Abstract
Brain-wide circuits that coordinate affective and social behaviours intersect in the amygdala. Consequently, amygdala lesions cause a heterogeneous array of social and non-social deficits. Social behaviours are not localized to subdivisions of the amygdala even though the inputs and outputs that carry social signals are anatomically restricted to distinct subnuclear regions. This observation may be explained by the multidimensional response properties of the component neurons. Indeed, the multitudes of circuits that converge in the amygdala enlist the same subset of neurons into different ensembles that combine social and non-social elements into high-dimensional representations. These representations may enable flexible, context-dependent social decisions. As such, multidimensional processing may operate in parallel with subcircuits of genetically identical neurons that serve specialized and functionally dissociable functions. When combined, the activity of specialized circuits may grant specificity to social behaviours, whereas multidimensional processing facilitates the flexibility and nuance needed for complex social behaviour.
Survival often depends on the individual’s ability to integrate into a social group. Social behaviour is coordinated by a distributed network of brain areas that include the amygdala, an ancient structure that is already present in reptiles1 and known to be a central hub for processing affective stimuli in many mammalian species. Compared with its role in assigning value (or valence) to environmental stimuli, the social functions of the amygdala are less well understood. Social deficits caused by temporal lobe lesions that included the amygdala were first reported more than a century ago2. However, the specific contribution of the amygdala to social behaviour is still debated, as a series of lesion studies, based on increasingly refined techniques, failed to converge on a core function for this structure. Instead, amygdala lesions produced a full list of social deficits, including faulty face and gaze processing and changes in aggression, submission, trust or social status, which in combination result in context-inappropriate social behaviours and disadvantageous social judgements (TABLE 1; also see REF.3). These deficits were often entangled with abnormalities in attention, arousal, defensive behaviours, salience or relevance detection and value-based decision-making (for example, see REFS4,5).
Table 1 |.
The effects of complete amygdala lesions on social behaviour in monkeys
| Deficit | Effect on social behaviour | Species | Age | Sex | Refs |
|---|---|---|---|---|---|
| Affective blunting (that is, reduced response to positive or negative social stimuli) | ↑ | Rhesus | Adult | Mixed | 2,99,100 |
| ↑ | Rhesus | Infant | Mixed | 101 | |
| Tameness towards humans | ↑ | Rhesus | Adult | Mixed | 2,22,99,102 |
| Social status | ↓ | Rhesus | Adult | Males | 100 |
| ↓ | Vervet | Adult | Males | 22 | |
| ↓ | Stump-tail | Adult | Males | 103 | |
| ↑ | Rhesus | Adult | Males | 100 | |
| Ø | Rhesus | Adult | Males | 104 | |
| Affiliation | ↓ | Vervet | Adult | Males | 22 |
| ↓ | Rhesus | Adult | Males | 105 | |
| Aggression | ↓ | Rhesus | Adult | Males | 4,106 |
| ↓ | Rhesus | Infant | Mixed | 107 | |
| ↑ | Rhesus | Adult | Males | 100,105 | |
| Submissive behaviours | ↓ | Rhesus | Adult | Mixed | 108 |
| ↑ | Rhesus | Adult | Mixed | 4,109,110 | |
| ↓ | Rhesus | Infant | Mixed | 111 | |
| Defensive behaviours | ↓ | Rhesus | Adult | Mixed | 4 |
| ↓ | Rhesus | Adolescent | Males | 37a | |
| ↑ | Rhesus | Infant | Mixed | 111–115 | |
| ↑ | Rhesus | Adult | Males | 105 | |
| Ø | Rhesus | Adult | Mixed | 116–118 | |
| Social interest | ↓ | Rhesus | Adult | Mixed | 106,109 |
| ↓ | Rhesus | Infant | Mixed | 23,110,119 | |
| ↑ | Rhesus | Adult | Males | 9,38b,118,120 | |
| Attending to social stimuli and responding adequately | ↓ | Rhesus | Infant | Mixed | 108,121,122 |
| Spontaneous interactions and reciprocation of social signals | ↓ | Rhesus | Infant | Mixed | 107 |
| ↓ | Rhesus | Infant | Mixed | 101,123 | |
| Sexual and maternal behaviours | ↑ | Rhesus | Adult | Mixed | 2,99,120,124 |
| ↓ | Rhesus | Infant | Females | 125 | |
| Autonomic and endocrine responses | ↓ | Rhesus | Infant | Mixed | 126 |
| ↓ | Rhesus | Adult | Males | 122,127 | |
| Context-inappropriate behaviours | ↑ | Various | Various | Mixed | 22,106,107,128 |
↑, increase; Ø, no change; ↓, decrease.
The lesion was restricted to the central nucleus of the amygdala.
Pharmacological inactivation only of the basolateral nucleus.
Cell type-specific circuit manipulations in the rodent amygdala have in the past 5 years identified clusters of neurons that, when activated or inactivated, produce or prevent select behaviours6. The shared genetic identity and connectivity of these neurons determines the role they play in diverse processes such as valence assignment7–9 or feeding10,11. Even more complex behaviours, such as prey pursuit and prey capture, have been shown to depend on subdivisions of the central nucleus of the amygdala (CeA), with each behaviour assigned to distinct clusters of neurons that project to different targets in the brainstem12.
In contrast to this mosaic of specializations in amygdala circuits is the observation that neurons in the amygdala have multidimensional response properties13. Indeed, when monitored during tasks that require the activation of multiple functions of the amygdala, each amygdalar neuron responds to multiple types of stimuli and multiple, often disparate, task parameters14–17. These and other similar observations suggest the presence of an organizational scheme in the amygdala in which individual neurons hold membership in multiple, functionally distinct, neural ensembles. The resulting multidimensional representations18,19 permit flexible strategies to navigate the complex, ever-changing social environment. In a broader perspective, multidimensional processing is not unique to the amygdala or to the social domain, but the amygdala illustrates this neurophysiological feature, which is shared with high-level associative cortices, particularly well.
In this Perspective, I aim to contrast established ideas of neural specializations with new concepts of multidimensional processing and to discuss from this viewpoint the foundation of social specializations in the primate amygdala. The emphasis on social behaviour is justified by the unusually large contribution of social factors to affective states coordinated by the amygdala. The emerging view is that multidimensional processing and functional compartmentalization are difficult to reconcile in the framework of simple psychological constructs such as attention, perception, valence or salience detection. In a higher-dimensional and more abstract framework, however, these two alternatives can be reconciled.
The amygdala in social behaviour
Is the amygdala necessary for normal social behaviour?
In order to establish the necessity of the amygdala for social behaviour, this section focuses on lesions and causal manipulations of the entire amygdala; however, studies addressing nucleus-specific lesions or pharmacological manipulation of the amygdala are also mentioned. The rich literature of neuroimaging and electrophysiological studies that implicate the amygdala in social behaviour has been reviewed elsewhere20,21.
The amygdala acquired its place in the inventory of brain areas necessary for social behaviour from the earliest descriptions of deficits that are caused by removal of the temporal lobes, including the amygdala2. One of the earlier lesion-based research studies claimed that the amygdala is necessary for survival22. In this study, adult vervet monkeys with bilateral amygdala lesions were returned after amygdalectomy to their social group to continue living in their natural environment. These monkeys showed no interest in others, did not behave according to their pre-surgical social status, ignored or failed to understand the signals of their peers, became isolated, defenceless and anorexic, and died in a few weeks. In later studies, as the amygdala lesions became more selective and the testing environments less naturalistic, the number and severity of the deficits in social behaviour shrank to approximately a dozen key features (TABLE 1). Only limited consensus can be extracted from the literature on amygdala lesions; the extent and developmental timing of the lesion, the socialization of the subjects preceding and following the lesion, and the tests used to quantify the social deficits have generated divergent and often contradictory outcomes23.
