The prefrontal cortex (PFC) appears to be important for processing both cognitive and motivational context information. Primate lateral PFC (LPFC) neurons are involved in cognitive context-dependent stimulus coding by responding differently to an identical stimulus according to the task situation. Such context-dependent LPFC activity appears to be supported by context-representing activity, observed also in LPFC neurons, in which the baseline activity differs as a function of the task. In LPFC, there are also neurons that code stimulus on the basis of motivational context. This motivational context is represented in differential baseline activity as a function of the reward context. In the orbitofrontal cortex (OFC), there are neurons that code stimuli depending on the motivational context as well as neurons that represent motivational context information. Furthermore, we found LPFC neurons that coded the stimulus depending on both the cognitive and motivational context, as well as LPFC neurons that represented both the cognitive and motivational context. For adaptive behavior, it is important to code the meaning of the environmental situation based on the context. While OFC is predominantly concerned with processing motivational context information, LPFC seems to play important roles in integrating the cognitive and motivational context for adaptive goal-directed behavior.
The prefrontal cortex (PFC) is important for higher-order cognitive as well as motivational operations (Luria 1980; Stuss and Benson 1986; Goldman-Rakic 1987; Fuster 1997). The term “context” may be the key word for understanding the cognitive and motivational function of PFC. To our knowledge, Pribram was the first researcher to refer to this idea, indicating that, “The PFC is involved in providing and maintaining a context, a temporal organization of brain events…. Patients and animals with prefrontal resection do not seem to know how to use the context” (Pribram 1971). Simulation studies by Cohen et al. (1996) indicate that PFC is indispensable for representing and monitoring task-relevant context information, consisting of a set of task instructions and/or the outcome of previous stimulus–response sequences, in complex goal-directed behavior.
The lateral PFC (LPFC) appears to be important for processing “cognitive” context information, which consists of the task situation and/or previous stimulus–response sequences. Clinical and lesion studies have indicated the importance of LPFC for performing a conditional discrimination task in which the correct response is not determined by an associative rule but is dependent on the context defined by the conditional cue (Petrides 1985, 1990; Murray et al. 2000). The orbitofrontal cortex (OFC), on the other hand, seems to be predominantly involved in “motivational” context information, which consists of the organism's present motivational state and information regarding the attractiveness/aversiveness of the past, present, or possible future reward/threat. Clinical studies have indicated that OFC patients show disturbances in evaluating the current situation depending on the motivational context with respect to possible outcomes (gain vs. loss) of the behavior. Thus, the patients tend to make decisions aiming at short-term gain, disregarding long-term consequences (Damasio 1994).
Here, we will discuss the functional significance of LPFC and OFC neurons that are involved in processing cognitive/motivational context information for adaptive goal-directed behavior.
Processing of Cognitive Context Information in Primate Lateral Prefrontal Neurons
To examine neuronal activities of primate LPFC in relation to processing cognitive context information, we trained monkeys on a conditional Go/No-go discrimination task (Sakagami and Niki 1994). In this task, there were 3 kinds of task conditions in which monkeys based their response on one of 3 different kinds of dimensions (color, shape, or position). The monkey faced a video monitor and initiated each trial by pressing a lever (Fig. 1a). Then a fixation spot was turned on whose color indicated which dimension (color, shape, or position) was relevant for the current trial and thus defined the task requirement (i.e., cognitive context). When the monkey continued pressing the lever for 1–2 s, a Discriminative Cue (colored geometric shape; circle, cross, stripes, or diamond shape with red, green, yellow, or purple color) was presented at one of 4 locations (right, left, up, or down) on the video monitor for 160 ms. Then there was a delay period of 1–2 s. After the delay period was over, which was indicated by the color change of the fixation spot, the monkey had to perform a Go (immediate lever release) or No-go (delayed lever release after the fixation spot returned to the original color) response depending on both the currently relevant dimension and the Discriminative Cue presented for the current trial. For example, a red cross or yellow diamond meant Go in the color condition, but No-go in the shape condition. Similarly, a green circle or purple stripes meant No-go in the color condition, but Go in the shape condition (see Fig. 1b).
