Who Else Is In There
Do bees feel anything? Do fish, crabs, octopuses? What the experiments show, what they cannot show, and why scientists now argue for precaution where the evidence runs out.
The headline that started this series said that the understanding that even bees have feelings is changing our view of consciousness.[1] It sounds far-fetched, and it may be. But the honest way to judge it is to ask what evidence there could be, for bees or for anyone else.
The first essay ended on the problem. Nobody observes experience in another being; we infer it. With other people the inference is easy: they are built like us and tell us what they feel. With patients who cannot speak it already strains, and doctors get it wrong about 40% of the time at the bedside. With animals there are no reports at all, only behaviour, brains and evolutionary history, and the further an animal is from us the weaker each of those becomes.
Philosophers were asked about it in their largest survey, in 2020.[2] Almost all accept that some cats are conscious, two-thirds that some fish are, and they split almost evenly on flies.
Opinion is not evidence. This essay goes through the evidence group by group, from mammals and birds to fish, octopuses, crabs and insects, and ends with the argument that has moved the debate in the last five years: that certainty is the wrong standard.
How one infers a mind
There are three kinds of evidence, and each has a weakness. Behaviour: an animal that guards an injury, avoids a place where it was hurt, or trades a reward against a cost behaves as a creature that feels would. But simple machines can be built to do all of those things, so behaviour alone shows that something is processed, not that it is felt. Brains: if a structure that supports experience in us is present in another animal, that counts in its favour. But insects, octopuses and birds have brains organised very differently from ours, and if consciousness depends on what a brain does rather than how it is built, the absence of a familiar part proves little. Evolution: if consciousness is old and useful, it may be widespread; if it is recent, narrow. Nobody knows which.
Jonathan Birch, Alexandra Schnell and Nicola Clayton argued in 2020 that asking whether an animal is “more” or “less” conscious is the wrong question. They proposed five dimensions along which consciousness could vary: how rich perception is, how rich the evaluation of good and bad is, how unified experience is at one moment, how connected across time, and how much awareness of oneself there is. On their framework there is no single scale; each species has its own “consciousness profile”.[3] A bee might score well on some dimensions and not at all on others.
The theories of Part I give different answers again. On integrated information theory consciousness is graded and common among living things; on higher-order theories it may need a capacity to monitor one’s own states that few animals clearly have. The same experiment can count for or against depending on which theory one reads it through.
The easy cases, and why they were not always easy
For most of the twentieth century, behaviourist science avoided talking about animal experience at all. The change came slowly, through experiments that were hard to explain otherwise. In 1970 Gordon Gallup let chimpanzees get used to a mirror, then anaesthetised them and marked their faces with red dye. On waking, they used the mirror to inspect and touch the marks; monkeys given the same test did not.[4] The “mark test” became the standard test of self-recognition, passed since by a few other species.
Birds were a harder case, because they lack the layered cerebral cortex that much of the human evidence points to. In 2020 Andreas Nieder and colleagues recorded single neurons in crows trained to report whether they had seen a faint stimulus. The neurons’ activity had two stages: an early one that tracked how strong the stimulus was, and a later one that tracked whether the crow reported seeing it, the same signature that in primates is taken as a marker of conscious perception. The authors concluded that the neural foundations of sensory consciousness “do not necessarily require a cerebral cortex”.[5]
Eight years before that, a group of neuroscientists had already proclaimed, at a conference in Cambridge in July 2012, that “the absence of a neocortex does not appear to preclude an organism from experiencing affective states”, and that “all mammals and birds, and many other creatures, including octopuses” possess the neurological substrates of consciousness.[6] The Cambridge Declaration was a statement of opinion, not a finding, and some found it too confident. But for mammals and birds, few scientists now dispute its conclusion.
Fish, and the argument from anatomy
Fish are where the disagreement becomes serious. In 2003 Lynne Sneddon and colleagues found nerve endings on the head of rainbow trout that respond to damaging pressure, heat above 40 °C and acid, much like the pain receptors of mammals. Trout whose lips were injected with acid breathed faster, took longer to resume feeding and behaved abnormally, effects that lasted long enough for the authors to suggest discomfort.[7]
Brian Key answered in 2016 with the strongest version of the opposing view. In people, he argued, the feeling of pain depends on specific structures in the cerebral cortex; fish have nothing equivalent, and so “fish lack the necessary neurocytoarchitecture, microcircuitry, and structural connectivity” for feeling pain.[8] Their responses would be nociception, the detection of damage, without the experience. The reply, from Sneddon, Birch and others, is the bird argument: if crows can have the markers of consciousness without a layered cortex, a missing cortex is not decisive.
