 637–643.Find this resource:
Anderson, J., Kennedy, H., and Martin, K. (2011a). Pathways of attention: Synaptic relationships of frontal eye field to v4, lateral intraparietal cortex, and area 46 in macaque monkey. Journal of Neuroscience 31: 10872–10881.Find this resource:
Anderson, E., Mitchell, J., and Reynolds, J. (2011b). Attentional modulation of firing rate varies with burstiness across putative pyramidal neurons in macaque visual area v4. Journal of Neuroscience 31: 10983–10992.Find this resource:
Averbeck, B., Latham, P., and Pouget, A. (2006). Neural correlations, population coding and computation. Nature Reviews Neuroscience 7: 358–366.Find this resource:
Bair, W., Zohary, E., and Newsome, W. T. (2001). Correlated firing in macaque visual area MT: Time scales and relationship to behavior. Journal of Neuroscience 21: 1676–1697.Find this resource:
Berens, P., Ecker, A., Gerwinn, S., Tolias, A., and Bethge, M. (2011). Reassessing optimal neural population codes with neurometric functions. Proceedings of the National Academy of Sciences of the USA 108: 4423–4428.Find this resource:
Bisley, J. and Goldberg, M. (2003). Neuronal activity in the lateral intraparietal area and spatial attention. Science 299: 81–86.Find this resource:
Bisley, J. and Goldberg, M. (2006). Neural correlates of attention and distractibility in the lateral intraparietal area. Journal of Neurophysiology 95: 1696–1717.Find this resource:
Bisley, J. and Goldberg, M. (2010). Attention, intention, and priority in the parietal lobe. Annual Review of Neuroscience 33: 1–21.Find this resource:
Bollimunta, A., Chen, Y., Schroeder, C., and Ding, M. (2008). Neuronal mechanisms of cortical alpha oscillations in awake-behaving macaques. Journal of Neuroscience 28: 9976–9988.Find this resource:
Boudreau, C., Williford, T., and Maunsell, J. H. R. (2006). Effects of task difficulty and target likelihood in area v4 of macaque monkeys. Journal of Neurophysiology 96: 2377–2387.Find this resource:
Boynton, G. (2009). A framework for describing the effects of attention on visual responses. Vision Research 49: 1129–1143.Find this resource:
Braun, J., Koch, C., and Davis, J. L. (2001). Visual Attention and Cortical Circuits (p. 313). Cambridge, Mass.: MIT Press.Find this resource:
Buschman, T. and Miller, E. (2007). Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science 315: 1860–1862.Find this resource:
Busse, L., Wade, A. R., and Carandini, M. (2009). Representation of concurrent stimuli by population activity in visual cortex. Neuron 64: 931–942.Find this resource:
Carandini, M., Heeger, D. J., and Movshon, J. A. (1997). Linearity and normalization in simple cells of the macaque primary visual cortex. Journal of Neuroscience 17: 8621–8644. (p. 341) Find this resource:
Carandini, M. and Heeger, D. J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience 13: 51–62.Find this resource:
Carrasco, M. (2011). Visual attention: The past 25 years. Vision Research 51: 1484–1525.Find this resource:
Chelazzi, L., Duncan, J., Miller, E. K., and Desimone, R. (1998). Responses of neurons in inferotemporal cortex during memory-guided visual search. Journal of Neurophysiology 80: 2918–2940.Find this resource:
Chen, Y., Martinez-Conde, S., Macknik, S., Bereshpolova, Y., Swadlow, H., and Alonso, J. (2008). Task difficulty modulates the activity of specific neuronal populations in primary visual cortex. Nature Neuroscience 11: 974–982.Find this resource:
Cohen, M. and Newsome, W. (2008). Context-dependent changes in functional circuitry in visual area MT. Neuron 60: 162–173.Find this resource:
Cohen, M. and Maunsell, J. H. R. (2009). Attention improves performance primarily by reducing interneuronal correlations. Nature Neuroscience 12: 1594–1600.Find this resource:
Cohen, M. R. and Maunsell, J. H. R. (2010). A neuronal population measure of attention predicts behavioral performance on individual trials. Journal of Neuroscience 30: 15241–15253.Find this resource:
Cohen, M. R. and Kohn, A. (2011). Measuring and interpreting neuronal correlations. Nature Neuroscience 14: 811–819.Find this resource:
Cohen, M. R. and Maunsell, J. H. R. (2011a). Using neuronal populations to study the mechanisms underlying spatial and feature attention. Neuron 70: 1192–1204.Find this resource:
