Neural Mechanisms for the Executive Control of Attention
Abstract and Keywords
The prefrontal cortex is a source of internal control of attention as it captures three important components of an executive controller. First, it provides top-down selection of neural representations through descending projections, This top-down input may act by increasing the synchrony of local neural populations, enhancing their connectivity, and boosting the transmission of information. Second, intelligent top-down control of behaviour requires integrating diverse information. Neural representations in prefrontal cortex capture this breadth of information: representing anything from the specific contents of working memory to abstract categories and rules. Third, through reciprocal connections with the basal ganglia, prefrontal cortex neurons are ideally situated to learn the ‘rules’ of behaviour that allow us to know what to attend to in a given situation. These connections may support an iterative, bootstrapping, process that allows for increasingly complex rules to be learned. The prefrontal cortex acts as a generalized executive controller, acting through mechanisms such as attention, to guide thoughts and behaviour.
Top-down, or ‘executive’ control is the ability to use previously acquired internal information to select a (typically unseen) goal, plan corresponding actions, and then keep thought and action ‘on task’ while achieving it. This is the core of intelligent, rational, behaviour—a brain that doesn’t just react to the world, but acquires information and uses it to act on the world in order to obtain future objectives.
To do so, the brain needs to deal with its limited capacity: The mental sketchpad where all this planning and organization occurs (known as ‘working memory’) is severely limited in its capacity. While we can store apparently unlimited amounts of items in a latent form (i.e. in long-term memory and as habits), many studies have shown that the neural mechanisms employed when we consciously ‘think’ can only hold a few (three to four) items simultaneously. The central challenge of executive control, then, is how finite cognitive resources are brought to bear on the information (sensory inputs, stored memories, action plans, strategies, etc.) that is currently important for the goal at hand and how potential distractions are excluded. When this is applied to the external world, we call it attention.
The distractions are often inherently salient stimuli (such as a police siren or a looming object). Our brains reflexively orient to such strong, ‘bottom-up’ sensory inputs because they often signal events that need an immediate response (like ducking). By contrast, the type of attention that is synonymous with executive control is called ‘top-down’ because it is based on acquired knowledge. We choose to pay attention to (p. 778) something (like a lecture or a clock on the wall) because we have learned it is important for achieving a goal.
This is the subject of this chapter. We review evidence that top-down attention signals originate from a brain region thought to be central to executive control, the prefrontal cortex. We will discuss candidate neural mechanisms that may mediate focusing of attention. Then, we will broaden the discussion to the neural mechanisms that allow the brain to learn what is important and worth attending to. But first, we begin with a discussion of why attention is needed in the first place: the limitation in cognitive capacity.
Selective Attention Compensates for Limited Cognitive Capacity
The finite resources of cognition have been well known since the classic George Miller paper describing the capacity of working memory as the ‘magic number’ of seven plus or minus two (Miller 1956). More recent work, using stimuli that cannot be easily combined or ‘chunked’, has lowered the magic number to three to four for the average adult human (Cowan 2001; similar to the average monkey, as we will see below). In a typical task, subjects are asked to hold a varying number of visual stimuli ‘in mind’ for a brief period and then report the contents (Luck and Vogel 1997). When the number of stimuli increase to the point that subjects make errors (they begin to lose one or more of the stimuli), their capacity has been exceeded. The exact capacity of a person varies from individual to individual; some can remember only 1–2 items and others can remember up to 7 (Vogel and Machizawa 2004; Vogel et al. 2005). An individual’s capacity is highly correlated with measures of fluid intelligence, reflecting the fact that these capacity limits are a fundamental restriction in high-level cognition (Engle et al. 1999; Fukuda et al. 2010). This makes sense: the more thoughts that can be simultaneously held ‘in mind’ and manipulated, the more associations, connections, and relationships can be made, and the more sophisticated thought can be.
Not only is attention the ‘gate-keeper’ to these finite resources, it may itself be the bottleneck. This was nicely illustrated by Edward Vogel and colleagues. They showed that much of the variability in cognitive capacity among individuals reflects differences in how well they can filter out distracting information (Vogel et al. 2005). Human subjects briefly saw two arrays, up to eight coloured squares, separated by a memory delay (Fig. 27.1a). On half of the trials one of the squares changed colour from the first array to the second. The subjects’ task was to report whether a change occurred. In order to test the impact of attention on the variance in capacity, the authors directed subjects to attend to one half of the display, indicating the colour change was more likely to occur in that half. Using electrophysiological and behavioural measures, Vogel and colleagues were able to determine the amount of information subjects were able to retain (Fig. 27.1b). They found that subjects with low capacities were ineffective in excluding irrelevant (p. 779) (p. 780) information that they were directed to ignore. In contrast, high-capacity individuals were not capable of storing more total information; they were just more effective in filtering out distracting information (Fig. 27.1c). In other words, it is not the capacity of individuals that varies but their ability to use attention to control the contents of working memory.
The close relationship between cognitive capacity and attentional capacity has been underscored by other studies as well. For example, adult humans can typically attend to, and track, about four moving objects at a time, a task that has no explicit working memory requirement but rather requires dividing attention (Pylyshyn and Storm 1988; Cavanagh and Alvarez 2005; Drew and Vogel 2008). This capacity is also correlated with their fluid intelligence (Oksama and Hyana 2004). In fact, even when items are absent (and therefore in working memory) there is an intimate relationship between attention and the ability to maintain those memories (Awh and Jonides 2001; Postle et al. 2004). For example, engaging spatial working memory improves visual recognition at remembered locations, similar to directing attention to items when they are actually present. When attention is shifted to a second, parallel, task, spatial memory in the primary task is compromised (Awh et al. 1998). The reverse is also true: holding items in working memory compromises attention. For example, Woodman and Luck asked humans to perform a visual search task (requiring directing attention) while concurrently performing a visual-spatial working memory task and found that increasing a person’s working memory load slowed their attentional search speed (Woodman and Luck 2004). Such interactions between working memory and attention suggest a singular capacity is tapped by both behaviours.
An attention bottleneck is also suggested by neurophysiological studies in monkeys that show that when capacity is exceeded, the information loss is in initial encoding of stimuli, not their subsequent retention. Buschman et al. (2011) trained monkeys on a task similar to that used by Vogel and colleagues (see above). Two arrays of two to five coloured squares were separated by a brief memory delay. In this case, one of the squares always changed colour and monkeys had to indicate which one by making an eye movement to it. When capacity was exceeded (as with humans, this occurred around four items) the loss of neural information about the squares was apparent right from the beginning of neural activation to the stimuli. That is, information was lost as the stimuli were being attended and encoded, well before the memory delay. Buschman et al. also found that the monkeys’ apparent overall capacity of four items was actually composed of two separate capacities of two and two in the right and left halves of visual space that were independent of each other. In other words, the ability to encode and retain an item on the right half of space, for example, was unaffected by items in the left half of space, regardless of how many items there were on the left (and vice versa). But adding even one more item on the same (right) side impaired performance. This suggests that the right and left cerebral hemispheres can independently process the two halves of visual space and have independent finite resources. The apparent split between the two hemispheres recalls some of the initial observations of humans who had their cerebral hemispheres split to control epilepsy. Without careful testing these subjects usually appear normal. Thus, there may be something of a split even in the intact brain. (p. 781)
Hemifield independence further underscores the close relationship between attentional capacity and general cognitive capacity. Neurophysiological studies, for example, have demonstrated that in visual cortex, attentional filtering is strongest (indeed, often only apparent) when the attended object or location is on the same half of visual space as the to-be-filtered item (Desimone and Duncan 1995). Only at the level of prefrontal cortex do attentional effects seem to bridge the vertical visual meridian (Everling et al. 2002) but, even then, there is a strong bias toward contralateral representation (Rainer et al. 1998a). Finally, the strongest evidence for human hemifield independence comes from divided-attention tasks like multiple object tracking, where attention can be split between the visual hemifields (Alvarez and Cavanagh 2005; Cavanagh and Alvarez 2005).
