이 장은 원문만 있습니다. 문장별 해석은 앞 장부터 차례로 붙입니다.
3장 · Working Memory in the Prefrontal Cortex — Importance of Delay-Period Activity in Working Memory
Delay-period activity has been observed in prefrontal neurons while monkeys performed behavioral tasks with a delay period. This activity often exhibits tonic sustained activation during the delay period. However, this activity can also exhibit tonic sustained suppression, or a gradually increasing, or gradually decreasing pattern. Since the definition of delay-period activity is activity observed during the delay period, delay-period activity can be observed in any brain area while the subject performs any behavioral task with delay. In fact, delay-period activity has been observed in the parietal cortex, the temporal cortex, the somatosensory cortex, and the premotor cortex. Delay-period activity has been observed in the primary visual cortex, the superior colliculus, the basal ganglia, the hippocampus, the thalamus, and even the spinal cord.
Although delay-period activity can be observed in any task with a delay, the information that must be maintained during the delay period can differ from task to task. Therefore, the information that neurons hold as delay-period activity can be different in different tasks. For example, when a monkey performs a delayed-response task, delay-period activity in the prefrontal cortex represents either the location of the visual cue or the direction of the response. In contrast, when a monkey performs a delayed matching-to-sample task, delay-period activity represents non-spatial physical features of the visual stimuli (e.g., color, shape). Prefrontal delay-period activity has been reported to represent a variety of information, such as tactile information, auditory information, task rules, task differences, expected reward, a numerical quantity, the relative distance between stimuli, timing, and the temporal order of stimuli, depending on the requirements of the task. This evidence strongly supports the notion that delay-period activity is an important component for understanding the neural mechanisms of working memory.
In addition, different brain areas participate in different functions and different types of information processing. Therefore, the information that neurons hold as delay-period activity is different in different brain areas. For example, in the prefrontal cortex, most delay-period activity encoded the location of the visual cue, while some encoded the direction of the response. On the other hand, in the parietal cortex, most delay-period activity encoded the direction of the saccade response or the direction of the arm response, although delay-period activity encoding sensory stimuli has also been reported Similarly, in the thalamic mediodorsal nucleus, most delay-period activity encoded the direction of the saccade response. This evidence supports the notion that delay-period activity is important for understanding working memory processes not only in the prefrontal cortex but also in other cortical and subcortical areas.
Exploring Neural Mechanisms of the Central Executive
Baddeley’s model of working memory is used to explain the results obtained from human psychological and neuropsychological studies and helps us to understand the mechanisms of various cognitive functions including language comprehension, thinking, reasoning, and decision-making. However, since his model is an abstract model, it is difficult to assign each of its components to a particular neural structure or brain area. However, the central executive is a neural component to coordinate the operations of multiple cognitive systems by integrating top-down and bottom-up signals to accomplish a specific goal. These functions are similar to those that have been proposed for the prefrontal cortex. Therefore, the central executive of Baddeley’s model is thought to reflect functions of the prefrontal cortex. In fact, the term “executive function” is often used to describe prefrontal functions. The prefrontal cortex is known to receive a variety of information by bottom-up signaling through cortico-cortical and cortico-subcortical connections and to control the operations of other cortical and subcortical areas by top-down signaling. Therefore, it is possible to understand the functions of the central executive by examining the neural mechanisms of prefrontal functions neurophysiologically using primate models.
Neural Mechanisms for Memory Control in the Prefrontal Cortex
As discussed above, Baddeley (2012) proposed a model of working memory, which is composed of one master component (the central executive) and three slave components (the phonological loop, the visuo-spatial sketch pad, and the episodic buffer). The function of the central executive is to coordinate and integrate the operation of the three slave components to achieve a current goal. The dual-task paradigm is often used to examine functions of the central executive. The dual-task paradigm is a behavioral procedure in which subjects are required to perform two different tasks simultaneously. When two tasks are performed under the dual-task condition, subjects often exhibit worse performance in one or both of the component tasks compared to when each component task is performed independently. This effect is known as dual-task interference. In human studies, this effect is considered to be caused by a limited capacity of cognitive resource, which corresponds to a task-general information-processing capacity in the brain that is shared by simultaneous cognitive tasks. The cognitive resource provides the workspace for information processing in a variety of cognitive functions. However, the capacity of the cognitive resource is limited. Therefore, when the subject needs to perform multiple cognitive tasks simultaneously, some neural system is required to allocate the cognitive resource to each task based on its demand for information processing.
