Report

A Role for the Macaque Anterior Cingulate Gyrus in Social Valuation

See allHide authors and affiliations

Science  01 Sep 2006:
Vol. 313, Issue 5791, pp. 1310-1312
DOI: 10.1126/science.1128197

Abstract

Complex human social interaction is disrupted when the frontal lobe is damaged in disease, and in extreme cases patients are described as having acquired sociopathy. We compared, in macaques, the effects of lesions in subdivisions of the anterior cingulate and the orbitofrontal cortices believed to be anatomically homologous to those damaged in such patients. We show that the anterior cingulate gyrus in male macaques is critical for normal patterns of social interest in other individual male or female macaques. Conversely, the orbitofrontal cortex lesion had a marked effect only on responses to mildly fear-inducing stimuli. These results suggest that damage to the anterior cingulate gyrus may be the cause of changes in social interaction seen after frontal lobe damage.

Normal patterns of human social interaction are disrupted after ventromedial frontal lobe damage (1). The absence of normal social behavior may be so extreme that patients are described as suffering from acquired sociopathy (2). The lesions are not restricted to ventromedial frontal cortex but encompass laterally adjacent orbitofrontal cortex and medially adjacent anterior cingulate cortex (ACC). Damage to just one anatomical subdivision may explain the patients' impaired social cognition. We assessed how selective lesions of ACC gyrus (ACCG), ACC sulcus (ACCS), or lateral orbital and ventral prefrontal cortex (PFv+o) affect the way macaques value social information.

The possibility that orbitofrontal damage underlies impaired social interaction has received attention not just as a result of patient studies but because, in macaques, circumscribed orbitofrontal lesions lead to altered emotional responsiveness to stimuli that normally induce mild fear (3, 4). The deficit may be due to an inability to predict the reinforcement consequences of a stimulus or of another individual. Orbitofrontal lesions impair visual discrimination reversal learning, which requires the modification of associations between stimuli and primary reinforcers (5). It has been argued that the flexible assignment of reinforcement values to stimuli is a prerequisite for emotion and social behavior (6). Patients with ventromedial frontal lesions also perform poorly on visual discrimination reversal tasks (7). On the other hand, neuroimaging studies have shown that the ACC is active when human participants engage in social interaction, although its contribution has been unclear (811). Despite their proximity, the connections of ACC and orbitofrontal cortex are distinct in both man and monkey, and so their roles in social behavior may be correspondingly distinct (1214).

The three lesions are summarized in Fig. 1 (and figs. S1 to S3). The PFv+o lesion was similar to one previously used to remove the principle target region of visual connections within the frontal lobe (15). The ventromedial region, which has connections with both PFv+o and ACCG regions (13), was excluded to better assess each area's independent contribution to social behavior and emotion. Three animals received each lesion, and performances were compared with those of four unoperated controls.

Fig. 1.

Summary of (A) intended ACCG lesion, (B) intended ACCs lesion (in both cases medial and lateral views of the brain are shown in top and bottom images, respectively), and (C) intended PFv+o lesion (ventral and lateral views are shown in top and bottom images, respectively). Lesion locations are shown on views adapted from the atlas of Paxinos et al. (29).

Measurements were made of latencies to pick up food items in the presence of fear-inducing stimuli (toy snakes) in experiment 1, social stimuli (short films of other macaques) in experiment 2, or neutral control objects (fig. S4b). The latencies indexed the macaques' assessment of the value of obtaining further information about the stimulus before reaching and reflected their relative valuation of the stimulus in contrast to the incentive value of the food. Normally macaques react fearfully to snake stimuli even in the absence of prior experience (16), but such responses to the type of stimuli used in experiment 1 have been shown to be disrupted by lesions of more-medial orbitofrontal cortex (3, 4). Male macaques prefer to view images of females or males of high social status, similar to those used in experiment 2, and will forgo small food rewards to do so (17, 18). Macaques may estimate dominance of even unknown males from their appearance because of its correlation with size, age, and maturity (19).

Animals were tested in the Wisconsin General Testing Apparatus (WGTA) (fig. S4a). On each trial, they were given 30 s to pick up a small food item. A 30-s intertrial interval preceded the onset of the next trial. On each day, animals were exposed to five different stimuli of possible social or emotional importance and 10 neutral objects that were placed in a plastic box beneath the food. The test was repeated for 4 days with the same stimuli and for 4 further days with a new stimulus set.

The four groups did not respond in the same way to the mild fear stimuli in experiment 1 (Fig. 2A). A three-way analysis of variance (ANOVA) comparing reaching latencies of the four groups to the two snake stimuli during four presentations of each revealed significant differences [group by stimulus interaction: F(3, 9) = 9.725, P = 0.003]. Although control animals were not always fearful of a static rubber snake, they were consistently reluctant to take food from above a moving toy snake, but the case was significantly different for both the PFv+o group [stimulus by group interaction (control or PFv+o): F(1, 5) = 31.141, P = 0.003] and the ACCs group [stimulus by group interaction (control or ACCS): F(1, 5) = 18.953, P = 0.007]. The ACCG animals varied in their responsiveness to the moving snake, with one individual, G2, within the control range, so there were no significant differences between ACCG and controls (P > 0.1). The difference between the ACCG and PFv+o group reached borderline significance [F(1, 4) = 7.548, P = 0.052].

Fig. 2.

Response to mild fear and social stimuli in experiment 1. (A) Median latency to retrieve food in the presence of a mild fear-inducing stimulus (static or moving snake). (B) Median latency to retrieve food in the presence of social stimuli in experiment 2, either unknown human actors (left) or macaques (right). (C) Median latency to retrieve food in the presence of neutral control stimuli. Symbols indicate scores for each individual.

The four groups also differed in responsiveness to the social stimuli in experiment 2, but now the effect was due to the distinctive pattern of behavior in the ACCG group. There was agreement among normal male controls as to which social stimuli of other macaques were more interesting (17); reaching latencies varied significantly depending on stimulus [F(1, 4) = 17.493, P = 0.018] (Fig. 2B). They were slower to pick up food in the presence of a large staring male, a female macaque with visible sexual perineal swellings, and a midsized macaque making affiliative lip-smacking gestures (Fig. 2B, right) than was the case with other less socially salient images. The same effect was not, however, seen in all groups [main effect of group: F(1, 5) = 14.113, P = 0.013] (Fig. 2B). The ACCG group remained uninterested in any of the images of other macaques, and the patterns of change in interest with stimulus type seen in the control group were absent (Fig. 2B). ACCG performance was significantly different to both that of controls [main effect of group: F(1, 5) = 14.113, P = 0.013] and that of the PFv+o group [main effect of group: F(1, 4) = 8.651, P = 0.042], although the contrast with the ACCS group did not reach significance (P > 0.1). By contrast, comparable modulations of reaching latencies were seen in PFv+o and ACCS groups as in the controls, and there were no significant differences between any of these three groups (P > 0.1). A complementary analysis of the expression of social behaviors found that ACCG animals produced fewer social responses than either control, PFv+o, or ACCS animals (Materials and Methods and fig. S5). Previous investigations have assessed macaques' emotional and social responsiveness by exposure to unfamiliar humans (5). Although PFv+o latencies were generally comparable to those of control animals, they were sometimes shorter in response to one of the two unfamiliar humans, and this prevented ANOVAs involving this group from reaching significance (Fig. 2B). Nevertheless a comparison of the ACCG and control groups once again revealed significant differences [main effect of group: F(1, 5) = 7.030, P = 0.045].

To further test the robustness of the distinction drawn between ACCG and PFv+o group's diminished responsiveness to mild fear and social stimuli, respectively, we used a four-way ANOVA to compare the two groups' latencies for the two exemplars of fear stimuli and the two most effective exemplars of social stimuli (female macaque and dominant male macaque) across the four presentations. A highly significant interaction between stimulus category, stimulus exemplar, and group was found on comparing ACCG and PFv+o groups [F(1, 4) = 23.494, P = 0.008]. The difference was only marginally significant [F(1, 4) = 4.954, P = 0.09] when ACCG and ACCS groups were directly compared.

