The Case for Insect Consciousness

For years, I was on the fence about the possibility of insects feeling pain — sometimes, I defended the hypothesis; ¹ more often, I argued against it. ²

Then, in 2021, I started working on the puzzle of how to compare pain intensity across species. If a human and a pig are suffering as much as each one can, are they suffering the same amount? Or is the human’s pain worse? When my colleagues and I looked at several species, investigating both the probability of pain and its relative intensity, ³ we found something unexpected: on both scores, insects aren’t that different from many other animals. 

Around the same time, I started working with an entomologist with a background in neuroscience. She helped me appreciate the weaknesses of the arguments against insect pain. (For instance, people make a big deal of stories about praying mantises mating while being eaten; they ignore how often male mantises fight fiercely to avoid being devoured.) The more I studied the science of sentience, the less confident I became about any theory that would let us rule insect sentience out. 

I’m a philosopher, and philosophers pride themselves on following arguments wherever they lead. But we all have our limits, and I worry, quite sincerely, that I’ve been too willing to give insects the benefit of the doubt. I’ve been troubled by what we do to farmed animals for my entire adult life, whereas it’s hard to feel much for flies. Still, I find the argument for insect pain persuasive enough to devote a lot of my time to insect welfare research. In brief, the apparent evidence for the capacity of insects to feel pain is uncomfortably strong. ⁴ We could dismiss it if we had a consensus-commanding theory of sentience that explained why the apparent evidence is irrelevant. But we don’t. And however we approach the issue in the absence of a consensus-commanding theory, insect sentience seems like a possibility we shouldn’t ignore. 

Capsaicin is the chemical that makes hot peppers hot. But to experience a capsaicin burn, you need capsaicin receptors. In humans and other mammals, that receptor is TRPV1, which is found in the sensory neurons present throughout our bodies. Knock out that receptor in mice and they don’t react aversively to capsaicin. Fruit flies lack these receptors and, as a result, can eat jalapeños without consequence. 

Last year, a team in Korea created transgenic flies that carried the human capsaicin receptor in their nociceptors — the neurons that detect negative stimuli. ⁵ In flies as in humans, these neurons are widely distributed throughout the body. When the researchers brushed capsaicin on normal fly larvae, nothing happened. When the researchers did the same to the transgenic larvae, they curled and rolled, thrashing as though in pain. ⁶

Then, the researchers turned their attention to adult flies. After denying them food for 18 hours, the team offered the flies a capsaicin-laced liquid. The normal ones ate readily. The transgenic flies sipped for just a moment before they “ran away, scurried around, and vigorously rubbed and pulled on their mouthparts with their front legs.” ⁷ As before, upping the concentration altered behavior, reducing the amount of time before “the nociceptive response,” or the reaction that looks, at least, like one driven by pain. The response was so strong that, eventually, the transgenic flies starved to death, refusing to eat any more of the capsaicin-laced substance.

There’s no doubt that insects have the capacity for nociception: their nervous systems can detect and encode negative stimuli. When they’re exposed to those stimuli, we see them exhibit behaviors that look a lot like a mammalian pain response — as in the case of the transgenic fruit flies given capsaicin. However, nociception isn’t always associated with pain. When you accidentally touch a hot stove, your withdrawal reflex is triggered before you consciously experience pain. One of the signals never even reaches your brain. It goes to the spinal cord and activates motor neurons there while a second, slower signal heads to the thalamus and cortex. So, how can we be sure that we aren’t dealing with mere nociception — rather than pain — in insects? 

The first argument is that negative responses in insects are modulated by a variety of drugs that relieve pain and distress in humans. This includes analgesics like gabapentin, which helps manage chronic neuropathic pain (caused by damage to the nervous system); ibuprofen, which helps reduce inflammatory pain; and, surprisingly, a benzodiazepine, which is typically prescribed for anxiety. Partly for this reason, the authors felt justified to conclude that fly “nociception features intricate pain sensation, sensitization, and modulatory neural circuits comparable to those in mammals.” In other words, an animal that’s just three or four millimeters long may feel something we all work hard to avoid. 

A second consideration is that the capsaicin study is hardly unique. Many researchers have used Drosophila — the fruit fly — as a model for human pain. Early work focused on establishing simple parallels. Like vertebrates, fruit flies respond differently to extreme temperatures when given antinociceptive drugs. Fruit flies also try to escape aversive stimuli; they jump both when they come into contact with a hot copper plate and when hit by a high-heat laser. More recent work has shown that, if you remove a fly’s leg, it becomes more sensitive to negative stimuli, a phenomenon known as neuropathic allodynia. So, for instance, 42°C is the usual temperature that will trigger an escape response in uninjured flies, but for the injured flies, the threshold drops to 38°C. Accordingly, fruit flies have been hailed as a valuable model for drug development: because they’re sufficiently similar to mammals, they can serve as a cheap way to screen possible analgesics for human use.