Nevertheless, a few reliable threads weave through the fabric of these observations. Bilateral destruction of the amygdala in adult male monkeys often causes a fall in social status and the erasure or reduction of their natural wariness of others, and may increase or reduce aggression as well as affiliative or submissive behaviours (TABLE 1). When engaged by social partners, animals that have undergone amygdalectomy can respond to social cues and produce meaningful social behaviours, but their responses are typically context-inappropriate. By contrast, neonatal lesions in the amygdala leave more subtle sequelae or none at all, suggesting that the amygdala is not required for social behaviour and that its absence can be compensated for by redundant circuits and neural degeneracy. For example, the amygdala and the anterior cingulate cortex both contribute to the perception, and perhaps even the production, of facial expressions24.
The literature on lesions suggests that the amygdala may not be strictly necessary for any aspect of social behaviour but, given its extensive connectivity, it is probably necessary to fine-tune and add nuance to social behaviour and co-opt non-social functions (such as value processing) to support social decisions. Indeed, functional imaging and neural recordings in both humans and non-human primates strongly support this view. According to these correlative studies, the amygdala plays a role in: social decision-making20; observational learning and vicarious reward processing (reviewed elsewhere25); face, gaze and eye-contact processing (reviewed in REF.21); the representation of social status26; the allocation of visual attention to social stimuli27; social anxiety28; the production of facial expressions24,29; social memory30,31; and the coordination of autonomic responses that are elicited by social stimuli32.
Is there a social–non-social division of labour among amygdalar nuclei?
Given that the nuclei of the amygdala have different developmental origins33–35, and connect via multiple processing loops to different cortical and subcortical areas36, it is conceivable that anatomically distinct areas of the primate amygdala are specialized to process social signals. However, this speculation awaits experimental verification.
There have been too few nucleus-specific lesion studies in primates to reveal a division of labour among the nuclei. In one study, adolescent monkeys with lesions restricted to the CeA reacted with fewer defensive behaviours to human intruders, showed reduced fear of snakes and had lower levels of stress hormones in the brain and in the periphery compared with control animals37. This outcome is expected given that direct projections of the CeA to the hypothalamus and brainstem coordinate the endocrine and energetic or homeostatic aspects of affective behaviour35.
It would be interesting to compare the outcome of CeA lesions with the outcome of lesions that are restricted to the basolateral nuclei of the amygdala. Indeed, the basolateral nuclei, which primarily include the lateral, basal and accessory basal nuclei, are not directly connected with autonomic and endocrine effectors; rather, they receive inputs from, and send projections to, cortical areas of the brain36. The primate amygdala, and in particular the lateral nucleus, receives fewer inputs from the thalamus than the rodent amygdala35,36, likely because of the dramatic expansion of the cortex in primates. The connectivity of the basolateral nuclei suggests that they evaluate the valence or social relevance of cortically processed stimuli and redistribute the outcome of this evaluation to multiple areas of the cortex. They also project to the central nuclei to initiate the appropriate autonomic responses. Selective basolateral lesions in primates have not been carried out; however, one study showed that pharmacological inactivation of the basolateral nuclei (by local infusion of the GABAA receptor agonist muscimol) led to a seemingly paradoxical increase in social behaviour that could not be explained by a reduction of social fear38. Pharmacological blockade of NMDA-type and AMPA-type glutamate receptors in the basolateral nuclei, however, did not increase social behaviours, likely because excitatory–inhibitory interactions in the microcircuitry of the basolateral nuclei are altered differently when GABAergic versus glutaminergic transmission is manipulated39. In summary, based on the literature discussing lesions of and inactivation of the amygdala alone, there is no clear evidence for or against the idea that the primate amygdala is compartmentalized for social behaviours.
Neural specializations for social functions.
In primates, the social deficits caused by amygdala lesions are entangled with non-social deficits, such as the absence of snake fear or reward devaluation40,41. These deficits can appear to be more robust and reproducible than social deficits, perhaps because they are easier to detect and quantify compared with spontaneous social interactions. It is also possible that social deficits are a domain-specific manifestation of broader cognitive malfunctions. In this scenario, a monkey who fails to discriminate between unconditioned, negative or positive stimuli (for example, between snakes and food) will also fail to attach value or valence to appeasing or aggressive social displays.
One review42 made a compelling case that domain-specific social functions are independent from domain-general cognitive functions in the anterior cingulate cortex; that is, the authors cited literature that showed that social functions are not merely a special case of more general functions. Their analysis showed that the social and non-social functions in the anterior cingulate are anatomically and functionally separable, and that further local subdivisions process ‘self ‘ versus ‘other’. It is interesting to consider whether the social and non-social functions of the amygdala can also be detangled.
Social-specific regions of the amygdala could possibly be defined by clusters of face-responsive neurons that exchange signals with neurons in the face patches of the temporal cortex43. However, unlike the face patches in the temporal cortex, face-responsive neurons in the amygdala do not appear to be clustered in any particular nucleus or subnuclear region44. This point is the case for all other classes of neurons with social correlates, such as the neurons that respond to eye contact45, parts of the face46 or social status26. Although a few studies have reported that neurons with particular response properties (some of which indicate social function) tend to be more abundant in certain nuclei or nuclear subregions of the amygdala47–50, overall there is no obvious and reliable clustering of neurons with social functions in the primate amygdala.
If the anatomical differences between amygdalar nuclei are not reflected in the response properties of individual neurons, then specificity may not be present at the level of individual neurons. It is possible that specificity emerges at a higher level of organization, such as the level of dynamic ensembles of neurons51,52. Indeed, the local field potentials recorded with geometrically configured electrode arrays confirm that subnetworks that are spatially restricted to nuclear subdivisions are present in the primate amygdala53.
Of compartments and species.
Many studies report on elegant region-specific and cell type-specific causal manipulations of the rodent amygdala, a subset of which target social behaviour54,55. These studies do not have a primate equivalent, partly because the same optogenetic and chemogenetic techniques have only recently been transferred to primates, and partly because social neuroscience studies in these species have focused on different sensory domains. The typical studies in rodents manipulate olfaction, and the outcome of these manipulations suggests that the social hub of the rodent amygdala is the medial nucleus, where inputs from the vomeronasal organ are processed56. By contrast, in primates, olfactory inputs are processed throughout the amygdala57 but olfaction plays only a small role in social recognition. Indeed, primates do not have a vomeronasal organ, and for social recognition they rely primarily on visual signals that are processed throughout the amygdala, including the medial nucleus44,45.
Even in rodents, the functional compartmentalization of the amygdala is more evident in non-social experimental contexts58–61, such as the differentiation of appetitive and aversive stimuli. The compartments are defined by the genetic identity of their neurons, their connectivity and the behaviours that are altered by their selective activation or inactivation (recently reviewed in REFS9,13). Importantly, the exquisite details gained from cell type-specific circuit dissection of the rodent amygdala, such as differential encoding of valence by the lateral and medial subdivisions of the CeA, do not align with single-unit activity recorded during valence discrimination9. Indeed, neither the genetic identity, connectivity nor response preference for appetitive or aversive stimuli alone is sufficient to label a neuron as having positive or negative valence bias. Different combinations of these features — that is, neurons with multidimensional response properties — can generate relatively heterogeneous neuronal ensembles that may transmit valence preferences more reliably than neurons that are defined based on a single dimension9.
How many jobs can a neuron hold?
The stimuli and task demands processed by the monkey amygdala are so diverse that almost any task can induce stimulus-related or task-related amygdalar responses (FIG. 1). This diversity reflects the intersection of multiple sensory and cognitive loops in the amygdala. In the social domain, neurons in the amygdala respond to bodies, faces, facial expressions, eyes, gaze direction (reviewed in REF.21), social status26 and the observed or expected behaviours of social partners50. The most common non-social neural correlates in the amygdala include responses to unconditioned stimuli of all sensory modalities across a wide spectrum of valence62,63; conditioned sensory stimuli that predict appetitive or aversive outcomes64,65; economic decisions66,67; and abstract information about cognitive context68,69.
