The case for hemispheric lateralization of the human amygdala in fear processing

A systematic review by Sehlmeyer et al. (2009) suggests that findings on the laterality of amygdala activity in fear conditioning studies remain inconclusive [10]. Specifically, nine studies reported bilateral amygdala activation, eight detected left-lateralized activation, and another eight detected right-lateralized activation. To address whether amygdalae play lateralizing functions in fear conditioning, however, fear acquisition and expression must be disambiguated methodologically.

For instance, Petrovic et al. (2008) found that learning positive or negative affective values for faces was associated with the left amygdala activation. Specifically, the authors used a reinforcement learning algorithm that assigned values to faces as a function of a historical pairing/un-pairing with aversive stimuli, and they found that the left amygdala activation correlated with the output of the reinforcement learning model that expressed change in face values [21]. Similarly, Phelps et al. (2001) found significant left amygdala activation related to instructed fear learning [22], and Olsson et al. (2007) found that observational fear learning also preferentially activated the left amygdala [23]. Delgado et al. (2008) found that cognitive re-evaluation of the CS to reduce the conditioned response correlated with left amygdala activation [24, 25]. The BLA is reciprocally connected with the hippocampus and medial prefrontal cortices [5], and it has been particularly implicated in fear acquisition [1]. Given the correlation between the cognitively mediated fear learning and left amygdala activation, and the primary role of the left amygdala in cognitive-mediated emotional processes (Fig. 1c-top), we propose that the left amygdala (particularly the BLA) is specialized for cognitive-mediated fear acquisition (Fig. 1c-bottom).

In contrast to the studies described above, Cheng et al. (2003) and Knight et al. (2005) sought to separate fear expression-related brain activity during the fear conditioning process by focusing on the autonomic fear response, and found that right amygdala activity was more closely tied to conditioning- related autonomic fear responses [26, 27]. To further dissociate fear expression from acquisition, Cheng et al. (2006) categorized CS+ trials based on whether they elicited a conditioned SCR. Specifically, they categorized the CS+ trials into those trials in which the subjects learned to associate the CS+ with an aversive stimulus and produced a conditioned SCR (CS+ response), and into those trials in which the subjects learned the association but did not exhibit a conditioned SCR (CS+ nonresponse). They found that right amygdala activation was significantly increased for ‘CS+ response’ but not for ‘CS+ nonresponse’ trials [28]. Notably, our selective thermal ablation of the right amygdalohippocampal complex in two epilepsy patients resulted in improved PTSD symptoms, including reduced SCR to arousing stimuli, alleviation of hyperarousal, and better tolerance of cognitive therapy [29]. Subsequent additional cases of highly selective right amygdalotomy in epilepsy patients have confirmed relief of PTSD symptoms, including decreased response to trauma reminders and reduced hyperarousal (J. T. Willie, unpublished results). The CMA are a primary source of amygdalar output, with direct projections to regions involved in autonomic responses [3]. Accordingly, given the correlation between conditioning-related autonomic response and right amygdala activation, and prominent relief of hyperarousal associated with right amygdala removal, we propose that the right amygdala (especially output mediated by CMA) may be particularly important for sensory-mediated emotional processes, including fear expression (Fig. 1c-bottom).

Fear extinction is the process of reducing a previously acquired CR following repeated presentation of the CS in the absence of the aversive stimulus. It is an active learning process that ultimately leads to the formation of a consolidated extinction memory (Fig. 1a, b). We re-evaluated studies included among prior systematic/meta-reviews of fear extinction studies of healthy humans using fMRI [10, 30, 31]. While left and/or right amygdala activations were reported in individual studies, previous reviewers drew no conclusions regarding amygdala laterality during extinction learning. Overall, fourteen studies reported significant activation in either amygdala during extinction learning (Table 1) [32,33,34,35,36,37,38,39,40,41,42,43,44,45], with ten studies reporting significant activation exclusively on the left amygdala [32,33,34,35,36,37,38,39,40,41] and four studies reporting bilateral amygdala activation [42,43,44,45]. None of the studies indicated a greater right amygdala activation during fear extinction (Table 1). Thus, while both amygdalae participate in neural processes in extinction learning, the left amygdala may play the more dominant role.