Then came a fish in front of a mirror. In 2019 Masanori Kohda and colleagues reported that cleaner wrasse, small reef fish that pick parasites off larger fish, went through every stage of the mark test, and tried to scrape off a coloured mark by rubbing their body against the bottom, but only when they could see it in a mirror, and not when the mark was transparent.[9] The journal noted that the paper had received both positive and negative reviews: either a fish is self-aware, or the mark test does not measure what it was thought to. In 2023 the same group showed that wrasse that had passed the test attacked photographs of strangers but not photographs of themselves, and did not attack composites with their own face on a stranger’s body, which they took as evidence of a mental image of their own face.[10]
Without a backbone
Octopuses are the invertebrates most people find easy to credit with a mind, for their large brains and their reputation for solving problems, discussed on this blog in Many Ways to Be Clever. Robyn Crook tested pain directly in 2021. Octopuses given a painful injection of acetic acid in one chamber of a tank later avoided that chamber; given a local anaesthetic in another, they came to prefer it; without pain, they had no preference. They groomed the injection site, and stopped when it was anaesthetised. The paper concluded that octopuses “are likely to experience the affective component of pain”: not just the signal, the unpleasantness.[11]
Crustaceans had a long reputation as reflex machines; lobsters are still boiled alive in many kitchens. In 2009 Robert Elwood and Mirjam Appel gave hermit crabs small electric shocks inside their borrowed shells. Shocked crabs were more likely to abandon their shell, but less likely to do so if it was a shell of a species they preferred, and they took a new shell offered to them more quickly.[12] A reflex would not weigh the shock against the quality of the shell; the crabs did.
In 2021 the British government asked a team at the London School of Economics, led by Jonathan Birch, to review the evidence on octopuses, squid, crabs, lobsters and their relatives. The team set out eight criteria, among them nociceptors, brain regions that integrate information, responses changed by painkillers, trade-offs like the hermit crabs’, and learning to avoid harm, and rated the confidence that each group met each one.[13] They found strong evidence of sentience in octopuses and in true crabs, substantial evidence in hermit crabs and in lobsters and crayfish, and weaker evidence, mostly for lack of studies, in shrimps.[14] The Animal Welfare (Sentience) Act 2022 followed the review: for its purposes, “animal” means any vertebrate other than humans, “any cephalopod mollusc” and “any decapod crustacean”.[15]
What the bee experiments show
Insects are the frontier, and bees are the best-studied insects. A honeybee has fewer than a million neurons, yet it navigates, learns abstract rules such as “same” and “different”, and tells its nestmates where food is by dancing. Four experiments carry most of the weight of the claim about bee feelings.
Pessimism. In 2011 Melissa Bateson and colleagues trained honeybees to associate one smell with a reward and another with punishment, then shook half of them vigorously, to simulate an attack by a predator. Presented with mixtures in between, the shaken bees were more likely to treat them as bad news; they also had lower levels of dopamine, octopamine and serotonin. In people, negative moods go with exactly this bias, and the authors suggested that honeybees “could be regarded as exhibiting emotions”.[16]
Optimism. In 2016 Cwyn Solvi, Luigi Baciadonna and Lars Chittka ran the reverse. Bumblebees trained that one colour meant sugar and another nothing were given an unexpected drop of sugar before the test; they then approached an ambiguous colour more readily, and after a simulated predator attack resumed feeding sooner. A drug that blocks dopamine abolished both effects.[17] The authors called these “positive emotion-like states”.
Pain. In 2022 Matilda Gibbons and colleagues gave bumblebees a choice of feeders, some heated to 55 °C, marked by colours. Bees avoided the hot feeders when the cool ones had rich sugar, but used them more and more as the cool ones got poorer, a trade-off guided by learned colours and so made in the brain. The authors concluded that this is “consistent with a capacity for pain experiences in insects”.[18]
Play. Also in 2022, Hiruni Samadi Galpayage Dona and colleagues gave bumblebee colonies an arena with small wooden balls, which they could ignore on their way to food. Forty-five bees rolled balls 910 times, individually between 1 and 117 times over the experiment, though rolling them never led to food. Younger bees rolled more, as young animals play more, and bees came to prefer the colour of a chamber where they had found balls they could roll.[19] The authors argued that the behaviour met the accepted criteria for play.