The third and fourth considerations are evolutionary. We know that pain emerged at some point in biological history, as we can feel it. If we had a strong reason to think that pain requires some particular trait that insects lack, we might guess that insect behavior is driven by mere nociception. But we don’t. In the absence of a clear alternative, we should assume that their nociceptive responses are accompanied by pain, as that’s the simplest explanation for the powerful motivational force of nociception. 

We can also consider what our best theories about the function of pain suggest about its distribution in the animal kingdom. One such theory is that pain is adaptive because it provides fitness-relevant information to an organism in a way that’s motivationally salient. Put differently: “Ouch!” says “That’s bad for you!” while encouraging you to get away. Simultaneously, pain teaches: remembering what caused it discourages you from making the same mistake twice. These capacities seem particularly valuable once organisms become capable of agency, making decisions about how to navigate environments full of opportunities and threats. ⁸ Understood in this way, positive and negative feelings serve as useful tools for mobile animals that face complex tradeoffs: they facilitate choice in the moment and over time, motivating animals to seek the fitness-enhancing and avoid the fitness-compromising. Insofar as this picture is plausible, it’s a reason to think that sentience is widely distributed in the animal kingdom — insects included. 

Finally, it’s telling that insects are good pain models. By design, a model system doesn’t have every feature of the target system — or even most of them. Models are much simpler than whatever they’re modeling. But we use pain models to understand the causes and consequences of aversive experiences, and it would be a strange model of pain that had no such experiences at all. Indeed, we think that mice, rabbits, dogs, and pigs are good pain models partly because we take them to have the same negative states we do. And if we think that they have those states because of the striking parallels between their neurobiology and behavior and our own, then it’s reasonable to use those same parallels as evidence when assessing Drosophila. Either fruit flies aren’t a good model, in which case lots of researchers are wasting their time, or fruit flies are a good model, which is a reason to think they can feel pain.

If you find all these considerations unsatisfying, it’s worth considering what would do the trick. Otherwise, there’s a risk of perpetually moving the goalposts. We’ve seen this happen before. Some scientists argue that pain requires the neural structures we find in humans, such as a cortex (the so-called “no cortex, no cry” hypothesis). However, it’s now clear that many traits are multiply realizable — that is, different structures can produce the same capacities. So, some skeptics have insisted that sentience requires certain levels of integration in the brain. But new research has made it clear that fruit flies have brains that are more tightly integrated than anyone thought. (See the stunning FlyWire Connectome.) A standard fallback is to insist that, as we learn more about insects, we’ll be able to explain their reactions to negative stimuli without appealing to the subjective experience of pain. However, as our understanding of the human nervous system improves, we’ll eventually be able to explain human behavior without appealing to the subjective experience of pain. And that, of course, won’t show that we can’t suffer. 

Still, it’s fair to be wary of bold claims. So, let’s step back and think about methodology. How should we assess whether animals are sentient? Maybe we’ve gotten off on the wrong foot, and some earlier choice is now leading us astray.

I submit, however, that the opposite is true. Thinking about methodology shows that it doesn’t matter where we start our inquiry: regardless, we have reason to give insects the benefit of the doubt.

Sentience, we’ve said, is the capacity to have positive and negative feelings. Or, in more technical jargon, it’s the capacity for valenced conscious states. It’s clear that at least some insects, such as fruit flies and bees, have valenced states. Entomologists test for the presence of these states using cognitive bias tests, which involve training animals to associate one stimulus (like the color red) with a reward and another stimulus (like the color blue) with something aversive. Then, the animals are presented with an ambiguous stimulus (like the color purple). Relative to baseline, bees rewarded before encountering the ambiguous stimulus are more likely to approach it, whereas bees given something aversive are more wary. In other words, the insects are updating their assessment of how friendly their environment is, and adjusting their behavior accordingly. Researchers interpret this change in disposition as akin to optimism or pessimism — in other words, valenced states. 

But — as with the distinction between pain and nociception — we might not want to assume that valenced states are necessarily conscious. Maybe conscious valenced states only occur in some organisms, with the valenced states in other organisms all being nonconscious. So, how would we go about assessing consciousness in insects?

