Fish Feel Pain
Comprehensively rebutting Defending Feminism, Key, and the other fish pain skeptics
1 Introduction
(henceforth DF) is an interesting and thoughtful writer, who wrote a semi-viral post arguing that fish (and by extension shrimp and insects) don’t feel pain. Her arguments are nothing new—they’ve been made since around 2002, and rejected by most consciousness researchers. In this post, I thought I’d go through the abundant evidence for fish consciousness and explain why her arguments—and, by extension, the arguments of fish pain skeptics like James Rose and Brian Key—are wildly unpersuasive. While I will get into the weeds in this article, I will note: how you behave shouldn’t be significantly influenced by your evaluation of these arguments. There are quite a number of experts who think fish feel pain. It would be wildly irresponsible to have credence above 90% that fish don’t feel pain. But even a 10% chance that fish feel pain—and that we annually painfully slaughter a population roughly ten times the number of humans who have ever lived—is enough to make it a serious issue. Given the mind-bending scale of the harm we inflict on fish, even a modest chance that they feel pain is enough.
In section 2, I’ll discuss the evidence for fish consciousness. Then, in sections 3 and 4, I’ll explain why both of the major steps in the argument against fish consciousness are totally unpersuasive. In section 5, I’ll conclude.
2 The evidence that fish feel pain
(Note: this section focuses on fish, but see here for my arguments for shrimp sentienc and here for a video I did with a doctor rebutting the main arguments against fish pain).
How do we decide if a creature experiences pain? You can’t see another’s pain, and other creatures have brains very different from our own (more on that later). But if a creature behaves in various ways that make far more sense if they experience pain than if they don’t, then you should think they probably feel pain. It’s unlikely that they’d behave in many different wa as if they were in pain if they didn’t really feel pain. One such set of criteria indicative of pain comes from a report from Lynne Sneddon—fish meet every criteria.
There’s a vast body of literature documenting evidence of fish sentience, so I’ll just list some of the most impressive findings and address objections. One study by Sneddon (2012) found that fish who were in pain sought out a location where they were given a painkiller, even if they otherwise liked that location less. This makes sense if they feel pain—of course they’d seek out painkillers if they’re hurting—but is otherwise surprising.
If fish are given unpleasant electric shocks in some particular location, they learn to avoid that location, but if they haven’t been fed for a long period of time, then they’re willing to re-enter that location to get food.
Being in pain distracts people from fear. If I’m scared of spiders, and you start torturing me, I probably won’t be focused on the scary spider in the room. Fish avoid unfamiliar objects, but this effect is eliminated when they’re in intense pain. Fish also like to hang out with their friends, but if shocks get too intense when they’re near their friends, then they don’t seek out their friends. However, if given morphine, this effect is cancelled out. Various fish are also capable of transitive inference and pass the mirror test. When harmed, fish try to get away and struggle, but this effect is muted by analgesia.
Various studies have found that when acid is injected into the lips of fish, they rock back and forth, rub their lips against the side of the tank, and eat less. However, when given an analgesic, this anomalous behavior stops. Once again, this is exactly what we’d expect if they felt pain, but their pain was diminished by the analgesic. They didn’t show less general activity—only less pain activity.
Now, DF claims that this study “failed to replicate,” and uses this as evidence of a broader replication crisis in the field of fish consciousness, citing a follow-up study by Newby and Stevens. But this is much too hasty. As Sneddon (2009) notes in a letter to the editor, Newby and Stevens used different acid doses and placed the fish in a different environment where rocking was less natural and stress was greater. Previous experiments have already shown that acid doses as high as the ones used by Newby and Stevens completely destroy fish pain-censors, so it’s no wonder that feeding onset wasn’t delayed.
Oher studies have suggested that though fish feel sharp, immediate, shooting pain, they don’t experience long-term, dull, aching pain of the kind humans do. Thus, it’s no surprise that when their nociceptors are completely destroyed, delayed feeding is eliminated.
DF blows a lot of smoke about the alleged severe replication crisis in fish pain research. Other than the study that “failed to replicate” when a totally different experimental design was put in place, the sources she lists don’t really discuss specific findings that didn’t replicate, but mostly just dispute the interpretation of some studies.
DF rightly notes that compassion might bias people to think that fish feel pain. But there are many biases that go in the opposite direction. Those who eat fish are reluctant to acknowledge their sentience for obvious reasons! Because of fish’s alien properties, anthropomorphizing might lead us to wrongly think they aren’t conscious. Bias can go either way. Taking into account the fact that historically, sentience in animals has been underestimated—with human infants and other mammals believed not to be conscious until recently—assessments of bias shouldn’t affect one’s judgments too much. One should simply consider the evidence.
