Report

Mnemonic Function of the Dorsolateral Prefrontal Cortex in Conflict-Induced Behavioral Adjustment

See allHide authors and affiliations

Science  09 Nov 2007:
Vol. 318, Issue 5852, pp. 987-990
DOI: 10.1126/science.1146384

Abstract

Our cognitive abilities in performing tasks are influenced by experienced competition/conflict between behavioral choices. To determine the role of the anterior cingulate cortex (ACC) and dorsolateral prefrontal cortex (DLPFC) in the conflict detection-resolution process, we conducted complementary lesion and single-cell recording studies in monkeys that were resolving a conflict between two rules. We observed conflict-induced behavioral adjustment that persisted after lesions within the ACC but disappeared after lesions within the DLPFC. In the DLPFC, activity was modulated in some cells by the current conflict level and in other cells by the conflict experienced in the previous trial. These results show that the DLPFC, but not the ACC, is essential for the conflict-induced behavioral adjustment and suggest that encoding and maintenance of information about experienced conflict is mediated by the DLPFC.

The Wisconsin Card Sorting Test (WCST) (1), which has been used routinely for neuropsychological assessment, is a suitable task to examine the mechanisms of the conflict-induced behavioral adjustment. In the WCST, the relevant rule and its frequent changes are not cued, and therefore the subjects face a conflict or competition between the potential matching rules. Patients with prefrontal cortex damage typically show impaired performance on the WCST and other tasks in which they should resolve conflict between potential rules or responses (24). We trained 16 macaque monkeys to perform a close analog of the WCST (figs. S1 and S2) (57). Four monkeys received bilateral aspiration lesions to the principal sulcus within the DLPFC, four other monkeys received bilateral lesions to the sulcal region within the ACC, and six other monkeys remained as unoperated controls (Fig. 1C). In the two remaining monkeys (M1 and M2), single-cell activity was recorded from the DLPFC.

Fig. 1.

Conflict-induced behavioral adjustments in control, DLPFC, and ACC groups. (A) The mean normalized STS in low-conflict (L) and high-conflict (H) trials (error bars indicate SEM). (B) The mean difference in STS, with respect to the second trial of HH versus LH trial sequences. Diamonds indicate the data from individual monkeys. (C) Black shadings show the intended lesion within the DLPFC (left) or ACC (right).

In each trial, the monkeys were required to select which one out of three different test items matched a sample, according to the currently relevant matching rule, under one of two different levels of conflict: low conflict and high conflict. In the high-conflict condition, the sample matched one of the test items in color and another test item in shape; however, it did not match the remaining test item in either color or shape (fig. S1). Therefore, the monkeys had to resolve the competition between two potential matching rules (i.e., matching by color and matching by shape) to select the target. In the low-conflict condition, the sample matched one of the test items in both color and shape, and it did not match the other two test items in either color or shape; thus, there was no conflict between matching rules. The low- and high-conflict trials were intermingled in a random order. Trial events and the feedback were similar in high- and low-conflict conditions (figs. S1 and S2).

We examined the possible role of the ACC and DLPFC in conflict-related behavioral adjustments by comparing the postoperative behavior of monkeys with bilateral lesions within the ACC (ACC group) or DLPFC (DLPFC group) (Fig. 1C) with the behavior of intact monkeys (control group). The resulting lesions were as intended in the ACC and DLPFC lesion groups (7) (figs. S8 and S9).

All groups successfully performed the WCST analog, as evidenced by their abilities to make numerous switches between matching rules per session(9.9 switches per daily session in the control group as compared to 9.6 and 9.6 switches in the ACC and DLPFC groups, respectively). For all groups, we observed that the level of conflict between matching rules influenced the monkeys' behavior in current as well as in subsequent trials.

With regard to current trials, we found that, in the control, ACC, and DLPFC groups, target selection was slowed by the presence of conflict (Fig. 1A): The speed of target selection (STS), which was calculated as the reciprocal of the time between the test item onset and the first screen touch, was slower in high-conflict trials. A three-way analysis of variance (ANOVA) [“monkey” (between-subject factor) × “conflict” (low/high; within-subject factor) × “group” (control/DLPFC/ACC; between-subject factor)] applied to the normalized STS in correct trials showed that there was a significant main effect of conflict (F1, 256 = 509, P < 1× 10–62), but the interaction between the conflict and group factors was not significant (F2, 256 = 1.9, P = 0.15) (7). When we used only one averaged (across sessions) normalized STS value for each monkey, no significant difference was seen in the magnitude of the STS modulation (i.e., the difference in STS between high- and low-conflict trials) between the control and ACC groups (t6.4 = 0.3, P = 0.5, t test) or between the control and DLPFC groups (t7.98 = 0.61, P = 0.4). These results indicate that the modulation of STS by the current conflict level was not different between control and lesion groups.

