The Neural Bases for Empathy

What This Research Found

Simone G. Shamay-Tsoory’s influential review presents definitive evidence that empathy is not a unitary capacity but comprises two dissociable neural systems that can be independently preserved or impaired. Published in The Neuroscientist and cited over 2,000 times, this paper established the neural architecture that explains one of the most confusing aspects of narcissistic relationships: how someone can understand your feelings with perfect accuracy while showing no compassion whatsoever.

The fundamental insight is that “understanding” and “feeling with” are separate neural operations. Before tracing the neural architecture, we must distinguish two fundamentally different empathic capacities: cognitive empathy, which involves understanding others’ mental states through mentalising and perspective-taking; and affective empathy, which involves actually sharing others’ emotional experiences through visceral resonance. These capacities feel related—we often assume that someone who understands our feelings will be moved by them. But the brain implements them through distinct networks that can function independently.

Cognitive empathy depends on the mentalising network. The temporoparietal junction (TPJ), superior temporal sulcus (STS), and medial prefrontal cortex (mPFC) support the capacity to represent others’ mental states—their beliefs, intentions, desires, and feelings. This network enables perspective-taking, theory of mind, and the attribution of mental states to others. When you consider what someone else might be thinking, when you infer their intentions from their behaviour, when you predict how they will respond to a situation, this network is active. Cognitive empathy is fundamentally computational: it represents others’ minds as objects of understanding.

Affective empathy depends on the emotional resonance network. The anterior insula (AI) and anterior cingulate cortex (ACC) support the capacity to actually feel what others feel—to experience visceral resonance with their emotional states. When you see someone in pain and wince yourself, when a friend’s grief creates a heaviness in your own chest, when another’s joy sparks your own smile, this network is converting perceived emotion into felt emotion. Affective empathy is fundamentally embodied: it creates shared experience rather than mere understanding.

The double dissociation proves these systems are genuinely independent. Shamay-Tsoory’s review synthesises evidence from patients with localised brain damage that provides the gold standard for establishing neural distinctions. Damage to the AI impairs affective empathy while sparing cognitive empathy—patients can accurately identify what others feel but do not share those feelings viscerally. Damage to the TPJ impairs cognitive empathy while sparing affective empathy—patients may feel emotional contagion from others’ states without understanding them. This double dissociation proves that the cognitive-affective distinction is not merely conceptual but reflects genuinely separable neural mechanisms. You can have one without the other.

Integration of these systems enables full empathic functioning. Healthy empathy requires both components working together: accurate understanding of others’ states (cognitive) coupled with appropriate emotional resonance (affective). The systems interact through neural connectivity—cognitive understanding can modulate affective responses, and affective resonance can motivate cognitive engagement. When integration breaks down, characteristic patterns of empathy failure emerge.

Why This Matters for Survivors

If you have experienced narcissistic abuse, this research provides a neurological framework for understanding patterns that may have left you bewildered, self-doubting, and questioning your own reality.

The person who hurt you understood exactly what they were doing. Cognitive empathy—the capacity to read your mental states, predict your reactions, and understand your vulnerabilities—was likely intact or even enhanced in your abuser. They knew when you were hurt. They could tell when you were reaching your breaking point. They understood precisely what would wound you most deeply. This was not a failure of perception. The TPJ-mediated mentalising circuits were processing your emotional signals with accuracy—perhaps frightening accuracy.

Their understanding was never converted into caring. Affective empathy—the capacity to actually feel your suffering as aversive, to experience visceral resonance with your pain—was the missing component. The anterior insula that should convert perceived distress into felt distress was not generating that signal. They knew you were suffering. That knowledge simply did not feel like anything to them. There was no internal brake, no automatic compassion, no visceral discomfort that would motivate them to stop causing harm.

This explains why explaining never worked. When you tried to make them understand how their behaviour affected you, you were addressing the wrong system. Their problem was never cognitive—they understood. Their problem was affective—they did not feel. No explanation, no matter how articulate, no matter how emotionally delivered, can create affective resonance in neural circuits that are not generating it. You were not failing to communicate. You were speaking to a system that was already functioning while the system that should have responded was silent.

The insight they showed made their cruelty more devastating, not less. Cognitive empathy without affective empathy produces a particularly harmful combination. Their accurate reading of your mental states was deployed strategically rather than compassionately. They knew what you feared and used it against you. They understood your vulnerabilities and exploited them. Their insight served manipulation, not connection. The very capacity that should have enabled intimacy became a weapon.