DF next argues that various items of behavioral evidence point against fish pain, writing, claiming, for instance, “crustaceans like shrimp are known to engage in autophagy, or eating their own body parts.”
First, shrimp don’t eat their own body parts except in exceptional circumstances, such as when the limb is degenerating; it’s not surprising evolution would cause them to do that. Second, animals often behave in unpredictable ways in response to injury—when Mike Tyson is hit in the head, he acts like nothing is happening. Behavioral responses to pain are often complicated and highly context sensitive.
Third, at most this would show shrimp aren’t conscious—it wouldn’t apply to fish. Fourtb, the kind of mutilation seen in shrimp in stressful conditions also often occurs in mammals who often scratch and bite themselves in stressful situations. Laboratory rats and mice sometimes chew off limbs or tails after nerve damage, and humans bite off body parts in exceptional circumstances.
RF’s major argument against the pro-fish pain evidence is that one can have complex behavior without consciousness. Thus, the behavioral indicators cited don’t clearly indicate consciousness. But this misstates the pro-sentience argument. Of course one can, in principle, do the things that fish do without being conscious. You could program a robot to respond to anesthetic—and as recent developments in AI have shown, even get it to speak—without it being conscious.
The evidence comes from the fact that in scenario after scenario, fish respond exactly the way you’d expect them to if they felt pain. If a creature felt pain you’d expect them to:
Have sensors for external damage (nociceptors).
Have central processing that takes in signals from their nervous system.
Respond to analgesic drugs.
Seek out analgesic drugs, but only when in pain.
Display behavioral response to pain that is muted by analgesic drugs.
Have physiological responses to harmful stimuli.
Make trade-offs between pain and reward.
Be distracted from fear by intense pain.
This is precisely what we observe! And while it’s possible that evolution produced some kind of non-conscious signal that produces identical behavior to pain, such a thing is unlikely. If a creature didn’t feel pain, it’s unlikely it would respond to analgesics, seek out analgesic drugs, and get distracted by bodily damage. In other words, if you see that in many different ways, a creature behaves as if it’s in pain, and it has a brain that processes damage-signals, the most reasonable assumption is that it feels pain.
It’s also possible that your complex behavior is produced despite you not feeling pain, but I don’t think that’s very likely!
Probability theory tells us that if X would be evidence of Y, then not X must be evidence against Y. For example, if a person’s DNA being at the crime scene is evidence they committed the crime, then their DNA being absent must be evidence they didn’t do it (though not necessarily proof). If fish didn’t respond to analgesics, that would clearly be evidence they didn’t feel pain. So the fact that they do must be evidence that they do feel pain!
Lastly, DF cites Diggles for the rather ridiculous claim that:
Slime molds are alleged to exhibit learning and problem-solving behaviors… Indeed, such behavior is not restricted to microorganisms from the animal kingdom, given that plants make sounds when stressed by dehydration.
Essentially, all of this means that if the criteria used by Birch et al. were universally applied, there is a high chance that few, if any, organisms would fail to meet their threshold for “some evidence of sentience”.
First of all, no, slime molds don’t have very impressive behavior. Their “problem-solving behaviors,” discussed in the papers Diggles cites include that they can orient towards food and find a shorter path to it using their external sensors. This is not complicated problem solving! The extent of their “learning” is that when exposed to damaging stimuli, they absorb some of it and avoid it in the future. They’re not doing rocket-science!
The claim that the Birch criteria would find evidence of sentience in slime molds is particularly absurd—the Birch criteria are as follows:
1) possession of nociceptors; 2) possession of integrative brain regions; 3) connections between nociceptors and integrative brain regions; 4) responses affected by potential local anaesthetics or analgesics; 5) motivational trade-offs that show a balancing of threat against opportunity for reward; 6) flexible self-protective behaviours in response to injury and threat; 7) associative learning that goes beyond habituation and sensitisation; 8) behaviour that shows the animal values local anaesthetics or analgesics when injured.
While fish meet many of the criteria, slime molds don’t! They don’t have nociceptors, integrated brain regions, nor do they respond to analgesics. It would be particularly surprising if they had integrative brain regions, seeing as they do not have brains! They are complicated machines. And the fact that plants sometimes make sounds when dehydrated doesn’t tell us they’re conscious, any more than the fact that a rusty door makes sounds when opened means it’s in pain. It’s hard to fathom how one could think that the Birch criteria imply sentience in slime molds if one has so much as read the Birch criteria!
Lastly, DF suggests that shrimp nociceptors haven’t been found yet, seeming to think this gives evidence that shrimp don’t suffer. But nociceptors are very hard to find! We didn’t find them in fish until 2003. There’s strong behavioral and neurological evidence for decapod (shrimp, lobster, crab, etc) nociceptors, even though we haven’t yet found them, and nociceptors are quite ubiquitous across the animal kingdom.
Thus, I think DF’s replies to the arguments for fish pain are totally unpersuasive. There is very strong behavioral evidence that fish feel pain. We should think they don’t feel pain only if given strong evidence to the contrary. As we’ll see in the next two sections, we don’t have such evidence—the argument against fish pain is unsuccessful.
3 Is a cortex needed for human pain?
DF’s big argument against fish feeling pain is that fish don’t have a cortex. In humans, she claims, a cortex is needed for consciousness. Fish don’t have any analogous structure. Thus, they don’t feel pain. In her words:
Pain is a product of the brain.
In particular, for pain to exist in an organism, the brain of the organism must sustain the sorts of neurological features that allow for consciousness and pain perception. In humans, this is the cortex, which performs the signal amplification and integration necessary for any particular nerve firing to be felt as pain.
Fish and invertebrates lack those neurological structures that can perform those tasks. They lack a cortex, and no “substitute” part of the fish or invertebrate brain can perform the task of signal amplification or integration necessary for the firing of a nerve to be felt as pain.
Therefore, fish do not feel pain, and neither do animals that similarly lack those neural structures.
In this section, I’ll explore the first controversial premise—that pain in humans requires a cortex, and argue it’s false. Then, in the next, I’ll discuss whether other animal brains could produce pain in a different way.
There are many different theories about how pain is produced in humans. Some include:
Cortex-centric theories, according to which the cortex is necessary.
Mid-brain centric theories, according to which the midbrain is necessary.
Other theories, according to which pain is produced in a complex way involving the mid-brain and the cortex and perhaps other structures.
I think the evidence pretty soundly points away from 1., which is needed for the Key thesis. In favor of mid-brain centric theories:
It’s much easier to produce pain by stimulating the mid-brain than by stimulating the cortex. In fact, it’s very tricky to generate pain by stimulating the cortex.
Disrupting the part of the midbrain called the PAG is particularly devastating to consciousness.
Various humans and animals without a cortex seem to still be conscious.
In favor of cortex-centric theories:
In humans when they’re in pain, the cortex is active.
Lesions that eliminate important parts of the cortex often eliminate pain.
In some cases, electrical stimulation of the cortex can produce pain.
I find the arguments for cortex-centric theories be totally unconvincing for establishing with high confidence the necessity of the cortex. First of all, all they would illustrate, at best, is that the cortex plays a key role in pain. They would not demonstrate sufficiency. Stimulating the fingers can get a person to type and finger activity correlates with typing, but that doesn’t mean that one can only type with their fingers, or that fingers alone, without brain input, are responsible for typing.
Second of all, as already mentioned, it’s generally a lot easier to produce pain by stimulating the midbrain than the cortex. Stimulation arguments seem to point in favor of the midbrain theory.
Third, as Bjorn Merker, a leading researcher favoring midbrain theories argues, the lesion studies are dicey—the cortex might generate pain by transmitting pain signals, rather than producing pain. Thus, knocking it out might eliminate pain, just as knocking the power cord out of your computer will prevent you from typing, even though a power cord isn’t needed in principle for typing. Also, the lesion studies generally only lesion one area, when full lesion of two areas is needed for methodological adequacy. In addition, in one case after a cortical lesion, a patient began feeling far more intense pain.
Fourth, the fact that the cortex is active when humans are in pain doesn’t tell us it’s necessary for pain. The cortex is active most of the time, and the mere fact that the cortex is doing something when we’re in pain doesn’t mean that it’s responsible. Merker also disputes the sources Key cites for the correlation, noting, for instance, that of seventeen sources Key cites for the claim that gamma ray oscillations underlie pain “A full twelve of these references do not even mention pain or feelings with one passing word.”
While Key suggests that destroying part of the fish midbrain doesn’t eliminate nociceptive responses, as Merker notes, the study he cites says exactly the opposite—that nociceptive responses stop until the brain region is regenerated.
Overall, the evidence is at best inconclusive, and certainly does not definitively point in the direction of the cortex being needed. But the death-blow for cortex-based theories is that in lots of cases consciousness appears to survive removal of the cortex.
One such class of cases involves hydrancephaly, a brain condition where the cortex, as well as much of the brain, is mostly or entirely destroyed. If Key is correct, people with hydrancephaly should not be conscious. But children with hydrancephaly often appear to be conscious, roughly like other severely mentally disabled children. Often their symptoms aren’t obvious for the first few months of life. They can laugh, cry, show preference for certain stimuli, recognize their parents, track objects, laugh at happy songs and cry at sad ones, and learn to play with a toy by throwing a switch.
Key—and DF—give two classes of reply. First they suggest that there is still some remaining cortical tissue, but this ignores autopsies from the aforementioned studies showing the tissue is entirely gone or dead and non-functional. Second, they suggest maybe the children aren’t really conscious. This strikes me as rather extraordinary; if you are going to suggest that someone who behaves relevantly like a severely mentally disabled child, playing with toys, recognizing parents, displaying sadness, and so on, is not conscious, there had better be some very convincing argument for this. But there is not.
While these children are severely impaired, that is on account of missing the majority of their brains!
When only the cortex is removed, such as in experiments with rats, impairment is often quite a lot more minor. Rats without a cortex continue to play mostly normally, and it’s not obvious from their behavior that they’re missing a cortex. Merker notes that such rats struggle against the insertion of a feeding tubs, whine piteously, push at it with their forepaws, attempt to bite the syringe and experimenter’s hand, and lick the injection site. The notion that such creatures aren’t conscious strikes me as rather absurd.
Lastly, if a cortex is needed for pain, as Key and DF suggest, this implies that octopi and other cephalopods aren’t conscious. Key readily admits this, though he thinks birds are conscious because they have a functional analogue of a cortex (Merker argues that his thesis implies absurdly that even birds don’t feel pain, but this gets into complex neuroscience questions that are above my paygrade). But the suggestion that octopi aren’t conscious seems deeply unlikely. Look at, in this video, how octopi solve a complicated maze using planning and ingenuity.
Octopi play, use tools, and are mischievous—shooting water at people they don’t like! They can continue to recognize those people even if the people change clothes, and octopi sometimes shoot water at the ceiling lights to turn them off. They’re probably about as smart and curious as dogs. Behaviorally, every study done on them looking at behavior we’d expect to find if they were in pain has found that they do display that behavior—responding and seeking out analgesic, for instance. They’re likely self aware. They can even remember what they had for breakfast, where they ate it, when they ate it, and whether they’d previously seen or smelled a prey item. Evidence has even suggested that they dream.
As Dinits notes, the Key argument implies that various animals like cichlids and rays that play and pass the mirror test aren’t really conscious. It thus implies that for nearly every behavior associated with consciousness, a creature without consciousness performs the same function, in almost exactly the same way! This even holds for behaviors that seem to only fulfill conscious functions, like playing and dreaming. Such a result would be quite surprising! It would be particularly shocking if removing the parts of rat brains responsible for consciousness produced non-debilitating behavioral modification.
Thus, I think the first step in Key’s argument is very implausible. There is not compelling neuroscientific evidence for the cortex being responsible for pain in humans. The neuroscientific evidence supports the cortex playing a big role in typical experience of pain, but that doesn’t demonstrate that it’s needed for pain. Various case studies show pain without a cortex. At the very least, it seems abundantly clear that non-experts like me and DF shouldn’t be extremely confident in a theory as contentious as the theories that claim the cortex is necessary for pain. It seems more plausible either that the midbrain is the main driver of pain or that pain is generated by some complex combination of different brain regions, with both the midbrain and cortex typically playing a role.
4 If the cortex is needed for human pain, is it needed for pain in general?
Suppose that, contrary to what I argued in the last section, the cortex is needed for pain in humans. Should we thereby think that it’s needed for pain in all animals?
On its face, this inference would seem very dubious. In humans, having teeth is needed for eating solid food, but that doesn’t mean birds without teeth can’t eat solid food. Feathers are needed for birds to fly, but that doesn’t mean insects and bats without feathers can’t fly. Very different structures often perform the same function in different animals. Thus, even if fish don’t have the brain region that gives rise to consciousness in humans, it would be wildly premature to suggest they aren’t conscious. This is especially true given how long ago we branched off evolutionarily and the evidence for consciousness in very different creatures from us without cortices (octopi, cuttlefish, and fish).