Activities of many LPFC neurons in response to the Discriminative Cue were found to be dependent on the cognitive context. For example, the LPFC neuron shown in Figure 1b showed activations whenever a Go-indicating cue was presented irrespective of the cue's physical properties. Thus, the activity of this neuron was related specifically to its behavioral significance (Go or No-go) determined by the context in terms of which dimension was currently relevant. Such context-dependent stimulus coding activity was observed in 249 (75.9%) out of 328 neurons (among 680 task-related neurons) that showed cue-related activity.
Similar LPFC neurons, which code stimuli depending on the cognitive context, have subsequently been reported in multitask situations (White and Wise 1999; Asaad et al. 2000; Hoshi et al. 2000; Genovesio et al. 2005; Johnston and Everling 2006). For example, in the study by Asaad et al. (2000), in which the monkey was trained on 3 different kinds of discrimination tasks (object, spatial, associative), LPFC neurons responded differently to an identical discriminative cue depending on the context of the current task.
To attain such context-dependent stimulus coding, context information must somehow be represented in the brain. LPFC seems to be actively involved in this process. Activities before the presentation of the cue (precue activities) were previously dismissed as background noise while they are presently receiving much attention because they show task relatedness in many brain areas. In LPFC, we found neurons that were concerned with the cognitive context as could be measured by their precue baseline activities. For example, the neuron shown in Figure 1c showed differential precue baseline activity depending on which dimension was currently relevant. This neuron showed stronger precue activity in the position condition, than in the color or shape condition (we also observed neurons that increased their precue activity selectively for color or shape conditions). Among 154 neurons that showed cue-related activity and could be examined in all the 3 kinds of task conditions, context-representing (differential precue) activity was observed in 27 (17.5%) neurons.
Cognitive context-representing neuronal activities (showing differential precue baseline activities) have also been reported in LPFC in multitask situations (Asaad et al. 2000; Wallis et al. 2001) and in the Wisconsin card sorting test analog (Mansouri et al. 2006). For example, in the study by Mansouri et al. (2006), in which the monkey was required to match a sample to one of 3 test items by either color or shape, LPFC neurons showed differential precue baseline (as well as postcue) activities depending on the relevant rule (color or shape), although the relevant rule was not explicitly indicated to the monkey and the monkey had to find the rule on the basis of feedback to the response.
Human neuroimaging studies also indicate that LPFC is concerned with coding the situation depending on the cognitive context. For example, Durston et al. (2002) showed that LPFC showed activity changes at the time of the response depending on the context of the subjects' previous responses. Thus, during the Go/No-go discrimination task, when No-go trials were preceded by either 1, 3, or 5 Go trials, LPFC demonstrated stronger activation at the time of No-go response as the number of preceding Go trials increased. There are also many neuroimaging studies indicating that LPFC represents the cognitive context (Bunge et al. 2003; Sakai and Passingham 2003, 2006). For example, Sakai and Passingham (2003) showed that different LPFC subareas were differentially activated depending on the task requirements and that polar PFC was concerned with the control of such context-representing operations. Thus, both human neuroimaging and monkey neurophysiological studies clearly indicate that LPFC plays significant roles in processing cognitive context information.