Fig. 1 |. Alternative functional organization schemes of the primate amygdala.
a | One view of primate amygdala organization is that specialized neurons, which are narrowly tuned to a single stimulus or task parameter, form clusters that follow, to some extent, the anatomical target or origin of inputs and outputs, respectively. b | An alternative view of primate amygdala organization is that multidimensional neurons are distributed quasi-equally across the nuclei of the amygdala. As indicated in the key, the colours represent a subset of the known stimulus categories that activate the amygdala16,17,30,64 with the exception of visceral inputs, for which there is limited evidence. The proportion of neurons that respond to each type of input is unknown. AB, accessory basal nucleus; B, basal nucleus; CeA, central nucleus of the amygdala; L, lateral nucleus; Me, medial nucleus.
Multidimensionality (or mixed selectivity) is a reliable feature of neural responses in the primate amygdala that was documented even in early recordings70; however, the concepts associated with multidimensionality and its computational importance came into focus only in the last decade. These new concepts are reframing our understanding of the amygdala. Indeed, in humans, non-human primates and rodents alike, many neurons in the amygdala respond to more than one stimulus dimension and task parameter. That is, the activity of a single amygdala neuron typically conveys information about several stimulus dimensions or task parameters. For example, in the monkey amygdala, the same neurons that encode reward value predicted by fractal images also encode the dominance status of familiar individuals26. In rodents, the same neurons encode valence and active versus passive behaviours when multiple responses can be given to the same conditioned stimulus16. When monkeys perform a task with four variables (specifically, attention to alerting cues, learning the value associated with different stimuli, discrimination of social and non-social stimuli and discrimination between individuals), a large proportion of neurons in the amygdala respond to combinations of two, three or all four variables17. The proportion of neurons selective for all four task variables exceeded the level that would be expected by chance (that is, the level expected if the probabilities of a neuron coding for variable A were independent of the probabilities of it coding for variable B), suggesting that a non-random combinatorial scheme gives rise to multidimensional selectivity. Multisensory neurons (that represent an elemental level of multidimensionality) in the primate amygdala follow the same rule: the probability of single neurons responding to multiple sensory modalities is higher than expected by chance62. Neurons in the human amygdala also show multidimensional properties: a subset of neurons encode the identity of a visual stimulus and also memory for that stimulus71. In light of these findings, I proposed that neurons in the amygdala that respond to faces or eyes should qualify as ‘face cells’ or ‘eye cells’ only if their responses remain selective for faces and eyes in multiple behavioural contexts and after testing with a broad array of stimuli (FIG. 1).
Multidimensional processing
Defining multidimensional selectivity.
It is important to consider whether observations accumulated to date are sufficient to define what a ‘dimension’ is. The term dimension emerged from the computational and analytical methods used to document and quantify this phenomenon, not from theoretical and conceptual advances that preceded experiments and analytical approaches. It appears that any cognitive variable that reliably accounts for variations in the firing rate of single neurons can serve as a dimension. For example, a visual stimulus, an operant behaviour, a reward, the predicted outcome associated with a stimulus, the kinematics of a motor act, abstract variables (such as time, context and memory), and even the level of engagement of the subject in the task can function as a dimension. Nevertheless, to classify neurons as multidimensional, it is not sufficient to document that they respond to more than one stimulus or task parameter. In technical terms, a dimension can be any arbitrary axis in a multidimensional space (analogous to the ‘x’, ‘y’ and ‘z’ axes in three-dimensional space) as long as the axes are orthogonal (uncorrelated). By representing each dimension on a separate axis, one can build a multidimensional space.
An important and yet only partially resolved issue is whether populations of multidimensional neurons encode linear or non-linear combinations of the different inputs. The terms ‘multidimensional’ and ‘mixed’ selectivity have been used interchangeably; these terms do not map onto the linear versus non-linear combination of dimensions that characterize the responses of single neurons, although such a distinction would be desirable. I use the term multidimensional to suggest a large number of non-linear combinations given that linear combinations are likely to saturate with the addition of each new dimension. Indeed, in the social domain, linear combinations of selectivity are difficult to capture, but let us consider a hypothetical neuron that responds to the intensity of a threatening facial expression and the age of the threatening individual; the older the threat-emitting individual and the more intense the threat, the higher the firing rate of the neuron. More often, neurons that respond to individual faces and facial expressions suggest non-linear combination stimulus dimensions44, as exemplified by the response properties of the neuron shown in FIG. 2.
Fig. 2 |. Face-responsive neurons show non-linear combinations of selectivity.
a–d | This neuron was recorded from the basal nucleus of the amygdala in response to the passive viewing of face images that represented specific combinations of identity and facial expression. The stimuli in each row of images depict three facial expressions displayed by the same individual. The three columns correspond to appeasing (lip-smack), neutral and aggressive (threat) facial expressions. Rasters (top) and peri-event time histograms (bottom) are shown below each stimulus face and their analysis indicates that this neuron responded to the appeasing facial expression of the monkeys in parts a and b and to the threatening expressions of the monkeys in parts c and d. Analysis of variance showed that there was a significant interaction (P < 0.001) between identity and facial expression. Parts a–d reprinted with permission from REF.44, American Physiological Society.
The distinction between linear and non-linear combinations of selectivity (or mixed selectivity) is important because neurons with non-linear combinations of selectivity can generate higher-dimensional representations than neurons with linear combinations of selectivity19.
The importance of non-linear mixed selectivity emerged from the analysis of neural responses during multi-alternative tasks, such as the analysis of prefrontal neural activity recorded from monkeys that were trained to recognize or recall specific sequences of images as part of a sequence-learning task72. These neurons showed both linear and non-linear mixed selectivity for task parameters and also for multiple stimulus features18. When the authors removed the non-linear component of mixed selectivity from every neuron, they could decode every stimulus dimension or task parameter from the population separately using binary decoders. When they removed the linear components (that is, the traditional stimulus selectivity or task selectivity of each neuron), there was sufficient residual information in the population to decode stimuli and task variables at a rate above chance. Based on this and other similar studies, populations of neurons with non-linear, mixed selectivity were argued to generate high-dimensional, abstract representations that are optimal to support flexible behaviours19. The reasoning behind this argument is that non-linear mixing allows a downstream area (an area that receives the output from the area that contains the high-dimensional representation) to easily read the required information (akin to a linear decoder)19. Unlike linear mixing, non-linear mixing allows choices that depend on latent variables and/or on operations such as exclusive–or (that is, stimulus 1 and not stimulus 2 is rewarded in context A, and stimulus 2 and not stimulus 1 is rewarded in context B). An excellent illustration of this principle was experimentally implemented in a multi-alternative task that distilled the problem to its essence68. In this study, amygdalar neurons were recorded as monkeys learned that one set of conditioned stimuli were rewarded and another set predicted an aversive outcome. This rule held only for a block of trials and was reversed in the subsequent block, but nothing explicit signalled to the monkeys which rule applied during each block. Monkeys used each block of trials as a cognitive context to flexibly and efficiently switch between the appetitive and defensive responses that were appropriate for each context. Note that, in this example, neurons in the amygdala and orbitofrontal cortex encoded the cognitive context (akin to an abstract representation) of the task.
Latent and abstract variables may be difficult to track experimentally, but they are often the critical elements that inform social decisions. Social behaviour relies more often than not on decisions that depend on context or the ongoing pattern of events and interactions, as suggested by the neural responses shown in FIG. 2. Given the hierarchical nature of macaque societies, the status of the displaying individuals determines how substantial and actionable an appeasing or threatening facial expression might be. A mild threat, or even a stern stare from a high-ranking individual, is more consequential than a full-blown threat from an individual of low status. It is more likely than not that the social status and facial expressions would show non-linear mixing. The extent to which neurons in the amygdala show non-linear mixed selectivity remains to be quantified and compared across nuclei, tasks and species. Such analyses are possible because there is robust evidence for mixed selectivity in the amygdala in several multi-alternative tasks that combine multiple behaviours, multiple stimulus categories or decision options16,17,68.
Advantages of multidimensional processing.
Multidimensional processing contributes to and benefits from learning. Mixed selectivity arises not only from random connectivity (see below) but also as a consequence of learning73. This fact was demonstrated by adding Hebbian learning to a model of the prefrontal cortex (built with random connectivity), which resulted in a higher level of mixed selectivity and a better match between the model and the neural data recorded from the prefrontal cortex of monkeys74. Conversely, mixed selectivity allows learning that would enable rapid switching between behavioural options18,68.