Andrew Barron and Colin Klein had argued in 2016 that insects have the capacity for “the most basic aspect of consciousness: subjective experience”. In vertebrates, they pointed out, structures in the midbrain build an integrated model of the animal’s body in space, from its own point of view, and the insect brain has structures that do the same job; if those suffice for experience in us, they may in insects, and the capacity would go back to the Cambrian, over 500 million years.[20]
The language of the papers is worth noticing. “Emotion-like states”. “Consistent with a capacity for pain”. “Could be regarded as exhibiting emotions”. The researchers claim that bees have internal states that work the way emotions work in us, change judgement, depend on the same brain chemicals and can be traded against rewards. They do not claim to have shown that it feels like anything to be a bee. Each result has a reading without experience: a state of the nervous system that biases decisions, as a thermostat’s setting biases a heater. The headline’s “bees have feelings” drops the word “like” that the papers are careful to keep. What has changed is not that someone proved bees feel, but that the behavioural case against it has grown thin: everything that used to be offered as evidence of feeling in mammals has now been found, in some form, in bees.
Where the evidence stops
If bees, why not worms, or plants, or bacteria? Not everything that responds to its surroundings is a candidate. Plants sense light, gravity and damage, and some respond to being touched; a group of “plant neurobiologists” has argued that they may be conscious. Lincoln Taiz and colleagues replied in 2019, drawing on Todd Feinberg and Jon Mallatt’s comparison of nervous systems, that only three groups of animals meet the criteria for consciousness: vertebrates, including fish; arthropods, including insects and crabs; and cephalopods, such as octopuses and squid. For plants, without neurons or brains, they judged the chance “effectively nil”.[21]
That list matches the boundary drawn by the 2024 declaration below, and it is drawn by the evidence available as much as by biology: worms, spiders and most insects other than bees and flies have hardly been tested. The absence of evidence there is mostly absence of experiments.
From proof to precaution
The argument that has changed the debate is not about any one experiment. It is that waiting for proof is the wrong standard, because proof may never come, not even for other people. In April 2024 a group of scientists and philosophers launched the New York Declaration on Animal Consciousness. Its three points are short enough to quote almost in full: there is “strong scientific support” for conscious experience in mammals and birds; the evidence indicates “at least a realistic possibility of conscious experience in all vertebrates (including reptiles, amphibians, and fishes) and many invertebrates (including, at minimum, cephalopod mollusks, decapod crustaceans, and insects)”; and “when there is a realistic possibility of conscious experience in an animal, it is irresponsible to ignore that possibility in decisions affecting that animal.”[22] Its first signatories include Birch, Chalmers, Chittka, Crook and Elwood; by November 2026 it had 614 signatures, from researchers and others signing in a personal capacity.
Birch developed the argument in The Edge of Sentience in 2024.[23] Where there is a realistic possibility of sentience backed by evidence, he calls the animal a “sentience candidate”, and argues that the response should be proportionate: precautions that are reasonable given the risk and the cost, with the public involved in deciding them, rather than either nothing until proof arrives or a ban on everything that might feel. The same framework, he argues, applies to human patients and to AI.
The law has moved in the same direction, at its own pace. The European Union treaty already required policy to pay full regard to animal welfare “since animals are sentient beings”.[24] Portugal changed its Civil Code in 2017 to give animals a legal status of their own, no longer simply things, recognising “a sua natureza de seres vivos dotados de sensibilidade”, their nature as living beings endowed with sensibility.[25] The British act names octopuses and crabs, and allows other invertebrates, insects among them, to be added by regulation.
The objections are serious too. A precaution has costs, and they fall on real people: fishers, farmers, researchers. A “realistic possibility” is a judgement, not a measurement, and the same reasoning could be stretched to anything that behaves in complex ways. And a declaration signed by people who already think animals are conscious is not the same as a consensus. Those who sign reply that the alternative, treating the absence of proof as proof of absence, would, applied consistently, deny experience to every animal, and that the cost of that error falls on the animals.
Why the numbers make it matter
The question would be academic if few animals were involved. The opposite is true, and the animals for whom the evidence is weakest are the ones involved in the largest numbers.