Processing of Motivational Context Information in Primate Lateral Prefrontal Neurons
LPFC plays important roles not only in cognitive but also in motivational operations (Stuss and Benson 1986; Watanabe 1998). Thus, LPFC neurons are responsive to reward delivery as well as to the occurrence of errors (Watanabe 1989). Reward-expectancy–related neurons are also observed in LPFC (Watanabe 1996; Leon and Shadlen 1999). In our recent study, we showed that LPFC neurons were also involved in coding the stimulus dependent on the motivational context (Watanabe et al. 2002). We trained monkeys in a delayed reaction task in which an instruction cue indicated the presence or absence of reward. The monkey faced a panel on which a rectangular window, a circular key and a holding lever were vertically arranged (Fig. 2a,b). The window contained one opaque screen and one transparent screen with thin vertical lines. The monkey first depressed the lever for 10–12 s. Then, the opaque screen was raised for 1 s as an instruction cue, revealing a piece of food or an empty food tray (Fig. 2a, visible food). The presence or absence of food predicted the presence or absence of the reward after the response. Monkeys were also trained in a Cued food version (Fig. 2b). Either a red or green color cue was presented on the key for 1 s as an instruction cue. A red cue predicted reward delivery, whereas a green cue predicted no-reward delivery. After a delay of 5 s, a white light appeared on the key as a go signal. When the monkey released the hold lever and pressed the key after the go signal, the monkey was rewarded or not rewarded depending on the previously presented instruction cue. Reward and no-reward trials alternated pseudorandomly. The monkey had to press the key even in no-reward trials to advance to the next trial. Pieces (about 0.5 g) of raisin, sweet potato, cabbage, or apple were used as food rewards. The same reward was used for a block of about 50 trials, so that the current reward would define the motivational context for the block of trials. The reward type was changed after about every 50 trials.
In response to an instruction cue indicating absence of reward, we found that some LPFC neurons not only predicted the absence of reward but also represented more specifically which kind of reward would be omitted in a given trial. Figure 2c shows such an example. This LPFC neuron, examined during the Visible food version of the task, showed activations only in no-reward trials during the cue period. Considering only no-reward trials, this neuron showed the highest firing rate for cabbage and the lowest firing rate for raisin as the reward type in a block of trials (monkeys preferred cabbage most and raisin least), despite the fact that the monkey was shown the same empty food tray as an instruction cue in all blocks. These neurons seem to code the motivational significance of the cue depending on the context information concerning which kind of reward may be delivered in subsequent trials. Motivational context-dependent stimulus coding activity was observed in 72 (79.1%) out of 91 neurons (among 230 task-related neurons) that showed cue/delay-related activity and could be examined across 3 different kinds of rewards.
We also found neurons in the LPFC that represented the motivational context (Watanabe et al. 2002). These neurons showed reward-discriminative tonic baseline activities. Figure 2d shows an example of an LPFC neuron, examined in the Cued food version of the task, which showed differential precue baseline activity depending on the reward type used in a block of trials. Thus, during precue periods this neuron showed the highest firing rate in the cabbage reward block and the lowest firing rate in the raisin reward block (this neuron also showed motivational context-dependent stimulus coding after cue presentation, despite the fact that the same color cues were used as the instruction cue on any reward block). Differential precue activity in relation to representing the motivational context was observed in 28 (38.9%) out of the 72 neurons that showed motivational context-dependent stimulus coding activity.
Baseline neuronal activity in primate LPFC was reported to be modified by the monkey's arousal level (Lecas 1995). Hasegawa et al. (2000) observed LPFC neurons whose precue activity reflected the monkey's performance level in the past trial, or predicted the performance level in the future trial. Such activity in LPFC may be concerned with representing the motivational context.
Neuroimaging studies also indicate that LPFC is concerned with coding the situation depending on the motivational context. For example, in a study by Nieuwenhuis et al. (2005), LPFC similarly responded to different trial outcomes during a monetary gambling task depending on the motivational context (here, the range of possible outcomes from which the current outcome was selected). Thus, LPFC as well as other reward-sensitive brain areas were activated to a comparable degree by the best outcomes in each condition (such as a large gain in the “win” condition or no loss in the “loss” condition), although there was a large difference in the objective value of these outcomes.