What all this suggests is that cognitive capacity limitations are not due to limitations in working memory per se. Rather, it reflects a fundamental limitation in the number of separate items that can be represented simultaneously in neural activity, particularly in an active state that is accessible to high-level cognition. But why exactly is this format of neural coding so limited in capacity? One hint comes from the mounting evidence that information encoding during cognitively demanding tasks may depend on the temporal dynamics between neurons, a format that has a natural limitation in bandwidth. It has long been known that the electrical potentials that reflect the summed activity of many neurons show a wide range of rhythmic synchronized oscillations (from 1 to 100 Hz), often called ‘brain waves’. A number of brain areas in monkeys, humans, and rodents show increases in oscillations during cognitively demanding tasks (Tallon-Baudry et al. 1998; Lee et al. 2005; Jensen et al. 2007; Pesaran et al. 2008). Information may be encoded by alignment of spikes from individual neurons to specific phases of neuronal population oscillations (O’Keefe 1993; Hopfield and Herz 1995; Konig et al. 1995; Laurent 2002; Mehta et al. 2002; Brody et al. 2003; Lee et al. 2005; Fries et al. 2007; Montemurro et al. 2008; Kayser et al. 2009). For example, in the rat hippocampus, spatial information may be encoded at specific phases of ongoing population theta-oscillations (O’Keefe 1993; Mehta et al. 2002; Dragoi and Buzsaki 2006). This has led to the hypothesis that multiple items are simultaneously held ‘in mind’ by multiplexing them at different phases of population oscillations (Lisman and Idiart 1995; Jensen and Lisman 2005). In other words, the mechanisms for conscious thought ‘juggle’ separate items by oscillating them out of phase of one another. This has an inherent capacity limitation because, presumably, only so much information can fit within an oscillatory cycle (that is, only a few ‘balls’ can be juggled at once). Evidence for this multiplexing was recently reported by Siegel et al. (2009): when monkeys held multiple objects in working memory, prefrontal neurons encoded information about each object at different phases of an ongoing, ~32 Hz, oscillation. More work is needed, but all this suggests that making thoughts conscious may depend on generation of oscillatory rhythms and the precise temporal relationships between them and the spiking of neurons representing the conscious thoughts. In short, attention may gate access to brain waves.
The central problem of attention, then, is the same as that of executive control in general: How does the brain’s executive mechanism figure out what information is (p. 782) important enough to be in this active state? We will consider that later in this review. But first we have to find out where the executive may be.
Finding the ‘Top’ in the Top-Down Control of Attention
The term ‘executive’ implies a hierarchy with a certain brain area or areas providing top-down control to other less-executive brain areas. A good place to look for this is the frontal-parietal network, which is thought to play a central role in attention (Corbetta et al. 1993, 1995, 1998; Coull and Nobre 1998; Donner et al. 2000, 2002; Nobre et al. 2002). In humans, there are regions in the parietal cortex (specifically within the intraparietal sulcus) and prefrontal cortex (including the human analogue of the monkey frontal eye fields in the precentral sulcus) that show increased blood flow when attention is shifted or focused (Corbetta et al. 1993; Coull and Nobre 1998; Corbetta and Shulman 2002; Liu et al. 2003). Damage to these brain regions can cause deficits in attention (Eglin et al. 1991; Knight et al. 1995; Knight 1997).
In monkeys, both top-down and bottom-up shifts of attention modulate neural activity in the parietal (particularly in the lateral intraparietal area, or LIP) and frontal cortex (both the lateral prefrontal cortex, LPFC and the frontal eye fields, FEF; Bichot and Schall 1999; Hasegawa et al. 2000; Bisley and Goldberg 2003; Buschman and Miller 2007; Johnston and Everling 2009; Moore et al. 2009). Neurons in these areas respond more vigorously to visual stimuli that are task relevant (and therefore must be attended) and show little or no activity to irrelevant stimuli (that must be ignored; Rainer et al. 1998b; Everling et al. 2006). A highly influential theory, Desimone and Duncan’s Biased Competition Model, suggests that this occurs because neural representations of different stimuli compete for activation by inhibiting one another (Desimone and Duncan 1995). Top-down or bottom-up signals add extra neural energy to the to-be-attended stimuli. This tips the balance of the competition and results in the attended representations winning the competition.
But which is on top, the prefrontal or parietal cortex? Evidence suggests that the parietal cortex is more involved in the bottom-up capture of attention whereas the prefrontal cortex is more central to top-down executive control of attention. LIP seems to have a saliency map—a topographic map of the visual field where each position is weighted by the saliency, or noticeability, of a stimulus at that location. Neurons in LIP reflect the attentional priority of stimuli in their receptive field (Bisley and Goldberg 2003) and are known to respond transiently to flashed stimuli, which can automatically draw attention (Bisley and Goldberg 2006). Further, LIP neurons reflect the target location of a pop-out stimulus very quickly, at about 80 ms after the onset of the stimulus array (Ipata et al. 2006). By contrast, PFC neurons carry more information about a task-relevant stimulus than a salient, but irrelevant, visual stimulus (Hasegawa et al. 2000). Inactivating the (p. 783) lateral PFC with muscimol disrupts top-down visual search tasks, but not simple detection tasks (Iba and Sawaguchi 2003).
Many of the neurophysiological studies of PFC and LIP have employed single electrodes to study single neurons in each area alone. This approach is valuable for learning the properties of each but it makes it difficult to determine the relative roles of different areas. Thus, while both top-down and bottom-up attention shifts are reflected in both frontal and parietal cortex, it is possible that the neural effects seen in one or the other may have been computed in one and simply inherited by the other. To help sort this out, Buschman and Miller recorded from multiple electrodes simultaneously in PFC, FEF, and LIP of monkeys (Buschman and Miller 2007) while they performed a visual search task. This allowed for precise comparison of the timing of neural effects in each area. The logic is straightforward: An area with a shorter latency to show an effect (e.g. reflect a top-down shift of attention) is more likely to be the source of that signal than one with a longer latency.
In visual search, subjects search a visual field for a particular target stimulus (see Fig. 27.2; Treisman and Gelade 1980; Duncan and Humphreys 1989; Wolfe et al. 1989). When distractors (non-target stimuli) all differ from the target in a single dimension (Fig. 27.2a, top row) the target will stand out, or ‘pop-out’, from distracting stimuli, capturing attention in an automatic bottom-up manner. By contrast, when the distractors differ from the target in more than one dimension, and do so independently from one another (Fig. 27.2a, bottom row), the target no longer automatically grabs the subject’s attention based on its inherent qualities. Instead, it must be selected by the subject’s knowledge of which stimulus is the target (by top-down mechanisms). This results in an overall slower search speed and, generally, the time to find the target is a function of the number of total items in the search array (Treisman and Gelade 1980). We trained monkeys to alternate between the easy ‘pop-out’ condition and the difficult ‘search’ condition (Fig. 27.2). As has been found in humans, we found search time to be longer and more variable in search than pop-out (see also Iba and Sawaguchi 2003). Simultaneous neural recordings revealed that when attention was automatically captured by a salient stimulus in the bottom-up, pop-out, condition, the shift of attention appeared with a shorter latency in LIP than in the LPFC and FEF (Fig. 27.2b). This suggests the bottom-up attention signals flowed anteriorly in the brain from the parietal to the frontal cortex. By contrast, we found the opposite pattern of latencies for the top-down condition, when monkeys had to find the target based on their knowledge, rather than its salience. In that case, neural signals reflecting the shift of attention to the target appeared with a shorter latency in the frontal cortex (both LPFC and FEF) than LIP (Buschman and Miller 2007; Fig. 27.2b). Taken together, these results suggest that when attention is captured by external stimuli, selection of the target is ‘bottom-up’: fed forward from LIP (possibly as part of a saliency map) to frontal cortex. In contrast, during the internal direction of attention, the signals flow in the opposite direction: ‘top-down’ from the frontal lobe and fed back to the parietal cortex. Similar results have recently been found in humans (Li et al. 2010).
Other evidence for a frontal source of top-down signals comes from the work of Tirin Moore and colleagues (see Clark, Noudoost, Schafer, and Moore (chapter 13), (p. 784) this volume). Their work focuses on FEF, a portion of the posterior PFC that seems to be responsible for voluntary eye movements. They found that stimulating FEF neurons at sub-threshold levels (i.e. levels that would not elicit an eye movement) induces attention-like effects in V4 neurons: responses were increased for V4 neurons with receptive fields that overlap with the stimulated FEF neurons (Moore and Armstrong (p. 785) 2003). It was as if the stimulation elicited a top-down attention that acted on visual cortex. Further, microstimulation in FEF will also boost the animal’s behavioural discriminability at the target location, suggesting the allocation of attentional resources to that location (Moore and Fallah 2001, 2004). These results suggest that FEF, a PFC subarea known to play a role in generating volitional eye movements, plays a direct role in top-down attention.
Neural Synchrony and Attention
We previously discussed a possible role for oscillatory activity in holding multiple items in mind and how this may explain why cognition is capacity-limited. Oscillations seem to be involved in attention as well. Above, we mentioned how attentional selection occurs when to-be-attended stimuli gain a competitive edge over other representations (Desimone and Duncan 1995). This could occur by simply raising the level of overall activity of neurons. But there is mounting evidence that another way to boost neural representations is by synchronizing neural activity. Oscillations are a good way to do that.
Synchrony has been proposed to boost neural representations because spikes arriving simultaneously at downstream neurons have a greater impact than unsynchronized spikes (Aertsen et al. 1989b; Usrey and Reid 1999; Salinas and Sejnowski 2001; Fries 2005). Coincidence of spikes from multiple neurons converging on a post-synaptic neuron has a super-additive effect (Aertsen et al. 1989a; Usrey and Reid 1999; Engel et al. 2001; Salinas and Sejnowski 2001; Fries 2005). Therefore, if sensory neurons tuned to the same stimulus synchronize their firing, that stimulus will be more strongly represented in downstream areas (as its impact on those neurons is enhanced). In this fashion, local synchrony may help the brain to improve its signal-to-noise ratio while, at the same time, reducing the number of spikes needed to represent a stimulus (Aertsen et al. 1989a; Tiesinga et al. 2002; Siegel and Konig 2003). This mechanism seems ideal for focal attention which involves enhancing some stimulus representations at the expense of others.