Jonathan Birch, a philosopher and expert in nonhuman sentience, identifies three approaches to navigating this problem: the theory-heavy, theory-neutral, and theory-light. ⁹ All three are attempts to solve the fundamental problem that philosophers of mind have profound disagreements about which theory of consciousness is best. The “theory-heavy” approach tries to get around this by taking all the main theories of consciousness — global workspace theory, higher-order thought theory, integrated information theory, etc. — and assessing whether insects satisfy the conditions that each theory details. If insects wouldn’t count as conscious according to any of them, that would rule them out.

Predictably, though, theories of consciousness disagree deeply about what consciousness requires. Some theories claim that consciousness consists of higher-order mental representations of other mental states — roughly, you’re conscious when you’re in a mental state of which you’re aware. These theories vary in terms of their implications about animals, but in some versions, consciousness is probably limited to humans. Others suggest a view like the one outlined in this essay, where consciousness is widely distributed and includes many invertebrates. These include Bjorn Merker’s midbrain theory, which holds that subjective experience consists of an integrated simulation of an animal’s body in its environment. This kind of processing happens in the human midbrain, with comparable structures present across much of the animal kingdom. Still others, like integrated information theory, conclude that we should attribute sentience to nearly everything. 

Since we don’t know which theory of consciousness is correct — or even what credence to assign to each of them — we don’t know what to infer from insects satisfying, or failing to satisfy, the requirements of a given theory. One thing we can say, however, is that some not-clearly wrong theories imply that insects are conscious. In other words, we don’t know enough to rule insects out.

The theory-neutral approach tries to address this uncertainty from a different angle. Instead of starting with theories of consciousness, it encourages us to look for principles that don’t rely on any theory at all. One such principle is Isaac Newton’s second rule of scientific reasoning, which says that, all else equal, like effects probably have like causes. In the context of animal sentience, this means that, absent a strong reason to believe otherwise, we should assume that similar behaviors are most likely to be produced by similar mechanisms. This is the approach taken by Michael Tye, a philosopher at the University of Texas at Austin, who argues that insects probably are conscious, based on the kinds of behaviors we discussed when thinking about Drosophila.

The risk of the theory-neutral approach is that it can prove too much: without some assumptions about consciousness, broad principles like “similar behaviors are probably produced by similar mechanisms” could support the thesis that plants and LLMs are conscious. In the case of plants, I’m skeptical of that result, though not in a way that undermines Newton’s rule: I tend to think the behaviors plants exhibit aren’t similar enough to warrant the inference. In any case, whether or not the theory-neutral approach needs some nuance, there’s something to the idea that we should look for principles that let us move beyond our deep uncertainties about consciousness. And if we do, it’s once again difficult to rule out insects.

Finally, the theory-light approach tries to thread the needle between the theory-neutral and theory-heavy approaches: unlike the theory-neutral approach, it makes substantive claims about consciousness; unlike the theory-heavy approach, it doesn’t assume that some current theory of consciousness is correct. Instead, the theory-light approach offers “the facilitation hypothesis,” which just says that the subjective experience of consciousness facilitates a cluster of cognitive abilities that tend to occur together. So, if we can get evidence of that cluster of abilities, we can get evidence for consciousness.

Proponents of different theories of consciousness might disagree on what, exactly, these consciousness-linked abilities are, but most can agree on at least a few core contenders. These include trace conditioning, where subjects learn to associate stimuli that are separated in time. (So, as opposed to having a red light shine during an electric shock, the light might shine a second after the shock.) Another example is rapid reversal learning, where subjects learn one association (red light = shock) and then learn a contrary one after just a few examples (now, blue light = shock). Critically, these abilities seem to kick in only when (human) subjects are consciously aware of the associations between stimuli that they’re learning — that is, consciousness seems to play an important role in enabling us to do these things at all. And while any one such ability might not be strong evidence of consciousness, identifying a whole group of them would be strong evidence indeed.

It’s unclear whether the theory-light approach is the alternative it’s supposed to be: depending on how we interpret it, it may be a gussied-up version of the theory-neutral approach (if the facilitation hypothesis is genuinely theory-neutral) or a version of the theory-heavy approach in disguise (if, in fact, the facilitation hypothesis assumes some theory of consciousness). However, we don’t need to worry about those interpretive questions for the time being. Instead, what matters is that the theory-light approach seems to imply — once again — that insect sentience is on the table. Birch points to a growing body of evidence that bees display the core capabilities facilitated by conscious experience. In The Edge of Sentience, he contends that there’s a “realistic possibility” that insects can feel pain.

We went down this methodological rabbit hole as a way to check ourselves. Did we make some mistake at the outset? Maybe insect sentience seems plausible when we start by thinking about insects starving to avoid capsaicin-laced food, but wouldn’t if we started elsewhere. As we’ve seen, this isn’t true. Wherever we start, insect sentience is on the table. That’s a reason to give these animals the benefit of the doubt. We don’t know enough not to.