This was the main objection given to Key, and he did not have a good reply. His response was that in cases of other structures that produce physical functions, we understand the mechanism by which they do that. Because we understand the mechanism by which one might fly, we can see how bats might do it without feathers. But critics have not explained a mechanism by which fish might have consciousness without a cortex. It’s merely, he claims, a just-so story—no coherent mechanism is explained.
This strikes me as an extremely unpersuasive reply. If you claim that “structure X is needed for function Y,” and your only evidence for this is that according to an extremely contentious theory, in humans X produces Y, the burden of proof is not on your opponents to provide another functional way to produce Y!
As an analogy, we don’t know how to produce non-carbon-based life, but it would be irresponsible to claim with extreme confidence that carbon is needed for life, based on the fact that no one has yet formulated a proposal for how to give rise to life without carbon. If we had only investigated the human eye, and didn’t understand how it worked, it would not be responsible to confidently declare that nothing could see without a human eye! Before we invented the airplane, when we only saw flight produced by wings, it would not have been responsible to hold that flight absent flappable-wings is impossible.
Second, even granting the cortex produces consciousness, we have no idea how! There are many different theories! It’s not obvious why consciousness exists at all or what structures are needed to sustain it. If you don’t know how X produces Y, it’s wildly irresponsible to suggest that things like X are needed for Y—that would be like declaring that feathers are needed for flight before we understand the aerodynamic principles relevant to flight.
Third, of the various structural explanations for why the cortex might produce consciousness in humans, many of them imply that other structures in fish might produce consciousness. Jonathan Birch suggested in an email that depending on which theory of consciousness one has, they might think that though the cortex gives rise to consciousness in humans, different structures might result in consciousness in other animals: the hippocampal homologue in fish, the vertical lobe in cephalopods, and the mushroom body in insects. He also suggested that various theories like global workspace theory, recurrent processing theory, and the perceptual reality monitoring theory “are all plausible examples of theories that implicate the cortex in mammals but could be realized in other ways in other non-mammalian species.” Merker also suggested the tectum might be relevantly functionally equivalent to the cortex.
If we really took Key’s argument seriously, then, seeing as he thinks behavior isn’t a guide to experience, even if we came across a creature that could speak and function like a young post-verbal human child, so long as it didn’t have a cortex, we’d deny that it was conscious. The argument, therefore, provides much too restrictive a set of criteria.
Key’s argument is an outrageous leap. He claims:
The cortex is needed for consciousness in humans.
We know exactly why it does that.
No other structure in animals can do that, nor might there be another structure that does something else that gives rise to consciousness.
I find every one of these extremely doubtful (interestingly, as Merker notes, the structural criteria that Key cites seem to imply that a creature can’t feel pain of any sort unless it feels dull, aching pain in addition to sharp, shooting pain). None seem obvious enough to rest a confident declaration in non-consciousness in fish on! It’s no wonder that Key’s argument is an extreme minority position among relevant researchers, including those with no vested interest in fish consciousness.
5 Conclusion
In this essay, I have argued against DF—and more broadly Key and Rose—who argue fish don’t feel pain. I’ve argued there’s a wealth of behavioral evidence that’s at least suggestive of pain in fish. Their argument against fish pain relies on the notion that a thing can’t be conscious without a cortex. However, it’s quite dubious because:
The neuroscientific evidence doesn’t support cortical necessity.
Rats and humans without a cortex still appear to display conscious activity.
It implies even octopi and animals who can recognize themselves in the mirror aren’t conscious.
Even if a cortex is needed for pain in humans and other mammals, pain could very well be produced in fish in some other way.
The structures present in fish and invertebrates, on various theories, could plausibly support consciousness by doing the things that the cortex does to produce consciousness, even if the cortex is needed in humans.
Together, these are enough to make me about 80% sure that fish feel pain. I’m not certain—there’s some chance that they behave like they’re in pain without being in pain—but I’m pretty confident. The arguments against fish pain strike me, and the majority of relevant experts, as quite weak. Fish merit serious moral consideration! At the very least, we do not have the certainty needed to morally license us to treat them as unfeeling automata. Our current policy of erecting mass fish torture farms and suffocating them by the trillions is extremely ethically risky, and very likely seriously morally wrong.
Yeah, the evidence for fish pain is very strong — I wrote a piece on this topic for Vox a while back! https://www.vox.com/future-perfect/23639475/pescetarian-eating-fish-ethics-vegetarian-animal-welfare-seafood-fishing-chicken-beef-climate
Fishes suffer up to 22min of intense pain when taken out of water, this is a fact