Cohen, M. R. and Maunsell, J. H. R. (2011b). When attention wanders: How uncontrolled fluctuations in attention affect performance. Journal of Neuroscience 31: 15802–15806.Find this resource:
Connor, C. E., Gallant, J. L., Preddie, D. C., and Van Essen, D. C. (1996). Responses in area v4 depend on the spatial relationship between stimulus and attention. Journal of Neurophysiology 75: 1306–1308.Find this resource:
Connor, C. E., Preddie, D. C., Gallant, J. L., and Van Essen, D. C. (1997). Spatial attention effects in macaque area v4. Journal of Neuroscience 17: 3201–3214.Find this resource:
Connors, B. and Gutnick, M. (1990). Intrinsic firing patterns of diverse neocortical neurons. Trends in Neurosciences 13: 99–104.Find this resource:
Contreras, D. and Palmer, L. A. (2003). Response to contrast of electrophysiologically defined cell classes in primary visual cortex. Journal of Neuroscience 23: 6936–6945.Find this resource:
Cook, E. P. and Maunsell, J. H. R. (2004). Attentional modulation of motion integration of individual neurons in the middle temporal visual area (MT). Journal of Neuroscience 24: 7964–7977.Find this resource:
David, S., Hayden, B., Mazer, J., and Gallant, J. (2008). Attention to stimulus features shifts spectral tuning of v4 neurons during natural vision. Neuron 59: 509–521.Find this resource:
Dicarlo, J. J. and Maunsell, J. H. R. (2000). Form representation in monkey inferotemporal cortex is virtually unaltered by free viewing. Nature Neuroscience 3: 814–821.Find this resource:
Eriksen, C. and Yeh, Y. (1985). Allocation of attention in the visual field. Journal of Experimental Psychology: Human Perception and Performance 11: 583–597.Find this resource:
Espinosa, I. and Gerstein, G. (1988). Cortical auditory neuron interactions during presentation of 3-tone sequences: Effective connectivity. Brain Research 450: 39–50.Find this resource:
Fox, M., Snyder, A., Vincent, J., and Raichle, M. (2007). Intrinsic fluctuations within cortical systems account for intertrial variability in human behavior. Neuron 56: 171–184.Find this resource:
Fries, P. (2009). Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annual Review of Neuroscience 32: 209–224.Find this resource:
Ghose, G. M. and Maunsell, J. H. R. (2002). Attentional modulation in visual cortex depends on task timing. Nature 419: 616–620. (p. 342) Find this resource:
Ghose, G. M. and Maunsell, J. H. R. (2008). Spatial summation can explain the attentional modulation of neuronal responses to multiple stimuli in area v4. Journal of Neuroscience 28: 5115–5126.Find this resource:
Ghose, G. M. (2009). Attentional modulation of visual responses by flexible input gain. Journal of Neurophysiology 101: 2089–2106.Find this resource:
Ghose, G. M. and Bearl, D. (2010). Attention directed by expectations enhances receptive fields in cortical area MT. Vision Research 50: 441–451.Find this resource:
Gilbert, C. and Sigman, M. (2007). Brain states: Top-down influences in sensory processing. Neuron 54: 677–696.Find this resource:
Gregoriou, G., Gotts, S., Zhou, H., and Desimone, R. (2009). High-frequency, long-range coupling between prefrontal and visual cortex during attention. Science 324: 1207–1210.Find this resource:
Grill-Spector, K., Knouf, N., and Kanwisher, N. (2004). The fusiform face area subserves face perception, not generic within-category identification. Nature Neuroscience 7: 555–562.Find this resource:
Gu, Y., Liu, S., Fetsch, C., et al. (2011). Perceptual learning reduces interneuronal correlations in macaque visual cortex. Neuron 71: 750–761.Find this resource:
Gutnisky, D. and Dragoi, V. (2008). Adaptive coding of visual information in neural populations. Nature 452: 220–224.Find this resource:
Haenny, P. E. and Schiller, P. H. (1988). State-dependent activity in monkey visual cortex, I: Single cell activity in v1 and v4 on visual tasks. Experimental Brain Research 69: 225–244.Find this resource:
Heeger, D. J. (1992). Normalization of cell responses in cat striate cortex. Visual Neuroscience 9: 181–197.Find this resource:
Heeger, D. J. (1993). Modeling simple-cell direction selectivity with normalized, half-squared, linear operators. Journal of Neurophysiology 70: 1885–1898.Find this resource:
Heeger, D. J., Simoncelli, E. P., and Movshon, J. A. (1996). Computational models of cortical visual processing. Proceedings of the National Academy of Science (USA) 93: 623–627.Find this resource:
Herrington, T. and Assad, J. (2009). Neural activity in the middle temporal area and lateral intraparietal area during endogenously cued shifts of attention. Journal of Neuroscience 29: 14160–14176.Find this resource:
Hussar, C. and Pasternak, T. (2009). Flexibility of sensory representations in prefrontal cortex depends on cell type. Neuron 64: 730–743.Find this resource:
Janssen, P. and Shadlen, M. N. (2005). A representation of hazard rate of elapsed time in macaque area lip. Nature Neuroscience 8: 234–241.Find this resource:
Khayat, P., Niebergall, R., and Martinez-Trujillo, J. C. (2010). Attention differentially modulates similar neuronal responses evoked by varying contrast and direction stimuli in area MT. Journal of Neuroscience 30: 2188–2197.Find this resource:
Kohn, A. and Smith, M. (2005). Stimulus dependence of neuronal correlation in primary visual cortex of the macaque. Journal of Neuroscience 25: 3661–3673.Find this resource:
Kohn, A., Zandvakili, A., and Smith, M. (2009). Correlations and brain states: From electrophysiology to functional imaging. Current Opinion in Neurobiology 19: 434–438.Find this resource:
Komiyama, T., Sato, T., O’Connor, D. H., et al. (2010). Learning-related fine-scale specificity imaged in motor cortex circuits of behaving mice. Nature 464: 1182–1186.Find this resource:
Laberge, D. (1983). Spatial extent of attention to letters and words. Journal of Experimental Psychology: Human Perception and Performance 9: 371–379.Find this resource:
Lavie, N. and Tsal, Y. (1994). Perceptual load as a major determinant of the locus of selection in visual attention. Perception & Psychophysics 56: 183–197.Find this resource:
Leber, A. (2010). Neural predictors of within-subject fluctuations in attentional control. Journal of Neuroscience 30: 11458–11465. (p. 343) Find this resource:
Lee, D. K., Itti, L., Koch, C., and Braun, J. (1999). Attention activates winner-take-all competition among visual filters. Nature Neuroscience 2: 375–381.Find this resource:
Lee, J. and Maunsell, J. H. R. (2009). A normalization model of attentional modulation of single unit responses. PLoS One 4: e4651.Find this resource:
Lee, J. and Maunsell, J. H. R. (2010a). Attentional modulation of MT neurons with single or multiple stimuli in their receptive fields. Journal of Neuroscience 30: 3058–3066.Find this resource:
Lee, J. and Maunsell, J. H. R. (2010b). The effect of attention on neuronal responses to high and low contrast stimuli. Journal of Neurophysiology 104: 960–971.Find this resource:
Li, X. and Basso, M. (2008). Preparing to move increases the sensitivity of superior colliculus neurons. Journal of Neuroscience 28: 4561–4577.Find this resource:
Liu, T. and Mance, I. (2011). Constant spread of feature-based attention across the visual field. Vision Research 51: 26–33.Find this resource:
Luck, S. J., Chelazzi, L., Hillyard, S. A., and Desimone, R. (1997). Neural mechanisms of spatial selective attention in areas v1, v2, and v4 of macaque visual cortex. Journal of Neurophysiology 77: 24–42.Find this resource:
Martinez-Trujillo, J. C. and Treue, S. (2002). Attentional modulation strength in cortical area MT depends on stimulus contrast. Neuron 35: 365–370.Find this resource:
Maunsell, J. H. R. and Cook, E. P. (2002). The role of attention in visual processing. Philosophical Transactions of the Royal Society of London B 357: 1063–1072.Find this resource:
Maunsell, J. H. R. and Treue, S. (2006). Feature-based attention in visual cortex. Trends in Neurosciences 29: 317–322.Find this resource:
McAdams, C. J. and Maunsell, J. H. R. (1999). Effects of attention on orientation-tuning functions of single neurons in macaque cortical area v4. Journal of Neuroscience 19: 431–441.Find this resource:
McCormick, D., Connors, B., Lighthall, J., and Prince, D. (1985). Comparative electrophysiology of pyramidal and sparsely spiny stellate neurons of the neocortex. Journal of Neurophysiology 54: 782–806.Find this resource:
Mitchell, J. F., Sundberg, K., and Reynolds, J. H. (2007). Differential attention-dependent response modulation across cell classes in macaque visual area v4. Neuron 55: 131–141.Find this resource:
Mitchell, J. F., Sundberg, K., and Reynolds, J. H. (2009). Spatial attention decorrelates intrinsic activity fluctuations in macaque area v4. Neuron 63: 879–888.Find this resource:
Moran, J. and Desimone, R. (1985). Selective attention gates visual processing in the extrastriate cortex. Science 229: 782–784.Find this resource:
Motter, B. C. (1993). Focal attention produces spatially selective processing in visual cortical areas v1, v2, and v4 in the presence of competing stimuli. Journal of Neurophysiology 70: 909–919.Find this resource:
Motter, B. C. (1994). Neural correlates of feature selective memory and pop-out in extrastriate area v4. Journal of Neuroscience 14: 2190–2199.Find this resource:
Navalpakkam, V. and Itti, L. (2007). Search goal tunes visual features optimally. Neuron 53: 605–617.Find this resource:
Nienborg, H. and Cumming, B. (2010). Correlations between the activity of sensory neurons and behavior: How much do they tell us about a neuron’s causality? Current Opinion in Neurobiology 20: 376–381.Find this resource:
Nienborg, H., Cohen, M., and Cumming, B. (2012). Decision-related activity in sensory neurons: Correlations among neurons and with behavior. Annual Review of Neuroscience 35: 463–483.Find this resource:
Nirenberg, S. and Latham, P. (2003). Decoding neuronal spike trains: How important are correlations? Proceedings of the National Academy of Sciences of the USA 100: 7348–7353. (p. 344) Find this resource:
O’Connor, D. H., Fukui, M. M., Pinsk, M. A., and Kastner, S. (2002). Attention modulates responses in the human lateral geniculate nucleus. Nature Neuroscience 5: 1203–1209.Find this resource:
Olson, C. R. (2001). Object-based vision and attention in primates. Current Opinion in Neurobiology 11: 171–179.Find this resource:
Parker, A. J. and Newsome, W. T. (1998). Sense and the single neuron: Probing the physiology of perception. Annual Review of Neuroscience 21: 227–277.Find this resource:
Poulet, J. and Petersen, C. (2008). Internal brain state regulates membrane potential synchrony in barrel cortex of behaving mice. Nature 454: 881–885.Find this resource:
Ress, D., Backus, B. T., and Heeger, D. J. (2000). Activity in primary visual cortex predicts performance in a visual detection task. Nature Neuroscience 3: 940–945.Find this resource:
Reynolds, J. H., Chelazzi, L., and Desimone, R. (1999). Competitive mechanisms subserve attention in macaque areas v2 and v4. Journal of Neuroscience 19: 1736–1753.Find this resource:
Reynolds, J. H., Pasternak, T., and Desimone, R. (2000). Attention increases sensitivity of v4 neurons. Neuron 26: 703–714.Find this resource:
Reynolds, J. H. and Chelazzi, L. (2004). Attentional modulation of visual processing. Annual Review of Neuroscience 27: 611–647.Find this resource:
Reynolds, J. H. and Heeger, D. (2009). The normalization model of attention. Neuron 61: 168–185.Find this resource:
Roelfsema, P. R., Lamme, V. A. F., and Spekreijse, H. (1998). Object-based attention in the primary visual cortex of the macaque monkey. Nature 395: 376–381.Find this resource:
Roelfsema, P. R., Lamme, V., and Spekreijse, H. (2004). Synchrony and covariation of firing rates in the primary visual cortex during contour grouping. Nature Neuroscience 7: 982–991.Find this resource:
Roy, A., Steinmetz, P., Hsiao, S., Johnson, K., and Niebur, E. (2007). Synchrony: A neural correlate of somatosensory attention. Journal of Neurophysiology 98: 1645–1661.Find this resource:
Saalmann, Y., Pigarev, I., and Vidyasagar, T. (2007). Neural mechanisms of visual attention: How top-down feedback highlights relevant locations. Science 316: 1612–1615.Find this resource:
Saenz, M., Buracas, G. T., and Boynton, G. M. (2002). Global effects of feature-based attention in human visual cortex. Nature Neuroscience 5: 631–632.Find this resource:
Saenz, M., Buracas, G. T., and Boynton, G. M. (2003). Global feature-based attention for motion and color. Vision Research 43: 629–637.Find this resource:
Sapir, A., D’Avossa, G., McAvoy, M., Shulman, G., and Corbetta, M. (2005). Brain signals for spatial attention predict performance in a motion discrimination task. Proceedings of the National Academy of Sciences of the USA 102: 17810–17815.Find this resource:
Schwartz, O. and Simoncelli, E. P. (2001). Natural signal statistics and sensory gain control. Nature Neuroscience 4: 819–825.Find this resource:
Scolari, M. and Serences, J. (2009). Adaptive allocation of attentional gain. Journal of Neuroscience 29: 11933–11942.Find this resource:
Scolari, M., Byers, A., and Serences, J. (2012). Optimal deployment of attentional gain during fine discriminations. Journal of Neuroscience 32: 7723–7733.Find this resource:
Shadlen, M. N. and Newsome, W. T. (1996). Motion perception: Seeing and deciding. Proceedings of the National Academy of Sciences USA 93: 628–633.Find this resource:
Shadlen, M. N. and Newsome, W. T. (1998). The variable discharge of cortical neurons: Implications for connectivity, computation, and information coding. Journal of Neuroscience 18: 3870–3896.Find this resource:
Smith, M. and Kohn, A. (2008). Spatial and temporal scales of neuronal correlation in primary visual cortex. Journal of Neuroscience 28: 12591–12603.Find this resource:
Spitzer, H., Desimone, R., and Moran, J. (1988). Increased attention enhances both behavioral and neuronal performance. Science 240: 338–340. (p. 345) Find this resource:
Spitzer, H. and Richmond, B. J. (1991). Task difficulty: Ignoring, attending to, and discriminating a visual stimulus yield progressively more activity in inferior temporal neurons. Experimental Brain Research 83: 340–348.Find this resource:
Steinmetz, P. N., Roy, A., Fitzgerald, P. J., Hsiao, S. S., Johnson, K. O., and Niebur, E. (2000). Attention modulates synchronized firing in primate somatosensory cortex. Nature 404: 187–190.Find this resource:
Thiele, A., Pooresmaeili, A., Delicato, L., Herrero, J., and Roelfsema, P. (2009). Additive effects of attention and stimulus contrast in primary visual cortex. Cerebral Cortex 19: 2970–2981.Find this resource:
Thut, G., Nietzel, A., Brandt, S., and Pascual-Leone, A. (2006). Alpha-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection. Journal of Neuroscience 26: 9494–9502.Find this resource:
Tolhurst, D. J., Movshon, J. A., and Dean, A. F. (1983). The statistical reliability of signals in single neurons in cat and monkey visual cortex. Vision Research 23: 775–785.Find this resource:
Treue, S. and Maunsell, J. H. R. (1996). Attentional modulation of visual motion processing in cortical areas MT and MST. Nature 382: 539–541.Find this resource:
Treue, S. and Martinez-Trujillo, J. C. (1999a). Reshaping neuronal representations of visual scenes through attention. Current Psychology of Cognition 18: 951–972.Find this resource:
Treue, S. and Martinez-Trujillo, J. C. (1999b). Feature-based attention influences motion processing gain in macaque visual cortex. Nature 399: 575–579.Find this resource:
Urbach, D. and Spitzer, H. (1995). Attentional effort modulated by task difficulty. Vision Research 35: 2169–2177.Find this resource:
Vaadia, E., Haalman, I., Abeles, M., et al. (1995). Dynamics of neuronal interactions in monkey cortex in relation to behavioural events. Nature 373: 515–518.Find this resource:
Vogels, R. and Orban, G. A. (1994). Activity of inferior temporal neurons during orientation discrimination with successively presented gratings. Journal of Neurophysiology 71: 1428–1451.Find this resource:
Williford, T. and Maunsell, J. H. R. (2006). Effects of spatial attention on contrast response functions in macaque area v4. Journal of Neurophysiology 96: 40–54.Find this resource:
Womelsdorf, T., Anton-Erxleben, K., Pieper, F., and Treue, S. (2006). Dynamic shifts of visual receptive fields in cortical area MT by spatial attention. Nature Neuroscience 9: 1156–1160.Find this resource:
Womelsdorf, T. and Fries, P. (2007). The role of neuronal synchronization in selective attention. Current Opinion in Neurobiology 17: 154–160.Find this resource:
Zhou, H. and Desimone, R. (2011). Feature-based attention in the frontal eye field and area v4 during visual search. Neuron 70: 1205–1217.Find this resource:
Zohary, E., Celebrini, S., Britten, K. H., and Newsome, W. T. (1994). Neuronal plasticity that underlies improvement in perceptual performance. Science 263: 1289–1292.Find this resource: