Prefrontal–Amygdala Pathways for Object and Social Value Representation

Abstract

This special focus article was prepared to honor the memory of our National Institutes of Health colleague, friend, and mentor Leslie G. Ungerleider, who passed away in December 2020, and is based on a presentation given at a symposium held in her honor at the National Institutes of Health in September 2022. In this article, we describe an extension of Leslie Ungerleider's influential work on the object analyzer pathway in which the inferior temporal visual cortex interacts with the amygdala, and then discuss a broader role for the amygdala in stimulus–outcome associative learning in humans and nonhuman primates. We summarize extant data from our and others' laboratories regarding two distinct frontal–amygdala circuits that subserve nonsocial and social valuation processes. Both neuropsychological and neurophysiological data suggest a role for the OFC in nonsocial valuation and the ACC in social valuation. More recent evidence supports the possibility that the amygdala functions in conjunction with these frontal regions to subserve these distinct, complex valuation processes. We emphasize the dynamic nature of valuation processes and advocate for additional research on amygdala–frontal interactions in these domains.

INTRODUCTION

Several lines of investigation emerged from the “two visual systems” concept put forth by Leslie Ungerleider and Mortimer Mishkin (Ungerleider & Mishkin, 1982). Work on the ventral visual stream, or “object analyzer pathway,” spawned investigations into how object information could become attached to affect, reward, and action selection. Many experimental studies in macaques pointed to interactions of the inferior temporal visual cortex and the amygdala as being essential for object–reward learning (Gaffan, Gaffan, & Harrison, 1988; Spiegler & Mishkin, 1981), and to interactions of the inferior temporal visual cortex with both the OFC and the ventral lateral prefrontal cortex (VLPFC) being essential for visually cued response selection, including visual–visual and visuomotor learning and retrieval (Browning & Gaffan, 2008; Browning, Easton, & Gaffan, 2007; Bussey, Wise, & Murray, 2002; Tomita, Ohbayashi, Nakahara, Hasegawa, & Miyashita, 1999; Parker & Gaffan, 1998). The role of the inferior temporal cortex, OFC, and VLPFC in the learning and use of abstract rules to guide behavior has been reviewed elsewhere (Eldridge, Hines, & Murray, 2021; Mansouri, Freedman, & Buckley, 2020).

This article presents an extension of this seminal work into the area of value-based decision-making in social and nonsocial contexts, with an emphasis on the amygdala. Humans and monkeys make decisions about what to value independently of decisions about who to value. Decisions about what to value typically involve nutrient selection, including foods and fluids, and the selection of other resources that are important to animals. Decisions about who to value include who to trust, who to cooperate with, who to compete with, who to mate with, and so on. Knowing what and who to value requires learning about the cues and signs that predict beneficial outcomes. When navigating a sensory-rich environment, it is necessary to be able to distinguish a conspecific from a rewarding object, not only by visual discrimination and categorization alone, but also through relative valuation. The experimental work in monkeys is guided by the knowledge that the inferior temporal cortex is reciprocally related to the basolateral amygdala, which is, in turn, reciprocally connected to the ventral and medial sectors of the PFC. The basolateral amygdala also gives rise to projections to the central nucleus of the amygdala, which projects to the lateral hypothalamus, basal forebrain, and periaqueductal gray, among other regions. Both connections from classic, feature-sensitive parts of the inferior temporal cortex (Webster, Ungerleider, & Bachevalier, 1991), as well as face patches, reach the amygdala (Liu et al., 2022; Grimaldi, Saleem, & Tsao, 2016).

We suggest that diverging pathways from the amygdala to distinct areas of the PFC guide decisions regarding distinct targets of action (the “who” or “what” in question). We first address the evidence supporting the amygdala's role in associative learning and choice behaviors in both social and nonsocial contexts. We then describe how the amygdala and two regions of the frontal cortex, specifically the OFC and the ACC, functionally interact to represent what to value separately from who to value. For a recent review on the contributions of prefrontal–amygdala interactions to behavior and cognition more broadly, we refer readers to Murray and Fellows (2022).