Support for this idea comes from observations that attention correlates with increased spiking synchrony in visual cortex (Fries et al. 2001; Womelsdorf et al. 2006) and somatosensory cortex (Steinmetz et al. 2000; Bauer et al. 2006). For example, when a monkey’s attention was directed to a particular visual stimulus, neurons in area V4 with receptive fields encompassing the attended stimulus showed increased synchrony in the gamma band (30–90 Hz) and a reduction in low frequency (<17 Hz) synchronization (Fries et al. 2001). Synchrony can also enhance neural processing by putting the brain and the external world in lockstep. Lakatos et al. (2008) presented monkeys with a stream of sequential visual and auditory stimuli and found that when monkeys attended to the visual or auditory stream, LFPs and spikes in visual cortex synchronized to the rhythm of the attended stream and not to the rhythm of an unattended stream. (p. 786)
Synchrony between regions may also regulate and sculpt communication between brain areas, helping top-down signals find their intended neural representations. If two brain areas oscillate in phase they are more likely to influence one another. If they are out of phase, they are less likely to influence each other. This has led to the suggestion that inter-areal synchrony could be used to flexibly change the effective connection between regions (Bressler 1996; Engel et al. 2001; Salinas and Sejnowski 2001; Fries 2005). Support for this notion comes from observations that inter-areal oscillatory coherence between ‘cognitive’ regions (such as LIP or FEF) and sensory areas (such as MT or VT) has been found to increase with attention (Saalmann et al. 2007; Siegel et al. 2008; Gregoriou et al. 2009).
Another role of oscillations in top-down attention comes from Buschman and Miller (Buschman and Miller 2007, 2009), who found evidence for a role in controlling when top-down attention is shifted. During top-down visual search, there was a greater increase in ~25 Hz oscillations in the frontal-parietal network relative to bottom-up pop-out. These oscillations corresponded well with behavioural and neural observations that the locus of attention shifted about every 40 ms (40 ms = 25 Hz) during the search. This suggests a relationship between the oscillations and shifts in attention. One hypothesis is that each period of the oscillation encapsulates a shift in attention. To test this, a decoding approach was used to determine how well the locus of attention could be determined in the spiking activity of neurons in the FEF. Decoding was best when the analysis window was centred on and synchronized to each 25 Hz oscillatory cycle on each trial (relative to using time windows fixed to an external event, like the behavioural response). This supports our hypothesis: shifts of attention during search were synchronized to the 25 Hz LFP oscillations (Buschman and Miller 2009). These oscillations could have been extrinsically or intrinsically generated: either reflecting a mechanism specifically generated to regulate the timing of attentional shifts, or, alternatively, it could have been generated by the process of serially attending to different locations in a rhythmic manner. Either way, these oscillations were synchronized across frontal and parietal cortex (Buschman and Miller 2007), and would be ideal for providing a ‘timing’ or ‘clocking’ signal that helps coordinate shifts of attention across different brain areas (Buschman and Miller 2010). This is analogous to the bus clock on a computer that coordinates the timing of operations in the computer’s many different circuits.
Deciding What to Attend: Neural Mechanisms for Executive Control
Now we take up the central question we raised earlier: How does the brain determine what is important and needs attending? For bottom-up attention, selection is more straightforward: our brain has evolved to provide more neural energy to salient stimuli that are loud, looming, sudden, etc. However, navigating complex situations to achieve (p. 787) long-term goals cannot rely on uncoordinated reactions to the environment. Rather, this must be orchestrated ‘top-down’ from within oneself. And to put it simply, you can’t play this game without learning the rules.
Rules are central to our ability to coordinate thought and action and direct them toward a goal. Virtually all long-term, goal-directed behaviours are learned, and thus depend on a cognitive system that can acquire and represent elaborate representations that reflect all the information needed to achieve a goal: what outcomes are possible, what actions have been successful at achieving them or similar goals in the past, information stored in long-term memory, and, of course, what things in the environment require our attention. Consider the set of rules invoked when we dine in a restaurant, such as ‘wait to be seated’, ‘order’, and ‘pay the bill’. These rules give us an idea about what to expect and what is expected of us when we try a new restaurant (for example, paying attention to the waiter to hear the specials). Thus, rules are needed to orchestrate processing in diverse brain regions along a common, internal theme. The challenge is a model that can explain this neurobiologically, without resorting to a homunculus. Over the rest of this review, we will try to offer one.
Rules and the prefrontal cortex
The PFC seems anatomically well situated to play a role in rule learning. It receives and sends projections to most of the cerebral cortex (with the exception of primary sensory and motor cortices) as well as the hippocampus, amygdala, cerebellum, and, most importantly for our model, the basal ganglia (abbreviated as BG; Porrino et al. 1981; Amaral and Price 1984; Amaral 1986; Selemon and Goldman-Rakic 1988; Barbas and De Olmos 1990; Eblen and Graybiel 1995; Croxson et al. 2005). Thus, the PFC seems to be a hub of cortical processing, able to synthesize a wide range of external and internal information and also exert control over much of the cortex. Although different PFC subdivisions have distinct patterns of interconnections with other brain systems (e.g. lateral—sensory and motor cortex; orbital—limbic), there are prodigious connections both within and between PFC subdivisions, ensuring a high degree of integration of information (Pandya and Barnes 1987; Barbas and Pandya 1989; Pandya and Yeterian 1990; Barbas and Pandya 1991; Petrides and Pandya 1999). Such a dense network of connections could allow PFC to act as a large associative network for detecting and storing associations between diverse events, experiences, and internal states (Fig. 27.3).
There is a large amount of evidence supporting the role of the frontal cortex in rule learning and use (for reviews see Wise et al. 1996; Miller and Cohen 2001). Neurophysiological studies in animals and imaging studies in humans have shown that the PFC has many of the needed attributes. First, the neurons sustain their activity across short, multisecond memory delays (Pribram et al. 1952; Fuster and Alexander 1971; Fuster 1973; Funahashi et al. 1989; Miller et al. 1996). This ‘working memory’ property is crucial for goal-directed behaviour, which, unlike ‘ballistic’ reflexes, typically extends over time and allows associations to be formed between items that are not (p. 788) simultaneously present. Second, the PFC is highly plastic. After training, a large proportion of neurons in the monkey PFC acquire selectivity for the task contingencies (typically one third to one half of the population; White and Wise 1999; Asaad et al. 2000; Wallis et al. 2001; Mansouri et al. 2006). A computational model by Rigotti et al. (Rigotti et al. 2011) argued that the PFC needs large proportions of neurons with mixed selectivity. Random connections result in arbitrary mixes of external and internal information. These broad representations endow the PFC with a large, perhaps unlimited, capacity to learn new rules. It may also endow cognitive flexibility by allowing the PFC brain to re-utilize the same pool of neurons for different tasks. This has been supported by observations that large proportions of PFC neurons can ‘multitask’ and play a role in representing disparate, independent categorical rules (Cromer et al. 2010; Roy et al. 2010).
How the PFC may acquire rule information is considered next.
Rule learning and dopamine
Rule learning requires feedback about which behaviours have been successful. The brain must strengthen co-activations that are successful at achieving a goal (rewarded) while breaking associations that are ineffective. This needs to be guided by feedback so that (p. 789) relevant events and predictive relationships can be distinguished from spurious coincidences. This guidance appears to come in the form of a ‘reinforcement signal’, thought to be provided by DA neurons in the midbrain.
Dopaminergic neurons are located in both the ventral tegmental area and the substantia nigra, pars compacta (Schultz et al. 1992; Schultz et al. 1997; Schultz 1998), and show activity that directly corresponds to the reward prediction error signals suggested by models of animal learning. Midbrain DA neurons send heavy projections into both the frontal cortex and the striatum, the main input of the BG. The projections into the frontal cortex show a gradient connectivity with heavier inputs anteriorly that drop off posteriorly, suggesting a preferential input of reward information into the PFC relative to posterior cortex (Thierry et al. 1973; Goldman-Rakic et al. 1989). Evidence suggests that neither strengthening nor weakening of synapses in the striatum by long-term potentiation or depression can occur without DA input (Calabresi et al. 1992, 1997; Otani et al. 1998; Kerr and Wickens 2001). After training, DA neurons in the midbrain will learn to increase activity to an unexpected stimulus that directly predicts a reward: the event ‘stands in’ for the reward (Schultz et al. 1993). DA neurons will now respond to the predictive event when it is unexpected, but will no longer respond to the actual, now expected, reward event.