Multidimensional selectivity enables the same neurons to be deployed for multiple processes, such that a relatively small number of neurons (only 6 million in the monkey amygdala) can contribute to a large number of cognitive and behavioural functions. Indeed, the number of dissociable outputs of the amygdala may be small compared with the extraordinary richness of the inputs. Consider, for example, the large variety of external stimuli (for example, images, sounds and touch), cognitive processes (such as memory and decision-making) and social events (including conflict or deep attachment) that trigger, via the amygdala, the same judgement of valence (that is, positive or negative valence), the same behavioural responses (for example, approach or avoidance) or the same autonomic responses (for example, changes in heart rate, blood pressure or pupil dilatation). In the framework of multilayered artificial neural networks, multidimensional neurons are quintessential examples of what is ‘hiding’ in the hidden layer. Indeed, dimensionality expansion by adding layers is a critical step performed in deep neural networks to enhance the accuracy of classifications (that is, of decoding)75.
Multidimensional processing may also be advantageous to link the basic building blocks of complex behaviours. For example, the social importance of facial expressions exchanged between monkeys depends on the identity of individuals engaged in facial signalling. Accordingly, face-responsive neurons in the amygdala respond to various combinations of identity and facial expression44 (FIG. 2). In hierarchical societies, each individual has a social status (that is, a rank) and the difference in rank between the signal-emitting individuals and the signal-receiving individuals conveys meaning to the social exchange. It is therefore likely that, in addition to identity and facial expressions, multidimensional neurons also respond to the status differential between social partners. Indeed, neurons in the amygdala do encode the status of individuals26 and the responses of these neurons span the social–non-social divide, as they also respond to rewards. Similarly, social attention76 and spatial attention15 are often processed concurrently because individuals of social interest often change location in space. Under these conditions, neurons in the amygdala may be multidimensional and signal, at the same time, particular individuals and their locations.
Another benefit of multidimensional processing is that, during natural social interactions, it may link neural activity related to ‘self ‘ to neural activity related to ‘other’. Most social processes that are subjected to neurophysiological scrutiny are set up as open-loop situations (for example, monkeys or humans passively viewing images of facial expressions). This approach captures only the social perception component of social behaviour. By contrast, in real-life interactions, social signals are exchanged in perpetual receiving–emitting cycles that are more accurately modelled by closed loops. In a closed loop, the outcome of the current action becomes the input for the next cycle. If the same population of neurons detect and evaluate the incoming social signals and also trigger or coordinate reciprocal behavioural responses, then the loops between ‘self ‘ and ‘other’ can be closed. Indeed, neurons in the amygdala and the motor areas of the anterior cingulate cortex show activity related to both the perception and the production of facial expressions24. Likewise, the amygdala contains neurons that respond to the facial expression of others and also neurons that respond to the facial expressions of self77. It is unknown, but highly likely (given the high probability that multidimensional processing occurs in the amygdala), that the same neurons respond to both.
Finally, one can add to these advantages of multidimensional processing an advantage that has been often neglected: interfacing interoceptive (or viscerosensory) inputs with affective states. Visceral signals arrive at the amygdala via ascending autonomic pathways from the nucleus of the solitary tract, parabrachial nuclei, the hypothalamus and the insula36, and probably generate the background on which other ongoing neural processes are grafted78. For example, the insula and amygdala in hungry animals bias the visual cortex towards processing visual food cues79. Viscerosensory inputs may contribute to global, brain-wide patterns of activity that incorporate into a brain state the state of the periphery and generate coactivity patterns among areas that are not directly interconnected.
Multidimensional processing in other brain areas and cognitive domains.
The high-dimensional representation generated by ensembles of neurons with multidimensional selectivity19 in the amygdala suggests an overlap in the role of these structures in creating mental states that drive complex behaviours, whether social or non-social68. The presence of multidimensional processing in the amygdala, which often captures abstract representations, raises the question of whether the role or roles of the amygdala can be described in terms of simple psychological constructs, such as valence, arousal or relevance detection. The answer seems to be ‘probably not’, because these constructs cannot describe the full dimensionality of these mental states and fail to capture latent, non-observable variables. Neuroscientists have not yet invented a lexicon to adequately denote abstract mental states80 that morph through various shapes in high-dimensional space. One can speculate that high-dimensional, abstract representations not only allow the selection of behaviourally useful combinations of stimuli but may also facilitate the generation of novel combinations of stimuli that could be useful in new, previously unrehearsed situations. If high-dimensional representations hold the potential for novel combinations, useful to solve new problems, then they might be essential for imagination and creativity.
In retrospect, multidimensional selectivity, with or without the explicit use of this term, has been reported in many brain areas and in multiple behavioural contexts. For example, single neurons in the human parietal cortex show mixed encoding of moving body parts, body side and cognitive strategy (that is, imagined versus executed) applied to movement81. A thorough review of the neurophysiological literature, from the perspective of multidimensional selectivity, will probably conclude that unidimensional responses, such as receptive fields, are the exception and not the rule in the brain, and that unidimensional neurons are found mainly in primary sensory areas. Even in the mouse primary visual cortex, however, neurons show multidimensional responses, combining visual information with information about the behavioural and motivation state of the animal82.
A three-dimensional analogy for such a dynamically changing shape of multidimensional representations is the image of a murmuration of starlings: hundreds or thousands of birds flying together in a cloud that perpetually shifts shape and in which the momentary position of each bird depends on the strength and range of interactions with other nearby birds83. Each starling can be considered a ‘neuron’, as a point in three-dimensional space where each axis represents the firing rate in response to three different stimuli or task variables. For example, the three axes could represent the increasing proximity of a social partner, the perceived or predicted aggressive intent of this partner and the facial expressions produced by the receiver. As the social interaction unfolds, this cloud of points moves contiguously through a sequence of states instantiated by the same population of neurons.
Although the emerging framework of multidimensional neural responses and high-dimensional, abstract representation is ideally suited to explore the neural basis of social behaviour, these concepts may inform our perspective in other cognitive domains, too. Certain aspects of the sensory transduction of early sensory processes notwithstanding, it is almost impossible to image how more complex functions of the brain can be carried out efficiently without multidimensional processing.
Large- and small-scale circuits
The contribution of brain-wide circuits.
The intersection of multiple brain-wide processing loops in the amygdala (FIG. 3) may provide the anatomical basis for multidimensional processing. Human studies of resting-state connectivity84, the task-dependent coactivation of the amygdala with other brain areas (reviewed in REF.20), and the application of graph theory to anatomical findings85 have suggested that the amygdala is one of the subcortical hubs of the brain86, and specifically the quintessential social hub20. Indeed, the amygdala is always on the list of structures in definitions of the social brain87.
Fig. 3 |. Multidimensional selectivity arises from the intersection of brain-wide circuits.
The amygdala (represented here by the circle in the centre of the image) receives inputs from, and sends outputs to, numerous brain areas, giving rise to a complex web of intranuclear and internuclear connections (as indicated by the coloured lines) and closing circuits that carry out quasi-dissociable functions. The six main functions represented in the ring around the central circle were chosen because they are obligatory components of social behaviours; a truncated list of the brain areas that are connected to the amygdala36 and involved in the specified function is given in each box. The list of connected areas is by no means complete and does not reflect the level of detail gained from anatomical studies. The multicoloured circles represent multidimensional neurons in the amygdala; each colour represents a stimulus from the brain areas highlighted in the same colour.
The internal structure and connectivity of the amygdala is best understood from a developmental perspective. Unlike other multinucleate structures, the nuclei of the amygdala originate from different compartments of the developing telencephalon33–35. Neurons in the basolateral nuclei (that is, the lateral, basal and accessory basal nuclei) and in all cortical areas originate from the pallium; by contrast, the central and medial nuclei, together with the extended amygdala, develop from the subpallium (which also gives rise to the caudate, putamen and globus pallidus). This distinction is important because the basolateral and centromedial nuclei share cell types, circuit motifs and patterns of connectivity with the cortex and the basal ganglia, respectively.