To better understand the potential predominance of the left amygdala in extinction, we examined the specific methodologies employed in previous studies (Table 1) [32,33,34,35,36,37,38,39,40,41,42,43,44,45]. Specifically, five studies identified significant and exclusive left amygdala activation during various phases (i.e., early [38], late [32, 33], and whole phases [36, 40]) of extinction by contrasting CS+ with CS−. Schiller et al. (2013) and Apergis-Schoute et al. (2014), focusing on the left amygdala as the region of interest, showed increased activation by contrasting CS+ with CS− in the early phase of extinction learning [35, 37]. Martynova et al. (2020) explored functional connectivity between each amygdala and various brain regions following extinction and found the left amygdala to exhibit stronger functional connectivity [41]. Linnman et al. (2012) observed that left amygdala activation decreased from early to late phases of extinction [34]. Molapour et al. (2015) found that left amygdala activation reflected the differentiation of facial features during extinction [39]. Even among the four studies reporting bilateral amygdala activation, there were still indications of focal asymmetries. Specifically, Gottfried et al. (2004) found a broader activation in the left amygdala (i.e., lateral and medial nuclei) compared to the right amygdala (i.e., lateral nucleus) [43]. Knight et al. (2004) observed an initially increased and decreased activation in the right and left amygdala, respectively, but right amygdala activation quickly returned to pre-extinction levels, whereas left amygdala activation persisted across the whole phase of extinction [44].

This predominance of left amygdala activation across extinction studies was unlikely to have resulted from confounding methodology such as asymmetric sensory presentation in the aversive stimuli. Specifically, among the studies that reported exclusively left amygdala activation, three studies used symmetric presentations (including bilaterally presented auditory white noise [33], viewing film-clips [36], and mechanical rectal distension [38]), four studies delivered electric shocks to the right wrist/foot [35, 37, 39, 41], and one delivered electric shocks to either the right or the left wrist [40]. Among the studies with varying combinations of left and right aversive stimuli, left amygdala activation was nevertheless consistently observed.

With respect to amygdala nuclei, different neural populations in the BLA may play critical and distinct roles in fear acquisition and extinction. In rodents, the lateral nucleus of the BLA was often associated with fear acquisition, while the basal nucleus exhibited a mixed population of neurons (some contributed to fear acquisition, while others contributed to extinction learning) [1]. Overall, phases of acquisition and extinction of fear responses (i.e., fear and safety learning) reflect similar cognitive-mediated processes, leading us to propose that the left amygdala, and especially the left BLA, may be particularly critical for both fear acquisition and extinction learning (Fig. 1c-bottom).

LeDoux & Pine (2016) proposed a ‘two systems’ framework to help better understand fear and anxiety responses: one set of neural circuits predominantly controlling subconscious physiology and reflexive behavior in response to threats, and a second set for generating conscious awareness and evaluation of fear and anxiety. The first system, composed of subcortical areas such as the amygdala, hippocampus, nucleus accumbens (NAcc), and bed nucleus of the stria terminalis (BNST), is proposed to operate subconsciously. The second system, primarily involving cortical areas such as the lateral and medial prefrontal cortices, features the circuits through which conscious experiences emerge [46]. Studies indicate that the ventromedial prefrontal cortex (vmPFC) reduces amygdala activity (to inhibit fear expression and promote extinction), while the dorsal anterior cingulate (dACC) activates amygdala (to increase fear expression and oppose extinction) [47]. Furthermore, the dorsolateral prefrontal cortex (dlPFC) is involved in working memory and emotion regulation [48, 49] and is a crucial region in the PTSD pathological circuitry [8]. Intriguingly, White et al. (2023) proposed a laterality role of dlPFC in the expression and regulation of anxiety: the left dlPFC contributes to anxiety regulation by supporting more effortful cognitive strategies (i.e., maintenance processes of working memory), while the right dlPFC contributes to anxiety regulation by supporting more autonomic emotion processes (i.e., top-down filtering of negative affective information) [50]. Indeed, dlPFC is a still emerging therapeutic target for PTSD using repetitive transcranial magnetic stimulation (rTMS) [51, 52]. Boggio et al. (2010) performed rTMS on either left or right dlPFC of PTSD patients in a double-blind, placebo-controlled phase II trial. They found a significant improvement in mood after left rTMS and a substantial reduction in anxiety following right rTMS [53].

