Allostasis

The central claim: allostasis is what a brain is for

Added with the Barrett (2017) ingest, and it reorganizes this page. Where the wiki’s earlier sources treat allostasis as one process the body does, Barrett treats it as the thing brains are for, and derives interoception, affect and emotion from it in that order.

A brain did not evolve for rationality, happiness or accurate perception. All brains accomplish the same core task: to efficiently ensure resources for physiological systems within an animal’s body… so that an animal can grow, survive and reproduce.

Her definitional emphasis matters and sharpens the one at the top of this page: allostasis “is not a condition of the body, but a process for how the brain regulates the body according to costs and benefits” — and “efficiency” means anticipating needs and satisfying them before they arise. It is a budget, run forward.

The derivation from there is three steps, and the order is the argument:

1. To regulate a system you must model it (Conant & Ashby’s cybernetic principle: “every good regulator of a system must be a model of that system”). So the brain runs an internal model of the body in the world.
2. The model must be built from the body’s perspective. Modelling the world “accurately in some detached, disembodied manner would be metabolically reckless” — the internal model costs 20% of the body’s energy budget (Raichle, 2010), and long-range connections are the most expensive part of it. So the model includes the statistical regularities of the internal milieu, not just the external world. That is interoception, and it is at the core of the model rather than an addition to it.
3. Affect is the budget made conscious. Interoceptive sensations are “usually experienced as lower dimensional feelings of affect,” so valence and arousal are “basic features of consciousness” — and, importantly, “not unique to instances of emotion.” See core-affect.

This is the wiki’s deepest available answer to a question its other sources answer at shallower levels. Craig’s answer to why is the body in emotion is anatomical (there is an afferent pathway and it ends in the insula). Seth’s is computational (the body is a hidden cause the brain must infer). Barrett’s is prior to both and teleological: interoception exists because a brain that must budget a body has to model it. Everything experiential is downstream of the budget.

Barrett’s examples of anticipatory regulation are worth keeping because they make the “before they arise” clause concrete: the heart beats harder and vessels constrict in advance of standing up; you drink before dehydration; alpha-amylase-rich saliva is secreted pre-emptively when the body needs glucose — “even just imaging food causes glucose secretion.”

The consequence for the wiki’s anatomy

If allostasis is the core task, then the regions that implement it (amygdala, ventral striatum, insula, OFC, ACC, mPFC — collectively visceromotor regions) are “usually assumed to contain the circuits for emotion,” and on Barrett’s reading that assumption is the field’s foundational mistake. They are not emotion circuits that happen to touch the body. They are body-budgeting circuits that the field has been calling emotion circuits. The whole three-way convergence the wiki records on salience-network, core-affect and visceromotor-areas gets reinterpreted this way — see default-mode-network.

As the motivational endpoint of interoceptive regulation (Farb et al. 2015)

Used in Farb et al. (2015) to name the motivational endpoint of interoceptive regulation. Emotional valence attached to the simulation-map (aversive/negative valence signaling deviation from adaptive ranges) motivates allostasis — physiological or behavioral change to restore homeostatic ranges.

Mostly automatic

Much allostasis proceeds through autonomic self-regulatory physiology, occurring internally without conscious choice (e.g., pupil/blood-vessel dilation in response to luminance or emotional relevance) — a form of physiological active-inference that Farb et al. note begins even before birth.

Not the whole story

Not all motivated behavior is allostatic. Hedonic and pragmatic goals (sensation-seeking to distract from low mood, self-caffeinating against fatigue, riding roller-coasters, spicy food) can deliberately sacrifice physiological balance to achieve other ends. Interoceptive regulation is therefore framed more broadly than homeostatic demand alone — as any action reshaping the sensory signals constituting the interoceptive simulation.

A dissenting definition: stress as failure-to-reset, with allostatic load as its effect (Levine, via Payne et al. 2015)

Payne, Levine & Crane-Godreau (2015) use allostatic load as a foil, and the contrast is worth keeping because it is the only source here that treats allostasis as downstream of something else rather than as the fundamental process.

Levine (1977/1986) defines stress negatively: as the inability of the complex dynamical system of the autonomic nervous system to recover to normal functionality — a dysfunctional dynamical state of the core-response-network. He then reads McEwen & Wingfield’s allostatic load as the effect of being stuck in that state, not the state itself: the “wear and tear” is what accumulates because the system failed to reset, and “leaves the exact nature of the stress response itself still undefined.” The striking corollary Payne et al. draw: a fully functional CRN “will not accumulate allostatic load in response to challenging environmental circumstances” — load is a symptom of a dynamical failure, not an inevitable cost of adaptation.