SOCIAL AND NONSOCIAL VALUATION: AMYGDALA

Converging evidence points to a role for the amygdala in ascribing value to conspecifics and objects associated with rewarding outcomes. One clinical case—patient S.M., who has Urbach–Wiethe disease, a rare genetic condition that results in a constellation of phenotypic traits, including congenital calcification of the amygdalae—has been instrumental in our understanding of amygdala function in humans. In the social domain, S.M. exhibits decreased gaze to the eye region of faces in free-viewing paradigms (Kennedy & Adolphs, 2011; Adolphs et al., 2005), and this is true for monkeys with selective amygdala damage as well (Dal Monte, Costa, Noble, Murray, & Averbeck, 2015). Selective amygdala damage in monkeys has further been shown to eliminate viewing preferences for faces (real or illusory) altogether (Taubert et al., 2018). S.M. also exhibits decreased interpersonal space in her social interactions (Kennedy, Gläscher, Tyszka, & Adolphs, 2009) and is more trusting of others than a comparison group (Adolphs, Tranel, & Damasio, 1998). In a similar vein, monkeys with selective amygdala lesions are more likely to approach unfamiliar conspecifics relative to intact monkeys (Emery et al., 2001). Finally, fMRI activations in human amygdala reflect social rank and amygdala volume is positively related to the ability to learn social hierarchy (Kumaran, Melo, & Duzel, 2012); amygdala volume in monkeys has been reported to be related to an individual's social status (Noonan et al., 2014). The precise role of the amygdala in social cognition has not been specified. We return to this topic later in the Social Valuations section.

In paradigms of nonsocial, value-based decision-making, S.M. and another participant with amygdala damage showed differences from controls in switching behavior on a stimulus–reward reversal task and showed a pronounced change in activations of the ventral PFC during learning (Hampton, Adolphs, Tyszka, & O'Doherty, 2007). In reinforcement learning paradigms, monkeys with amygdala lesions show both a slowed learning rate and a reduced consistency of choices relative to controls (Basile et al., 2023; Costa, Dal Monte, Lucas, Murray, & Averbeck, 2016). In a study of patients with Urbach–Wiethe disease (Rosenberger et al., 2019), in a nonsocial learning task, learning about positive outcomes was intact, whereas learning about negative outcomes was impaired.

Few studies have examined the contributions of the amygdala to both social and nonsocial decision-making in the same subjects. In a neurophysiological study of social and nonsocial value in nonhuman primate amygdala, the same neuronal ensembles that encoded social hierarchical rank were also found to encode information about rewards and outcome value more broadly (Munuera, Rigotti, & Salzman, 2018). These findings indicate that individual neurons in the primate amygdala encode value representations in a domain-general way and may participate in appropriate outcome selection in both social and nonsocial contexts. In addition to studies involving patient S.M., mentioned above, a group of patients with Urbach–Wiethe disease were found to invest more money than controls (van Honk, Eisenegger, Terburg, Stein, & Morgan, 2013) and to have trouble learning from prior experiences about who to trust (Rosenberger et al., 2019) in variations of economic trust games. The findings are consistent with a role for the amygdala in both social and nonsocial decision-making. The findings also suggest the possibility that, in contexts that involve learning about the value of conspecifics, the amygdala may be playing a crucial role in learning about who to trust by weighting prospective losses over prospective gains. In other words, an intact amygdala may be necessary for driving adaptive defensive social responses during periods of uncertainty.