Significant reductions in ACCG reaching latency were only observed on trials involving social stimuli but not when control objects were presented (Fig. 2C). Although group differences were found in responses to the first set of neutral junk objects and moving inanimate visual patterns [F(9, 27) = 2.817, P = 0.046], these transpired to be due to quicker latencies in the PFv+o group in comparison with those of controls [F(1, 5) =30.075, P = 0.003] and not quicker latencies in the ACCG group (P > 0.1). Significant interactions [F(1, 5) > 6.686, P < 0.05] between group (ACCG versus control) and stimulus (neutral versus each of the four most effective social stimuli) confirmed the specificity of the ACCG effect to social stimuli.

The orbitofrontal involvement in emotional responses to stimuli, such as those that induced mild fear (Fig. 2A), may be related to its role in representing expectations about the reinforcement outcomes associated with stimuli because lesions also impair visual discrimination reversal learning (37, 20). Unlike the orbitofrontal cortex, the ACCG contribution to social behavior may be unrelated to the representation and learning of outcome expectations. Indeed, both ACCG and ACCS groups were unimpaired on a reversal learning test similar to the one previously used with orbitofrontal lesions (P > 0.1) (Fig. 3).

Fig. 3.

Visual discrimination reversal learning task. (A) Animals were presented with two images on the screen. Selection of one stimulus led to a reward, whereas selection of the other did not. Animals learned the task until they performed above a criterion of 90% correct performance. The reward contingencies were then reversed, and animals had to relearn the new stimulus reward association to a criterion of 90%. Animals completed nine full reversals after initial learning. (B) Mean (±SEM) number of errors to criterion per reversal.

It has previously been observed that macaques spend less time in proximity with one another after large ACC lesions (21). The present results suggest that this may be a consequence of a lowered valuation of social information. The specificity of the ACCG lesion was apparent when its effects were compared with those of the PFv+o lesion on several tests involving fear and social stimuli. The ACC region critical for mediating the valuation of social stimuli appears to be the ACCG immediately rostral and dorsal to the genu of the corpus callosum and includes areas 32 and rostral area 24. The ACCG lesion did not include subcallosal area 25 ventral to the genu of the corpus callosum (Fig. 1 and fig. S2), which has been linked with mood change and depression (22). The ACCG lesions sometimes disrupted parts of ACCS, but the damage was neither complete nor bilateral. Moreover, it is difficult to attribute the deficit to the ACCS because complete ACCS lesions only affected the mild fear task significantly and not the social task; a pattern of change opposite to that seen in the ACCG group. A note of caution is warranted; although ACCG and ACCS groups exhibited opposite patterns of change when contrasted with controls, often only marginally significant differences were found when the two groups were directly compared with one another, perhaps because of the small group sizes and some interindividual variability that might be related to damage to adjacent white matter. Despite these caveats, the absence of consistent impairment after ACCS lesions should not be overlooked. The ACCS lesion is not simply ineffective and is known to cause clear changes in reward guided choice and decision-making (23).

Signal changes have been recorded in a homologous area when human subjects perform tasks requiring consideration of other individuals or engagement in social interaction (811). The prisoner's dilemma game, ultimatum game, and theory of mind tasks used in neuroimaging studies are complex, and changes are recorded beyond the ACCG, making it difficult to ascertain its particular importance. ACCG changes have also been correlated with autistic social interaction impairments (24). The present results demonstrate an essential and causal role for ACCG in valuing social information when the implications of another individual's presence must be taken into account before acting.

Normal emotion and social behavior is dependent on the integrated interactions of several brain areas, including orbitofrontal cortex and amygdala (25, 26), but the present study suggests the ACCG has an important complementary role. Orbitofrontal cortex may be essential for emotion and social behavior as a consequence of its involvement in the representation of the outcome expectancies that guide choice and learning (4, 5, 20) and its general role in encoding the relative economic values of a wide range of both social and nonsocial stimuli (4, 27). Damage to any one of these interconnected areas early in development may lead to widespread neurobiological and behavioral changes (28).

Supporting Online Material

www.sciencemag.org/cgi/content/full/313/5791/1310/DC1

Materials and Methods

Figs. S1 to S5

References

References and Notes

View Abstract

Navigate This Article