Their empathy was conditional, not absent. Research shows that individuals with this pattern can activate affective empathy under specific conditions—when motivated, when instructed, when the other person provides narcissistic supply. This explains the whiplash you experienced: moments of genuine-seeming connection followed by inexplicable coldness. Both responses were neurologically real. The difference was not capacity but motivation. When caring served their needs, the affective circuits engaged. When it did not, they went silent.

Clinical Implications

For psychiatrists, psychologists, and trauma-informed healthcare providers, Shamay-Tsoory’s research has direct implications for assessment, diagnosis, and treatment planning.

Assessment must distinguish cognitive from affective empathy. Standard empathy measures that combine both components into a single score may miss critical dysfunction. A patient may score in the normal or even above-average range overall while showing profound affective empathy deficits masked by intact or enhanced cognitive empathy. Clinicians should employ measures that separately assess perspective-taking abilities (Interpersonal Reactivity Index perspective-taking subscale, Reading the Mind in the Eyes test) and emotional resonance (IRI empathic concern subscale, observation of physiological and facial responses to others’ distress). Discrepancies between cognitive and affective scores are diagnostically significant—they indicate the preserved-cognitive/impaired-affective pattern characteristic of narcissistic and antisocial presentations.

Observational assessment during sessions provides crucial data. Does the patient accurately describe your emotional state while showing no corresponding affective response? Can they articulate what a family member must have felt during a conflict while remaining completely unmoved? Do they demonstrate sophisticated understanding of others’ perspectives that seems deployed for strategic advantage rather than genuine connection? These clinical observations, informed by the cognitive-affective distinction, reveal patterns that standardised measures may miss. The patient who “passes” empathy tests may nonetheless show the characteristic clinical profile of understanding without feeling.

Treatment targets must match the specific deficit. Cognitive empathy deficits, when present, may respond to explicit mentalising training—structured exercises in perspective-taking, feedback on accuracy of mental state attribution, practice at considering alternative interpretations of others’ behaviour. Affective empathy deficits require fundamentally different approaches. Since the deficit is in embodied, visceral resonance rather than intellectual understanding, interventions must target the body. Interoceptive awareness training, mindfulness practices that enhance felt-sense, body-based therapies that strengthen the connection between bodily states and emotional awareness—these may strengthen anterior insula function over time. Compassion meditation specifically targets the generation of caring feelings and may partially address affective empathy deficits.

Expectations for change should be realistic and communicated clearly. The motivational challenge is significant. Patients rarely seek treatment for empathy deficits because others’ suffering does not register as a problem requiring solution. When narcissistic patients enter treatment, it is typically because external circumstances have become intolerable—occupational failure, relationship loss, legal consequences—not because they wish to become more caring. Sustained therapeutic work over years might produce modest improvements in affective empathy, but profound transformation is rare. Clinicians should not promise outcomes the evidence does not support, while remaining open to whatever growth is possible.

Countertransference monitoring is essential. Patients with preserved cognitive empathy and impaired affective empathy may appear highly engaged in treatment while deploying their perspective-taking abilities strategically. They may seem to understand therapeutic concepts perfectly, provide insight-laden reflections, and create the impression of productive work—all while fundamentally unchanged in their affective relationship to others. The clinician’s felt sense that something is missing, that understanding is not translating into genuine caring, is important data. Supervision and peer consultation help identify when sophisticated performance is substituting for authentic transformation.

Broader Implications

This research extends beyond individual therapy rooms to illuminate patterns across relationships, families, organisations, and society.

The Architecture of Calculated Cruelty

The cognitive-affective distinction explains why certain forms of harm are possible. Random cruelty requires no empathy at all. But targeted cruelty—knowing precisely where to strike for maximum damage, understanding exactly what will devastate versus merely hurt, predicting how the victim will respond—requires accurate cognitive empathy. When this understanding is uncoupled from affective empathy, it becomes a tool rather than a bridge. The narcissist’s abuse is often exquisitely crafted because they understand you well enough to craft it. The same capacity that should enable intimacy enables devastation.

Why “Making Them Understand” Fails

Survivors often believe that if they could just explain their pain clearly enough, their abuser would finally get it and change. This hope reflects an assumption that the abuser’s harmful behaviour stems from cognitive failure—not understanding the impact of their actions. Shamay-Tsoory’s research reveals why this hope is typically futile. The cognitive system is often functioning normally. Understanding is already present. What is missing is the affective resonance that would make understanding matter. You cannot explain someone into feeling. No words, no matter how precise or passionate, can generate activity in neural circuits that are not engaged.