The central executive in Baddeley’s model is thought to be this neural system that allocates a limited capacity of memory to each slave component depending on the demand necessary for its performance. By doing this, the central executive can control the performance of the slave components. Baddeley (1986) did not assign the central executive to any particular brain area. However, shifting of attention or the mental set, updating and monitoring of information, and inhibition of prepotent responses are typical examples of executive functions, and the prefrontal cortex is known to be an essential brain area for such executive functions. Further, impairment caused by prefrontal damage often produces executive dysfunction. Thus, the functions of the prefrontal cortex are closely linked to the functions of the central executive proposed by Baddeley. Therefore, the prefrontal cortex could act as the central executive and control the operations of other cortical systems. However, it is not yet clear how the prefrontal cortex acts as the central executive and how a prefrontal neural mechanism allocates a limited capacity of memory to each of multiple simultaneous tasks depending on their needs for information processing.
Watanabe and Funahashi (2014, 2015) examined this issue by comparing prefrontal neural activities using a dual-task paradigm. They used a spatial working memory task and a spatial attention task. In the spatial working memory task, which is the same as the ODR task, monkeys were required to remember the location of a visual cue that was briefly presented in the peripheral visual field and then perform a memory-guided saccade to that location at the end of the delay period. In the spatial attention task, monkeys were required to attend to a small white circle presented on a monitor and make a quick lever-release response when they detected a change in the color of the circle (from white to red). In the dual-task condition, the spatial working memory task was introduced while monkeys performed the spatial attention task. It is well known that the dorsolateral prefrontal cortex is an essential brain area for both memory and attention tasks and that many neurons in this area are activated during performance of these tasks. Therefore, simultaneous performance of these two tasks could produce an interference effect, because both tasks require information processing based on a common neural resource. It is expected that, if a dual-task interference effect is observed behaviorally, the delay-period activity observed in the memory task might be affected and there may be a difference in the magnitude of activities between the single-task condition of the memory task and the dual-task condition.
Watanabe and Funahashi (2014, 2015) observed that the behavioral performance of the memory task was impaired in the dual-task condition and that greater impairment was observed in the memory task as the difficulty of the attention task increased. Thus, the dual-task interference effect observed in the memory task increased with an increase in the difficulty of the attention task. These results indicate that similar dual-task interference can be observed in both monkeys and human subjects. Next, they analyzed prefrontal single-neuron activities while monkeys performed both the memory task and the attention task under the dual-task condition and while monkeys performed the memory task alone under the single-task condition. They showed that both the memory task and the attention task recruited the activation of largely overlapping prefrontal neural populations. The degree of overlap of the neural populations recruited by these two tasks was correlated with the strength of the dual-task interference effect observed in behavioral studies. A higher degree of overlap in the recruitment of neural populations was observed when a greater interference effect was observed in the behavioral response. In addition, the magnitude of delay-period activity that represented spatial information of the visual cue in the memory task was significantly attenuated in the dual-task condition compared with the single-task condition. Further, a greater attenuation of delay-period activity was observed when the difficulty of the attention task increased in the dual-task condition. These results indicate that the ability of a prefrontal neural population to represent task-relevant information decreases in proportion to the increased demand of the concurrent counterpart task and that the dual-task interference effect originates in the simultaneous and overloaded recruitment of common prefrontal neural populations by the two tasks.
Watanabe and Funahashi (2014) also explored the temporal dynamics of the competitive interaction between the memory task and the attention task in the dual-task condition. When monkeys performed only the attention task, the size of the neural population encoding information regarding the attention cue increased after the attention cue was presented, and its magnitude was maintained until the end of the attention task. Similarly, when monkeys performed only the spatial memory task, the size of the neural population encoding information regarding the memory cue increased after the memory cue was presented, and its magnitude was maintained during the delay period until the end of the memory task. In the dual-task condition, the size of the neural population encoding the attention cue showed similar temporal patterns as when only the attention task was introduced. On the other hand, although the size of the neural population encoding the memory cue transiently increased when the memory cue was presented, it was significantly attenuated during the delay period compared with the single-task condition in the memory task. However, the size of this neural population significantly increased (“reawakening”) at the end of the attention task, and this increase was maintained until the end of the memory task.
These results show that the locus of the dual-task interference effect is the competitive and overloaded recruitment of overlapping neural populations in the prefrontal cortex by the two concurrent tasks. These results also suggest that the capacity of information processing by a single neuron is limited to a fixed level, can be flexibly allocated or reallocated between two concurrent tasks based on their needs, and enhances behavioral performance when allocated to a task. The concurrent elevation, attenuation, and reawakening of the size of neural populations observed in the dual-task condition suggest that the cognitive resource for information processing in the prefrontal cortex is divided and allocated to the two tasks based on their needs.