This is a genuine disagreement of level with the sources above, though not a hard contradiction:

• For McEwen/Sterling and Barrett, allostasis (and its load) is the primary frame — the anticipatory regulation is the thing, and wear-and-tear is what chronic over-regulation costs.
• For Levine, the primary frame is the dynamical state of the autonomic system, and allostatic load is a second-order consequence of that state going wrong.

The two need not conflict — a defender of allostatic load could simply relabel Levine’s “stuck dynamical state” as chronic allostatic over-activation, which is close to what McEwen means. But Levine’s insistence that the state is (a) discrete, (b) fully reversible in principle, and (c) not determined by the external situation gives it a flavour the allostatic-load literature lacks: reversibility by intervention rather than management of an accumulating cost. Filed as a minority reading held by parties with a therapeutic stake (see somatic-experiencing), and resting on Levine’s 1977 thesis rather than on data — but a clean statement of the alternative to reading allostasis as bedrock.

A precision-weighting account (Seth & Friston 2016)

Seth & Friston (2016) recast the homeostasis/allostasis distinction within predictive coding: whether the system falls back on a low-level homeostatic reflex (e.g., hypoglycemia triggering glucose mobilization via precise interoceptive prediction errors) or engages allostatic behaviour (deliberately preparing and eating a meal) depends on the confidence (precision) placed in deeper expectations about future action — not two separate systems, but different points on the same precision-weighted hierarchy. This gives Farb et al.’s largely descriptive homeostasis/allostasis pairing an explicit computational mechanism.

The social turn: allostasis run outward (Theriault et al. 2021)

Theriault, Young & Barrett (2021) take the allostatic frame in a direction the wiki’s other sources do not — outward, to other people. If a brain regulates its body prospectively to avoid the metabolic cost of error, and if other people are the largest unpredictable part of a human’s environment, then regulating the predictability of one’s social environment is itself an allostatic problem. Conforming to others’ expectations keeps them predictable, which keeps prediction error (and its metabolic cost) low — so social conformity is derived as an allostatic strategy, and its felt form is the sense-of-should. The engine is metabolic-cost-of-prediction-error, the explicit statement of the energetic premise this page mostly leaves in the background.

The paper leans on Schulkin’s social allostasis (“anticipatory regulation of the internal milieu,” already in this page’s citations via Schulkin 2011) and on the same Sterling/Barrett lineage, so it is continuous with the Barrett (2017) reading above rather than a rival. What it adds is that the social environment is a regulable allostatic variable, and that affect — the felt consequence of the budget — is what makes social unpredictability aversive. This is the wiki’s first use of allostasis to explain a social motivation rather than a bodily feeling.

The clinical turn: allostatic self-efficacy (Stephan, via Khalsa et al. 2018)

The Khalsa et al. (2018) roadmap carries the precision account one step further, into psychiatry. On the hierarchical-Bayesian reading it adopts (from Stephan et al. 2016 and Petzschner et al. 2017), reactive homeostatic control sits at the bottom of the hierarchy and prospective allostatic control modulates the homeostatic set-points from above — and crucially, belief precision sets the force and pace of corrective action: “the tighter the expected range of bodily state, the more vigorous the elicited regulatory action.” Stephan’s allostatic self-efficacy is the metacognitive belief about one’s own capacity to regulate the body; persistent unresolved dyshomeostasis reads upward as low self-efficacy over bodily control, whose psychological expression is fatigue and depression. This is where allostasis stops being a homeostatic-physiology construct and becomes a candidate mechanism of mental illness — see computational-psychiatry. The roadmap is candid that the evidence for the whole hierarchical-Bayesian picture is “indirect so far.”

The narrow, control-theoretic definition (Petzschner et al. 2021)

Worth recording beside Barrett’s expansive reading, because it is the definition a modeller would use and it is noticeably thinner. Petzschner et al. (2021): allostasis is “the process of achieving stability, or homeostasis, by dynamically adjusting homeostatic setpoints.” It is one of two ways to make a reflex arc flexible — the other being predictive homeostasis, which leaves the set-point where it is and temporarily moves the state away from it in anticipation. See homeostasis for the three-way table.

Two things follow.

First, on this reading allostasis is subordinate to homeostasis — a mechanism for achieving it — where Barrett makes it the premise from which homeostasis, interoception and affect are all derived. Not a contradiction (Barrett is making a teleological claim about what brains are for; Petzschner et al. a mechanical one about what the set-point does), but the wiki should not let Barrett’s usage silently become the definition. The gap between them is roughly the gap between “the brain is a body budget” and “the set-point is a variable.”

Second, both flexible modes depend on machinery the field has barely formalized: anticipating a perturbation requires a model of how internal states evolve in time. See forecasting, where that gap is recorded — and note that it sits directly under the clinical claims on computational-psychiatry, which are claims about mis-set anticipation.