Despite the proposed cortical-subcortical dichotomy in the ‘two systems’ framework, conscious and subconscious circuits clearly operate interdependently, with the amygdala still contributing to the feeling of fear by generating behavioral and physiological arousal responses that interact with conscious awareness through reciprocal connection with cortical regions [46]. Indeed, direct low-amplitude electrical stimulation in the amygdala first induces subconscious cardiac and respiratory changes, with conscious emotion perceptions (normally associated with a threat) emerging with repeated, prolonged, or higher-intensity stimulation [54,55,56]. Furthermore, stimulation targeting lateral and medial prefrontal cortices could regulate emotions by indirectly interacting with the amygdala [57]. With subcortical-cortical circuits providing integrated physiological-behavioral responses with varying degrees of conscious awareness, hemispheric specialization in humans provides for the possibility that the right amygdala is more involved in subconscious reflexive responses to threat, while the left amygdala is more embedded in systems that consciously appraise salience to contextualize feelings of fear and anxiety. Eliciting physiological changes out of context from an actual threat (as when activating the amygdala via direct electrical stimulation) might be experienced more variably. This could explain in part the observation that right amygdala stimulation more often elicits fear and anxiety, whereas left amygdala stimulation can elicit joy or happiness [54].

By evaluating crucial findings from human amygdala neuroimaging studies using fear conditioning and extinction paradigms (Fig. 1c-bottom), together with the lateralized finding in more general emotional processes from the prior lesion and neuroimaging studies (Fig. 1c-top), we have proposed that the right amygdala (and particularly the right CMA) specializes in sensory-mediated fear expression, while the left amygdala (especially the left BLA) specializes in cognitive-mediated fear acquisition and extinction learning. However, several challenges remain for future investigations.

First, the reasons and neural mechanisms for lateralized amygdala activity during fear-related emotional processes remain debated [58,59,60,61,62]. McMenamin & Marsolek (2013) hypothesized that asymmetric signals in left and right amygdala could be categorized into potential causes that were not mutually exclusive: language-related differences, masking-related differences, habituation-rate differences, and gender effects [63]. More recently, Ocklenburg et al. (2022) suggested that lateralized amygdala functions are determined by temporal characteristics, emotional valence, and perceptual properties [11]. Such categorizations may frame future investigations to systematically control patient and task variables for additional insights into the causes of lateralized amygdala activity during fear processing.

Second, human neuroimaging studies have lacked causality and faced other technical limitations. While rodent studies have elucidated causal amygdala-based neural mechanisms in fear conditioning [2, 3, 64], these provide little insight into the hemispheric laterality of higher cognitive functions in humans due to fundamental differences between animal and human neurobiology [65]. Human fMRI has been limited by the temporal profile of amygdala response, inadequate anatomical resolution, and insufficient power to identify subtle responses [15, 30, 66, 67]. Attempts to overcome such limitations include larger sample sizes, such as in a study using neuroimaging data from 601 humans during a well-validated fear conditioning paradigm, in which Wen et al. (2022) affirmed human amygdala involvement in fear conditioning and extinction [15]. Alternatively, human intracranial electroencephalography recordings provide direct evidence of brain activity with high spatiotemporal resolution [68]. A recent study showed increased amygdala activation for CS+ versus CS− using human intracranial recording [69], but additional work is needed to characterize lateralized network activity across phases of learning and behavior. Ultimately, causal relationships of human amygdala function may be studied further using direct intracranial electrical stimulation or highly focal ablation. For instance, rodent studies found that amygdala electrical stimulation could decrease the burying behavior during re-exposure to the CS [70, 71] and alleviate fear conditioning-induced alterations in synaptic plasticity in the amygdala-cortical pathway [72]. Early-phase clinical trials of continuous amygdala electrical stimulation (intended to blockade amygdala function) in treatment-refractory PTSD showed promising clinical improvements [73, 74], and noninvasive neurostimulation approaches that indirectly target amygdala have shown early promising effects in the treatment of PTSD [57, 75,76,77,78]. Likewise, selective thermal ablations encompassing right amygdala improved PTSD symptoms, particularly hyperarousal [29, 79].