Institutional and Organisational Selection

Organisations often inadvertently select for the precise empathy pattern that enables exploitation. Leaders who can read the room, predict stakeholder reactions, navigate political dynamics, and anticipate opposition demonstrate high cognitive empathy. These skills are rewarded with promotion and power. But if this cognitive empathy is uncoupled from affective empathy, the leader who understands people perfectly may manipulate them ruthlessly. Due diligence for leadership positions should include assessment of affective empathy—not just understanding others, but caring about them. Otherwise, organisations systematically elevate individuals who excel at reading others while remaining unmoved by their welfare.

The Limits of Education and Awareness

Well-meaning interventions often assume that teaching perspective-taking will improve behaviour. Diversity training, empathy education, social-emotional learning—many such programmes implicitly target cognitive empathy, helping participants understand others’ experiences and perspectives. Shamay-Tsoory’s research suggests why such interventions may fail or even backfire in individuals whose deficit is affective rather than cognitive. Teaching perspective-taking to someone who already reads minds accurately but deploys that reading for advantage rather than connection may simply improve the sophistication of their manipulation. Interventions must specifically target affective empathy—a far more challenging goal requiring sustained, embodied, experiential approaches rather than cognitive instruction.

Intergenerational Transmission of Empathy Patterns

Children learn empathy by being empathised with. The mirror neuron systems and anterior insula circuits that mediate affective empathy develop through being mirrored—through having caregivers attune to and reflect back the child’s emotional states. When a narcissistic parent does not mirror (because their own affective empathy circuits do not generate the signals that would motivate mirroring), the child’s empathy architecture develops abnormally. They may develop enhanced cognitive empathy (hypervigilantly tracking the unpredictable parent’s mental states for survival) while their affective empathy circuits remain underdeveloped. The parent’s neural architecture shapes the child’s neural architecture through the relational environment of development.

Political and Social Manipulation at Scale

Leaders who can accurately read collective emotional states while remaining personally unmoved can manipulate populations with surgical precision. They know what the crowd fears, hopes, resents, and desires—and they deploy this knowledge strategically without genuine concern for those they lead. The populist leader who seems to “get” the common person’s frustrations while systematically acting against their interests demonstrates this pattern at scale. Cognitive empathy serves the reading; absent affective empathy removes the caring. Understanding this pattern helps populations recognise when apparent attunement is strategic rather than genuine.

Limitations and Considerations

No research is without limitations, and responsible engagement with this framework requires acknowledging several caveats.

The cognitive-affective distinction, while validated, is necessarily simplified. Real neural processing involves extensive overlap and interaction between systems. The TPJ and AI are not isolated modules but nodes in interconnected networks. Empathy emerges from dynamic coordination across multiple regions, and the clean distinction between cognitive and affective is more useful than it is neurally precise.

Lesion studies have inherent limitations. Patients with brain damage often have lesions affecting multiple regions, making attribution to specific areas challenging. Lesion patients may develop compensatory strategies that mask deficits. And generalising from brain-damaged populations to healthy individuals with trait-level empathy variation requires caution.

The motivational versus capacity question remains partially unresolved. Studies showing that narcissists can activate affective empathy when instructed raise the question of whether their deficit is structural (cannot feel) or motivational (will not feel unless required). The clinical and ethical implications differ significantly depending on which is primary. The truth is likely that both factors contribute in varying proportions across individuals.

Individual variation is substantial. The framework describes general patterns; specific individuals may show different configurations of preserved and impaired empathy components. Assessment must be individualised rather than assumed from diagnostic categories. Some narcissists may have more genuine affective empathy impairment; others may have largely intact circuits that they simply do not deploy.

Neuroplasticity means the architecture is not fixed. While this research maps empathy’s neural basis, it does not imply that patterns are permanent. The same plasticity that allowed maladaptive patterns to develop theoretically allows therapeutic change. However, such change typically requires sustained effort over years and genuine motivation—both of which the disorder itself often precludes.

How This Research Is Used in the Book

This research is cited in Chapter 10: Building the Maze (previously titled “Diamorphic Scales” or “Neural Scales”) to explain the neural basis of the narcissistic empathy deficit and why cognitive empathy often remains intact while affective empathy is impaired.

The chapter opens by distinguishing the two empathic capacities that develop differently in narcissism:

“Before tracing the neural architecture, we must distinguish two fundamentally different empathic capacities.”

This framing, drawn directly from Shamay-Tsoory’s work, establishes the cognitive-affective distinction that runs through the chapter’s analysis of how empathy circuits develop—or fail to develop—in the narcissistic brain.

The chapter emphasises the neural independence of these systems:

“Affective and cognitive empathy rely on partially distinct subnetworks.”

This finding is crucial for understanding why someone can read your emotions perfectly while remaining utterly unmoved. The narcissist who knows exactly what you feel yet shows no compassion is not failing to perceive—their cognitive empathy network (TPJ-mediated) is processing your signals. What fails is the conversion to feeling (AI-mediated).

The chapter applies Shamay-Tsoory’s double dissociation evidence to explain the characteristic narcissistic pattern:

“Damage to AI impairs affective empathy while sparing cognitive empathy. Damage to TPJ impairs cognitive empathy while sparing affective empathy.”

This double dissociation—the gold standard for establishing neural separability—proves that cognitive and affective empathy are not just conceptually distinct but genuinely independent in the brain. One can be lost while the other is preserved. This explains why the narcissist’s understanding of your inner world is so accurate: their TPJ-mediated cognitive empathy may be functioning normally or even enhanced by the hypervigilant attention to others’ states they developed in childhood. What is missing is the AI-mediated affective resonance that should convert understanding into caring.

The citation supports the book’s broader argument that narcissism involves selective morphing of neural systems: the empathy circuits developed to survive an environment that demanded mind-reading (strengthening cognitive empathy) while offering no model for heart-feeling (leaving affective empathy underdeveloped or selectively suppressed). The same developmental environment that created hypervigilance and validation addiction also created this characteristic empathy pattern—preserved understanding, impaired feeling.

Historical Context

Shamay-Tsoory’s 2011 review appeared at a crucial moment in the evolution of empathy research. While earlier theorists had proposed distinctions between types of empathy at the conceptual level, the neural evidence remained scattered and sometimes contradictory. Neuroimaging studies had identified multiple brain regions associated with empathy tasks, but the field lacked an integrative framework that could explain why certain patterns of empathy failure occurred together or separately.

The key contribution of this review was synthesising lesion evidence to demonstrate double dissociation. Neuroimaging shows correlations—certain regions activate during empathy tasks—but correlation does not establish necessity. Lesion studies establish causation: if damage to a region eliminates a capacity, that region is necessary for that capacity. Shamay-Tsoory’s integration of lesion evidence showing that AI damage selectively impairs affective empathy while TPJ damage selectively impairs cognitive empathy provided the causal proof that earlier correlation-based research could not offer.

This work built on several research traditions. The conceptual distinction between cognitive and affective empathy had roots in developmental psychology and clinical observation. The identification of the TPJ as critical for theory of mind emerged from research on mentalising. The role of the anterior insula in emotional processing had been established through interoception research. Shamay-Tsoory’s contribution was showing how these pieces fit together into a coherent architecture—and proving through lesion evidence that the architecture was not just conceptual but neural.

The timing was also significant for clinical application. Understanding of personality disorders was evolving beyond purely behavioural descriptions toward neurocognitive frameworks. The cognitive-affective distinction proved immediately useful for understanding why narcissistic and antisocial individuals showed such puzzling combinations of insight and indifference. They were not simply “low on empathy”—they had a specific pattern of preserved cognitive empathy with impaired affective empathy, and this pattern now had a neural explanation.

The paper continues to be influential, cited approximately 200 times per year, and remains the standard reference for the neural separability of empathy components. Subsequent research has refined the model’s details while confirming its core architecture.

Further Reading

• Decety, J. & Jackson, P.L. (2004). The functional architecture of human empathy. Behavioral and Cognitive Neuroscience Reviews.
• Ritter, K. et al. (2011). Lack of empathy in patients with narcissistic personality disorder. Psychiatry Research.
• Hepper, E.G. et al. (2014). Moving Narcissus: Can narcissists be empathic? Personality and Social Psychology Bulletin.
• Fan, Y. et al. (2011). Is there a core neural network in empathy? An fMRI based quantitative meta-analysis. Neuroscience & Biobehavioral Reviews.
• Singer, T. et al. (2004). Empathy for pain involves the affective but not sensory components of pain. Science.
• Shamay-Tsoory, S.G. et al. (2009). Two systems for empathy: A double dissociation between emotional and cognitive empathy in inferior frontal gyrus versus ventromedial prefrontal lesions. Brain.