Despite different patterns of cell migration, neurons of the same origin become more extensively and reciprocally interconnected than neurons of different origins. Indeed, the basolateral nuclei receive and send projections from and to the vast majority of cortical areas (reviewed in detail elsewhere36) and contribute as much as, if not more than, the central nuclei to amygdalar output36. The central and medial nuclei establish bidirectional connections with an equally numerous list of subcortical structures, including the thalamus, the hypothalamus and the autonomic and homeostatic centres in the brainstem88–90. Overall, a complex web of intranuclear and internuclear connections (illustrated in FIG. 3) integrates the signals processed by the cortical and subcortical loops that intersect in the amygdala91. Multidimensional processing that emerges from the crosstalk between these loops reflects the spectrum of information processed in the amygdala. Accordingly, a neuron that responds to salient, attention-inducing stimuli, as well as to faces and reward magnitude, may process inputs from the network of cortical and subcortical areas implicated in visual attention92, from the temporal and prefrontal face-processing areas93,94 and from reward circuits95.
The contribution of microcircuits.
The internal architecture of the amygdala constrains the response selectivity of its component neurons. Most cells in the central nuclei are GABAergic and establish inhibitory interactions characteristic of basal ganglia circuits, whereas the cell types (defined based on their molecular, morphological, connectional and functional properties) in the basolateral nuclei are highly similar to those in the cortex33,96. These neurons in the basolateral amygdala establish excitatory–inhibitory interactions that seem to recapitulate the interaction patterns between cortical neurons. These interactions shape response patterns along three axes: response magnitude, response duration (promoting phasic or tonic responses) and response polarity (leading to a decrease or increase of the firing rate)48,62. For example, a neuron in the monkey amygdala may discriminate between social and non-social stimuli through its response duration (exhibiting phasic responses to faces and tonic responses to objects) and, at the same time, discriminate between individual faces through differences in the amplitude of the phasic response48. These three types of spike-train variations have been shown to discriminate between sensory modalities62. Although no relationship was found between a particular spike-train metric and a specific sensory modality (for example, it was not the case that all visual responses were phasic and all tactile responses tonic) in this study, multisensory neurons in the amygdala ‘voted’ for each sensory modality with a different spike-train metric.
Inhibition and disinhibition play crucial roles in tuning the selectivity of amygdalar neurons, by blocking, reducing or enhancing (through disinhibition) the relative contributions of certain inputs97 and thus dynamically altering the output of small circuits. The duration of responses elicited by each stimulus might also be modified through inhibition (BOX 1).
Box 1 |. A circuit account for the polarity, size and length of multidimensional responses.
Most amygdalar neurons are multimodal Spike density functions of a typical amygdala neuron receiving excitatory inputs for visual, auditory and tactile response, based on data in REF.62, are shown (see the figure, part a).
In addition to being multisensory, the prototypical amygdalar neuron is also multidimensional as it responds with phasic increases and decreases in activity to alerting stimuli, such as the fixation cue (light-green arrow), sensory stimuli (red, green and purple arrows, representing visual, tactile and auditory stimuli, respectively) and reward (blue arrows) (see the figure, part a).
The patterns of activity seen in this neuron might arise from interactions among different cell types; specifically, the phasic decrease in firing rate at the appearance of the fixation icon (see the figure, part a, light-green arrow) might be explained by activation of GABAergic inhibitory interneurons expressing parvalbumin (PV1 neurons; see the figure, part b) that directly inhibit the soma129,130.
After successful fixation, one of the sensory stimuli is delivered (see the figure, part a, arrows at time 0). Here, again, inhibition probably helps shape this neuron’s distinctive firing responses to different stimuli. For example, the phasic decrease in response to the tactile stimulus (see the figure, part a, green trace) could be explained by the activation of the same PV1 neurons that directly inhibit the soma. visual stimuli might activate the proximal dendrites of the pyramidal cell, causing a phasic elevation of the firing rate (see the figure, part a, red trace), and also activate the PV1 neuron that, through somatic inhibition, shortens the maximal response duration. PV1 neuron-mediated inhibition does not reduce the response to baseline because the visual inputs also activate a second set of GABAergic inhibitory interneurons (PV2 neurons) that inhibit a third type of inhibitory interneuron, somatostatin-positive (SST) neurons (see the figure, part b)129,130. the SST neurons tonically inhibit the distal dendrites and the removal of this inhibition, via disinhibition from PV2 neurons, can explain the tonic elevation of firing rate in response to the visual stimulus after the phasic response subsides (around 0.6 s after stimulus delivery; see the figure, part a, red trace). The auditory stimulus causes a small increase in firing rate in the first third of the trial (connection not shown) followed by a decrease in the latter two-thirds of the trial (see the figure, part a, purple trace; each trial lasts 1 s), potentially owing to weak inhibition of the distal dendrite by an SST neuron.
After the sensory stimuli were removed (see the figure, part a, indicated by the second vertical dashed line), the same neuron signalled each juice drop received as a reward, perhaps by activating excitatory neurons (see the figure, part a, blue arrows) that directly contact the pyramidal neuron.
This is only one example of the many interactions that can be set up in simplified circuits. Given that multidimensional neurons can use phasic–tonic distinctions in their responses, the ensemble activity and, implicitly, the specificity of the representation instantiated by that ensemble, can change substantially when the phasic responses decay but the tonic responses are ongoing.
For example, multidimensional neurons were commonly recorded from the basolateral nuclei of the amygdala while the monkey was presented with eight different visual, tactile and auditory stimuli62. A simplified, hypothetical circuit that generally accounts for the observed changes in firing rate of such multidimensional neurons showed how, even without the inevitable complexities added by the effects of acetylcholine and serotonin, the same neuron could be part of different ensembles that signal attention, sensory stimuli and reward (BOX 1). This simple circuit also helped to reveal the importance of considering the temporal dimension in neural activity: specifically, at two different points, the same neuron (or ensemble) can represent markedly different information about stimuli or context.
Conclusions
The goal of this Perspective has been to discuss the role of the primate amygdala in social behaviour in light of growing evidence that its component neurons have multidimensional responses. There is no evidence for social specializations in the primate amygdala at the nuclear or single-neuron level. On the contrary, it seems that the same neurons contribute to multiple neural ensembles that contain different information at different times. These ensembles may contain abstract or high-dimensional representations to enable flexibility and both social and non-social behaviours. Multidimensional selectivity at the level of the individual neuron therefore reflects the sequential activation–inactivation of brain-wide circuits as the behaviour unfolds.
At first sight, this organizational scheme appears at odds with the known neuroanatomy (that is, the localized destination of inputs and origin of outputs) in primates and the highly compartmentalized view of the rodent amygdala that has emerged from recent cell type-specific circuit dissections. The experimental evidence that supports each view depends, to a large extent, on the advantages and limitations of the approach and techniques used to generate the data. For example, single-cell and ensemble recordings in human and non-human primates are agnostic of the genetic and neurochemical identity of the recorded neurons (although some non-genetic tools recently became available98), whereas cell type-specific tasks in rodents rarely test other sensory modalities or use multi-alternative tasks to reveal multidimensional processing.
The road to better understanding of the amygdala goes through improved technologies and, ideally, technologies that can be shared across species. Note, however, that technological complexity without theoretical sophistication will only generate more data instead of a deeper understanding of the amygdala. The difficulty in answering the question ‘what is the role of the amygdala in social behaviour?’ illustrates the need to move away from defining the role of this complex hub in terms of simple psychological constructs. Progress is likely to come from developing theoretical and computational constructs that can account for both the specificity and flexibility of multidimensional representations at the level of neural populations and their dynamics.
Acknowledgements
The author thanks U. Rutishauser and J. Minxha for helpful comments on an earlier versions of this manuscript and A. Martin for assistance. Supported by grants NIH/NIMH R01 MH121009 and NIH/NIMH P50MH100023.
Footnotes
Competing interests
The author declares no competing interests
Peer review information
Nature Reviews Neuroscience thanks S. Chang, L. Malkova and R. Paz for their contribution to the peer review of this work.