Nonsocial Valuations

Early studies of amygdala function in macaques pointed to a role for this region in linking objects with reward. Tasks such as win-stay, lose-shift, and object reversal learning required animals to flexibly make and break stimulus–reward associations, and were therefore used to assess this associative function. Although early studies in macaques found that aspirative lesions of the amygdala led to impairments in win-stay, lose-shift (Spiegler & Mishkin, 1981), and object reversal learning (Jones & Mishkin, 1972), these findings were overturned when monkeys with selective, excitotoxic (fiber-sparing) amygdala lesions were studied using the same tasks (Izquierdo & Murray, 2007; Stefanacci, Clark, & Zola, 2003). The reason for the difference in outcome of the studies with excitotoxic versus aspirative lesions is likely the inadvertent disruption of cortico-thalamic and cortico-cortical fibers by amygdala aspirative lesions (Muñoz, Mishkin, & Saunders, 2009; Goulet, Doré, & Murray, 1998; Murray, 1992). This realization led to further exploration of the possible contribution of the amygdala to object–reward association using other tasks, and to a reliance on excitotoxic lesion methods, especially for the study of subcortical structures like the amygdala.

To probe more deeply into the possible role of the amygdala in reward learning, we employed the reinforcer devaluation task (Málková, Gaffan, & Murray, 1997)—referred to hereafter as the devaluation task—which had previously been used in work with rodents with and without amygdala lesions (Hatfield, Han, Conley, Gallagher, & Holland, 1996; Balleine & Dickinson, 1991; Adams & Dickinson, 1981). Unlike visual discrimination learning and reversal tasks, which require monkeys to link objects with food delivery and can probably be acquired by several mechanisms, the devaluation task requires monkeys to link objects with food value and tests whether they have the capacity to adapt to changes in food value.

In the devaluation task, monkeys are allowed to choose between pairs of neutral objects wherein each object covers a unique food reward in a small well located directly beneath the object. Monkeys first complete a visual discrimination task, which gives them the opportunity to learn to associate each object with a specific food reward. Half the rewarded objects cover one type of food, and the other half cover a different type of food. The key experimental manipulation is selective satiation: Animals are given ad libitum access to one type of food until they stop eating. Probe tests conducted after selective satiation reveal the ability of monkeys to link objects with the current, updated values of the foods. Importantly, each trial of the probe test involves a different pair of objects; monkeys are offered a choice between two different rewarded objects, one overlying the devalued food and the other overlying the nondevalued food. Thus, choices in the probe test are made before any relearning can occur.

Intact monkeys that are pre-fed with one of the two foods will shift their choices away from the object associated with the devalued food, in favor of the remaining object that predicts the nondevalued food. Monkeys with bilateral amygdala lesions show less of a shift away from objects that predict the devalued food (Izquierdo & Murray, 2007; Machado & Bachevalier, 2007; Málková et al., 1997). Thus, monkeys with amygdala lesions are impaired in linking object information with the current value of the food. Control procedures have ruled out alternative interpretations of the data. Monkeys with amygdala lesions have good visual discrimination abilities, their food preferences are unchanged (cf. Agustín-Pavón, Parkinson, Man, & Roberts, 2011; Machado & Bachevalier, 2007), and their satiety mechanisms are intact. The latter point is underscored by the fact that monkeys with amygdala lesions shift their choices away from the devalued food following selective satiation in a “food-only” control task (i.e., a task in which monkeys make visual choices between devalued and nondevalued foods).

Because of the strong evidence for an amygdala contribution to stimulus–reward-value associations, we continued to use this task to probe the neural circuits underlying nonsocial valuations. Throughout this article, we will use “valuation” as a shorthand for “representation of value.” Later, in the Nonsocial Valuation: Amygdala–Prefrontal Interactions section, we discuss a set of findings exploring the idea that the amygdala needs to interact with the PFC to acquire and update information about the value of environmental stimuli.

Social Valuations

As reviewed above, monkeys exhibit several behaviors that might be classified as indexing “social valuation.” Macaques view the eye region of the face of conspecifics longer than other face parts (Dal Monte et al., 2015; Leonard, Blumenthal, Gothard, & Hoffman, 2012), will preferentially view faces over objects (Taubert et al., 2018), and will forego water (Deaner, Khera, & Platt, 2005) or delay food retrieval (Rudebeck, Buckley, Walton, & Rushworth, 2006) to view visually presented images of conspecifics. All these findings strongly suggest that primates value the opportunity to gather information about conspecifics, and the latter two studies provide direct evidence in support of this idea. Under the umbrella of social valuation, we include related aspects of social cognition, including the orientation of selective attention to socially relevant visual stimuli, affective reactions and responses to those stimuli, subjective assessments of those stimuli, and inherent interest in social signals.

