What Suffering Is Made Of
The question is not “Can they reason?” nor “Can they talk?” but “Can they suffer?” Jeremy Bentham, Introduction to the Principles of Morals and Legislation (1789)
How to ask a crab whether it hurts
How can we know whether an animal can suffer when it cannot tell us? Bentham asked the question in 1789, and in 2024 Birch provided the means to answer it. His eight criteria grew out of a review he led for the British government into cephalopod molluscs and decapod crustaceans. That review became the basis for the UK’s Animal Welfare (Sentience) Act of 2022, which extended legal protection to octopuses, crabs, and lobsters. Sentience, in the report’s words, is “the capacity to have feelings.”
The eight criteria grade evidence across four neurobiological and four behavioural markers, and they correlate with sentience across species yet are independent of any theory of consciousness. The criteria tell us whether to worry and how much, but they do not explain why.
The neurobiological markers ask whether the capacity is there.
- Nociceptors are receptors that respond to noxious stimuli.
- Integrative brain regions are hubs that combine information from multiple sources.
- Integrated nociception is the connection between the two, so that damage signals travel beyond the reflex arc.
- Analgesia is the capacity to alter the response to noxious stimuli: the system can be turned down.
The behavioural markers probe what the animal does with the capacity.
- Motivational trade-offs weigh injury against competing needs, as hermit crabs do when they stay in an inferior shell to avoid a shock.
- Flexible self-protection is nocifensive behaviour: tending wounds, guarding a limb, avoiding the place where it was hurt.
- Analgesia preference is choosing the drugged food over the plain when injured.
- Associative learning is how a neutral cue comes to predict harm.
The mind as a city
Consciousness, in the systems view, is a set of interacting components that are independent of the brain region(s) involved:
- Arousal/readiness decides whether the system is awake at all.
- Relay/synchrony keeps distant regions synchronized, so that what happens in one place can reach another.
- Coordination combines information from different sources into a single picture.
- Sensing supplies sensory data from the world.
- Scene construction stitches the sensory feed into a coherent present.
- The access buffer broadcasts a piece of content to the whole system, or lets it stay local.
- Valuation/salience assigns positive and negative value: it decides what matters.
- Self-context maintains the self to whom the scene happens.
- Report/action turns processed content into behaviour, decisions, and speech.
- The world interface is the boundary with the world: the body’s sensors and actuators.
Components are connected by content links (solid lines), which carry information, and modulatory links (dashed lines), which alter what content is promoted or suppressed through neuromodulators. Because the systems view specifies functions and their connections rather than a cortex or thalamus, it applies across species.
Two roads to the same structure
Birch’s criteria were derived empirically, from animal welfare research, and are deliberately atheoretical. The systems view is a theory of consciousness, derived from clinical evidence and cross-species neuroscience. The two derivations are independent, yet they converge on the same structure. That convergence is evidence the structure is real.
Birch’s criteria map to the systems view of consciousness as follows:
| Criterion | Kind | What it tests | Systems view |
|---|---|---|---|
| Nociceptors | Neurobiological | Does it detect damage? | Sensing |
| Integrative brain regions | Neurobiological | Does it combine information? | Coordination |
| Integrated nociception | Neurobiological | Are detection and integration connected? | Content link: sensing → coordination |
| Analgesia | Neurobiological | Can the response be turned down? | Modulatory links |
| Motivational trade-offs | Behavioural | Does it weigh injury against needs? | Valuation/salience |
| Flexible self-protection | Behavioural | Does it protect itself adaptively? | Report/action |
| Analgesia preference | Behavioural | Does it seek relief? | Valuation/salience |
| Associative learning | Behavioural | Does it learn that a cue predicts harm? | — |
Associative learning does not map to any component in the systems view, because it is the whole system reorganizing in response to experience.
Components are not sentience
Various critics argue that the criteria are too easy to satisfy. Unconscious humans have nociceptors, integrative brain regions, integrated nociception, and respond to analgesia: all four neurobiological markers yet no consciousness. The systems view explains why: in an unconscious human, arousal/readiness is not active. They are necessary conditions, not sufficient ones.
A single component can function in isolation, as it does in a spinal reflex, because the components are independently controllable. This is also why Birch’s five-of-eight threshold is a screening device, not a proof. It errs on the side of caution, which is right for welfare decisions, but it cannot tell a system that has the structure from one that merely has the components.
Sentience-washing
When an artificial agent draws on human data, it can perform the behaviours that appear to be sentience markers without possessing the underlying capacity. Birch calls this sentience-washing.
The systems view explains why: such pretence satisfies the behavioural criteria without the neurobiological ones, because the physical components are absent. A model that has no sensing, no coordination, and, most importantly, no valuation cannot suffer, however convincingly it talks about pain.
Can they suffer?
Birch developed his criteria to answer Bentham’s question. The crab cannot tell us whether it hurts, but the systems view says that suffering is an aversive state that requires arousal/readiness and valuation/salience. If that structure is present, suffering is possible.