In short, the activity of dopaminergic neurons corresponds to a teaching signal that says, ‘Something good happened and you did not predict it, so remember what just happened so you can predict it in the future.’ Alternatively, if a reward is expected, but not received, the signal provides feedback that whatever behaviour was just taken is not effective in getting rewarded. This teaching signal is thought to guide the learning of the associations that are the building blocks of rules. If these reward signals affect connections that were recently active, and therefore likely involved in recent behaviour, then the result may be to help strengthen reward-predicting associations within the network, while reducing associations that do not increase benefits. In this way, the brain can learn what rules are effective in producing desirable outcomes.
Top-down control depends on a balance between different styles of learning
Normal learning has to balance different demands. One might expect that the greatest evolutionary benefit would be gained from learning as quickly as possible—adapting at a faster rate than competing organisms lends a definite edge, whereas missed opportunities can be costly (even deadly). However, there are also disadvantages to learning too quickly: one loses the ability to integrate across multiple experiences to form a generalized, less error-prone prediction. Take the classic example of one-trial learning: conditioned taste aversion. Many of us have had the experience of eating a particular food and then becoming ill for an unrelated reason. However, in many cases, the person develops an aversion to that food, even though the attribution is erroneous. (p. 790) Extending learning across multiple episodes allows organisms to detect the regularities of predictive relationships and leave behind spurious associations and coincidences. Further, networks that learn at a slower rate also tend to be more stable. Artificial neural networks with small changes in synaptic weights at each learning episode converge very slowly. Networks with large synaptic weight changes can quickly capture some patterns; however, the resulting networks tend to be more volatile and exhibit erratic behaviour. This is due to the fact that a high learning rate can overshoot minima in the error function, even oscillating between values on either side of the minima, but never reaching the minima (for more information on artificial neural networks, see Hertz et al. 1991; Dayan and Abbott 2001).
In addition to avoiding errors, slower, more deliberate learning also provides the opportunity to integrate associations across many different experiences to detect common structures across them. It is these commonalities that form abstractions, general principles, concepts, and symbolisms that are the medium of the sophisticated, ‘big-picture’ thought needed for truly long-term goals. Indeed, this is fundamental to proactive thought and action. Generalizing among many past experiences gives us the ability to generalize to the future, to imagine possibilities that we have not yet experienced—but would like to—and given the generalized rules, we can predict the actions and behaviours needed to achieve our goal. In addition, abstraction may aid in cognitive flexibility, because generalized representations are concise (by definition), lacking the details of more specific representations. These compressed representations should make it easier to maintain multiple generalized representations within a given network and ease switching between them, particularly in contrast to when representations are elaborate and detailed.
Given the advantages and disadvantages associated with both forms of learning, the brain must balance the obvious pressure to learn as quickly as possible with the advantages of slower learning. One possible solution to this conundrum comes from O’Reilly and colleagues, who suggested that fast learning and slow learning systems interact with one another (McClelland et al. 1995; O’Reilly and Munakata 2000). Studying the consolidation of long-term memories, McClelland et al. (McClelland et al. 1995) specifically suggested that fast plasticity mechanisms within the hippocampus are able to quickly capture new memories while ‘training’ the slower-learning cortical networks. In this way, the brain is able to balance the need to initially grasp new memories with the advantages of a generalized, distributed representation of long-term memories. The idea is that the hippocampus is specialized for the rapid acquisition of new information; each learning trial produces large weight changes. The output of the hippocampus will then repeatedly activate cortical networks that have smaller weight changes per episode. Continued hippocampal-mediated reactivation of cortical representations allows the cortex to gradually connect these representations with other experiences. That way, the shared structure across experiences can be detected and stored, and the memory can be interleaved with others so that it can be readily accessed.
In fact, this architecture (fast learning in more primitive, non-cortical structures training the slower, more advanced, cortex) may be a general brain strategy; in (p. 791) addition to being suggested for the relationship between the hippocampus and cortex, it has also been proposed for the cerebellum and cortex (Houk and Wise 1995). This makes sense: the first evolutionary pressure on our cortex-less ancestors was presumably toward faster learning, whereas only later did we add on a slower, more judicious and flexible cortex. We propose that rule learning occurs through similar interactions between the PFC and the basal ganglia, a set of subcortical structures that is anatomically and functionally related to the PFC, as discussed next.
Interactions between different styles of plasticity in the PFC and basal ganglia
The BG is a collection of subcortical nuclei that, similar to the PFC, have a high degree of cortical convergence. Cortical inputs arrive largely via the striatum (which includes both the caudate and the putamen); are processed through the globus pallidus, the subthalamic nucleus (STN), and the substantia nigra; and are then directed back into the cortex via the thalamus (see Fig. 27.3). The segregated nature of BG inputs is maintained throughout the different nuclei such that the output from the BG (via the thalamus) is largely to the same cortical areas that gave rise to the initial inputs into the BG (Selemon and Goldman-Rakic 1985). Crucially, the frontal cortex receives the largest portion of BG outputs, suggesting a close collaboration between these structures (Middleton and Strick 1994, 2000, 2002). Lesions of the striatum produce impairments in learning new operant behaviours (or concrete rules) and show that damage to different parts of the striatum generally causes deficits similar to those caused by lesions of the area of the cortex that loop with the affected region of the striatum (Divac et al. 1967; Goldman and Rosvold 1972). For example, lesions of the regions of the caudate associated with the frontal cortex result in cognitive impairments. This all suggests that reciprocal connections between the BG and PFC play a significant role in PFC (and therefore cognitive) function.
Interestingly, the input of midbrain DA into the striatum is much heavier than that of the PFC, by as much as an order of magnitude (Lynd-Balta and Haber 1994). Further, DA neurons make connections close to the synapse that striatal neurons form with cortical neurons. By contrast, DA inputs to the cortex synapse out on the dendrites. Thus, DA may play a strong role in gating plasticity in the striatum whereas DA may have a more subtle influence in cortex, in shading, not gating, plasticity. This may reflect the trade-off (discussed above) between the demands of fast plasticity (in the striatum) versus slow plasticity (in the PFC). The result of these different learning styles was suggested by a computational model by Daw et al. (2006).
Daw et al. contrasted learning in the striatum with that in the PFC. They suggested that the rules represented in the PFC are the entire logical structure of a task in a tree-like model from initial state to goal achievement. Behaviours begin in an initial state with two or more possible response alternatives. Choosing one response leads to (p. 792) another state with new response alternatives, with this process continuing throughout the task, ultimately leading to a reward. The PFC is able to capture this entire tree structure, essentially providing the animal with an internal model of the entire task. This endows the characteristics of sophisticated goal-directed behaviour. It allows prediction of long-term outcomes by chaining together short-term predictions (direct associations) into multistep long-range predictions. It also allows mental flexibility. Chaining together predictions on the fly allows the system to flexibly react to changing circumstances. Further, a change in the value of a goal propagates back through the tree-model, changing which choices might be made.
In contrast to the PFC, the BG is thought to represent acquired information with a cache, not a tree-like model system. That is, the striatum learns not the entire structure of a task; rather it learns the most valuable alternative at each decision point in isolation. It is as if the BG learns each fork in the road whereas the PFC learns the whole route. The BG cache system is computationally simple (and therefore fast) but it is inflexible because the learning is divorced from any change in the outcome. This may explain why the BG is associated with inflexible habit learning.
We suggest that these two representational styles result from differences in plasticity in the striatum vs. PFC: namely, fast, DA-gated plasticity in the striatum vs. slower plasticity in the PFC that is DA-shaded. Support for different speeds of plasticity in each comes from an experiment by Pasupathy and Miller (2005). Monkeys were trained to associate a visual cue with a directional eye movement (Fig. 27.4a). Learning occurred over a period of approximately 60 trials (Fig. 27.4b), after which the associations were reversed and the animals had to re-learn the new associations. This allowed Pasupathy and Miller to study how single neurons in the prefrontal cortex and striatum learned (and re-learned) these associations during the trial (Fig. 27.4c/d). Neural activity in the striatum showed rapid, almost bi-stable, learning-related changes in the timing of selectivity (Fig. 27.4e). This is in contrast to the PFC where changes were much slower, with selective responses slowly advancing across trials (Fig. 27.4e). These results support the hypothesis that rewarded associations are first identified by the striatum, the output of which ‘trains’ slower learning mechanisms in the PFC.
Thus, the relationship between the BG and PFC may be similar to the relationship between the hippocampus and cortex as suggested by O’Reilly (discussed above). As the animal learns specific stimulus–response associations they are quickly acquired by the basal ganglia which, in turn, slowly train the prefrontal cortex. In this case, the fast (strong weight changes) plasticity in the striatum is better suited for the rapid formation of associations between a specific cue and response. However, as noted above, fast learning tends to be error prone, and indeed, striatal neurons began predicting the forthcoming behavioural response early in learning when that response was often wrong. By contrast, the smaller weight changes in the PFC may have allowed it to accumulate more evidence and arrive more slowly and judiciously at the correct answer. As has been proposed for the hippocampus and cortex, the fast striatal plasticity may be more suited for a quick stamping-in of immediate, direct associations (a cache system). By contrast, the slow PFC plasticity may be suitable for building elaborate rule representations that (p. 793) gradually link in more information (i.e. tree-like representations). The slower PFC plasticity may also be critical for finding the commonalities and regularities among the simpler representations acquired by the striatum that are the basis for abstractions and general principles (see above).