Third, the implications of lateralized amygdala function upon downstream targets of amygdala innervation are poorly understood. In rodent studies, the directly innervated targets of the amygdala serve more specialized functions [80]: hippocampus for the memory of emotional events [81]; dorsal and ventral striatum for motor learning and approach-avoidance, respectively [82,83,84]; and bed nucleus of the stria terminalis (BNST) for processing of autonomic and somatic responses related to anxiety [85]. Further studies on the functional lateralization in these target areas in humans during fear processing may be vital to understanding how lateralized amygdala function impacts behavior.

Focal lesions, direct electrical stimulation, neuroimaging, and electrophysiological recordings comprise critical pillars for understanding human amygdala functions. We have shown early evidence of amygdala functional lateralization through these modalities. However, there is still a lack of comprehensive understanding of the amygdala laterality during fear processing (Fig. 1c). To overcome the limitations and gaps, we recommend (1) probing amygdala-related network dynamics during each phase of fear conditioning/extinction with high spatiotemporal resolution and (2) leveraging focal amygdala interventions such as direct electrical stimulation or thermal ablation to dissect causality. Clinical intracranial electrophysiological recording provides a unique opportunity to study these dynamics and causal functions in humans with high spatiotemporal resolution [68]. Specifically, individual amygdala nuclei (e.g., BLA and CMA), downstream targets (e.g., hippocampus, striatum, and BNST), and upstream cortices (e.g., vmPFC, dACC, and dlPFC) are relatively accessible during intracranial investigations in epilepsy patients [86, 87] and can be studied across the phases of fear conditioning in well-validated tasks [69]. Such patients can also undergo acute interventions such as direct electrical stimulation and focal radiofrequency ablations that causally alter amygdala-related circuit activity [29, 73, 74, 86, 88]. Other emerging non-invasive opportunities include high spatiotemporal resolution recording using ultra-high-field MR neuroimaging [89, 90], as well as focal neuromodulation techniques such as transcranial focused ultrasound [91] and temporal interference electrical stimulation [92]. Indeed, a powerful approach would entail within-subject neuroimaging recording to inform network hypotheses, followed by directed electrophysiological recording, electrical stimulation, focal ablation, and behavioral characterizations. Initially carrying out such studies in the context of treating epilepsy patients with and without comorbid PTSD will help demonstrate safety and feasibility, but this also provides a clear stepping stone to clinical studies and personalized interventions for PTSD patients in a manner analogous to that emerging for other psychiatric conditions [93].

The dysregulation of fear processing is a key feature of PTSD [6, 7]. Psychophysiological studies indicate that alterations in fear conditioning and extinction are likely to be involved in the development and/or maintenance of PTSD [94,95,96,97,98]. Abnormal fear conditioning responses observed in PTSD include enhanced acquisition of the fear response to the CS+ (or CS−; i.e., fear generalization), reduced fear extinction, or a combination of these processes [95, 97]. Neuroimaging studies have revealed abnormal brain activity patterns in PTSD at each stage of fear conditioning and extinction learning: the fear ‘accelerators’ (e.g., amygdala and dACC) fail to disengage while the ‘decelerators’ (e.g., vmPFC and hippocampus) fail to engage in PTSD [99,100,101].