References
- 1.Laberge F, Mühlenbrock-Lenter S, Grunwald W & Roth G Evolution of the amygdala: new insights from studies in amphibians. Brain Behav. Evol. 67, 177–187 (2006). [DOI] [PubMed] [Google Scholar]
- 2.Brown S & Schafer EA XI. An investigation into the functions of the occipital and temporal lobes of the monkey’s brain. Philos. Trans. R. Soc. London 179, 303–327 (1888). [Google Scholar]
- 3.Adolphs R What does the amygdala contribute to social cognition? Ann. N. Y. Acad. Sci. 1191, 42–61 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Meunier M, Bachevalier J, Murray EA, Málková L & Mishkin M Effects of aspiration versus neurotoxic lesions of the amygdala on emotional responses in monkeys: amygdala lesions and emotional responses. Eur. J. Neurosci 11, 4403–4418 (1999). [DOI] [PubMed] [Google Scholar]
- 5.Wassum KM & Izquierdo A The basolateral amygdala in reward learning and addiction. Neurosci. Biobehav. Rev 57, 271–283 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mobbs D et al. Viewpoints: approaches to defining and investigating fear. Nat. Neurosci 22, 1205–1216 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kim J, Zhang X, Muralidhar S, LeBlanc SA & Tonegawa S Basolateral to central amygdala neural circuits for appetitive behaviors. Neuron 93, 1464–1479.e5 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Li B Central amygdala cells for learning and expressing aversive emotional memories. Curr. Opin. Behav. Sci 26, 40–45 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Pignatelli M & Beyeler A Valence coding in amygdala circuits. Curr. Opin. Behav. Sci 26, 97–106 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cai H, Haubensak W, Anthony TE & Anderson DJ Central amygdala PKC-δ+ neurons mediate the influence of multiple anorexigenic signals. Nat. Neurosci 17, 1240–1248 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Douglass AM et al. Central amygdala circuits modulate food consumption through a positive-valence mechanism. Nat. Neurosci. 20, 1384–1394 (2017). [DOI] [PubMed] [Google Scholar]
- 12.Han W et al. Integrated control of predatory hunting by the central nucleus of the amygdala. Cell 168, 311–324.e18 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Headley DB, Kanta V, Kyriazi P & Paré D Embracing complexity in defensive networks. Neuron 103, 189–201 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Shabel SJ & Janak PH Substantial similarity in amygdala neuronal activity during conditioned appetitive and aversive emotional arousal. Proc. Natl Acad. Sci. USA 106, 15031–15036 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Peck CJ, Lau B & Salzman CD The primate amygdala combines information about space and value. Nat. Neurosci 16, 340–348 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kyriazi P, Headley DB & Pare D Multi-dimensional coding by basolateral amygdala neurons. Neuron 99, 1315–1328.e5 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Putnam PT & Gothard KM Multidimensional neural selectivity in the primate amygdala. eNeuro 10.1523/ENEURO.0153-19.2019 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Rigotti M et al. The importance of mixed selectivity in complex cognitive tasks. Nature 497, 585–590 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Fusi S, Miller EK & Rigotti M Why neurons mix: high dimensionality for higher cognition. Curr. Opin. Neurobiol 37, 66–74 (2016). [DOI] [PubMed] [Google Scholar]
- 20.Bickart KC, Dickerson BC & Feldman Barrett L The amygdala as a hub in brain networks that support social life. Neuropsychologia 63, 235–248 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Rutishauser U, Mamelak AN & Adolphs R The primate amygdala in social perception — insights from electrophysiological recordings and stimulation. Trends Neurosci 38, 295–306 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kling A, Lancaster J & Benitone J Amygdalectomy in the free-ranging vervet (Cercopithecus aethiops). J. Psychiatr. Res 7, 191–199 (1970). [DOI] [PubMed] [Google Scholar]
- 23.Bachevalier J, Málková L & Mishkin M Effects of selective neonatal temporal lobe lesions on socioemotional behavior in infant rhesus monkeys (Macaca mulatta). Behav. Neurosci 115, 545–559 (2001). [DOI] [PubMed] [Google Scholar]
- 24.Livneh U, Resnik J, Shohat Y & Paz R Self-monitoring of social facial expressions in the primate amygdala and cingulate cortex. Proc. Natl Acad. Sci. USA 109, 18956–18961 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Burgos-Robles A, Gothard KM, Monfils MH, Morozov A & Vicentic A Conserved features of anterior cingulate networks support observational learning across species. Neurosci. Biobehav. Rev 107, 215–228 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Munuera J, Rigotti M & Salzman CD Shared neural coding for social hierarchy and reward value in primate amygdala. Nat. Neurosci 21, 415–423 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Minxha J et al. Fixations gate species-specific responses to free viewing of faces in the human and macaque amygdala. Cell Rep 18, 878–891 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Klumpp H & Fitzgerald JM Neuroimaging predictors and mechanisms of treatment response in social anxiety disorder: an overview of the amygdala. Curr. Psychiatry Rep 20, 89 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gothard KM The amygdalo-motor pathways and the control of facial expressions. Front. Neurosci 8, 43 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Rutishauser U et al. Representation of retrieval confidence by single neurons in the human medial temporal lobe. Nat. Neurosci 18, 1041–1050 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kamiński J et al. Persistently active neurons in human medial frontal and medial temporal lobe support working memory. Nat. Neurosci 20, 590–601 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Laine CM, Spitler KM, Mosher CP & Gothard KM Behavioral triggers of skin conductance responses and their neural correlates in the primate amygdala. J. Neurophysiol 101, 1749–1754 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Swanson LW & Petrovich GD What is the amygdala? Trends Neurosci 21, 323–331 (1998). [DOI] [PubMed] [Google Scholar]
- 34.Sah P, Faber ESL, Lopez De Armentia M & Power J The amygdaloid complex: anatomy and physiology. Physiol. Rev 83, 803–834 (2003). [DOI] [PubMed] [Google Scholar]
- 35.Pessoa L, Medina L, Hof PR & Desfilis E Neural architecture of the vertebrate brain: implications for the interaction between emotion and cognition. Neurosci. Biobehav. Rev 107, 296–312 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Amaral DG, Price JL, Pitkanen A & Carmichael TS in The Amygdala: Neurobiological Aspects of Emotion, Memory, and Mental Dysfunction (ed. Aggleton JP) 1–66 (Wiley-Liss, 1992). [Google Scholar]
- 37.Kalin NH, Shelton SE & Davidson RJ The role of the central nucleus of the amygdala in mediating fear and anxiety in the primate. J. Neurosci 24, 5506–5515 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Wellman LL, Forcelli PA, Aguilar BL & Malkova L Bidirectional control of social behavior by activity within basolateral and central amygdala of primates. J. Neurosci 36, 8746–8756 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Forcelli PA, Wellman LL & Malkova L Blockade of glutamatergic transmission in the primate basolateral amygdala suppresses active behavior without altering social interaction. Behav. Neurosci 131, 192–200 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Málková L, Gaffan D & Murray EA Excitotoxic lesions of the amygdala fail to produce impairment in visual learning for auditory secondary reinforcement but interfere with reinforcer devaluation effects in rhesus monkeys. J. Neurosci 17, 6011–6020 (1997). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Chudasama Y, Izquierdo A & Murray EA Distinct contributions of the amygdala and hippocampus to fear expression: effects of amygdala and hippocampal lesions on emotion. Eur. J. Neurosci 30, 2327–2337 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Apps MAJ, Rushworth MFS & Chang SWC The anterior cingulate gyrus and social cognition: tracking the motivation of others. Neuron 90, 692–707 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Grimaldi P, Saleem KS & Tsao D Anatomical connections of the functionally defined “face patches” in the macaque monkey. Neuron 90, 1325–1342 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Gothard KM, Battaglia FP, Erickson CA, Spitler KM & Amaral DG Neural responses to facial expression and face identity in the monkey amygdala. J. Neurophysiol 97, 1671–1683 (2007). [DOI] [PubMed] [Google Scholar]