To examine the neural circuits underlying social valuation, Rudebeck and colleagues (2006) employed a task that could be easily implemented in the laboratory, and provided a quantitative measure. Specifically, the task allowed macaques to approach and retrieve food and—at the same time—to view short videos of conspecifics, moving patterns, or other items. On each trial, only one video was presented, and the dependent measure was food-retrieval latency. Thus, each trial presented a conflict between whether to engage in immediate food retrieval or delay food retrieval to view the videos. Controls exhibited longer food-retrieval latencies on trials with videos of conspecifics relative to trials with videos of moving patterns (Rudebeck et al., 2006). We discuss the circuitry underlying this behavior later in the Social Valuation: Amygdala–Prefrontal Interactions section.

NONSOCIAL VALUATION: AMYGDALA–PREFRONTAL INTERACTIONS

The amygdala does not function in isolation. Using the same logic of disconnection studies that had been employed by Ungerleider and Mishkin (Ungerleider & Mishkin, 1982), as well as other neuropsychologists, we probed the neural circuits that were essential for performing the devaluation task. Crossed-surgical disconnections involve removal of one structure in one hemisphere and a different structure in the other hemisphere. Typically, the two regions possess intrahemispheric but not interhemispheric connections. As a result, the crossed lesions produce a functional disconnection of the two brain regions (in the case of interhemispheric connections between the two regions, a section of the corpus callosum is also required). Using crossed-surgical disconnection experiments in monkeys, we found that the amygdala needs to functionally interact with both the OFC (Baxter, Parker, Lindner, Izquierdo, & Murray, 2000) and the medial portion of the mediodorsal (MD) thalamus (Izquierdo & Murray, 2010) in performing this function. These studies indicate that monkeys care about the value of the food and that the amygdala, medial MD thalamus, and OFC need to interact in this process to update food value. Bilaterally symmetrical damage to the OFC (Rudebeck, Saunders, Lundgren, & Murray, 2017; Rudebeck, Saunders, Prescott, Chau, & Murray, 2013; Machado & Bachevalier, 2007; Izquierdo, Suda, & Murray, 2004) or medial MD thalamus (Mitchell, Browning, & Baxter, 2007) also produces impairments in updating stimulus–outcome value, as do reversible pharmacological inactivations of these regions (Wicker, Turchi, Malkova, & Forcelli, 2018; Murray, Moylan, Saleem, Basile, & Turchi, 2015; West et al., 2012). Notably, damage to VLPFC, which, like OFC, receives visual sensory inputs from inferotemporal cortex, yields no such impairment (Rudebeck et al., 2017; Baxter, Gaffan, Kyriazis, & Mitchell, 2009).

Devaluation tests have been performed in humans (Howard & Kahnt, 2017; Tricomi & Lempert, 2015; Tricomi, Balleine, & O'Doherty, 2009; Valentin, Dickinson, & O'Doherty, 2007). In the only human lesion study of its kind, carried out by Reber and colleagues (2017), human participants with and without damage to the orbital and medial frontal cortex combined were tested in a manner similar to that described for monkeys. Participants learned that specific foods were associated with images on a monitor and that they could earn the foods by selecting those images. After learning the image–food pairings, the participants were sated on one food, reported their ratings on how much they valued the foods, and then made choices between the images on a series of trials. After selective satiation, participants with damage to the orbital and medial aspects of the frontal cortex reported a marked decrease in subjective pleasantness ratings for the sated food. In other words, like sated monkeys and control participants, sated participants with orbital and medial frontal cortex no longer preferred the sated food. Unlike controls, participants with orbital and medial frontal cortex damage, like monkeys with orbital frontal cortex (or amygdala) lesions, also continued to choose the objects linked to the sated food. Interestingly, the patients failed to reduce the choices that they no longer deemed valuable by self-report, revealing a disconnection of knowledge and behavior. These results are in line with those based on functional imaging, and indicate a role for OFC in representing not only the value of expected rewards but also their identity (Howard & Kahnt, 2021).