This view of neural processing has been corroborated by the results of recent single-neuron studies that investigated the neural basis of visual short-term memory for multiple locations and multiple objects. These studies showed that, when two locations or two objects were presented as memoranda within the same visual hemifield, the stimulus selectivity of prefrontal neurons was significantly attenuated relative to when the two stimuli were presented separately in different hemifields. These results suggest that the limited capacity of cognitive resources is the result of the limited computational capacity of single neurons. Dual-task interference is likely to originate in the competitive and overloaded recruitment of common neural populations by two concurrent tasks. Although the neural mechanisms for dividing and allocating a common and limited-capacity cognitive resource in the prefrontal cortex are not known, it is very important to identify these mechanisms when we try to understand the neural mechanisms of executive control by the prefrontal cortex and the neural system that is responsible for the central executive in the working memory model.
Neural Mechanisms for Monitoring Information Contents in the Prefrontal Cortex
In Baddeley’s model of working memory, the central executive plays the most important function among the different components. Miyake et al. (2000) listed three separable executive functions: shifting of attention and the mental set, updating and monitoring of information, and inhibition of prepotent responses. When the prefrontal cortex functions as the central executive by controlling the operations of other cortical and subcortical neural systems using top-down signaling, it must monitor concurrent information processes within itself and other brain areas.
Humans can distinguish between what they remember and do not remember. This ability to monitor one’s own memory state is an important feature of human cognition and is referred to as metamemory. Human neuropsychological studies and neuroimaging studies have revealed that the prefrontal cortex participates in metamemory processes. The neural mechanism of metamemory function is not yet clear, partly because it is difficult to conduct neurobiological studies on metamemory using animals. However, increasing attention has been paid to exploring the neural substrates of metamemory. Neurophysiologists have recently entered this field in the search for the neuronal correlates of uncertainty associated with decision-making and for the neural mechanisms of metamemory itself. Studies of the neural mechanisms of metamemory functions in the prefrontal cortex could lead to important clues for understanding the neural mechanisms by which the central executive monitors information processes in slave components.
It has been believed that metacognition, which is the ability to monitor and access one’s own memory processes, can only be found in humans, since only humans can use language and express their own introspective experience using language. However, recent behavioral studies have provided evidence that rhesus monkeys can monitor whether or not they remember target information that is necessary to correctly perform subsequent tests. This ability is thought to be functionally analogous to human metamemory. The metamnemonic ability of monkeys has been examined under an experimental paradigm with two important features. First, difficult memory test conditions are included to ensure that the monkey often experiences uncertainty, such as uncertainty regarding whether or not information that is necessary to make a correct response is remembered. Second, the monkey is given a chance to cope with uncertain trials by selecting an escape option that can be introduced immediately before, simultaneously with, or immediately after a memory test. This escape option leads to a little more favorable result (e.g., getting a small amount of reward) than failure in the test (e.g., getting no reward), while correct performance on the test yields the most favorable result (e.g., getting a large amount of reward). The basic idea underlying these procedures is that, if the monkey can monitor its own memory state, the monkey can be predicted to select the escape option in trials in which a correct answer is uncertain. Consistent with this prediction, when monkeys perform a memory task that consists of a mixture of trials with various degrees of difficulty, they tend to select the escape option more frequently in more difficult trials, in which they are considered to experience mnemonic uncertainty (lower degree of confidence) more often than in easier trials. Moreover, by selectively taking the escape option in uncertain trials, monkeys can increase the reward acquisition ratio, compared to when they are required to perform the same task without an escape option (forced choice conditions).
Human studies have indicated that the prefrontal cortex plays an important role in metamemory. The metamnemonic capacity of monkeys can be examined with the use of an appropriate behavior. Neurophysiological studies have shown that prefrontal neurons exhibit directionally-selective delay-period activity, which has been considered to be a neural correlate of a mechanism for the short-term maintenance of spatial information in working memory. Based on these findings, Tanaka and Funahashi (2012, 2016) examined whether and how prefrontal delay-period activity is related to subjective confidence in working memory. Monkeys were trained to perform a modified oculomotor delayed-response task, in which the monkey was required to make a saccade to a memorized location (memory test condition). The difficulty of the memory test was controlled by varying the number of distractors that were presented during the delay period. A feature of this task was the introduction of a choice period between the end of the delay period and the start of the response period. In the choice period, the subject was sometimes allowed to choose to either accept the memory test or decline the test and select the escape option, and sometimes forced to take the memory test. When the subject chose to take the memory test, it was required to make a memory-guided saccade to the location where the visual cue had been presented during the cue period. A correct response was rewarded with a drop of juice, while no reward was given for an incorrect response. When the subject declined the memory test and selected the escape option, it performed a visually-guided saccade to the visual target presented on the monitor, and only 30% to 50% of correct saccades were rewarded. Under these conditions, the monkeys' performance in the memory test condition was significantly better when they chose to take the memory test by themselves than when they were forced to do so. At the same time, the monkeys tended to decline the memory test and chose the escape option more often when the difficulty of the memory test increased. These results agree with the prediction that the monkey can be predicted to select the escape option in trials in which a correct answer is uncertain, and indicate that the monkey can monitor its own working memory state. Middlebrooks and Sommer (2011) also confirmed that monkeys have metamnemonic capacity in working memory.