The conflict hypothesis (810) posits that the assessment of conflict and the ensuing adjustment in control should result in enhanced resolution of conflict and improved behavioral choices in subsequent circumstances where conflict arises yet again. Consequently, in order to examine whether monkeys' performance level improved when faced with repeated high-conflict situations, we compared the monkeys' behavior in high-conflict trials that followed low-conflict trials (LH trials) with that in high-conflict trials that followed high-conflict trials (HH trials). Only correct trials that were preceded by correct trials were considered in this analysis.

We observed that, in both the control and ACC groups, the STS of the second trial in HH pairings was significantly faster than that of the second trial in LH pairings; however, no such significant difference between HH and LH pairings was seen in the DLPFC group. A three-way ANOVA [“monkey” × “previous conflict” (HH/LH) × “group”] applied to the STS showed a significant main effect of previous conflict (F1, 256 = 8.4, P = 0.004) and no interaction between monkey and previous-conflict factors or among monkey, previous-conflict, and group factors (P > 0.05). However, there was a significant (F2, 256 = 3.4, P = 0.03) interaction between the previous-conflict and group factors. A follow-up post-hoc test (Tukey test) showed that there was a significant difference between the control and DLPFC groups (P < 0.05) but not between the control and ACC groups (P > 0.9). When we used only one averaged (across sessions) value for each monkey, there was a significant difference in the magnitude of the STS modulation (i.e., the difference in STS between HH and LH trials) between the control and DLPFC groups (t7.9 = 2.44, P = 0.02, one-tailed t test) but not between the control and ACC groups (t5.6 = 0.03, P = 0.5). Figure 1B shows the mean STS modulation in the control, ACC, and DLPFC groups, as well as in the two additional monkeys (without lesions) used in the single-cell recording study (M1 and M2). These results show that, in intact monkeys and also in monkeys with lesions within the ACC, the STS was modulated by the conflict experienced in the previous trial but that this modulation was impaired after lesions within the DLPFC.

Having observed that the conflict-induced behavioral adjustment was impaired in the DLPFC lesion group, we explored the underlying neural processes by recording single-cell activity in the DLPFC of two monkeys (M1 and M2). Both monkeys successfully performed the WCST analog (fig. S2) even with the additional requirement of eye fixation, and they accrued more than 10 changes in the relevant rule per day (7). For both monkeys, the STS was significantly slower in high-conflict trials in all three response directions (fig. S3A), and the STS in HH trials was significantly faster than that of LH trials in all three response directions (fig. S3B).

We found that the DLPFC cell activity differed between low- and high-conflict trials in the sample and decision periods (11). In contrast to the lesion study, in the recording study (7) (fig. S2), because different samples were consistently used in the high- and low-conflict trials (fig. S4), the monkeys could know the conflict level of the trial as soon as the sample was presented. This design allowed us to compare the activity between the two conflict levels while the eye position was fixed and before the monkey had initiated its motor response. The example cell shown in Fig. 2 showed higher activities in the low-conflict trials during the sample period, regardless of the direction of the upcoming motor response. The activity difference between low- and high-conflict trials was not related to the sample identity, because higher activities were consistently observed for each of the three samples used in low-conflict conditions, and lower activities were seen for each of the seven samples in high-conflict conditions. It was also independent from the relevant rule, because the activity was higher in low-conflict conditions irrespective of the relevant rule (fig. S5).

Fig. 2.

DLPFC cell activity represented currently experienced conflict. The leftmost peri-stimulus time (PST) histogram shows activities in low-conflict (black) and high-conflict (red) trials (bin width, 55 ms). Each column shows activities in low- and high-conflict trials that required the application of the same matching rule and responses in the same direction. Left and right vertical broken lines indicate the sample onset and the onset of test items, respectively. Samples presented in each conflict condition are shown above individual PST histograms (bin width, 20 ms). Bar graphs at bottom show the mean activity in the sample period. Only correct trials were included. The recording area covered both dorsal and ventral banks of the principal sulcus.