The same diagram, promoted to a general theory of illness (Bonaz et al. 2021)

Worth recording as a fact about how the construct is being used, not as a new claim about what it is. Bonaz et al. (2021) reproduce Stephan, Petzschner and colleagues’ allostatic self-efficacy circuit directly as their Figure 1 (adapted from Stephan et al. 2016’s Figure 7) — the same hierarchical schematic the section above holds as an account of fatigue and depression — and then draw a lightning bolt on it marking “interoceptive dysfunction,” with arrows radiating to psychiatric disorders, neurological disorders, addiction, chronic visceral pain, neurodegenerative disease, autism, pelvic disorders, arthritis and comorbid conditions.

So a diagram introduced to explain two psychiatric symptoms is now carrying the pathophysiology of most of psychosomatic medicine. Two observations:

• The efferent floor gets filled in. The version on this page runs down to visceromotor cortex. Bonaz et al.’s runs further: agranular visceromotor cortices (ACC, AIC, subgenual, OFC) project both to viscerosensory cortices (as corollary discharge/efference copy) and down to visceromotor and interoceptive hypothalamic and brainstem centres, which drive homeostatic reflexes that are “enslaved” by those descending allostatic policies. The wiki now has a name for that whole efferent hierarchy: central-autonomic-network.
• The generality is the risk. A model that accommodates IBS, Parkinson’s, hypermobility, addiction and depression equally well accommodates them by being schematic. Bonaz et al. offer no parameterization that would distinguish the cases, and the wiki should hold the Figure-1 framing as an organizing picture with real architectural commitments rather than as a mechanism that has been shown to explain any particular condition. computational-psychiatry is where the version with actual parameters lives, and Petzschner et al. are candid that “there is little empirical work testing such models’ predictions.”

Energy in and energy out (Quigley et al. 2021)

Quigley et al. (2021) restate Barrett’s derivation compactly — she is a co-author — so the theoretical content adds nothing to the section above. What it adds is scope on the behavioural side, and the wiki should record it because every other source here treats allostasis as regulation of intake.

The body is “the brain’s effector and sensor,” and the balance it maintains is resource intake and outflow. Movements supporting intake: foraging, ingestion, breathing, circulation. Movements removing waste: “breathing, circulation, elimination, and defecation.” Excretion is an allostatic function, and the review makes it an exemplar rather than an aside — see urinary-interoception, which is the wiki’s only material on it.

Two smaller points worth keeping:

• Homeostasis is granted more autonomy here than on Bonaz et al.’s reading. Quigley et al.: homeostasis “can operate independent of predictive or anticipatory regulation by the brain, but can also be modulated by the brain when necessary,” providing local fine-tuning and error correction “when predictive regulation fails to precisely match local needs.” Compare Bonaz et al., where homeostatic reflexes are “enslaved” by descending allostatic policies. Not a contradiction — a difference in how much the periphery is allowed to do on its own — but the wiki should not let either phrasing become the settled picture. See homeostasis.
• The mediators are multiple and cross-modal. Ghrelin, leptin and lactate act both as direct CNS endocrine signals and indirectly via vagal and spinal visceral afferents; cytokines (IL-1β, IL-6) signal through the carotid body. Quigley et al. read this multiplicity as evidence of multisensory-integration within interoception — the energetic state is estimated by combining channels, not read off one. The corresponding open question they pose is the first of their Outstanding Questions: how the brain uses signals from across different organ systems to estimate current energetic status at all.

Three sources, three depths — and a tension worth keeping

The wiki now holds allostasis at three levels of ambition, and they nest rather than conflict:

source	allostasis is…
Farb et al. (2015)	the motivational endpoint of interoceptive regulation — what felt valence is for
Seth & Friston (2016)	a position on a precision-weighted hierarchy, continuous with homeostatic reflex
Barrett (2017)	the core task of any brain — the premise from which interoception, affect and emotion are derived

The one live tension is with Farb et al.’s “Not the whole story” section above. Farb et al. insist not all motivated behaviour is allostatic: roller-coasters, spicy food and self-caffeination deliberately sacrifice physiological balance for hedonic or pragmatic ends, so interoceptive regulation must be framed more broadly than homeostatic demand.

Barrett’s framework has no obvious room for that exception — if a brain’s core task is the body budget, non-allostatic motivation is either not real or is allostatic under a longer time horizon. She does not address it, and Farb et al. do not address her. This is not a hard contradiction (a defender would say the roller-coaster is an allostatic investment in exploration, which Barrett’s “an animal thrives when it has sufficient resources to explore the world” arguably licenses), but it is the place where the strong reading of allostasis-as-everything would have to do work it has not done here. Recorded as an open seam, since the wiki holds both sources and neither engages the other.