All that said, I still find myself staring at an insect — just yesterday, it was a caterpillar — wondering whether there’s anything going on in there. Could that little creature really be sentient? Is there any serious chance that it can feel pain?

When I find myself having such a thought, several others now follow. 

First, brain size is largely driven by body size (as larger bodies take more neurons to map) not cognitive complexity. So, size doesn’t matter in itself; what matters is what brains can do. And little brains can do some very impressive things. Bees, for example, have abstract concepts like “same” and “different,” produce play behavior (as do flies) and even engage in metacognitive processes like avoiding difficult choices. We’ve seen ants use tools and bees, flies, crickets, and cockroaches display sophisticated learning abilities (including social learning). And as mentioned earlier, while we used to think that insect brains weren’t well-integrated, with little cross-talk among disparate regions; connectomic studies, which are based on maps of all the connections in the Drosophila brain, suggest extensive sensory integration within and among the various regions.

Second, my personal intuitions about sentience are driven by fairly arbitrary heuristics. For instance, I often think they’re based on some kind of “sufficiently like me” test. But animals can be unlike me in all sorts of ways and still be sentient — and, of course, what feels “sufficiently like me” varies based on many factors that don’t have much to do with neurobiology. It’s wild to consider how much My Octopus Teacher did to make octopuses relatable — and partly for that reason, seem sentient. No one has done My Octopus Teacher for an insect yet, but when it happens, I’d guess that intuitions will shift accordingly.

Likewise, my intuitions about sentience feel all too sensitive to size, numerosity, and perceived importance. Imagine that ants were as large and rare as rhinos, roaming distant plains. Even if their brains were the same size, I’d find it very difficult to deny that they’re sentient given their behavioral repertoires. This suggests that my judgments are being driven by insects being small, thus hard to observe; plentiful, thus expendable; and proximate, thus irritating. Not the best tests for the presence of pain.

Third, I’ve realized that my gut endorses some vague argument like this: Insects just don’t matter. But if they were sentient, they would. So, they must not be sentient.

That, of course, is a bad line of reasoning. We don’t learn facts by consulting our ethical intuitions. And it’s helpful — for me, anyway — to call that out explicitly. When I detach the idea of insect sentience from its moral significance — that is, when I consider the possibility completely isolated from any level of concern for nonhuman pain — it seems much more plausible to me that insects can hurt. And if so, I shouldn’t shy away from that conclusion just because of its possible moral consequences.

I hasten to add: even if insects are sentient, it doesn’t follow that we should let kids die of mosquito-borne diseases to avoid managing mosquito populations. We need to kill insects to prevent illness, produce food, protect property, and achieve lots of other legitimate goals. We use them as research tools, ecosystem service providers, and waste recyclers. However, we manage and use many sentient animals for many purposes. If we try to treat those animals humanely in the process, then we should begin to do the same for insects. Moral consideration is just that — consideration. We can think that insects deserve better than the status quo without thinking that they should dominate our moral deliberations.

We know very little about insect welfare. The project of figuring out how to improve the way we treat them has only just begun. The fledgling field of insect welfare science needs support to ensure that we get answers to foundational questions: What kinds of deaths are least painful for insects? What sublethal threats do they face? How can we distinguish the more from the less severe? What matters to them most at different life stages? 

That effort will also require some honest self-reflection. I’ve been broken by the image of a sow in a gestation crate, but I rarely feel much sympathy for insects. And while I’m willing to act on the argument for giving insects some consideration, my resolve can waver; I don’t have the visceral responses that steady me elsewhere. Even as someone who’s devoted most of his professional life to animal-focused work, insects test my capacity to care. So, it’s one thing to determine how to reduce the risk of causing unnecessary pain; it’s another to decide it’s worth prioritizing that reduction.

Nevertheless, I’ve sometimes been surprised by my own reactions. I grimaced through my first reading of that capsaicin study. And after hearing about a friend’s struggle to keep her mantises alive — she was rearing them for a study and they were dying in droves — I found myself weeping for those animals. We can care more than we might think.

My call, then, is for a little attention to be paid to the least of these, for a little caution toward the small and seemingly negligible. If we have ethical concerns about how we treat animals, then we can give some slice of that concern to insects. I’m not arguing against Drosophila in pain research, spraying pesticides, or any other way of using or managing insects. I’m only saying: if animal welfare is on our radar, then insects’ welfare should be there too. We should support research that can tell us more about it. And given what we’re going to do to insects, we should treat them better along the way.