When we applied a two-way ANOVA (“conflict” × “rule”) to the activity in correct trials of 146 recorded cells, we found that 22 cells (15%) and 30 cells (21%) showed a significant (P < 0.05) main effect of conflict in the sample and decision periods, respectively. Only two (sample period) and four (decision period) of the cells showed a significant interaction between the rule and conflict factors, indicating that the activity difference between low- and high-conflict trials was independent from the relevant rule (fig. S5). The activity was higher in high-conflict trials in some cells (5/22 and 14/30 cells in the sample and decision periods, respectively) and higher in low-conflict trials in others (fig. S5). Further analyses showed that the activity difference between low- and high-conflict trials was not due to selectivity for physical features of the sample stimuli (supporting online material).

The modulation of behavior in the current trial by the level of conflict experienced in the previous trial requires a purported system to hold the memory of experienced conflict across trials. The neural representation of such a memory is expected to appear in the period preceding the actual start of each trial. Thus, we also compared the cellular activity between LH and HH trials in the fixation period (a period in which the eye and hand positions of the monkey were fixed). A two-way ANOVA (“previous conflict” × “rule”) showed that out of the 146 cells, 15 cells had a significant (P < 0.025) main effect of previous conflict or its interaction with rule. This number of cells is significantly (P = 0.004, binomial test) greater than the number of cells that could be obtained by chance. The activities of two examples of these cells are depicted in Fig. 3, A and B. The activity in the fixation period was higher in HH trials in 4/15 cells and higher in LH trials in 11/15 cells. Figure S6 shows the mean normalized activities of the 15 cells in preferred (i.e., whichever trial sequence, LH or HH, that had the higher activity for that cell) and nonpreferred trial sequences. At the cell population level, the activity difference between LH and HH trials could be seen during and even before the fixation period. These results suggested that DLPFC cell activities maintained, across trials, the information of the previously experienced conflict level.

Fig. 3.

DLPFC cell activity represented conflict experienced in the previous trial. (A and B) Activities in LH trials (blue) and those in HH trials (pink) are shown for two cells [results are shown for one cell in (A) and for the other cell in (B)]. The mean activities are aligned at sample onset. Only activities in correct trials that were preceded by correct trials were included. The P values show the significance level of activity difference in the fixation period between HH and LH trials (bin width, 55 ms).

Our results showed marked similarity in conflict-induced behavioral adjustment between monkeys and humans (8, 1216). Previous studies (810, 1723) have (i) shown activation of the ACC when human subjects face a conflict between behavioral choices and (ii) reported impairment of conflict-related behavioral modulations in patients with ACC lesions (24), which suggests the involvement of the ACC in the conflict detection-resolution process. However, humans with lesions that include the ACC do not necessarily show impairment in conflict-related behavioral modulations (25). We found that lesions within the ACC of the monkey did not impair the behavioral modulations that occurred after experiencing conflict in the current or previous trial. Our results indicate that, even without a functional ACC, the behavioral effects of experienced conflict can still be realized.

We observed that lesions within the DLPFC impaired the conflict-induced behavioral modulations. This suggests that in contrast to the ACC, the DLPFC plays a crucial role in mediating the behavioral effects of experienced conflict. Previous studies have shown that lesions in the principal sulcus cortex lead to impairments in short-term maintenance of spatial information (2628). We show that the indispensable role of principal sulcus cortex in cognition extends beyond short-term storage of spatial information to support conflict-induced behavioral adjustment.

We found that DLPFC cells represented the conflict level in the current trial independently from other aspects of task, suggesting that the currently experienced conflict is encoded in DLPFC cell activity as a distinct variable. These results provide support to the hypothesis (17, 18) that the DLPFC is involved in the detection of conflict.

We also found that the DLPFC cell activity maintained information regarding previously experienced conflict, and, in turn, the conflict-induced behavioral adjustment was impaired by DLPFC lesion. Based on these complementary findings, we can conclude that this region of the DLPFC is necessary for conflict-induced behavioral adjustment, and we posit that the cellular activities that we have observed in the DLPFC represent the conflict level and bring this information to bear over time to the following trial to be used in adjustment of cognitive control (29). In a changing environment, we need to optimize the usage of our limited cognitive resources, and the DLPFC might support an adaptive and dynamic tuning of our cognitive control processes by maintaining information of recent cognitive challenges.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1146384/DC1

Materials and Methods

SOM Text

Figs. S1 to S9

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article