Support for the specific vs. generalized trade-off between the striatum and the PFC during learning comes from Antzoulatos and Miller (2011), who recorded from multiple electrodes in the lateral prefrontal cortex and dorsal striatum while animals learned two categories of stimuli. Each day, monkeys learned to associate novel, abstract dot-based categories with a right vs. left saccade (Fig. 27.5a and b). Early on, when they could acquire specific stimulus–response associations, striatum activity was an earlier predictor of the corresponding saccade (Fig. 27.5c). However, as the number of exemplars was increasing, and monkeys had to form abstractions to classify them, PFC began predicting the saccade associated with each category before the striatum (Fig. 27.5d). Thus, it seems that the striatum was leading the acquisition early on when behaviour could be supported by simple stimulus–response learning. However, when the abstraction requirements exceeded that of the simple striatum cache representations, the PFC (p. 794) took over. In this case the slower-learning, associative activity of PFC is ideal for the integration of stimulus properties over many exemplars, allowing for a generalized ‘concept’ of categories to be learned. This dual-learner strategy allows the animal to perform optimally throughout the task—early on striatum can learn associations quickly while later in the task, when learning associations is no longer viable, prefrontal cortex guides behaviour.
The interactions of the PFC and the BG might explain several aspects of goal-directed learning and habit formation. The initial learning of a complex operant task invariably (p. 795) begins with the establishment of a simple response immediately proximal to reward (i.e. a single state). Then, as the task becomes increasingly complex with more and more antecedents and qualifications (states and alternatives) the PFC shows greater involvement. It facilitates this learning via its slower plasticity, allowing it to stitch together the relationships between the different states. This is useful because uncertainty of the correct action at any given state adds across the many states within a complex task. Thus, in complex tasks the ability of reinforcement to control behaviour is lessened with the addition of more states. However, model-building in the PFC may provide the overarching infrastructure—the thread weaving between states—that facilitates learning of the entire course of action. Many tasks will remain dependent on the PFC and the models it builds, especially those requiring flexibility (e.g. when the goal often changes or there are multiple goals to choose among) or when a strongly established behaviour in one of the states (e.g. a habit) is incompatible with the course of action needed to obtain a specific goal. However, if a behaviour, even a complex one, is unchanging, then the actions at each state are constant and, once these are learned, control can revert to a piecemeal caching system in the BG. That is, the behaviour becomes a ‘habit’ and it frees up the more executive PFC model-building system for behaviours requiring the flexibility it provides.
Primates, especially humans, can engage in elaborate goal-directed behaviours. Plus, we can be creative and unique in finding new goals and strategies to pursue them. This means that the mechanisms that build the PFC rule representations must have a corresponding ability for open-ended growth. We propose that the anatomical loops through the PFC and BG support this via recursive, bootstrapping interactions, as we will discuss next.
Recursive processing and bootstrapping in corticoganglia loops
‘Bootstrapping’ is the process of building increasingly complex representations from simpler ones. The recursive nature of the loops between the BG and PFC may lend itself to this process. As described earlier, anatomical connections between the PFC and BG seem to suggest a closed loop—channels within the BG return outputs, via the thalamus, into the same cortical areas that gave rise to their initial cortical input (Fig. 27.3). This anatomy seems well suited for recursive processing. That is, the neural representations that are results from PFC–BG interactions are fed back into the loop as fodder for further learning. In this manner, new experiences can be added onto previous ones, linking in more information to build more elaborate rule representations. It can allow the discovery of commonalities among more experiences and thus more high-level concepts and principles. Indeed, a hallmark of human intelligence is the propensity for us to ground new concepts in familiar ones because it seems to ease our understanding of novel ideas—we learn to multiply by serial addition, exponentiation by serial multiplication, etc. (p. 796)
The frontal cortex–BG loops suggest an auto-associative type network, similar to that seen in the CA3 of the hippocampus. The looping back of outputs on the inputs allows the network to learn to complete (i.e. recall) previously learned patterns given a degraded version or a subset of the original inputs (Hopfield 1982). In the hippocampus, this network has been suggested to play a role in the formation of memories. The PFC–BG loops are heavily influenced by dopaminergic inputs, and therefore may be more goal-oriented than hippocampal learning and memory. Indeed, the cortical–BG loops may also explain the DA reward prediction signals. As previously described, midbrain DA neurons respond to earlier and earlier events in a predictive chain leading to a reward. Both the frontal cortex and the striatum send projections into the midbrain DA neurons, possibly underlying their ability to bootstrap to early predictors of reward (however, although this is suggestive, it is still unknown whether these descending projections are critical for this behaviour).
The loops may also explain another important aspect of goal-directed behaviour: the stringing together of sequences of thought and action. A key feature of auto-associative networks is their ability to learn temporal sequences of patterns and thus make predictions. This feature relies on the activity pattern being fed back into the network with a temporal delay, allowing the next pattern in the sequence to arrive as the previous pattern is fed back, building an association (Kleinfeld 1986; Sompolinsky and Kanter 1986). Inhibitory synapses in the pathways through the BG may add the temporal delay needed as they have a slower time constant than excitatory synapses (Couve et al. 2000). A second way to add lag is through a memory buffer. The PFC is well known for this type of property; its neurons can sustain their activity to act as a bridge for learning contingencies across several seconds, even minutes. The introduction of lag into the recursive loop through either (or both) mechanism(s) may be enough to tune the network for sequencing and prediction. This would seem to be key to the development of tree-like rule representations that describe an entire sequence of goal-directed actions.
Summary: The PFC as a General Executive Controller
We have reviewed why we need attention (our very finite cognitive resources) and shown that the PFC is a major source of the top-down attention signals that select the sensory (and other) information important for our current goal. This may occur by the top-down signals boosting the representations of the to-be-attended stimuli by raising their activity and/or by synchronizing the activity of those neurons so that they have a greater impact on downstream neurons. In addition, oscillatory coherence in and between brain areas may help route traffic throughout cortex, control or signal when attention is shifted, and, by playing a role in juggling multiple active neural (p. 797) representations, could explain why we have a cognitive capacity limitation in the first place. Finally, we discussed the neural mechanisms that allow the PFC, along with the BG, to learn the rules of the game that determine what is potentially important and in need of selection. To complete the circle, we need to address one more issue: How (and why) do rule representations in the PFC result in top-down selection?
Miller and Cohen (Miller and Cohen 2001) argued that rule representations in the PFC are not arbitrary, esoteric descriptions of a task’s logical structure. Rather, the PFC represents rules in a particular format: as a map of the cortical pathways needed to perform the task (‘rulemaps’) (Fig. 27.6). In other words, the tree-like set of a task’s rules in the PFC is also a tree-like map of the neural pathways in and between other brain regions that need to be activated to engage in the current task. In a given situation, cues about context and other current external and internal information activate and complete the corresponding PFC rulemap. Activation of the rulemap (which can be sustained, if needed) sets up bias signals that feed back to other brain areas, affecting sensory systems as well as the systems responsible for response execution, memory retrieval, and emotional evaluation. The aggregate effect is the selection of neural circuits that guide the flow of neural activity along pathways that establish the proper mappings between inputs, internal states, and outputs to reach the goal. It is as if the PFC is a conductor in a railroad yard and learns a map that it uses to guide trains (neural activity) along the right tracks (neural pathways). And when these signals act on sensory systems, we call it top-down attention. (p. 798)
Supported by the Intelligence Advanced Research Projects Activity (IARPA) via Department of the Interior (DOI) contract number D10PC20023. The US Government is authorized to reproduce and distribute reprint for governmental purposes notwithstanding any copyright annotation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of IARPA, DOI, or the US Government.