We present observations that should sharpen the framework for exploring amygdala-related circuits. Greater recognition of amygdala-dependent functional lateralizations during fear acquisition, expression, and extinction has important clinical implications. A recent case series of two PTSD patients reported significantly increased anxiety after a left-sided stellate ganglion nerve blockade, and the subsequent reduction of anxiety through a right-sided nerve blockade [102]. The amygdala has been increasingly pursued as a target for intervention in patients affected by PTSD [29, 73, 74, 88], but possibly without sufficient consideration of laterality and intervention parameters. Concordant with our model (Fig. 1c), we reported that highly selective stereotactic laser ablation encompassing the right amygdala and anterior hippocampus in two epilepsy patients profoundly ameliorated PTSD symptoms (particularly hyperarousal) [29]. By contrast, other individual case reports found PTSD onset [103] or exacerbation [104] after left amygdalohippocampectomy. We recently performed awake highly selective stereotactic amygdalotomy (sparing the hippocampus) on the right hemisphere in the setting of ipsilateral medial temporal lobe epilepsy, which provided immediate and persistent relief of PTSD symptoms. By contrast, awake left amygdalotomy in another subject caused immediate and persistent exacerbation of baseline anxiety (J. T. Willie, unpublished results). Another study found that epilepsy surgery for right amygdala lesions improved postsurgical mood, whereas right hippocampus resection was associated with worsening mood, which further supports the importance of discriminating the independent roles of the amygdala from hippocampus [105]. Although such surgical observations were made in the setting of epilepsy, our framework and observations lead to testable predictions with respect to amygdala-based interventions for PTSD: (1) right amygdala ablation or functional blockade should alleviate PTSD and/or anxiety by counteracting fear expression and hyperarousal, which may permit a more adaptive state of extinction/safety learning via left amygdala function. By contrast, (2) left amygdala ablation or blockade would impair extinction/safety learning and exacerbate PTSD and/or anxiety, especially when the right amygdala remains hyperactive and maintains hyperarousal.

Cognitive behavioral therapy, a first-line approach for PTSD, involves repeated exposure and desensitization to trauma reminders. The mechanism of exposure-based treatment is firmly rooted in understanding fear extinction (safety learning) from animal and human studies of fear processing [3, 6, 106, 107]. Indeed, multiple neurobiological circuits may contribute to exposure-based treatment response [57]. Additionally, pharmacological and non-exposure-based treatments for PTSD can provide benefits [108,109,110], suggesting further complexity of mechanisms of PTSD maintenance versus recovery. This aligns with the notion that PTSD is a heterogeneous disorder in which clinical signs and symptoms span multiple symptom clusters [111]. Such symptom heterogeneity suggests that any single neurobiological model is unlikely to capture the spectrum of mechanisms that underlie PTSD semiology. Recognizing the challenges of dissecting mechanisms contributing to broad diagnostic categories, neuropsychiatric disorders are increasingly studied dimensionally with transdiagnostic objective markers, such as hyperarousal. Indeed, the National Institute of Mental Health (NIMH) Research Domain Criteria (RDoC) framework promotes the consideration of mental health and psychopathology through the lens of major domains of neurobehavioral functions [112].

By evaluating key findings from human neuroimaging and lesion studies, we propose a framework for understanding hemispheric lateralization of amygdala-mediated fear processing (Fig. 1c-bottom). Specifically, we propose that the right amygdala is more associated with sensory-mediated fear expression (autonomic responses), while the left amygdala is more associated with cognitive-mediated fear acquisition and extinction learning (fear and safety learning). These hypotheses will require further empirical validation and refinement using integrated high-resolution neuroimaging, intracranial recordings, and targeted interventions (e.g., focal pharmacological, ablative, or neuromodulatory approaches). Ultimately, our framework highlights the importance of considering the anatomic specialization of fear processing and the therapeutic implications of hemispheric lateralization in fear and anxiety-related disorders, such as PTSD.