- 45.Mosher CP, Zimmerman PE & Gothard KM Neurons in the monkey amygdala detect eye contact during naturalistic social interactions. Curr. Biol 24, 2459–2464 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Rutishauser U et al. Single-unit responses selective for whole faces in the human amygdala. Curr. Biol 21, 1654–1660 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Nishijo H, Ono T & Nishino H Topographic distribution of modality-specific amygdalar neurons in alert monkey. J. Neurosci 8, 3556–3569 (1988). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Mosher CP, Zimmerman PE & Gothard KM Response characteristics of basolateral and centromedial neurons in the primate amygdala. J. Neurosci 30, 16197–16207 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Zhang W et al. Functional circuits and anatomical distribution of response properties in the primate amygdala. J. Neurosci. 33, 722–733 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Grabenhorst F, Báez-Mendoza R, Genest W, Deco G & Schultz W Primate amygdala neurons simulate decision processes of social partners. Cell 177, 986–998.e15 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Buzsáki G Neural syntax: cell assemblies, synapsembles, and readers. Neuron 68, 362–385 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Yuste R From the neuron doctrine to neural networks. Nat. Rev. Neurosci 16, 487–497 (2015). [DOI] [PubMed] [Google Scholar]
- 53.Morrow JK, Cohen MX & Gothard KM Mesoscopic-scale functional networks in the primate amygdala. bioRxiv 10.1101/2020.02.24.963587 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Wang F, Zhu J, Zhu H, Zhang Q, Lin Z & Hu H Bidirectional control of social hierarchy by synaptic efficacy in medial prefrontal cortex. Science 334, 693–697 (2011). [DOI] [PubMed] [Google Scholar]
- 55.Walum H & Young LJ The neural mechanisms and circuitry of the pair bond. Nat. Rev. Neurosci 19, 643–654 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Li Y et al. Neuronal representation of social information in the medial amygdala of awake behaving mice. Cell 171, 1176–1190.e17 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Resnik J & Paz R Fear generalization in the primate amygdala. Nat. Neurosci 18, 188–190 (2015). [DOI] [PubMed] [Google Scholar]
- 58.Duvarci S & Pare D Amygdala microcircuits controlling learned fear. Neuron 82, 966–980 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Janak PH & Tye KM From circuits to behaviour in the amygdala. Nature 517, 284–292 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Gafford GM & Ressler KJ Mouse models of fear-related disorders: cell-type-specific manipulations in amygdala. Neuroscience 321, 108–120 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Fadok JP, Markovic M, Tovote P & Lüthi A New perspectives on central amygdala function. Curr. Opin. Neurobiol 49, 141–147 (2018). [DOI] [PubMed] [Google Scholar]
- 62.Morrow J, Mosher C & Gothard K Multisensory neurons in the primate amygdala. J. Neurosci 39, 3663–3675 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Reitich-Stolero T & Paz R Affective memory rehearsal with temporal sequences in amygdala neurons. Nat. Neurosci 22, 2050–2059 (2019). [DOI] [PubMed] [Google Scholar]
- 64.Morrison SE & Salzman CD Re-valuing the amygdala. Curr. Opin. Neurobiol 20, 221–230 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Costa VD, Mitz AR & Averbeck BB Subcortical substrates of explore–exploit decisions in primates. Neuron 103, 533–545.e5 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Grabenhorst F, Hernadi I & Schultz W Prediction of economic choice by primate amygdala neurons. Proc. Natl Acad. Sci. USA 109, 18950–18955 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Averbeck BB & Costa VD Motivational neural circuits underlying reinforcement learning. Nat. Neurosci 20, 505–512 (2017). [DOI] [PubMed] [Google Scholar]
- 68.Saez A, Rigotti M, Ostojic S, Fusi S & Salzman CD Abstract context representations in primate amygdala and prefrontal cortex. Neuron 87, 869–881 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69.Bernardi S & Salzman CD The contribution of nonhuman primate research to the understanding of emotion and cognition and its clinical relevance. Proc. Natl Acad. Sci. USA 116, 26305–26312 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70.Nishijo H, Ono T & Nishino H Single neuron responses in amygdala of alert monkey during complex sensory stimulation with affective significance. J. Neurosci 8, 3570–3583 (1988). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71.Faraut MCM et al. Dataset of human medial temporal lobe single neuron activity during declarative memory encoding and recognition. Sci. Data 5, 180010 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 72.Warden MR & Miller EK Task-dependent changes in short-term memory in the prefrontal cortex. J. Neurosci 30, 15801–15810 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 73.Johnston WJ, Palmer SE & Freedman DJ Nonlinear mixed selectivity supports reliable neural computation. PLoS Comput. Biol 16, e1007544 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 74.Lindsay GW, Rigotti M, Warden MR, Miller EK & Fusi S Hebbian learning in a random network captures selectivity properties of the prefrontal cortex. J. Neurosci 37, 11021–11036 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 75.LeCun Y, Bengio Y & Hinton G Deep learning. Nature 521, 436–444 (2015). [DOI] [PubMed] [Google Scholar]
- 76.Pessoa L Emotion and cognition and the amygdala: from “what is it?” to “what’s to be done?”. Neuropsychologia 48, 3416–3429 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 77.Mosher CP, Zimmerman PE, Fuglevand AJ & Gothard KM Tactile stimulation of the face and the production of facial expressions activate neurons in the primate amygdala. eNeuro 3, 0182–0916 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78.Craig AD How do you feel? Interoception: the sense of the physiological condition of the body. Nat. Rev. Neurosci 3, 655–666 (2002). [DOI] [PubMed] [Google Scholar]
- 79.Burgess CR, Livneh Y, Ramesh RN & Andermann ML Gating of visual processing by physiological need. Curr. Opin. Neurobiol 49, 16–23 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80.Bernardi S et al. The geometry of abstraction in hippocampus and pre-frontal cortex. bioRxiv 10.1101/408633 (2018). [DOI] [Google Scholar]
- 81.Zhang CY et al. Partially mixed selectivity in human posterior parietal association cortex. Neuron 95, 697–708.e4 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82.Stringer C et al. Spontaneous behaviors drive multidimensional, brainwide activity. Science 364, eaav7893 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83.Bialek W et al. Statistical mechanics for natural flocks of birds. Proc. Natl Acad. Sci. USA 109, 4786–4791 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 84.Roy AK et al. Functional connectivity of the human amygdala using resting state fMRI. NeuroImage 45, 614–626 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 85.Mears D & Pollard HB Network science and the human brain: using graph theory to understand the brain and one of its hubs, the amygdala, in health and disease: graph theory, the brain, and the amygdala. J. Neurosci. Res 94, 590–605 (2016). [DOI] [PubMed] [Google Scholar]
- 86.Tomasi D & Volkow ND Association between functional connectivity hubs and brain networks. Cereb. Cortex 21, 2003–2013 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87.Insel TR & Fernald RD How the brain processes social information: searching for the social brain. Annu. Rev. Neurosci 27, 697–722 (2004). [DOI] [PubMed] [Google Scholar]
- 88.Mehler WR Subcortical afferent connections of the amygdala in the monkey. J. Comp. Neurol. 190, 733–762 (1980). [DOI] [PubMed] [Google Scholar]
- 89.Heimer L & Van Hoesen GW The limbic lobe and its output channels: implications for emotional functions and adaptive behavior. Neurosci. Biobehav. Rev 30, 126–147 (2006). [DOI] [PubMed] [Google Scholar]