Processing of Motivational Context Information in Primate Orbitofrontal Neurons
To examine OFC neuronal activity in relation to processing motivational context information, we trained monkeys in a delayed reaction task with different rewards (Hikosaka and Watanabe 2004). In this task, the monkey faced a panel on which a rectangular window, a circular key and a holding lever were vertically arranged (Fig. 3a). The monkey first held the lever depressed for 3 s, and a color instruction cue of either red or green light was presented for 1 s on the circular key. Then, there was a delay period of 5 s, after which a white light was presented on the key as a go signal and the monkey was required to press the key. Each set of 4 consecutive trials constituted one block within which 3 different kinds of rewards and one trial with no reward were given in a fixed order that differed from the monkey's reward preference. Thus, the first, second, and third red cue predicted the delivery of orange juice, water, and grape juice and the forth (green) cue predicted no-reward delivery (we also used food rewards that were delivered in the window located above the circular key). Even in no-reward trials, the monkey had to perform the key-press response to advance to the next trial. Reaction times of the monkey were significantly shorter in preferred than in less preferred reward trials (the monkey's preference was grape juice > orange juice > water > no-reward) even though the same red cue was presented at the first, second, and third trial within a block. This difference in reaction time indicates that the monkey had come to know the motivational context in terms of which reward would be delivered after the delivery of one kind of reward.
OFC neurons were shown to be involved in motivational context-dependent stimulus coding by responding differently to the same red instruction cue as a function of its motivational significance. For example, the neuron displayed in Figure 3b showed the lowest firing rate in grape reward trials and the highest firing rate in water reward trials among the first 3 trials within a block of 4 trials. The activity of this neuron after the red cue presentation appeared to be a continuation of the activity observed during the precue period. This type of neuron may be concerned with coding, not the cue's physical properties, but the cue's motivational meaning (attractiveness of the reward indicated by the cue), depending on the context of which reward would be delivered in the current trial. Such motivational context-dependent stimulus coding activity was observed in 21 (25.9%) out of 81 neurons (among 235 task-related neurons) that showed cue/delay-related activity.
We also found OFC neurons that were involved in representing the motivational context by showing precue baseline activity reflecting the reward type that would be delivered in the current trial. An example is shown in Figure 3c. (This is the same neuron as the one presented in Fig. 3b; here the activity is displayed in an expanded time scale.) The neuron showed precue activity changes according to the monkey's preference for each trial's outcome, showing stronger activity when the less preferred reward was expected. Thus, motivational context-dependent stimulus coding in this neuron was done, different from the stimulus coding in the LPFC neuron shown in Figure 2b, not by receiving context information from other neurons but by continuing its own precue activity.
In OFC, we observed another type of context-representing neurons that showed differential precue baseline activity. The example displayed in Figure 3d showed stepwise increased changes in precue activity from the least preferred reward trial toward the most preferred reward trial. It should be noted that 3 different kinds of rewards and no-reward were given in a fixed order that differed from the monkey's reward preference. This type of neuron did not show precue activity changes that reflected the preference of each trial's outcome, but showed stepwise changes (increase or decrease) in precue activity toward the trial with a particular outcome (usually the most or least attractive within a block of 4 trials). Motivational context-representing activity was observed in 32 (47.0%) out of 68 neurons that showed precue activity changes.
The first type of context-representing neurons (n = 14), which showed precue activity reflecting the preference of each trial's outcome, may represent the “short-range” motivational context of the current trial's expected reward. This information could be derived from the task situation in which the monkey could know the next reward type after the delivery of the previous reward. The second type of neurons (n = 18) that showed precue activity reflecting the expectancy of the specific kind of (most or least preferred) outcome within a block of 4 trials, may represent the “long-range” motivational context of the task situation, the most or least preferred reward being delivered every 4 trials.
Motivational context-dependent stimulus coding as well as motivational context-representing OFC activities have also been observed in other laboratories. For example, Tremblay and Schultz (1999) examined OFC neurons during a spatial delayed response task in which one of 3 different shape cues, each associated with a specific reward type, was presented either on the right or left side as a spatial cue. The same 2 shape cues were used continuously for a block of several tens of trials. OFC neurons responded differently to the identical shape cue depending on the motivational context of whether the reward associated with that shape cue was more or less preferable than the reward associated with the alternative shape cue.
In the Go/No-go discrimination task, they also found motivational context-representing OFC neurons that showed differential precue baseline activity depending on whether the monkey was rewarded or not in the previous No-go trial; in this task, correct No-go responses were rewarded in half of the trials and absence of reward in the previous No-go correct trial constituted a motivational context indicating an increased probability of reward delivery in the current trial (Tremblay and Schultz 2000).
Neuroimaging studies also indicate that OFC is involved in coding the situation depending on the motivational context. For example, Ursu and Carter (2005) showed that the affective impact of an outcome during a working memory task was modulated by the nature of the various possible outcomes. Thus, OFC responded differently to the same cue indicating the absence of monetary incentives, depending on the motivational context of whether the alternative outcome of the task performance was a 1$ gain or loss.
Integration of Cognitive and Motivational Context Information in Lateral Prefrontal Neurons
We recently found that LPFC neurons not only code the stimulus depending on both the cognitive and motivational context, but also represent both the cognitive and motivational context (Kobayashi et al. 2007). We trained monkeys on a memory-guided saccade task. A trial started with the onset of a fixation spot on a video monitor (Fig. 4a). During the monkey's fixation, a positional cue was presented at one of 2 diagonally opponent directions, selected out of 8 radially arranged candidate positions, in relation to the response field of the neuron that was being recorded. After the delay period, the fixation spot was turned off, and the monkey was required to make a saccadic eye movement toward the previously cued position. In this task, one cue position was associated with reward delivery, whereas the other cue position was not (monkeys were rewarded only when the eye movement response was directed to one predetermined position, but not when it was directed to the alternative position). The position that was associated with reward delivery was changed about every 60 trials without explicitly indicating the change to the monkey (Fig. 4b). We found LPFC neurons that coded the cue stimulus depending on both the cognitive and motivational context; the neuron in Figure 4c showed differential activity to the same (directional) cue depending on both the cognitive and motivational context of which direction (cognitive context) was associated with reward delivery (motivational context). Thus, this neuron showed higher firing rates to the same left cue presentation in the context of reward on the left than in the context of reward on the right. This neuron also showed higher firing rates to the same right cue presentation in the context of reward on the left than in the context of reward on the right.
Thus, coding the stimulus in this neuron was attained by integrating the spatially selective cue response with the (combined cognitive and motivational) context information that was also represented in the neuron's own precue baseline activity (this neuron showed higher firing rates during the precue period in the context of reward on the left). Among 189 task-related neurons, 36 (19.0%) of them showed precue activity changes representing both the cognitive and motivational context information, and 39 (41.0%) out of 95 neurons that displayed cue-related activity were concerned with coding the stimulus depending on both the cognitive and motivational context information.
Barraclough et al. (2004) reported that activity of some LPFC neurons during a free-choice task was modulated depending on both the monkey's (cognitive) choice and its (reward or no-reward; motivational) outcome in the previous trial. These neurons are also considered to be related to coding the situation depending on both the cognitive and motivational context.
Functional Significance of Processing Cognitive and Motivational Context Information in Prefrontal Neurons
For adaptive and flexible behavior, it could be more beneficial to code each stimulus according to the cognitive context in which the stimulus is presented, than to code only its physical properties. Indeed, processing of the cognitive context has been considered to be essential for goal-directed behavior (Miller and Cohen 2001). It has been proposed that the representation of the cognitive context is used to suppress competing behaviors and to coordinate execution over temporally extended periods (Cohen et al. 1996). Moreover, recent evidence amply demonstrates that LPFC neurons are involved in coding the stimulus depending on the cognitive context as well as representing the cognitive context (Watanabe 1986; Sakagami and Niki 1994; White and Wise 1999; Asaad et al. 2000; Wallis et al. 2001; Mansouri et al. 2006). Human neuroimaging studies also indicate the importance of LPFC in processing cognitive context information (Durston et al. 2002; Bunge et al. 2003; Sakai and Passingham 2003, 2006). Disorganization symptoms observed in schizophrenic patients are supposed to be related to impairments in processing cognitive context information in LPFC (McDonald et al. 2005).
It has been well documented that OFC, which is intimately connected with the limbic system, plays more important roles in motivational operations than LPFC (Fuster 1997; Rolls 1999). On the other hand, OFC is not thought to be directly involved in cognitive operations such as working memory because OFC neurons are rarely observed to show cognitive properties (Tremblay and Schultz 1999; Wallis and Miller 2003). Humans with OFC resections and monkeys with OFC lesions show disturbances in motivational, emotional, and social behavior (Stuss and Benson 1986; Damasio 1994; Rolls 1999). For the animal to survive, it may be particularly important to evaluate behavioral outcomes depending on the motivational context, with respect to the animal's nutrition condition and possible alternative outcomes, more so than to simply coding the presence or absence of reward. Primate OFC appears to have a crucial function in this evaluation process by coding the meaning of the stimulus depending on the motivational context.
We found 2 types of context-representing OFC neurons that showed precue baseline activities relating to short- or long-range reward expectancy. Short-range reward-expectancy neurons may be involved in adjusting the motivational level toward the current trial's outcome, whereas long-range reward-expectancy neurons would have an analogous function toward future outcomes. Lesion studies also indicated the importance of OFC in processing information regarding long-term incentive values (DeCoteau et al. 1997; Mobini et al. 2002). It may be that OFC patients cannot use long-range expectancy-related information, or cannot compromise between short- and long-range expectancies to make socially and economically appropriate decisions. For animals as well, the ability to develop long-range reward expectancies for food or mating partners, which are limited in space and time, would be essential for controlling behavior, and thus for survival.
Although neuronal activities related to the motivational context are not directly associated with correct task performance, they may be essential for detecting the congruency or discrepancy between expectancy and outcome, even in cases where no reward can be expected. Motivational context-processing OFC neurons may thus serve for the acquisition, maintenance, and modification of behavioral strategies according to the outcome of responses, and thus may support cognitive operations.
LPFC receives highly processed motivational information from OFC, and highly processed cognitive information from the posterior association cortices (Barbas 1993). We showed that LPFC neurons are involved in processing not only cognitive but also motivational context information (Watanabe 1996; Watanabe et al. 2002). Considering that LPFC is the center of higher cognitive activity (Goldman-Rakic 1987), what, then, is the functional significance of LPFC neurons relating to the motivational context? They may support cognitive operations, as context-processing OFC neurons do, by serving for the acquisition, maintenance, and modification of behavioral strategies according to the outcome of responses. More importantly, the functional significance of neuronal activities related to the motivational context in LPFC may lie in their interaction with neuronal activities related to the cognitive context.
For adaptive goal-directed behavior, it is essential to integrate cognitive and motivational information. In our previous studies on working memory, we showed that LPFC neurons take part in the integration of cognitive and motivational information because they exhibit enhanced activity related to working memory when the more preferred reward is used (Watanabe 1996; Kobayashi et al. 2002). Recently, it has been indicated that individual LPFC neurons can process both cognitive and motivational context information (Barraclough et al. 2004; Kobayashi et al. 2007). Thus, LPFC neurons appear to play important roles in integrating cognitive and motivational context information. The functional significance of LPFC may, thus, lie in representing both the cognitive and motivational context as well as in coding the situation depending on the integrated cognitive and motivational context to enable adaptive goal-directed behavior. It has been suggested that some mental disorders such as schizophrenia are associated with disturbances in processing context information in LPFC (MacDonald et al. 2005). Although many discussions have been concerned with deficits in cognitive context processing in these patients, it will be important to consider their symptoms also in relation to deficits in motivational context processing, or deficits in the integration of both the cognitive and motivational context.
The authors express their thanks to Drs K. Hikosaka, M. Shirakawa, S. Kobayashi, O. Hikosaka, J. Lauwereyns, R. Kawagoe, Y. Takikawa, M. Koizumi, K. Tsutsui, and H. Takenaka for their collaboration in conducting the experiments presented in this paper. The authors are also indebted to Drs J. Lauwereyns and K. Tsutui for their critical comments on the manuscript. These studies were supported by Grants-in-Aid for scientific research from Japanese Ministry of Education, Science, Sports and Culture to M.W. and M.S., a Grand-in-Aid for target-oriented research and development in brain science from Japan Science and Technology Corporation (JST) to M.W., a Grand-in-Aid for a Precursory research for embryonic science and technology from JST to M.S., and a Grant from Human Frontier Science Program to M.S.
Conflict of Interest: None declared.