In 2024 about 79 billion chickens were slaughtered for meat, along with 4 billion ducks, 1.5 billion pigs, 700 million sheep, 560 million goats, 500 million turkeys and 300 million cattle.[26] Fish are not counted at all in official statistics, only weighed. Alison Mood and Phil Brooke estimated from catch tonnages and typical body weights that 1.1 to 2.2 trillion wild fish were caught each year between 2000 and 2019, and that in 2019 wild fish were 87.5% of all vertebrates killed for food or feed.[27] Farmed fish killed for food rose nine-fold since 1990, to about 124 billion in 2019 (range 78–171 billion), now more than all farmed birds and mammals together; most, 70–72%, have no legal protection at slaughter.[28]
So the uncertainty about fish is not a footnote. If the sceptics are right, what happens to them matters only for its effects on people and ecosystems. If the New York Declaration is right that there is a realistic possibility, then the greatest number of beings whose welfare is at stake from human activity are fish, about whom science is least sure, and whom the law protects least.
What is known, and what is not
For mammals and birds, the case for conscious experience is strong, and few scientists dispute it. For fish, octopuses and crabs, the behavioural evidence now includes almost everything that was used to make the case for mammals: avoidance learned from pain, relief from painkillers, trade-offs, in fish even passing a mirror test. The remaining argument against is mainly anatomical, that they lack the brain structures that support feeling in us, and that argument has weakened as birds turned out to manage without them.
For bees, the evidence is newer and narrower but surprisingly consistent: moods that bias judgement and depend on dopamine, a trade-off between pain and reward, and play. It shows that bees have states that work like emotions. It does not show that they feel them; nothing could, with current methods, for any animal. The New Scientist headline goes beyond the evidence, and the papers it draws on do not. Its underlying point survives: the line between animals that might feel and animals that are just machinery has moved, and is still moving.
Behaviour that looks like feeling, with no way to check what is behind it, is also the whole of the case for consciousness in machines. A bee was not designed to look as if it felt; a language model is trained on human writing, which is full of the expression of feeling. The next essay asks what that does to the evidence.
This essay was researched and written with Claude Opus 5.5 (Anthropic): human specification, editorial direction and critical review; machine research, drafting and figure generation. The subject came from Luis, who found the New Scientist headline on bees far-fetched; the article itself could not be read, so nothing from it is used beyond its headline. Findings and quotations from research come from the abstracts of the papers cited, each checked against its PubMed record or, for papers not in PubMed, against Crossref and Semantic Scholar, except for the bumblebee play paper, read in full in the authors’ copy. The Cambridge and New York declarations, the British act, the Portuguese law and the fish estimates were read at their sources; the PhilPapers results on an archived copy of the results page (December 2023), because the live site blocks automated access; the land-animal numbers in Our World in Data’s processing of FAOSTAT. The hermit-crab paper, the LSE review and Birch’s book are described from their standard accounts and from the 2022 review by Crump and colleagues, not read in full. There is no model in this essay. scripts/consciousness.py draws the four figures from the published numbers, which it writes to docs/consciousness-results.json; the table in Figure 3 is my summary of the sources cited. The cover is drawn by the same script: one dot per billion animals killed for food in a year, at the midpoint of the fish estimates.
Authored by: Luis Matos Ferreira — Physicist, Developer, Writer
- Something It Is Like — a short guide to the series, and why the evidence thins out from people to machines (forthcoming).
- Somebody Home — human consciousness: patients, theories and the hard problem.
- Who Else Is In There — bees, fish, octopuses and crabs: what the experiments show (this essay).
- The Machine That Says It Feels — language models, the indicator tests, what is inside the models, and the companies (forthcoming).
- Many Ways to Be Clever — octopuses, bees, crows and language models: how different brains solve problems.
- Up to Us — free will, and whether consciousness is the real problem behind it.
- The Company We Keep — the microbes inside every animal, and what changed in ten years.
- The Complete Parts List — why knowing every part of a living thing does not explain it.
- New Scientist, “The understanding that even bees have feelings is changing our view of consciousness”, article 2593034, newscientist.com (headline only; the text was not accessible).
- D. Bourget & D. J. Chalmers (eds.), “The 2020 PhilPapers Survey”, results for target faculty: Other minds (1,092 answers), survey2020.philpeople.org.
- J. Birch, A. K. Schnell & N. S. Clayton, “Dimensions of Animal Consciousness”, Trends in Cognitive Sciences 24, 789–801 (2020), doi.org.