One study in monkeys that informs amygdala–frontal interactions in stimulus–reward valuation employed physiological (single-unit recording) methods combined with amygdala lesions. Monkeys first learned a value-based decision-making task. Monkeys chose between combinations of two images that each predicted different quantities of juice (0–8 drops of juice). Through trial and error, they learned to choose the image that predicted the greater amount of juice. Rudebeck and colleagues (2013) then recorded the activity of neurons in OFC and ACC before and after monkeys received amygdala lesions. OFC neurons showed a graded increase (or in some neurons, a decrease) in activity as a function of expected reward magnitude. These results were as expected, based on the work of other investigators (Padoa-Schioppa, 2011; Kennerley, Dahmubed, Lara, & Wallis, 2009). Following amygdala damage, monkeys were still able to select the image that predicted the greater amount of juice in the task. However, the percentage of neurons in OFC but not ACC that encoded outcome value decreased significantly. The loss of amygdala input affected not only the encoding of anticipated reward value before reward delivery, but also reward value when the monkey received the reward.

These findings show, at a minimum, that: (1) OFC neurons signal the value of anticipated and received resources and (2) that amygdala inputs to frontal cortex are essential for maintaining familiar stimulus–reward-value associations. These and other findings therefore suggest that the OFC is representing the value of the resources available in the environment based on visual sensory input that predicts those resources (e.g., when walking down the cereal aisle of the supermarket or passing by the green siren icon of Starbucks). Because the amygdala arose in early vertebrates and primate granular OFC arose in early primates, it seems likely that functions from older regions (e.g., amygdala) have been integrated into newer regions (e.g., granular OFC) to allow for adaptive strategies for foraging in complex or cluttered environments. Because granular OFC receives converging gustatory, olfactory, and visual sensory inputs, we propose that amygdala–OFC interactions in our primate ancestors conferred an advantage in using multisensory cues to identify and forage for resources (i.e., “what” to value).

Another study in monkeys that informs amygdala–cortical interactions involves fMRI combined with amygdala lesions. Hadj-Bouziane, Bell, Knusten, Ungerleider, and Tootell (2008) found that when monkeys were allowed to passively view images of monkey faces, the amygdala and inferior temporal cortex showed greater activation in response to facial expressions of emotion relative to neutral faces. They referred to this phenomenon as the “valence effect.” On the basis of a study in humans (Vuilleumier, Richardson, Armony, Driver, & Dolan, 2004) that reported lack of activations in visual cortex after medial temporal lobe damage, we decided to ask whether the amygdala was responsible. In our study, monkeys with selective amygdala lesions were shown the same fMRI task involving passive viewing of faces and, as predicted, we found a reduction in activations in visual inferior temporal cortex with facial expressions of emotion; in other words, a reduced valence effect as a result of amygdala damage was observed (Hadj-Bouziane et al., 2012).

Taken together, the results from the value-based decision-making study (Rudebeck, Mitz, Chacko, & Murray, 2013) and the face viewing study (Hadj-Bouziane et al., 2012) suggest the possibility that the amygdala is contributing to the maintenance of representations in cortex, especially representations of special biological importance that are updated in accord with current biological wants and needs. The effects of this influence would depend on what kind of information each of the target brain areas is representing. Loss of amygdala input removes these influences. Although primates without an amygdala behave normally in many circumstances, their behavior lacks a clear capacity to relate stimuli to their relative value at any given moment.

SOCIAL VALUATION: AMYGDALA–PREFRONTAL INTERACTIONS

In studying valuation systems, we wanted to move beyond object valuation to social valuation, and to extend our understanding of the neural circuitry supporting valuation systems. Several lines of evidence suggest that the ACC plays a role in social cognition and prosociality. Functional imaging studies in humans strongly implicate the ACC in social cognition, specifically when tracking and updating information about the motivations of others (Apps, Rushworth, & Chang, 2016; De La Vega, Chang, Banich, Wager, & Yarkoni, 2016). Relative to intact monkeys, monkeys with damage to the ACC spent decreased time in proximity with conspecifics (Hadland, Rushworth, Gaffan, & Passingham, 2003). In addition, monkeys with ACC lesions showed decreased latencies to retrieve food rewards in the presence of videos of social stimuli (Rudebeck et al., 2006). Furthermore, in a social reward allocation task in which an actor monkey can choose to give juice to a recipient monkey or withhold the juice, actor monkeys with bilateral, excitotoxic lesions of ACC were impaired in making prosocial choices (Basile, Schafroth, Karaskiewicz, Chang, & Murray, 2020). Altogether, the ACC seems to be important for mediating prosocial behaviors such as approach and reciprocity.

Given the foregoing information, together with knowledge that the amygdala is important for both social and nonsocial valuation, we asked whether amygdala–frontal interactions respected these domains (Pujara, Ciesinski, Reyelts, Rhodes, & Murray, 2022). To examine the contributions of amygdala interactions with OFC and ACC in object and social valuations, respectively, we employed two tasks and two groups of lesioned monkeys, compared with intact controls. For our social valuation task, we used the conflict task described earlier that pits “approach” to retrieve food against interest in viewing social or nonsocial videos, as in Rudebeck and colleagues (2006). For our nonsocial valuation task, we used the devaluation task, described in the Nonsocial Valuations section. Monkeys in the experimental groups received either crossed disconnection surgeries of the amygdala and OFC (A × OFC) or crossed lesions of the amygdala and ACC (A × ACC). In the social valuation task, a trial is initiated by the experimenter, who raises the opaque screen of the testing apparatus. The monkey had previously been trained to reach out and take a piece of food on top of a Plexiglas box once the screen is lifted. The latency at which the monkey takes the food is an indicator of relative interest in the presence of another salient stimulus, which in the current experiment were video clips of conspecifics (a submissive monkey, a staring dominant monkey, a monkey displaying an affiliative lip-smacking behavior, and a female “presenting” her perineum [a secondary sexual characteristic of many anthropoid species]) shown on a monitor directly behind the food.

Intact (control) monkeys in this task showed preferences for the social stimuli, taking longer to reach for the food in the presence of the social videos relative to nonsocial videos. By comparison, monkeys with disconnection of the amygdala and ACC were faster to reach for the food (Pujara et al., 2022). Monkeys with disconnection of the amygdala and OFC behaved like the controls, showing a longer latency to retrieve the food in the presence of socially relevant stimuli. The devaluation task yielded the converse pattern of results. On this nonsocial valuation task, monkeys with the disconnection of the ACC and amygdala performed as well as controls, whereas monkeys with the disconnection of the OFC and amygdala were impaired relative to controls. Analyzing the data from the two groups together revealed a double dissociation. In addition to replicating the findings of A × OFC circuitry in devaluation from Baxter and colleagues (2000), this study suggests that OFC and ACC connections with the amygdala may be specialized to subserve distinct valuation processes, perhaps specifically in contexts that require updating value based on changes in the internal representation of the outcome. One intriguing possibility is that, paralleling findings from the devaluation task, A × ACC disconnection would result in an insensitivity to changes in social rank.

A causal role for ACC–amygdala circuitry in social observational learning has been observed in rodents (Allsop et al., 2018); however, in monkeys and in humans, the causal role of this circuitry with respect to specific aspects of social cognition is not well understood. A recent study reveals that abnormal amygdala–ACC functional connectivity is associated with social communication impairments in participants with autism spectrum disorder (Yang et al., 2024). Still, more work needs to be done across species to determine the nature of these circuit-level interactions in the many aspects of social cognition, including social signaling, observational learning, dominance hierarchies, the development of alliances, and so on. One task that has been used to assess the neural correlates of social cognition in monkeys—the “reward allocation task” (Chang, Winecoff, & Platt, 2011)—measures the extent to which monkeys engage in prosocial reward allocation, which is thought to index prosociality or vicarious reinforcement. In this task, two monkeys—an actor and a recipient—view cues presented on a video monitor. The actor monkey gets to choose between visual cues that signal delivery of juice to self, a conspecific other, both, or neither. Importantly, the choices made by the actor monkey do not affect the likelihood of its earning a reward. For example, the actor might be offered a choice between cues signaling reward to the conspecific alone (Other), or to neither monkey (Neither); in this case, there is no benefit to the actor to provide reward to Other. When actors choose to reward the other monkey (Other) rather than no one (Neither), this is accepted as a prosocial behavior because the monkey expends effort to give juice to a conspecific. As a crucial control, monkeys do not make this choice if the juice is delivered to a nonsocial entity, such as a graduated cylinder, indicating that it is the presence of a conspecific that drives this “giving” behavior.

A series of studies using variations of this task have revealed that activities of neurons in the amygdala, OFC, and medial PFC (specifically the ACC) encode aspects of task performance (Chang et al., 2015; Chang, Gariépy, & Platt, 2013). It was found that, whereas amygdala neurons encoded reward amount but not social information, that is, information about the agent (Chang et al., 2015), and OFC neurons primarily signaled reward to self, ACC neurons encoded both other agents and reward information. Specifically, the ACC contained three separate populations of neurons that signaled social outcomes: one that signaled rewards to self, another that signaled rewards to a conspecific, and a third one with shared signaling that signaled reward to the self and the conspecific. Combined with causal evidence that the ACC alters prosocial performance on this task (Basile et al., 2020), the findings strongly implicate the ACC in learning about and representing other agents, and experiences shared with those agents.

To test the possible interaction between the amygdala and ACC during social decision-making, spike-field coherence was analyzed while monkeys performed a variant of the reward-allocation task (Dal Monte, Chu, Fagan, & Chang, 2020). Spiking activity of individual cells in each area was related to the local field potential oscillations in the other area. The main finding was that there was enhanced neural synchrony between the amygdala and ACC during prosocial decision-making. The timing of the coherence indicated it was unlikely to be responsible for generating the decision; instead, the authors suggested that the enhanced coherence might serve as a feedback mechanism during social learning that could be used to adjust future prosocial decisions.

Despite the findings outlined above, we suspect that the OFC plays an as yet undetermined role in social cognition (for review, see Elorette et al., 2021). For example, face patches have been found in the lateral orbital sulcus of the OFC (Troiani, Dougherty, Michael, & Olson, 2016; Tsao, Schweers, Moeller, & Freiwald, 2008). Recording studies suggest that these patches contain face-selective neurons that categorize social information from faces (Barat, Wirth, & Duhamel, 2018). It is therefore possible that an intra-PFC network involving subregions of medial frontal cortex and OFC encodes information that is relevant to social value and learning from social contexts. Additional work needs to be done to address the nature of these neural interactions during a wide variety of social cognitive processes in dynamic naturalistic settings involving dyads or groups.

SUMMARY AND CONCLUSIONS

The studies reviewed here indicate that the primate amygdala's interactions with the frontal cortex contribute to at least two of the many areas of behavior in which the amygdala is involved: foraging and social signaling (perhaps as a precursor to social approach). Although the amygdala is often characterized as a “fear module” (Öhman & Mineka, 2001), important for processing negative affect, our results point to a broader role for the amygdala in stimulus–outcome associative learning in human and nonhuman primates.

There is still more work to be done to determine whether the proposed pathways neatly subserve the complex ability to discern between what to value and who to value. Two aspects of social and nonsocial valuation deserve additional comment and, perhaps, additional study. The first is the dynamic nature of both social and nonsocial valuations. Valuations depend on the organism's internal state, the social context, and the experienced history of learned values that may change over time.

In the social domain, studies in macaques indicate that amygdala neurons are sensitive to the direct gaze of conspecifics (Gilardeau et al., 2021; Mosher, Zimmerman, & Gothard, 2014). Furthermore, in both humans and macaques, attention to faces gates amygdala responses (Minxha et al., 2017). These and other findings strongly suggest that the amygdala participates in social evaluation and dynamic social interactions with conspecifics. Future studies should evaluate the activity of both ACC and amygdala during dynamic social exchanges. Although macaques are often studied in the context of stable social relationships, it would be of interest to evaluate activity of neurons in these regions when social status or rank is manipulated (Champ et al., 2022).