- 90.Ernst M & Fudge JL A developmental neurobiological model of motivated behavior: anatomy, connectivity and ontogeny of the triadic nodes. Neurosci. Biobehav. Rev 33, 367–382 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 91.Pitkänen A & Amaral DG Organization of the intrinsic connections of the monkey amygdaloid complex: projections originating in the lateral nucleus. J. Comp. Neurol 398, 431–458 (1998). [DOI] [PubMed] [Google Scholar]
- 92.Bogadhi AR, Bollimunta A, Leopold DA & Krauzlis RJ Brain regions modulated during covert visual attention in the macaque. Sci. Rep 8, 15237 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 93.Freiwald WA The neural mechanisms of face processing: cells, areas, networks, and models. Curr. Opin. Neurobiol 60, 184–191 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 94.Hwang J & Romanski LM Prefrontal neuronal responses during audiovisual mnemonic processing. J. Neurosci 35, 960–971 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 95.Schultz W Neuronal reward and decision signals: from theories to data. Physiol. Rev 95, 853–951 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 96.Morgan JT & Amaral DG Comparative analysis of the dendritic organization of principal neurons in the lateral and central nuclei of the rhesus macaque and rat amygdala: dendritic arbors in rhesus and rat amygdala. J. Comp. Neurol 522, 689–716 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 97.Wang X-J & Yang GR A disinhibitory circuit motif and flexible information routing in the brain. Curr. Opin. Neurobiol 49, 75–83 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 98.Mosher CP et al. Cellular classes in the human brain revealed in vivo by heartbeat-related modulation of the extracellular action potential waveform. Cell Rep 30, 3536–3551.e6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 99.Klüver H & Bucy PC Preliminary analysis of functions of the temporal lobes in monkeys. Arch. Neurol. Psychiatry 42, 979 (1939). [DOI] [PubMed] [Google Scholar]
- 100.Rosvold HE, Mirsky AF & Pribram KH Influence of amygdalectomy on social behavior in monkeys. J. Comp. Physiol. Psychol. 47, 173–178 (1954). [DOI] [PubMed] [Google Scholar]
- 101.Bliss-Moreau E, Bauman MD & Amaral DG Neonatal amygdala lesions result in globally blunted affect in adult rhesus macaques. Behav. Neurosci 125, 848–858 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 102.Weiskrantz L Behavioral changes associated with ablation of the amygdaloid complex in monkeys. J. Comp. Physiol. Psychol 49, 381–391 (1956). [DOI] [PubMed] [Google Scholar]
- 103.Kling A & Cornell R Amygdalectomy and social behavior in the caged stump-tailed macaque (Macaca speciosa). Folia Primatol 14, 190–208 (1971). [DOI] [PubMed] [Google Scholar]
- 104.Machado CJ & Bachevalier J The impact of selective amygdala, orbital frontal cortex, or hippocampal formation lesions on established social relationships in rhesus monkeys (Macaca mulatta). Behav. Neurosci 120, 761–786 (2006). [DOI] [PubMed] [Google Scholar]
- 105.Machado CJ & Bachevalier J Behavioral and hormonal reactivity to threat: effects of selective amygdala, hippocampal or orbital frontal lesions in monkeys. Psychoneuroendocrinology 33, 926–941 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 106.Dicks D, Myers RE & Kling A Uncus and amiygdala lesions: effects on social behavior in the free-ranging rhesus monkey. Science 165, 69–71 (1969). [DOI] [PubMed] [Google Scholar]
- 107.Bliss-Moreau E, Moadab G, Bauman MD & Amaral DG The impact of early amygdala damage on juvenile rhesus macaque social behavior. J. Cognit. Neurosci. 25, 2124–2140 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 108.Kalin NH, Shelton SE, Davidson RJ & Kelley AE The primate amygdala mediates acute fear but not the behavioral and physiological components of anxious temperament. J. Neurosci 21, 2067–2074 (2001). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 109.Aggleton J & Passingham R Syndrome produced by lesions of the amygdala in monkeys (Macaca mulatta). J. Comp. Physiol. Psychol 95, 961–977 (1981). [DOI] [PubMed] [Google Scholar]
- 110.Malkova L, Mishkin M, Suomi SJ & Bachevalier J Long-term effects of neonatal medial temporal ablations on socioemotional behavior in monkeys (Macaca mulatta). Behav. Neurosci 124, 742–760 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 111.Thompson CI, Bergland RM & Towfighi JT Social and nonsocial behaviors of adult rhesus monkeys after amygdalectomy in infancy or adulthood. J. Comp. Physiol. Psychol 91, 533–548 (1977). [DOI] [PubMed] [Google Scholar]
- 112.Thompson CI, Schwartzbaum JS & Harlow HF Development of social fear after amygdalectomy in infant rhesus monkeys. Physiol. Behav 4, 249–254 (1969). [Google Scholar]
- 113.Prather MD et al. Increased social fear and decreased fear of objects in monkeys with neonatal amygdala lesions. Neuroscience 106, 653–658 (2001). [DOI] [PubMed] [Google Scholar]
- 114.Bauman MD, Lavenex P, Mason WA, Capitanio JP & Amaral DG The development of social behavior following neonatal amygdala lesions in rhesus monkeys. J. Cognit. Neurosci 16, 1388–1411 (2004). [DOI] [PubMed] [Google Scholar]
- 115.Bauman MD, Toscano JE, Mason WA, Lavenex P & Amaral DG The expression of social dominance following neonatal lesions of the amygdala or hippocampus in rhesus monkeys (Macaca mulatta). Behav. Neurosci 120, 749–760 (2006). [DOI] [PubMed] [Google Scholar]
- 116.Thompson CI & Towfighi JT Social behavior of juvenile rhesus monkeys after amygdalectomy in infancy. Physiol. Behav 17, 831–836 (1976). [DOI] [PubMed] [Google Scholar]
- 117.Izquierdo A, Suda RK & Murray EA Comparison of the effects of bilateral orbital prefrontal cortex lesions and amygdala lesions on emotional responses in rhesus monkeys. J. Neurosci 25, 8534 (2005). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 118.Machado CJ et al. Bilateral neurotoxic amygdala lesions in rhesus monkeys (Macaca mulatta): consistent pattern of behavior across different social contexts. Behav. Neurosci 122, 251–266 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 119.Moadab G, Bliss-Moreau E & Amaral DG Adult social behavior with familiar partners following neonatal amygdala or hippocampus damage. Behav. Neurosci 129, 339–350 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 120.Emery NJ et al. The effects of bilateral lesions of the amygdala on dyadic social interactions in rhesus monkeys (Macaca mulatta). Behav. Neurosci 115, 515–544 (2001). [PubMed] [Google Scholar]
- 121.Newman JD & Bachevalier J Neonatal ablations of the amygdala and inferior temporal cortex alter the vocal response to social separation in rhesus macaques. Brain Res 758, 180–186 (1997). [DOI] [PubMed] [Google Scholar]
- 122.Dal Monte O, Costa VD, Noble PL, Murray EA & Averbeck BB Amygdala lesions in rhesus macaques decrease attention to threat. Nat. Commun 6, 10161 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 123.Bliss-Moreau E, Moadab G, Santistevan A & Amaral DG The effects of neonatal amygdala or hippocampus lesions on adult social behavior. Behav. Brain Res. 322, 123–137 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 124.Kling A Effects of amygdalectomy and testosterone on sexual behavior of male juvenile macaques. J. Comp. Physiol. Psychol. 65, 466–471 (1968). [DOI] [PubMed] [Google Scholar]
- 125.Toscano JE, Bauman MD, Mason WA & Amaral DG Interest in infants by female rhesus monkeys with neonatal lesions of the amygdala or hippocampus. Neuroscience 162, 881–891 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 126.Raper J, Bachevalier J, Wallen K & Sanchez M Neonatal amygdala lesions alter basal cortisol levels in infant rhesus monkeys. Psychoneuroendocrinology 38, 818–829 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 127.Mitz AR, Chacko RV, Putnam PT, Rudebeck PH & Murray EA Using pupil size and heart rate to infer affective states during behavioral neurophysiology and neuropsychology experiments. J. Neurosci. Methods 279, 1–12 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 128.Amaral DG et al. The amygdala: is it an essential component of the neural network for social cognition? Neuropsychologia 41, 235–240 (2003). [DOI] [PubMed] [Google Scholar]
- 129.Letzkus JJ, Wolff SBE & Lüthi A Disinhibition, a circuit mechanism for associative learning and memory. Neuron 88, 264–276 (2015). [DOI] [PubMed] [Google Scholar]
- 130.Lucas EK & Clem RL GABAergic interneurons: the orchestra or the conductor in fear learning and memory? Brain Res. Bull 141, 13–19 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]