- G. G. Gallup Jr, “Chimpanzees: self-recognition”, Science 167, 86–87 (1970), doi.org.
- A. Nieder, L. Wagener & P. Rinnert, “A neural correlate of sensory consciousness in a corvid bird”, Science 369, 1626–1629 (2020), doi.org.
- P. Low et al., “The Cambridge Declaration on Consciousness”, proclaimed at the Francis Crick Memorial Conference, Cambridge, 7 July 2012, fcmconference.org.
- L. U. Sneddon, V. A. Braithwaite & M. J. Gentle, “Do fishes have nociceptors? Evidence for the evolution of a vertebrate sensory system”, Proceedings of the Royal Society B 270, 1115–1121 (2003), doi.org.
- B. Key, “Why fish do not feel pain”, Animal Sentience 1 (2016), doi.org.
- M. Kohda, T. Hotta, T. Takeyama et al., “If a fish can pass the mark test, what are the implications for consciousness and self-awareness testing in animals?”, PLoS Biology 17, e3000021 (2019), doi.org.
- M. Kohda, R. Bshary, N. Kubo et al., “Cleaner fish recognize self in a mirror via self-face recognition like humans”, Proceedings of the National Academy of Sciences 120, e2208420120 (2023), doi.org.
- R. J. Crook, “Behavioral and neurophysiological evidence suggests affective pain experience in octopus”, iScience 24, 102229 (2021), doi.org.
- R. W. Elwood & M. Appel, “Pain experience in hermit crabs?”, Animal Behaviour 77, 1243–1246 (2009), doi.org.
- J. Birch, C. Burn, A. Schnell, H. Browning & A. Crump, Review of the Evidence of Sentience in Cephalopod Molluscs and Decapod Crustaceans (LSE Consulting, November 2021).
- A. Crump, H. Browning, A. Schnell, C. Burn & J. Birch, “Sentience in decapod crustaceans: a general framework and review of the evidence”, Animal Sentience 7 (2022), doi.org.
- Animal Welfare (Sentience) Act 2022, section 5, legislation.gov.uk.
- M. Bateson, S. Desire, S. E. Gartside et al., “Agitated honeybees exhibit pessimistic cognitive biases”, Current Biology 21, 1070–1073 (2011), doi.org.
- C. Solvi, L. Baciadonna & L. Chittka, “Unexpected rewards induce dopamine-dependent positive emotion-like state changes in bumblebees”, Science 353, 1529–1531 (2016), doi.org.
- M. Gibbons, E. Versace, A. Crump et al., “Motivational trade-offs and modulation of nociception in bumblebees”, Proceedings of the National Academy of Sciences 119, e2205821119 (2022), doi.org.
- H. S. Galpayage Dona, C. Solvi, A. Kowalewska et al., “Do bumble bees play?”, Animal Behaviour 194, 239–251 (2022), doi.org.
- A. B. Barron & C. Klein, “What insects can tell us about the origins of consciousness”, Proceedings of the National Academy of Sciences 113, 4900–4908 (2016), doi.org.
- L. Taiz, D. Alkon, A. Draguhn et al., “Plants Neither Possess nor Require Consciousness”, Trends in Plant Science 24, 677–687 (2019), doi.org.
- K. Andrews, J. Birch, J. Sebo et al., “The New York Declaration on Animal Consciousness”, New York University, 19 April 2024, signature count 614 when read on 1 November 2026, sites.google.com/nyu.edu/nydeclaration.
- J. Birch, The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI (Oxford University Press, 2024).
- Treaty on the Functioning of the European Union, article 13, eur-lex.europa.eu.
- Lei n.º 8/2017, de 3 de março, estatuto jurídico dos animais, artigo 1.º, Diário da República, diariodarepublica.pt.
- FAO, FAOSTAT, “Producing or slaughtered animals”, world, 2024, via Our World in Data, “Yearly number of animals slaughtered for meat”, ourworldindata.org.
- A. Mood & P. Brooke, “Estimating global numbers of fishes caught from the wild annually from 2000 to 2019”, Animal Welfare 33, e6 (2024), doi.org.
- A. Mood, E. Lara, N. K. Boyland & P. Brooke, “Estimating global numbers of farmed fishes killed for food annually from 1990 to 2019”, Animal Welfare 32, e12 (2023), doi.org.
Comentários
Enviar um comentário