Tanaka and Funahashi (2016) examined single-neuron activities of prefrontal neurons while monkeys performed this metamemory task. Prefrontal neurons exhibited spatially selective delay-period activity regardless of whether the monkeys chose to take the memory test or to decline the test and take the escape option. However, an important observation was that the spatial selectivity of delay-period activity was significantly weaker when the monkeys chose to decline the memory test than when they chose to take the test. The reduction of spatial selectivity was not due to a decrease in the response to the preferred direction, but rather to an increase in the response to non-preferred directions. Since maintaining spatial information of the visual cue is important for performing this task and since delay-period activity has been considered to be a neural correlate of spatial working memory processes, the directional selectivity of delay-period activity and the maintenance of its strength are essential factors for monkeys to obtain the reward. A decrease in the spatial tuning strength of directional delay-period activity caused a reduction in the accuracy of spatial information regarding the visual cue. Thus, the quality of the spatial information represented by prefrontal delay-period activity is an important factor in the subject's choice regarding whether it chooses to take the memory test or the escape option. Therefore, spatially selective delay-period activity can serve as a source to estimate the subject’s own confidence in their working memory.
On the other hand, Middlebrooks and Sommer (2012) recorded single-neuron activity from the prefrontal cortex, the frontal eye field, and the supplementary eye field while monkeys performed a visual metamemory task using oculomotor responses. Monkeys were required to choose a visual target, which was presented at the same position as where the visual cue had been presented, by a saccade. In this task, monkeys first made a decision and reported it by a saccade. Next, they had to bet whether their decision was correct or not, by selecting one of two visual targets (high-bet target and low-bet target) using a second saccade. If the monkey made a correct decision and took a high-bet target, it obtained the maximum amount of reward. If the monkey made an incorrect decision and took a high-bet target, it received no reward during a 5-s timeout. If the monkey took a low-bet target, it received a small amount of reward regardless of whether it made a correct decision or not. They found that, although neural activity related to the decision and a bet was found in all three brain areas, the activity linking the decision to an appropriate bet was found exclusively in the supplementary eye field. Therefore, they concluded that the supplementary eye field is an important brain area for metacognitive processes and that the putative metacognitive activity began swiftly in the supplementary eye field during the decision stage and continued to the bet stage. Middlebrooks and Sommer (2011, 2012) used a task in which the monkeys needed to bet to reflect their confidence in their choice after they made a choice based on their memory. In contrast, Tanaka and Funahashi (2012, 2016) used a task in which the monkeys needed to choose a response after they estimated the confidence of their memory. Therefore, the difference between the tasks used may be responsible for the different activities in each brain area and the different contributions of different brain areas. Although these studies found some neural sources of metamemory processes, the neural mechanisms of metamemory processes and monitoring are not yet clear. Further studies are needed to identify which brain areas participate in metamemory processes, how the prefrontal cortex controls metamemory processes and how the prefrontal cortex monitors processes within itself and other areas.
Conclusions
Delay-period activity is usually defined as tonic sustained activation observed during the delay period. This definition is based on the temporal profile of the neuron’s discharge rate during the delay period. If we use this definition, delay-period activity can be observed not only in the prefrontal cortex but also in any brain area when the subject performs any task with a delay interval. However, since each brain area performs unique information processes using unique information, the information represented by delay-period activity may be different in different brain areas, even if these brain areas show the same temporal profile of delay-period activity. Therefore, a detailed analysis of the features of delay-period activity (e.g., information represented by this activity, selectivity, ratio of neurons having this activity) is critically important for understanding the mnemonic functions and the neural mechanisms of working memory in this brain area.
In the prefrontal cortex, delay-period activity is a neural correlate of the mechanism for temporarily maintaining information and a majority of this activity exhibits stimulus selectivity and encodes retrospective sensory (visual) information. Working memory consists of neural processes not only for the short-term active maintenance of information but also for manipulation of information. When executive functions operated by the prefrontal cortex are considered, neural processes for manipulating information would be more important than for maintaining information. Examining neural activities using a dual-task paradigm or a metamemory task, some important mechanisms related to the central executive have been identified. Although further studies are needed to elucidate how the prefrontal cortex controls memory resource and supervises other systems, detailed analysis of delay-period activity could provide important clue for these mechanisms.
이 글은 위 출처의 원문을 그 라이선스(CC BY 4.0)에 따라 옮겼습니다. 참고문헌 목록과 인용 번호, 그림 · 표 · 각주 · 수식 · 부록은 빼고 절별로 나누어 실었으며, 그 밖의 문장은 원문 그대로입니다. 원 저작자와 발행 기관은 이 화면을 만들거나 검토하지 않았고, 이곳의 해석과 낱말 풀이는 공식 번역이 아닙니다.