In the nonsocial domain, selective satiation is a temporary state that is believed to underlie, at least in part, dietary diversity. In this sense, flexible selection of nutrients, based on vision, is built into the system to ensure balanced intake of macronutrients. Although the physiological studies of Rudebeck and colleagues (2013) show that amygdala inputs to OFC are important for maintaining already established stimulus–reward-values in OFC, no one has yet systematically studied the activity of single neurons in OFC or basolateral amygdala while the changes in value are occurring (e.g., during the selective satiation process). Human fMRI studies suggest a reduced “activation” of OFC accompanies selective satiety when subjects consume liquid food (Kringelbach, O'Doherty, Rolls, & Andrews, 2003), and early macaque studies likewise emphasized a reduction in OFC neural activity with satiety when monkeys viewed food (Critchley & Rolls, 1996). More recently, however, a study examining the activity of a small number of neurons in OFC before and after monkeys had received a large bolus of fluid found that some neurons increased their activity, some decreased, and some had a change in the pattern of their activity while monkeys viewed cues that predicted the fluid (Bouret & Richmond, 2010). These findings suggest a need for additional study; indeed, recording from single neurons before, during, and after selective satiety would inform how changes in value are implemented in the brain, and how the dynamic changes are encoded.

Another area in need of study is the place of amygdala–frontal interactions in the context of brain evolution. As indicated earlier, we inherited the amygdala from our vertebrate ancestors. Yet ACC appeared in early mammals and granular OFC appeared in early primates. Changes in amygdala–frontal connections and interactions must have occurred in conjunction with these developments. Findings from an anatomical study in rhesus macaques indicate that the projections from the amygdala to the magnocellular part of the MD thalamic nucleus (MDmc) have some unique properties. As in sensory thalamic systems studied in rats and monkeys, large amygdalar terminals innervate excitatory relay and inhibitory neurons in the MDmc, forming synaptic triads thought to facilitate transmission to the cortex. In addition, however, some amygdala terminals also innervate MDmc neurons by surrounding and isolating large segments of their proximal dendrites, a unique pattern not seen in sensory systems (Timbie, García-Cabezas, Zikopoulos, & Barbas, 2020). Whether this is a primate specialization remains to be determined. Ultimately, it will be important to understand whether humans have evolved specializations of amygdala–frontal pathways for social cognition (Edmonds et al., 2024). A critical question that remains is what drives the allocation of attention to either social or nonsocial elements or vice versa at any given moment of one's day. If someone were to cross paths with a friend at Starbucks, what determines the “switch” from seeking the primary reward (coffee) to greeting the friend?

Corresponding author: Elisabeth A. Murray, Laboratory of Neuropsychology, NIMH, Building 49, Suite 1B80, 49 Convent Drive, Bethesda, MD 20892-4415, or via e-mail: murraye@mail.nih.gov.

Funding Information

This work is supported by the Intramural Research Program of the National Institute of Mental Health (ZIAMH002886 and ZIAMH002887 to E. A. M.).

Diversity in Citation Practices

Retrospective analysis of the citations in every article published in this journal from 2010 to 2021 reveals a persistent pattern of gender imbalance: Although the proportions of authorship teams (categorized by estimated gender identification of first author/last author) publishing in the Journal of Cognitive Neuroscience (JoCN) during this period were M(an)/M = .407, W(oman)/M = .32, M/W = .115, and W/W = .159, the comparable proportions for the articles that these authorship teams cited were M/M = .549, W/M = .257, M/W = .109, and W/W = .085 (Postle and Fulvio, JoCN, 34:1, pp. 1–3). Consequently, JoCN encourages all authors to consider gender balance explicitly when selecting which articles to cite and gives them the opportunity to report their article's gender citation balance.