Aertsen, A. M., Gerstein, G. L., Habib, M. K., and Palm, G. (1989a). Dynamics of neuronal firing correlation: Modulation of ‘effective connectivity’. Journal of Neurophysiology 61: 900–917.Find this resource:
Aertsen A. M. H. J., Gerstein, G. L., Habib, M. K., Palm, G., and Gochin, P. M. (1989b). Dynamics of neuronal firing correlation: Modulation of ‘effective connectivity’. Journal of Neurophysiology 61: 900–917.Find this resource:
Alvarez, G. A. and Cavanagh, P. (2005). Independent resources for attentional tracking in the left and right visual hemifields. Psychological Science 16: 637–643.Find this resource:
Amaral, D. G. (1986). Amygdalohippocampal and amygdalocortical projections in the primate brain. Advances in Experimental Medicine and Biology 203: 3–17.Find this resource:
Amaral D. G. and Price, J. L. (1984). Amygdalo-cortical projections in the monkey (Macaca fascicularis). Journal of Comparative Neurology 230: 465–496.Find this resource:
Antzoulatos, E. G. and Miller, E. K. (2011). Differences between neural activity in prefrontal cortex and striatum during learning of novel abstract categories. Neuron 71: 243–249.Find this resource:
Asaad, W. F., Rainer, G., and Miller, E. K. (2000). Task-specific activity in the primate prefrontal cortex. Journal of Neurophysiology 84: 451–459.Find this resource:
Awh, E. and Jonides, J. (2001). Overlapping mechanisms of attention and spatial working memory. Trends in Cognitive Sciences 5: 119–126.Find this resource:
Awh, E., Jonides, J., and Reuter-Lorenz, P. A. (1998). Rehearsal in spatial working memory. Journal of Experimental Psychology: Human Perception and Performance 24: 780–790.Find this resource:
Barbas, H. and De Olmos, J. (1990). Projections from the amygdala to basoventral and mediodorsal prefrontal regions in the rhesus monkey. Journal of Comparative Neurology 300: 549–571.Find this resource:
Barbas, H. and Pandya, D. N. (1989). Architecture and intrinsic connections of the prefrontal cortex in the rhesus monkey. Journal of Comparative Neurology 286: 353–375.Find this resource:
Barbas, H. and Pandya, D. N. (1991). Patterns of connections of the prefrontal cortex in the rhesus monkey associated with cortical architecture. In H. S. Levin, H. M. Eisenberg, and A. L. Benton (eds.), Frontal Lobe Function and Dysfunction (pp. 35–58). New York: Oxford University Press.Find this resource:
Bauer, M., Oostenveld, R., Peeters, M., and Fries, P. (2006). Tactile spatial attention enhances gamma-band activity in somatosensory cortex and reduces low-frequency activity in parieto-occipital areas. Journal of Neuroscience 26: 490–501. (p. 799) Find this resource:
Bichot, N. P. and Schall, J. D. (1999). Effects of similarity and history on neural mechanisms of visual selection. Nature Neuroscience 2: 549–554.Find this resource:
Bisley, J. W. and Goldberg, M. E. (2003). Neuronal activity in the lateral intraparietal area and spatial attention. Science 299: 81–86.Find this resource:
Bisley, J. W. and Goldberg, M. E. (2006). Neural correlates of attention and distractibility in the lateral intraparietal area. Journal of Neurophysiology 95: 1696–1717.Find this resource:
Bressler, S. L. (1996). Interareal synchronization in the visual cortex. Behavioural Brain Research 76: 37–49.Find this resource:
Brody, C. D., Hernandez, A., Zainos, A., and Romo, R. (2003). Timing and neural encoding of somatosensory parametric working memory in macaque prefrontal cortex. Cerebral Cortex 13: 1196–1207.Find this resource:
Buschman, T. J. and Miller, E. K. (2007). Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science 315: 1860–1862.Find this resource:
Buschman, T. J. and Miller, E. K. (2009). Serial, covert shifts of attention during visual search are reflected by the frontal eye fields and correlated with population oscillations. Neuron 63: 386–396.Find this resource:
Buschman, T. J. and Miller, E. K. (2010). Shifting the spotlight of attention: Evidence for discrete computations in cognition. Frontiers in Human Neuroscience 4: Article 194.Find this resource:
Buschman, T. J., Siegel, M., Roy, J. E., and Miller, E. K. (2011). Neural substrates of cognitive capacity limitations. Proceedings of the National Academy of Sciences USA 108: 11252–11255.Find this resource:
Calabresi, P., Maj, R., Pisani, A., Mercuri, N. B., and Bernardi, G. (1992). Long-term synaptic depression in the striatum: Physiological and pharmacological characterization. Journal of Neuroscience 12: 4224–4233.Find this resource:
Calabresi, P., Saiardi, A., Pisani, A., Baik, J. H., Centonze, D., Mercuri, N. B., Bernardi, G., Borrelli, E., and Maj, R. (1997). Abnormal synaptic plasticity in the striatum of mice lacking dopamine D2 receptors. Journal of Neuroscience 17: 4536–4544.Find this resource:
Cavanagh, P. and Alvarez, G. A. (2005). Tracking multiple targets with multifocal attention. Trends in Cognitive Sciences 9: 349–354.Find this resource:
Corbetta, M., Akbudak, E., Conturo, T. E., Snyder, A. Z., Ollinger, J. M., Drury, H. A., Linenweber, M. R., Petersen, S. E., Raichle, M. E., Van Essen, D. C., and Shulman, G. L. (1998). A common network of functional areas for attention and eye movements. Neuron 21: 761–773.Find this resource:
Corbetta, M., Miezin, F. M., Shulman, G. L., and Petersen, S. E. (1993). A PET study of visuospatial attention. Journal of Neuroscience 13: 1202–1226.Find this resource:
Corbetta, M. and Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. Nature Reviews Neuroscience 3: 201–215.Find this resource:
Corbetta, M., Shulman, G. L., Miezin, F. M., and Petersen, S. E. (1995). Superior parietal cortex activation during spatial attention shifts and visual feature conjunction. Science 270: 802–805.Find this resource:
Coull, J. T. and Nobre, A. C. (1998). Where and when to pay attention: The neural systems for directing attention to spatial locations and to time intervals as revealed by both PET and fMRI. Journal of Neuroscience 18: 7426–7435.Find this resource:
Couve, A., Moss, S. J., and Pangalos, M. N. (2000). GABAB receptors: A new paradigm in G protein signaling. Molecular and Cellular Neuroscience 16: 296–312.Find this resource:
Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences 24: 87–114; discussion 114–185.Find this resource:
Cromer, J. A., Roy, J. E., and Miller, E. K. (2010). Representation of multiple, independent categories in the primate prefrontal cortex. Neuron 66: 796–807. (p. 800) Find this resource:
Croxson, P. L., Johansen-Berg, H., Behrens, T. E. J., Robson, M. D., Pinsk, M. A., Gross, C. G., Richter, W., Richter, M. C., Kastner, S., and Rushworth, M. F. S. (2005). Quantitative investigation of connections of the prefrontal cortex in the human and macaque using probabilistic diffusion tractography. Journal of Neuroscience 25: 8854–8866.Find this resource:
Daw, N. D., O’Doherty, J. P., Dayan, P., Seymour, B., and Dolan, R. J. (2006). Cortical substrates for exploratory decisions in humans. Nature 441: 876–879.Find this resource:
Dayan, P. and Abbott, L. F. (2001). Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems (Computational Neuroscience). Cambridge, Mass.: MIT Press.Find this resource:
Desimone, R. and Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual Review of Neuroscience 18: 193–222.Find this resource:
Divac, I., Rosvold, H. E., and Szwarcbart, M. K. (1967). Behavioral effects of selective ablation of the caudate nucleus. Journal of Comparative and Physiological Psychology 63: 184–190.Find this resource:
Donner, T., Kettermann, A., Diesch, E., Ostendorf, F., Villringer, A., and Brandt, S. A. (2000). Involvement of the human frontal eye field and multiple parietal areas in covert visual selection during conjunction search. European Journal of Neuroscience 12: 3407–3414.Find this resource:
Donner, T., Kettermann, A., Diesch, E., Ostendorf, F., Villringer, A., and Brandt, S. A. (2002). Visual feature and conjunction searches of equal difficulty engage only partially overlapping frontoparietal networks. NeuroImage 15: 16–25.Find this resource:
Dragoi, G. and Buzsaki, G. (2006). Temporal encoding of place sequences by hippocampal cell assemblies. Neuron 50: 145–157.Find this resource:
Drew, T. and Vogel, E. K. (2008). Neural measures of individual differences in selecting and tracking multiple moving objects. Journal of Neuroscience 28: 4183–4191.Find this resource:
Duncan, J. and Humphreys, G. W. (1989). Visual search and stimulus similarity. Psychological Review 96: 433–458.Find this resource:
Eblen, F. and Graybiel, A. (1995). Highly restricted origin of prefrontal cortical inputs to striosomes in the macaque monkey. Journal of Neuroscience 15: 5999–6013.Find this resource:
Eglin, M., Robertson, L. C., and Knight, R. T. (1991). Cortical substrates supporting visual search in humans. Cerebral Cortex 1: 262–272.Find this resource:
Engel, A. K., Fries, P., and Singer, W. (2001). Dynamic predictions: Oscillations and synchrony in top-down processing. Nature Reviews Neuroscience 2: 704–716.Find this resource:
Engle, R. W., Tuholski, S. W., Laughlin, J. E., and Conway, A. R. (1999). Working memory, short-term memory, and general fluid intelligence: A latent-variable approach. Journal of Experimental Psychology: General 128: 309–331.Find this resource:
Everling, S., Tinsley, C. J., Gaffan, D., and Duncan, J. (2002). Filtering of neural signals by focused attention in the monkey prefrontal cortex. Nature Neuroscience 5: 671–676.Find this resource:
Everling, S., Tinsley, C. J., Gaffan, D., and Duncan, J. (2006). Selective representation of task-relevant objects and locations in the monkey prefrontal cortex. European Journal of Neuroscience 23: 2197–2214.Find this resource:
Fries, P. (2005). A mechanism for cognitive dynamics: Neuronal communication through neuronal coherence. Trends in Cognitive Sciences 9: 474–480.Find this resource:
Fries, P., Nikolic, D., and Singer, W. (2007). The gamma cycle. Trends in Neurosciences 30: 309–316.Find this resource:
Fries, P., Reynolds, J. H., Rorie, A. E., and Desimone, R. (2001). Modulation of oscillatory neuronal synchronization by selective visual attention. Science 291: 1560–1563.Find this resource:
Fukuda, K., Vogel, E., Mayr, U., and Awh, E. (2010). Quantity, not quality: The relationship between fluid intelligence and working memory capacity. Psychonomic Bulletin & Review 17: 673–679. (p. 801) Find this resource:
Funahashi, S., Bruce, C. J., and Goldman-Rakic, P. S. (1989). Mnemonic coding of visual space in the monkey’s dorsolateral prefrontal cortex. Journal of Neurophysiology 61: 331–349.Find this resource:
Fuster, J. M. (1973). Unit activity in prefrontal cortex during delayed-response performance: Neuronal correlates of transient memory. Journal of Neurophysiology 36: 61–78.Find this resource:
Fuster, J. M. and Alexander, G. E. (1971). Neuron activity related to short-term memory. Science 173: 652–654.Find this resource:
Goldman, P. S. and Rosvold, H. E. (1972). The effects of selective caudate lesions in infant and juvenile Rhesus monkeys. Brain Research 43: 53–66.Find this resource:
Goldman-Rakic, P. S., Leranth, C., Williams, S. M., Mons, N., and Geffard, M. (1989). Dopamine synaptic complex with pyramidal neurons in primate cerebral cortex. Proceedings of the National Academy of Sciences USA 86: 9015–9019.Find this resource:
Gregoriou, G. G., Gotts, S. J., 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:
Hasegawa, R. P., Matsumoto, M., and Mikami, A (2000). Search target selection in monkey prefrontal cortex. Journal of Neurophysiology 84: 1692–1696.Find this resource:
Hertz, J. A., Krogh, A. S., and Palmer, R. G. (1991). Introduction to Neural Computation Theory. Santa Fe, N.Mex.: Westview Press.Find this resource:
Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the National Academy of Sciences USA 79: 2554–2558.Find this resource:
Hopfield, J. J. and Herz, A. V. (1995). Rapid local synchronization of action potentials: Toward computation with coupled integrate-and-fire neurons. Proceedings of the National Academy of Sciences USA 92: 6655–6662.Find this resource:
Houk, J. C. and Wise, S. P. (1995). Distributed modular architectures linking basal ganglia, cerebellum, and cerebral cortex: Their role in planning and controlling action. Cerebral Cortex 5: 95–110.Find this resource:
Iba, M. and Sawaguchi, T. (2003). Involvement of the dorsolateral prefrontal cortex of monkeys in visuospatial target selection. Journal of Neurophysiology 89: 587–599.Find this resource:
Ipata, A. E., Gee, A. L., Gottlieb, J., Bisley, J. W., and Goldberg, M. E. (2006). LIP responses to a popout stimulus are reduced if it is overtly ignored. Nature Neuroscience 9: 1071–1076.Find this resource:
Jensen, O., Kaiser, J., and Lachaux, J. P. (2007). Human gamma-frequency oscillations associated with attention and memory. Trends in Neurosciences 30: 317–324.Find this resource:
Jensen, O. and Lisman, J. E. (2005). Hippocampal sequence-encoding driven by a cortical multi-item working memory buffer. Trends in Neurosciences 28: 67–72.Find this resource:
Johnston, K. and Everling, S. (2009). Task-relevant output signals are sent from monkey dorsolateral prefrontal cortex to the superior colliculus during a visuospatial working memory task. Journal of Cognitive Neuroscience 21: 1023–1038.Find this resource:
Kayser, C., Montemurro, M. A., Logothetis, N. K., and Panzeri, S. (2009). Spike-phase coding boosts and stabilizes information carried by spatial and temporal spike patterns. Neuron 61: 597–608.Find this resource:
Kerr, J. N. D. and Wickens, J. R. (2001). Dopamine D-1/D-5 receptor activation is required for long-term potentiation in the rat neostriatum in vitro. Journal of Neurophysiology 85: 117–124.Find this resource:
Kleinfeld, D. (1986). Sequential state generation by model neural networks. Proceedings of the National Academy of Sciences USA 83: 9469–9473.Find this resource:
Knight, R. T. (1997). Distributed cortical network for visual attention. Journal of Cognitive Neuroscience 9: 75–91.Find this resource:
Knight, R. T., Grabowecky, M. F., and Scabini, D. (1995). Role of human prefrontal cortex in attention control. Advances in Neurology 66: 21–34; discussion 34–26. (p. 802) Find this resource:
Konig, P., Engel, A. K., and Singer, W. (1995). Relation between oscillatory activity and long-range synchronization in cat visual cortex. Proceedings of the National Academy of Sciences USA 92: 290–294.Find this resource:
Lakatos, P., Karmos, G., Mehta, A. D., Ulbert, I., and Schroeder, C. E. (2008). Entrainment of neuronal oscillations as a mechanism of attentional selection. Science 320: 110–113.Find this resource:
Laurent, G. (2002). Olfactory network dynamics and the coding of multidimensional signals. Nature Reviews Neuroscience 3: 884–895.Find this resource:
Lee, H., Simpson, G. V., Logothetis, N. K., and Rainer, G. (2005). Phase locking of single neuron activity to theta oscillations during working memory in monkey extrastriate visual cortex. Neuron 45: 147–156.Find this resource:
Li, L., Gratton, C., Yao, D., and Knight, R. T. (2010). Role of frontal and parietal cortices in the control of bottom-up and top-down attention in humans. Brain Research 1344: 173–184.Find this resource:
Lisman, J. E. and Idiart, M. A. (1995). Storage of 7 short-term memories in oscillatory subcycles. Science 267: 1512–1515.Find this resource:
Liu, T., Slotnick, S. D., Serences, J. T., and Yantis, S. (2003). Cortical mechanisms of feature-based attentional control. Cerebral Cortex 13: 1334–1343.Find this resource:
Luck, S. J. and Vogel, E. K. (1997). The capacity of visual working memory for features and conjunctions. Nature 390: 279–281.Find this resource:
Lynd-Balta, E. and Haber, S. N. (1994). The organization of midbrain projections to the ventral striatum in the primate. Neuroscience 59: 609–623.Find this resource:
McClelland, J., McNaughton, B., and O’Reilly, R. (1995). Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychological Review 102: 419–457.Find this resource:
Mansouri, F. A., Matsumoto, K., and Tanaka, K. (2006). Prefrontal cell activities related to monkeys’ success and failure in adapting to rule changes in a Wisconsin card sorting test analog. Journal of Neuroscience 26: 2745–2756.Find this resource:
Mehta, M. R., Lee, A. K., and Wilson, M. A. (2002). Role of experience and oscillations in transforming a rate code into a temporal code. Nature 417: 741–746.Find this resource:
Middleton, F. A. and Strick, P. L. (1994). Anatomical evidence for cerebellar and basal ganglia involvement in higher cognitive function. Science 266: 458–461.Find this resource:
Middleton, F. A. and Strick, P. L. (2000). Basal ganglia and cerebellar loops: Motor and cognitive circuits. Brain Research Reviews 31: 236–250.Find this resource:
Middleton, F. A. and Strick, P. L. (2002). Basal-ganglia ‘projections’ to the prefrontal cortex of the primate. Cerebral Cortex 12: 926–935.Find this resource:
Miller, E. K. and Cohen, J. D. (2001). An integrative theory of prefrontal function. Annual Review of Neuroscience 24: 167–202.Find this resource:
Miller, E. K., Erickson, C. A., and Desimone, R. (1996). Neural mechanisms of visual working memory in prefrontal cortex of the macaque. Journal of Neuroscience 16: 5154–5167.Find this resource:
Miller, G. A. (1956). The magical number seven plus or minus two: Some limits on our capacity for processing information. Psychological Review 63: 81–97.Find this resource:
Montemurro, M. A., Rasch, M. J., Murayama, Y., Logothetis, N. K., and Panzeri, S. (2008). Phase-of-firing coding of natural visual stimuli in primary visual cortex. Current Biology 18: 375–380.Find this resource:
Moore, T. and Armstrong, K. M. (2003). Selective gating of visual signals by microstimulation of frontal cortex. Nature 421: 370–373.Find this resource:
Moore, T. and Fallah, M. (2001). Control of eye movements and spatial attention. Proceedings of the National Academy of Sciences USA 98: 1273–1276. (p. 803) Find this resource:
Moore, T. and Fallah, M. (2004). Microstimulation of the frontal eye field and its effects on covert spatial attention. Journal of Neurophysiology 91: 152–162.Find this resource:
Moore, T. L., Schettler, S. P., Killiany, R. J., Rosene, D. L., and Moss, M. B. (2009). Effects on executive function following damage to the prefrontal cortex in the rhesus monkey (Macaca mulatta). Behavioral Neuroscience 123: 231–241.Find this resource:
Nobre, A. C., Sebestyen, G. N., Gitelman, D. R., Frith, C. D., and Mesulam, M. M. (2002). Filtering of distractors during visual search studied by positron emission tomography. NeuroImage 16: 968–976.Find this resource:
Noudoost, B. and Moore, T. (2011). Control of visual cortical signals by prefrontal dopamine. Nature 474: 372–375.Find this resource:
O’Keefe, J. (1993). Hippocampus, theta, and spatial memory. Current Opinion in Neurobiology 3: 917–924.Find this resource:
Oksama, L. and Hyönä, J. (2004). Is multiple object tracking carried out automatically by an early vision mechanism independent of higher-order cognition? zapproach. Visual Cognition 11: 631–671.Find this resource:
O’Reilly, R. C. and Munakata, Y. (2000). Computational Explorations in Cognitive Neuroscience: Understanding the Mind. Cambridge, Mass.: MIT Press.Find this resource:
Otani, S., Blond, O., Desce, J. M., and Crépel F. (1998). Dopamine facilitates long-term depression of glutamatergic transmission in rat prefrontal cortex. Neuroscience 85: 669–676.Find this resource:
Pandya, D. N. and Barnes, C. L. (1987). Architecture and connections of the frontal lobe. In E. Perecman (ed.), The Frontal Lobes Revisited (pp. 41–72). New York: The IRBN Press.Find this resource:
Pandya, D.N. and Yeterian, E. H. (1990). Prefrontal cortex in relation to other cortical areas in Rhesus monkey: Architecture and connections. Progress in Brain Research 85: 63–94.Find this resource:
Pasupathy, A. and Miller, E. K. (2005). Different time courses of learning-related activity in the prefrontal cortex and striatum. Nature 433: 873–876.Find this resource:
Pesaran, B., Nelson, M. J., and Andersen, R. A. (2008). Free choice activates a decision circuit between frontal and parietal cortex. Nature 453: 406–409.Find this resource:
Petrides, M. and Pandya, D. N. (1999). Dorsolateral prefrontal cortex: Comparative cytoarchitectonic analysis in the human and the macaque brain and corticocortical connection patterns. European Journal of Neuroscience 11: 1011–1036.Find this resource:
Porrino, L. J., Crane, A. M., and Goldman-Rakic, P. S. (1981). Direct and indirect pathways from the amygdala to the frontal lobe in rhesus monkeys. Journal of Comparative Neurology 198: 121–136.Find this resource:
Postle, B. R., Awh, E., Jonides, J., Smith, E. E., and D’Esposito, M. (2004). The where and how of attention-based rehearsal in spatial working memory. Brain Research: Cognitive Brain Research 20: 194–205.Find this resource:
Pribram, K. H., Mishkin, M., Rosvold, H. E., and Kaplan, S. J. (1952). Effects on delayed-response performance of lesions of dorsolateral and ventromedial frontal cortex of baboons. Journal of Comparative and Physiological Psychology 45: 565–575.Find this resource:
Pylyshyn, Z.W. and Storm, R. W. (1988). Tracking multiple independent targets: Evidence for a parallel tracking mechanism. Spatial Vision 3: 179–197.Find this resource:
Rainer, G., Asaad, W. F., and Miller, E. K. (1998a). Memory fields of neurons in the primate prefrontal cortex. Proceedings of the National Academy of Sciences USA 95: 15008–15013.Find this resource:
Rainer, G., Asaad, W. F., and Miller, E. K. (1998b). Selective representation of relevant information by neurons in the primate prefrontal cortex. Nature 393: 577–579.Find this resource:
Rigotti, M., Rubin, D. B., Wang, X. J., and Fusi, S. (2011). Internal representation of task rules by recurrent dynamics: The importance of the diversity of neural responses. Frontiers in Computational Neuroscience 4: 24. (p. 804) Find this resource:
Roy, J. E., Riesenhuber, M., Poggio, T., and Miller, E. K. (2010). Prefrontal cortex activity during flexible categorization. Journal of Neuroscience 30: 8519–8528.Find this resource:
Saalmann, Y. B., Pigarev, I. N., and Vidyasagar, T. R. (2007). Neural mechanisms of visual attention: How top-down feedback highlights relevant locations. Science 316: 1612–1615.Find this resource:
Salinas, E. and Sejnowski, T. J. (2001). Correlated neuronal activity and the flow of neural information. Nature Reviews Neuroscience 2: 539–550.Find this resource:
Schultz, W. (1998). Predictive reward signal of dopamine neurons. Journal of Neurophysiology 80: 1–27.Find this resource:
Schultz, W., Apicella, P., and Ljungberg, T. (1993). Responses of monkey dopamine neurons to reward and conditioned stimuli during successive steps of learning a delayed response task. Journal of Neuroscience 13: 900–913.Find this resource:
Schultz, W., Apicella, P., Scarnati, E. and Ljungberg, T. (1992). Neuronal activity in monkey ventral striatum related to the expectation of reward. Journal of Neuroscience 12: 4595–4610.Find this resource:
Schultz, W., Dayan, P., and Montague, P. R. (1997). A neural substrate of prediction and reward. Science 275: 1593–1599.Find this resource:
Selemon, L. D. and Goldman-Rakic, P. S. (1985). Longitudinal topography and interdigitation of corticostriatal projections in the rhesus monkey. Journal of Neuroscience 5: 776–794.Find this resource:
Selemon, L. D. and Goldman-Rakic, P. S. (1988). Common cortical and subcortical targets of the dorsolateral prefrontal and posterior parietal cortices in the rhesus monkey: Evidence for a distributed neural network subserving spatially guided behavior. Journal of Neuroscience 8: 4049–4068.Find this resource:
Siegel, M., Donner, T. H., Oostenveld, R., Fries, P., and Engel, A. K. (2008). Neuronal synchronization along the dorsal visual pathway reflects the focus of spatial attention. Neuron 60: 709–719.Find this resource:
Siegel, M. and Konig, P. (2003). A functional gamma-band defined by stimulus-dependent synchronization in area 18 of awake behaving cats. Journal of Neuroscience 23: 4251–4260.Find this resource:
Siegel, M., Warden, M. R., and Miller, E. K. (2009). Phase-dependent neuronal coding of objects in short-term memory. Proceedings of the National Academy of Sciences USA 106: 21341–21346.Find this resource:
Sompolinsky, H. and Kanter, I. (1986). Temporal association in asymmetric neural networks. Physical Review Letters 57: 2861–2864.Find this resource:
Steinmetz, P. N., Roy, A., Fitzgerald, P. J., Hsiao, S. S., Johnson, K. O., and Niebur, E. (2000). Attention modulates synchronized neuronal firing in primate somatosensory cortex. Nature 404: 187–190.Find this resource:
Tallon-Baudry, C., Bertrand, O., Peronnet, F., and Pernier, J. (1998). Induced gamma-band activity during the delay of a visual short-term memory task in humans. Journal of Neuroscience 18: 4244–4254.Find this resource:
Thierry, A. M., Blanc, G., Sobel, A., Stinus, L., and Glowinski, J. (1973). Dopaminergic terminals in the rat cortex. Science 182: 499–501.Find this resource:
Tiesinga, P. H., Fellous, J. M., Jose, J. V., and Sejnowski, T. J. (2002). Information transfer in entrained cortical neurons. Network 13: 41–66.Find this resource:
Treisman, A. M. and Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology 12: 97–136.Find this resource:
Usrey, W. M. and Reid, R. C. (1999). Synchronous activity in the visual system. Annual Review of Physiology 61: 435–456.Find this resource:
Vogel, E. K. and Machizawa, M. G. (2004). Neural activity predicts individual differences in visual working memory capacity. Nature 428: 748–751. (p. 805) Find this resource:
Vogel, E. K., McCollough, A. W., and Machizawa, M. G. (2005). Neural measures reveal individual differences in controlling access to working memory. Nature 438: 500–503.Find this resource:
Wallis, J. D., Anderson, K. C., and Miller, E. K. (2001). Single neurons in the prefrontal cortex encode abstract rules. Nature 411: 953–956.Find this resource:
White, I. M. and Wise, S. P. (1999). Rule-dependent neuronal activity in the prefrontal cortex. Experimental Brain Research 126: 315–335.Find this resource:
Wise, S. P., Murray, E. A., and Gerfen, C. R. (1996). The frontal-basal ganglia system in primates. Critical Reviews in Neurobiology 10: 317–356.Find this resource:
Wolfe, J. M., Cave, K. R., and Franzel, S. L. (1989). Guided search: An alternative to the feature integration model for visual search. Journal of Experimental Psychology: Human Perception and Performance 15: 419–433.Find this resource:
Womelsdorf, T., Fries, P., Mitra, P. P., and Desimone, R. (2006). Gamma-band synchronization in visual cortex predicts speed of change detection. Nature 439: 733–736.Find this resource:
Woodman, G. F. and Luck, S. J. (2004). Visual search is slowed when visuospatial working memory is occupied. Psychonomic Bulletin & Review 11: 269–274.Find this resource: