Basic Neurosciences · Paper A
Neural correlates of empathy and interoception (insula and cingulate cortex)
Empathy is not one brain region — it is at least three separable systems that just happen to share a name. Feeling someone else's pain, understanding what they believe, and automatically grasping what their hand movement is reaching for are three different jobs done by three different networks: the anterior insula and anterior cingulate cortex for shared feeling, the temporoparietal junction and medial prefrontal cortexThe front part of the brain responsible for planning, judgement and controlling impulses.Read more → for reasoning about their mind, and the mirror neuron system (inferior frontal gyrus, inferior parietal lobule) for automatic action understanding. Interoception — the brain's own felt sense of its body — is the quiet foundation underneath the first of these, since sharing someone else's state means re-using your own bodily-feeling machinery a second time. What actually needs identifying is which of these three dissociable systems a given scenario is describing, not whether 'empathy' in general is involved.
Picture two people watching a video of a stranger accidentally slamming a car door on their own fingers. The first person winces, feels their own hand tense up, and says 'that must hurt so much' before they have even worked out anything else about the stranger. The second person can tell you, calmly and correctly, that the stranger is probably in pain and will want sympathy — but feels nothing themselves while saying it. Both people are doing 'empathy'. They are doing it with almost completely different parts of the brain.
Empathy is not one thing the brain does in one place. It splits into at least two separable routes. AFFECTIVE empathy is the automatic, embodied sharing of someone else's feeling — you wince because part of your own pain circuitry activates just from watching. COGNITIVE empathy is the explicit, effortful reasoning about what someone else believes, wants or feels — working it out, rather than catching it.
A third system, action understanding, lets you automatically grasp what a movement is for — reaching, grasping — without either feeling the mover's emotion or reasoning about their beliefs. Interoception (the brain's own sense of its owner's internal bodily state, such as heartbeat, hunger or a tight chest) turns out to be the quiet infrastructure underneath the first of these three systems: you can only feel someone else's pain as pain if you already have a working, felt sense of your own body's states to borrow the representation from.
Here is the trap almost every exam candidate falls into: treating empathy as if it lived in one spot, then guessing whichever structure they remember best. It does not work that way. Affective empathy runs mainly through the anterior insula and anterior cingulate cortex — together called the salience network (the pairing that flags emotionally or physically important events). Cognitive empathy runs mainly through the temporoparietal junction and medial — together called the mentalizing network, because they reason about someone else's mind.
Action understanding runs through the mirror neuron system, based in the inferior frontal gyrus and inferior parietal lobule. If a vignette describes shared feeling with no reasoning, or reasoning with no shared feeling, or bare motor imitation with neither, it is telling you exactly which of these three systems it wants.
Start from what interoception is actually for: the brain constantly tracks the state of its own organs and tissues to keep the body alive, and generates feelings in the first place from that tracking — a racing heart plus a locked jaw becomes 'I am angry' only once the brain has represented those bodily signals and labelled them. That representation is built in stages.
The posterior insula is the first cortical stop for ascending visceral and thermal signals, arriving as a raw, unintegrated map of the body's current state. The anterior insula re-represents that raw map as a conscious, felt experience, and — crucially for empathy — the same anterior insula activity that represents your own felt disgust or pain also activates when you watch someone else's face or body show it. Sharing a feeling, on this account, is not a separate emotional gift; it is your own interoceptive machinery firing a second time, triggered by someone else's state instead of your own.
Take the classic pain-empathy experiment: a participant receives a painful stimulus themselves in one condition, and simply watches a partner receive the same stimulus in another. The anterior insula and anterior cingulate cortex activate in both conditions, and the degree of overlap tracks how strongly the observer rates their own emotional reaction — this is the affective route doing its job.
Now change the task: instead of watching pain, the participant is asked to predict what the partner believes about a hidden object they have not seen moved. A completely different pair of regions — the temporoparietal junction and medial — activates instead, and this activity does not depend on whether the partner is in any distress at all.
Finally, if the task is just to watch the partner reach for a cup, the inferior frontal gyrus and inferior parietal lobule activate, coding the target and shape of the movement itself, again independent of either of the other two systems. One empathy question; three dissociable answers, depending on exactly what was measured.
| System | Core structures | What it actually does | Common trap |
|---|---|---|---|
| Affective empathy | Anterior insula, anterior cingulate cortex | Shares another's felt bodily/emotional state | Confusing it with reasoning about their beliefs |
| Cognitive empathy (mentalizing) | Temporoparietal junction, medial | Reasons explicitly about another's beliefs/intentions | Assuming it requires feeling anything |
| Action understanding | Inferior frontal gyrus (pars opercularis), inferior parietal lobule | Automatically simulates an observed movement's target | Equating 'mirror neurons' with emotional empathy |
| Threat/fear detection | Flags danger and reads fearful/angry faces | Selecting it for any 'emotion' vignette by reflex | |
| Reward-based compassion | Ventral striatum () | Generates a warm, approach-oriented response to suffering | Conflating it with the distress response it follows |
There is no scan that diagnoses 'reduced empathy' on its own; clinicians instead match a behavioural pattern to a known . Alexithymia (difficulty identifying and describing one's own emotions) is screened for with questionnaires such as the Toronto Alexithymia Scale, and interoceptive accuracy itself can be tested with simple tasks like heartbeat counting, where a person tracks their own heartbeats without feeling for their pulse.
Autism spectrum disorder is diagnosed clinically against DSM-5/ICD-11 criteria for persistent social-communication difficulty and restricted, repetitive behaviour, sometimes supported by structured observational tools such as the ADOS-2; imaging findings of altered mentalizing-network activity are a research finding about the diagnosis, not a diagnostic test in their own right. Psychopathic traits are assessed with structured interview- and history-based tools rather than imaging, precisely because reduced affective empathy with preserved cognitive empathy is a behavioural pattern, not a scan finding.
Management here means naming the right approach for each condition this system touches, not one treatment for 'low empathy' in general. For autism spectrum disorder, NICE guidance (CG142, adults) recommends structured psychosocial and psychoeducational support tailored to the individual's needs rather than a drug aimed at the empathy or mentalizing deficit itself, since no medication treats the core social-communication difficulty.
For , there is no dedicated first-line drug either; therapeutic approaches instead focus on building interoceptive and emotional-labelling skill directly, for example through mentalization-based or emotion-focused therapy. For the caregiver or clinician side of this system, compassion-focused training — deliberately cultivating a positive, reward-linked response rather than simply re-exposing someone to others' distress — has been proposed as a way to reduce the exhaustion of an overworked affective-empathy route, sometimes called compassion fatigue.
Before you pick a structure, ask yourself: is the vignette describing a FELT, shared bodily/emotional reaction (affective empathy), an explicit piece of REASONING about someone else's mind (cognitive empathy), or a purely MOTOR-level understanding of what a movement is for (action understanding)? Try explaining out loud why a patient could correctly predict what someone else believes while feeling absolutely nothing about their distress — if you can do that, you have the affective/cognitive split down.
Empathy is not one brain region but at least three dissociable systems — affective sharing (anterior insula, anterior cingulate cortex), cognitive reasoning about others' minds (temporoparietal junction, medial ), and motor-level action understanding (mirror neuron system) — built on an interoceptive foundation that runs from a raw bodily map (posterior insula) to a consciously felt one (anterior insula).