Hide table of contents

Executive Summary

In this, the first of three posts on features potentially relevant to invertebrate sentience, we assess 10 anatomical and evolutionary features and 5 types of noxious stimuli reactions. Here are some high-level takeaways:

  1. Neuron count and brain size are often over-emphasized in superficial discussions of sentience.
  2. Nociceptors (specialized peripheral sensory cells used by the body to detect potentially harmful stimuli) are found in a diverse range of animals including fruit flies, sea hares, and nematodes. The possession of nociceptors may be a necessary condition for painful experience, but it is not a sufficient condition.
  3. Centralized information processing of some kind is probably a necessary condition for consciousness.
  4. Physiological responses to noxious events don’t tell us much about valenced experience.
  5. Simple reactions to noxious events, such as immediate withdrawal, also don’t tell us much about valenced experience.
  6. More complex reactions to noxious events, such as long-term protective behavior, might tell us something about valenced experience.

Introduction and Project Overview

This post is the third in Rethink Priorities’ series on invertebrate[1] welfare. In the first post we examine some philosophical difficulties inherent in the detection of morally significant pain and pleasure in nonhumans. In the second post we discuss our survey and compilation of the extant scientific literature relevant to invertebrate sentience,[2] as well as the strengths and weaknesses of our approach to the subject. In this post we explain some anatomical, evolutionary, and behavioral features potentially indicative of the capacity for conscious experience in invertebrates. In the fourth post we explain some drug responses, motivational tradeoffs, and feats of cognitive sophistication potentially indicative of the capacity for conscious experience in invertebrates. In the fifth post we explain some learning indicators, navigational skills, and mood state behaviors potentially indicative of the capacity for conscious experience in invertebrates. In the sixth, seventh, and eighth posts, we present our summary of findings, both in narrative form and as an interactive database. In forthcoming work, to be published in late July, we analyze the extent to which invertebrate welfare is a promising cause area.

Anatomical and Evolutionary Features

Neuron count

Neurons are specialized, electrically excitable nerve cells that constitute the most important part of the brain and nervous system.[3] Neurons communicate with each other via synaptic connections to form densely interconnected webs. The adult human brain contains roughly 86 billion neurons, and each neuron averages around 7,000 synaptic connections.[4]

Neurons alone do not automatically generate conscious experience, and neurons are not themselves intrinsically morally valuable. We can imagine a lump of billions of neurons swirling around a laboratory jar with absolutely no conscious awareness. Conversely, we can imagine an alien or a computer program with zero neurons which nonetheless has the capacity for conscious experience. A high neuron count is thus neither a necessary nor sufficient condition on conscious experience.

Nevertheless, neuron count might be able to provide some modest evidence for conscious experience. Neuron count seems to be at least roughly correlated with cognitive ability, and cognitive sophistication of some degree is probably a necessary condition on conscious experience. Still, even granting the importance of cognitive ability to consciousness, there are important exceptions to the general rule that more neurons means more cognitive ability. Larger animals need more neurons just to coordinate movement and autonomic functions. Larger animals also require more neurons to innervate their larger muscles,[5] and larger animals tend to process larger sensory fields to interact with their larger world, which requires a greater number of neurons just to process the data at the same level of complexity as a smaller animal would. These additional neurons do not necessarily allow for a greater number of distinct behaviors. At least by some methods of counting, honey bees show more distinct behaviors then do some mammal species and perform better at color learning than any studied vertebrate species.[6]

Humans are the most cognitively sophisticated animal on the planet, but the African bush elephant has almost three times as many neurons. The difference between humans and elephants is that humans have a much higher concentration of neurons in the cerebral cortex. Thus neuron count in specific brain regions might be a more reliable guide to cognitive performance. (It is even sometimes claimed that cortical neuron number matches intuitive perceptions of moral value across animals.[7]) Unfortunately, ascertaining overall neuron counts for invertebrate species is hard enough; determining neuron count for specific brain regions would be even more difficult.[8] In short, neuron count alone does not tell us how the neurons are organized, how the neurons are used, or how many synaptic connections each neuron possesses. Thus neuron count alone should not be considered a particularly important feature for determining capacity for conscious experience.

Finally, it should be noted that brain size is importantly distinct from neuron count and that there is no close correlation between the two. Neuron densities vary dramatically across species. Grey parrots (Psittacus erithacus)[9] have roughly the same number of neurons as owl monkeys (Aotus trivirgatus)[10] despite brains only half as large, as measured by mass.[11] Humans have more neurons per kilogram than comparably sized animals, but far fewer neurons per kilogram than small fish and ants. Neuron sizes also differ significantly among animals. In general, smaller animals have smaller neurons, but there are notable exceptions. The neurons of the sea hare Aplysia californica, for instance, are so large they are visible to the naked eye.[12] For these reasons, brain size should not be used as a proxy for neuron count, and neuron count should not be construed as a measure of brain size.

Brain size

Intuitively, there is some plausibility to the claim that brain size is morally important. In biological organisms, the capacity for valenced experience almost certainly supervenes on various brain structures and functions. Destroy the brain, and you thereby destroy the capacity for valenced experience. Because the capacity for valenced experience, and consciousness more generally, is a complex phenomenon, it plausibly requires a complex brain. And, so it may seem, bigger brains are at least in general more complex.

Humans are the smartest animals on earth, but they do not have the biggest brains. Human brains average around 1.35 kilograms, whereas African elephants sport brains at least three times as massive (4.2 kilograms) and the brains of some large whales can reach an extraordinary 9 kilograms.[13] Of course, it should be expected that larger animals have larger brains. A better measure might be brain mass as a proportion of overall body mass. Here, again, though, humans do not come out on top, even among mammals. Human brains constitute roughly 2% of overall body mass. Shrew brains constitute nearly 10% of overall body mass.[14]

Another brain size metric often discussed in the literature is encephalization quotient (EQ). In lay terms, encephalization quotient is a measure of the deviation in brain size of a given species from a ‘standard’ species of the same taxon. Encephalization quotient thus attempts to correct for the effect of body size on brain size. Humans have an encephalization quotient between 7.4 and 7.8, indicating that their brains are roughly seven and a half times bigger than would be predicted by body mass alone.[15] This is larger than any other mammal. (The next highest is the bottlenose dolphin at 5.3.) So encephalization quotient correctly predicts which animal is most intelligent. Unfortunately, encephalization quotient fares less well as a predictor of relative intelligence across nonhuman species. For example, New World capuchin monkeys have a higher encephalization quotient than chimpanzees and gorillas despite their less sophisticated cognitive abilities.[16] More generally, for predictions of intelligence across nonhuman primates, Deaner et al. 2007 finds that absolute brain size does better than other measures that correct for body size, including encephalization quotient.[17]

Thus, neuroscientific reality is more complicated than the intuitive picture painted at the outset would suggest. Absolute brain size is probably an inadequate measure of neural complexity and a somewhat poor predictor of relative intelligence. Other brain size metrics, such as brain-mass:body-mass ratio and encephalization quotient, do slightly better, but the importance of these metrics probably pales in comparison to the importance of brain organization and architecture. Not all areas of the brain are equally important. Across species the size of certain brain regions may be more informative than overall brain size. But even when comparing the same comparably-sized brain regions across species, various cytoarchitectural differences, such as the extent of cortical folding, interneuronal distance, axonal conduction velocity, degree of myelination, and synaptic transmission speed, could plausibly be more important.


Nociception is the neural process of encoding and processing noxious stimuli. Nociception is accomplished by nociceptors, specialized peripheral sensory cells used by the body to detect potentially harmful stimuli. Nociceptors are activated by extreme temperatures, intense mechanical pressure, and toxic chemicals likely to cause physical injury. Activated nociceptors typically trigger so-called “nocifensive behavior,” including reflexive withdrawal, contraction, and conditioned escape. The ability to detect and respond to noxious stimuli is a highly useful trait. Unsurprisingly, nociception is widely conserved throughout the animal kingdom.[18] Even relatively simple creatures, like the nematode C. elegans, possess nociceptors.[19]

In humans, activated nociceptors are also associated with conscious pain. For that reason nociceptors are sometimes misleadingly called “the sensors of the pain pathway.”[20] However, conscious pain must be carefully distinguished from mere nociception.[21] Nociception operates below conscious awareness and thus does not have an attendant phenomenology. Nociception is fast and reflexive, but it is not normally associated with long-term memory. Conscious pain, on the other hand, with its attendant felt badness, tends to leave a memorial imprint. Because pain experiences are often stored in long-term memory, pain tends to induce long-term behavioral and motivational changes.[22] For long-lived animals in complex environments, pain is thus potentially more effective at protecting the animal from damage than mere nociception.

For the purposes of this report, we have divided this feature into two parts, a strict definition of nociception and a loose definition of nociception. An organism satisfies the strict definition of nociception when and only when actual nociceptors have been identified in the animal. An organism satisfies the loose definition of nociception when and only when it is capable of responding to potentially damaging stimuli, even in the absence of identifiable nociceptors. The distinction is important for two reasons. First, it’s possible that an organism could detect noxious stimuli with cells which are distinct from but homologous to nociceptors. Second, identifying nociceptors is a nontrivial scientific task. Scientists have yet to identify nociceptors even in some extremely complex animals, like elasmobranch fish (i.e., cartilaginous fish, such as sharks) that obviously have the capacity to detect and respond to noxious stimuli.[23]

Opioid-like receptors

Opioid receptors are a specialized group of receptor sites that bond with molecules of opioid drugs and endogenous opioid peptides.[24] In vertebrates there are four closely-related varieties of opioid receptors, the mu (MOR), delta (DOR), kappa (KOR), and somewhat more mysterious orphanin receptor (ORL). MOR, DOR, KOR, and ORL evolved from a single opioid unireceptor gene early in chordate evolution. This unireceptor likely developed from a proto-unireceptor gene around the the time arthropods and chordates split.[25] In the literature, functionally analogous but chemically different receptors in invertebrates are often called “opioid-like receptors,” and we follow that terminology here.[26]

The presence of opioid-like receptors is potentially important because opioids are analgesics (painkillers). Opioids are especially significant because, unlike other analgesics, opioids can reduce the affective (valenced) component of pain without a corresponding reduction in the sensory aspect (e.g., the burning/throbbing/cutting/stinging/aching quality) of pain.[27] The presence of opioid-like receptors is thus mild evidence that animals which possess them have the capacity for valenced experience. Unfortunately, the evidential status of opioid-like receptors is complicated by the fact that opioid-like receptors play many biological roles besides pain modulation. These diverse roles include “regulation of membrane ionic homeostasis, cell proliferation, emotional response, epileptic seizures, immune function, feeding, obesity, respiratory and cardiovascular control.”[28] For this reason, the presence of opioid-like receptors does not by itself tell us much about an organism’s capacity for subjective experience.

Centralized information processing

This feature refers to an entity’s ability to integrate disparate information from local sensory inputs into a unified understanding[29] of the world. In vertebrates this centralized processing occurs in the brain. Entities with different neurological architectures may integrate information differently or not at all. Plants, for instance, react to the environment and regulate their functioning through a network of cells, hormones and other growth regulators. Crucially, although the effects of these reactions may radiate outward to influence the rest of the organism, the reactions occur and are ‘processed’ locally. Hence, plants do not utilize centralized information processing.

Nervous system centralization is not an all-or-nothing feature. Most invertebrates fall somewhere between plants and vertebrates. Take the common octopus. Octopuses are notoriously intelligent, and much of their behavior seems to require organized planning and foresight. Yet fully two-thirds of their neurons are located in their arms, and the arms can move, taste, and touch independently.[30] Nonetheless, octopuses do possess centralized brain structures, with some experts in the field going so far as to call them “analogous to the human cerebral cortex.”[31]

In humans, centralized processing is associated with conscious experience while localized processing is not. Spinal cord reflexes are simple behaviors (such as a knee jerking when tapped) mediated by central nervous system pathways that lie entirely within the spinal cord. Our spinal reflexes operate before the sensations are consciously perceived, though the sensory information is also usually processed centrally in us. Spinal reflexes can still occur (without conscious perception) in people who are paralyzed from the neck down.[32]

Additionally, valenced experience allows for the possibility of a generalized utility function in which different pains and pleasures are weighed against each other in order to reach an overall decision.[33] Without centralized information processing, it’s not clear how such tradeoffs could be made.

Finally, according to the Integrated Information Theory of consciousness, consciousness just is suitably integrated information.[34] When the effective informational content of a system, mathematically defined in light of the system’s causal profile, is greater than the sum of the informational content of its parts, the system is said to carry integrated information. Integrated information of the relevant source is conscious, whether that integration occurs in a human brain or an insect ganglion. Integrated Information Theory is often considered the leading scientific theory of consciousness. If a view like Integrated Information Theory seems to be well-justified, then centralized information processing is an extremely important feature to examine when judging the probability that a nonhuman animal is conscious.

Vertebrate midbrain-like function

The midbrain[35] is the uppermost region of the brainstem, situated, as its name implies, between the forebrain and the hindbrain vesicles. In vertebrates the midbrain helps regulate alertness, body temperature, and sleep/wake cycles. It also contributes to motor control, hearing, and vision. Compared to more recent neural developments, such as the cerebral cortex, which is found only in mammals, the midbrain is phylogenetically old. It is conserved across virtually all vertebrate brain plans.

The midbrain has traditionally received less attention than other regions of the brain from researchers investigating the neural underpinnings of consciousness. For instance, the neocortex, the most recently evolved and (in humans) largest region of the cerebral cortex, is sometimes held to be, if not quite the seat of consciousness, than at least a necessary condition on valenced experience.[36] This is not an ad hoc view. Functional imaging studies show that, in humans, there is a correlation between the phenomenal intensity of pain and activity in the anterior cingulate cortex and the somatosensory cortex.[37] Hence, it was thought, any creature which lacks a neocortex thereby lacks the capacity for valenced experience. No matter how similar the behavior, the absence of a neocortex in a creature served as a defeater for the view that that creature experiences pain and pleasure.

Today this view is increasingly challenged in both experimental psychology and comparative neurology. For starters, evidence is emerging that, even in humans, a neocortex is not required for conscious experience.[38] More importantly, the absence of a neocortex doesn’t imply that there aren’t homologous brain regions performing the same role in other creatures.[39] Thus, some researchers have begun to look at other neural regions, such as the midbrain.

According to Andrew Barron and Colin Klein, a biologist and philosopher respectively at Macquarie University in Sydney, the evolutionary function of consciousness is to produce an integrated and egocentric spatial model to guide an animal as it navigates a complex environment. Barron and Klein argue that in vertebrates this function is mediated by the midbrain. Organisms that have a specialized brain region for the processing of spatial information and organization of movement are said to possess midbrain-like functioning. According to Barron and Klein, organisms that possess midbrain-like functioning, including insects, plausibly possess the same capacity for conscious experience as vertebrates.[40]

Last common ancestor with humans

Understanding the evolutionary distance[41] between humans and our target taxa may help us better gauge the likelihood that members of those taxa are conscious for two reasons. First, evolutionary distance serves as a rough proxy for overall similarity. Creatures more closely related to humans are more likely to share major anatomical structures and functions, including important neurological features which could underpin conscious experience. Second, if we could estimate when consciousness first emerged, knowing when the ancestors of humans and our target taxa diverged could be significant. If consciousness evolved relatively early in our evolutionary history, more creatures are likely to be conscious today. On the other hand, if consciousness is a rather recent evolutionary development, we would expect the distribution of consciousness to be more confined.

Investigating the evolutionary origins of conscious experience is no easy task. Nonetheless, the question has been addressed from a variety of perspectives in the scientific and philosophical literature.[42] Some researchers believe that consciousness evolved recently because conscious experience requires a particular style of advancing mental processing that is only likely to be found in cognitively sophisticated social mammals.[43] Other researchers date the evolution of consciousness as far back as the Cambrian Explosion, arguing that the complex predator-prey relationships that developed during this period led to a kind of perceptual and cognitive arms-race, culminating in conscious experience.[44]

Of course, drawing conclusions about the characteristics of animals alive today based on the characteristics of an ancestor that lived hundreds of millions of years ago is probably unjustified. Even if the ancestor was conscious, it’s possible that in certain lineages consciousness was selected against. (So even if all mollusks descended from a conscious ancestor, it’s possible that today’s cephalopod mollusks are conscious while today’s bivalve mollusks are not.) And even if the ancestor wasn’t conscious, it’s still possible that consciousness arose independently in different, later lineages.

Average lifespan

It seems that one of the most important evolutionary functions of valenced experience is to facilitate learning. Behaviors which improve evolutionary fitness, such as mating, tend to have a positive valence, while behaviors which reduce evolutionary fitness, such as eating spoiled meat, tend to have a negative valence. When these valences are paired with the appropriate behaviors often enough, an animal learns to pursue fitness-boosting behaviors and avoid fitness-reducing behaviors.

Short-lived creatures may be less likely to possess the capacity for valenced experience for two reasons. First, short-lived creatures may not have enough time to learn from pleasure and pain. Second, short-lived creatures are less likely to encounter novel situations, and thus pre-programmed, innate behavior may be a better strategy for maximizing fitness. Short-lived creatures need to start acting right away in life in order to survive.[45]

Two factors complicate the importance of this feature. First, lifespan is a measure of how long an organism is alive. However, life after fertility cessation (the generic term for what in humans is called ‘menopause’) doesn’t contribute to fitness. From an evolutionary perspective, the value of learning is that it helps an organism pass on its genes. Time to sexual maturity may thus seem like better evidence regarding the capacity for valenced experience. Unfortunately, time to sexual maturity doesn’t capture the huge variance among animals in number of reproductive cycles. Two creatures might both reach sexual maturity in 30 days, but if one creature dies after a single reproductive cycle while the other goes on to successfully reproduce for months afterward, the former will have less need for valenced experience than the later. Thus, the ideal metric is really something like ‘time to fertility cessation.’ Fortunately, average lifespan is an excellent proxy for this metric. The vast majority of animals die shortly after fertility cessation.[46]

Second, it is unclear at just what point average lifespan becomes evidence against the capacity for valenced experience. Octopuses don’t normally live more than a year or two, a far shorter lifespan than comparably intelligent vertebrates. Still, two years seems like plenty of time to encounter unexpected situations, and it is considerably longer than the mere month that typical fruit flies live. On the other hand, the probability of encountering novel experiences also depends on the type of environment in which an animal lives and the way in which the animal interacts with that environment.[47] Animals in tropical regions, which are highly variable, probably encounter more novel stimuli per unit of time than comparable animals in temperate zones.[48] Similarly, many migratory species (like monarch butterflies) must contend with a variety of environments and hence probably undergo a larger number of novel experiences than comparable non-migratory species. Thus, although the general point that short-lived creatures have less of a need for valenced experience may be sound, the importance of lifespan length for a given species cannot be assessed without appropriate background knowledge and context. And without at least a rough grasp of the cutoff where, ceteris paribus, valenced experience becomes more of a burden than a benefit, information about lifespan may not be particularly useful.[49]


Motility is the ability to move spontaneously and actively via the consumption of metabolic energy.[50] For our purposes, strict motility should be distinguished from passive locomotion, in which an organism utilizes the environment (e.g., wind or water currents) for transportation. Strict motility should also be distinguished from adaptations such as positive phototropism, in which the shoots of plants ‘move’ by growing toward light.

For motile organisms, the ability to spatially model the world (in some loose sense) and then incorporate one’s own movements into that model would seem to confer a fitness advantage. Representing the world and organizing sensations and self-movement into a special model are things that consciousness appears to facilitate in humans. Indeed, according to some researchers, the evolutionary function of consciousness is to create an integrated and egocentric spatial model to guide an organism as it moves through a complex environment.[51][52] Thus, there is some reason to think that motile organisms are more likely to be conscious than sessile (non-motile) organisms.

Distinct sleep-wake states

Adaptive inactivity is found throughout the biological world. In the plant kingdom, for instance, seeds will stay dormant until soil conditions are optimal for survival. Many unicellular organisms enter seasonal dormant states due to changes in temperature or moisture level. In response to predictable, unfavorable environmental conditions, many insect species undergo a period of diapause during some stage of development, which can last for weeks, even months. Many mammals, including some bats and many species of rodents, hibernate during the winter.[53]

Daily sleep cycles are one form of adaptive inactivity. Sleep reduces brain and body energy consumption, it increases behavioral efficiency, and it is thought to play a role in supporting brain plasticity, learning, and memory.[54] Sleep deprivation has a number of adverse effects. In humans it is associated with reduced cognitive abilities, and studies suggest it impairs both the immune system and the body’s ability to heal wounds.[55]

There is a great diversity of sleep patterns in the animal kingdom. Sleep length ranges from 2 to 20 hours a day. In general, carnivores tend to sleep longer than omnivores, who in turn sleep longer than herbivores. There is also great diversity regarding depth of sleep. In general, animals more at risk of predation awaken more easily than those animals not at risk of predation.[56]

Because sleep is one type of adaptive inactivity, and adaptive inactivity comes in many forms and is exhibited to various degrees, there is no sharp cutoff between behavior that qualifies as sleep and behavior that does not. Thus, evidence for sleep in invertebrates is open, to some extent, to interpretation. In vertebrates sleep is characterized by certain electrophysiological, energetic, and behavioral changes, such as “decreases in muscle tone, heart rate, breathing, blood pressure, and metabolic rate.”[57] Insofar as invertebrates display similar changes, the more likely it is that they experience recognizable sleep/wake cycles. Some of the potential points of similarity are behavioral sleep, detrimental health effects from sleep deprivation, related genes, and similar neurological features.[58]

Some have argued that sleep cycles are an indication of consciousness because there is a time when the creature is aware and a time when it is not.[59] Surely, however, this move is too quick. Jennifer Mather takes a more measured approach. She writes, “Sleep is another area in which the linkage of brain organization to behavior is obvious… However, linkage of sleep to consciousness must be more about the details of sleep than the possession of sleep itself. As is true for pain, the underlying physiology may be parallel but the behavioral manifestations more complex and indicating a higher order of control.”[60]

Noxious Stimuli Reactions

Physiological responses to nociception or handling

In humans, conscious pain states are associated with a number of autonomic physiological changes, such as elevated heart rate, pupil dilation, increased blood pressure, elevated respiratory rate, and increased body temperature. If similar changes are detected in nonhumans after exposure to noxious stimuli, that is modest evidence that those creatures also experience conscious pain states. Two standard caveats apply. First, in general, the more phylogenetically distant an organism is from humans, the less similar we should expect its physiological responses to conscious pain to be.[61] (For example, among earthworms, one common reaction to noxious stimuli is the secretion of copious amounts of nitrogenous mucus.[62]) Second, even in humans it is not always clear whether autonomic changes are a response to conscious pain or merely to nociception. It thus seems likely that organisms which do not experience conscious pain nevertheless undergo physiological changes in response to noxious stimuli. Hence, this feature is best used in the negative: if a biological organism musters no physiological response to noxious stimuli, that is strong evidence that the organism does not feel conscious pain.

Protective behavior

Protective behavior is a type of non-reflexive reaction to injury in which an injured animal attempts to guard, groom, or otherwise tend to the injured body part. Examples include limping, wound rubbing, wound licking, and wound guarding. These reactions are typically part of a long-term response to injury, measured in hours and days rather than seconds and minutes. In general, long-term responses to putatively painful experiences are better evidence for conscious experience than acute responses. Protective behavior, in our sense, must be carefully distinguished from reflexive reactions known (in humans) to operate subconsciously, such as grimacing, rapid withdrawal, postural adjustments, and some paralinguistic features of vocalization.[63]

In humans protective behavior is often mediated and controlled by the conscious sensation of pain. For example, a human will typically avoid using an injured limb for certain activities or types of movement that would aggravate the limb, but not other activities or types of movement that would not do so. Humans also modify the motion involved in those activities somewhat in order to engage in them without aggravating the limb. This is controlled in a nuanced way through the sensation of pain. In the absence of defeaters, similar behavior among nonhuman animals is evidence they too feel conscious pain.

Evidence for protective behavior among vertebrates is fairly extensive, and for some time it was thought that this type of behavior is restricted to vertebrates. For instance, Eisemann et al. (1984) write, “No example is known to us of an insect showing protective behavior towards injured body parts, such as by limping after leg injury or declining to feed or mate because of general abdominal injuries. On the contrary, our experience has been that insects will continue with normal activities even after severe injury or removal of body parts” (166).[64] The Eisemann review is still sometimes cited as evidence that insects (or invertebrates more generally) don’t exhibit protective behavior, for example in Tye 2016.[65] However, new evidence is beginning to undermine the old consensus. Coordinated grooming behavior in response to noxious stimuli has recently been reported in honeybees,[66] cockroaches,[67] and fruit flies.[68] It is as yet unclear whether these reactions constitute genuine protective behavior.[69]

Defensive behavior/fighting back

In the wild many animals face near-constant risk of attack from predators and/or conspecifics. Animals have evolved a diverse and complex repertoire of processes to respond to immediate and potential threats of this kind. Sea hares release a noxious mixture of ink and opaline[70] when attacked.[71] Crabs will self-amputate their claws if firmly grasped by predators.[72] Jellyfish deploy cnidocytes, explosive cells containing a paralyzing toxin, to protect themselves from predation. Collectively, these responses are known as defensive behavior.

Defensive behavior is modest evidence of negatively valenced emotional states like fear and anger. Felt fear may be an adaptive tool to teach animals to avoid situations that are dangerous, and felt anger may be an adaptive state when confrontation is inevitable.[73] On the other hand, defensive behavior is “phylogenetically old and exhibited by organisms throughout the animal kingdom” and thus may be mediated by mechanisms much simpler than conscious emotional states.[74]

Defensive behaviors should be distinguished from more passive adaptations, such as spines, armor, camouflage, mimicry, and preemptive avoidance (e.g., nocturnality), that help an animal reduce the chance of an attack. These defensive structures are active all or most of an animal’s life, and thus do not provide specific evidence for occurrent negatively valenced emotional states.

When a human is in pain, she often cries out. Many vertebrates do the same when exposed to putatively painful stimuli. Hence, groaning, whining, whimpering, yelping, screaming and other such vocalizations[75] might plausibly be considered modest evidence of conscious pain. In fact, vocalization has recently been taken to be a good metric of animal welfare in farmed pigs, cows, and chickens.[76]

We ought to view this feature with caution, however, for two reasons. First, owing to anatomical differences, we should not expect invertebrates to vocalize in the same manner as vertebrates, if at all. More importantly, although vocalization is associated with conscious pain in humans, it’s not clear that the vocalization is always caused by the conscious sensation of pain. At least some researchers regard pain-related vocalization in humans as reflexive and automatic, akin to one’s withdrawing her hand from a hot stove before she even realizes she is touching it.[77] Thus, noxious stimuli related vocalizations may be mediated by mechanisms much simpler than conscious experience.

Movement away from noxious stimuli

The ability to withdraw from potentially damaging stimuli is a basic evolutionary feature and is extremely well-conserved among motile organisms. This ability was probably one of the primary early advantages of motile organisms over sessile (non-motile) organisms.[78] Even many unicellular organisms are capable of moving under their own power toward attractants such as energy sources and away from repellents such as toxins. (This ability often takes the form of chemotaxis, movement under the influence of a chemical gradient.[79]) In humans, the withdrawal reflex can occur unconsciously and even in comatose patients. Because this feature is so easily satisfied, it probably reveals little about an organism’s capacity for valenced experience.


This essay is a project of Rethink Priorities. It was written by Jason Schukraft with contributions from Max Carpendale. Thanks to Kim Cuddington, Marcus A. Davis, Peter Hurford, Tegan McCaslin, Daniela Waldhorn, and Rachael Woodard for helpful feedback. If you like our work, please consider subscribing to our newsletter. You can see all our work to date here.


  1. Vertebrates constitute a subphylum in the phylum Chordata. Cladistically, it would be more precise to speak of ‘chordates’ and ‘non-chordates.’ In using the terms ‘vertebrates’ and ‘invertebrates’ we defer to common usage. However, the number of invertebrates in the phylum Chordata is trivial compared to the number of invertebrates outside Chordata, so common usage is not wholly inaccurate. ↩︎

  2. We use the terms ‘sentience,’ ‘phenomenal consciousness,’ and ‘subjective experience’ interchangeably. An organism is sentient just in case there is something it is like to be that organism. ‘Valenced experience’ denotes a proper subset of conscious experience in which experiences take on a positive or negative affect. All creatures with the capacity for valenced experience are necessarily sentient, but not all sentient creatures necessarily have the capacity for valenced experience. Note: ‘sentience’ gets used in different ways by different philosophical communities. In philosophy of mind, the term is normally used in its broad sense, to mean ‘phenomenal consciousness.’ (See, inter alia, this SEP article on animal consciousness.) In moral philosophy, the term is normally used in its narrow sense, to mean ‘valenced experience.’ (See, inter alia, this SEP article on the grounds of moral status.) We have adopted the philosophy of mind usage. ↩︎

  3. It is important to remember that, even in humans, neurons are not restricted to the brain. The human enteric nervous system stretches across the gastrointestinal tract and is capable of mediating reflex behavior independent of the brain or spinal cord. It contains about 100 million neurons. (It’s especially important to remember that brain neurons are a proper subset of overall neurons when considering invertebrates, which tend to have less centralized nervous systems. Two-thirds of an octopus’s neurons, for instance, are located in its tentacles.) ↩︎

  4. David A. Drachman. 2005. “Do We Have Brain to Spare?Neurology 64 (12). ↩︎

  5. This does not necessarily give them greater precision in movement; insects on average have similar numbers of distinct muscles in total. ↩︎

  6. Lars Chittka and Jeremy Niven. 2009. “Are Bigger Brains Better?Current Biology 19: R995–1008. ↩︎

  7. Those claims are sometimes partially retracted. And sometimes effective altruist organizations resolve some of the discrepancies between the original post and the data that led to the retraction. ↩︎

  8. This difficulty is compounded by the fact that invertebrate brains are structured differently than vertebrate brains. It’s not always clear which regions are homologous to which. ↩︎

  9. Seweryn Olkowicz, Martin Kocourek, Radek K. Lučan, Michal Porteš, W. Tecumseh Fitch, Suzana Herculano-Houzel, and Pavel Němec. 2016. “Birds have primate-like numbers of neurons in the forebrain.” Proceedings of the National Academy of Sciences 113, no. 26: 7255-7260. ↩︎

  10. Suzana Herculano-Houzel, Christine E. Collins, Peiyan Wong, and Jon H. Kaas. 2007. “Cellular scaling rules for primate brains.” Proceedings of the National Academy of Sciences 104, no. 9: 3562-3567. ↩︎

  11. I learned this fact (and many others) from Tegan McCaslin’s excellent (2019) “Investigation into the relationship between neuron count and intelligence across differing cortical architectures” for AI Impacts. ↩︎

  12. They measure roughly 1 mm in diameter. Leonid L. Moroz. 2011. “Aplysia.” Current Biology 21: PR60-R61. ↩︎

  13. See Table 1 in Gerhard Roth and Ursula Dicke. 2005. “Evolution of the Brain and Intelligence.” Trends in Cognitive Science 9: P250-257. ↩︎

  14. See Figure 2 in Roth and Dicke 2005. ↩︎

  15. Encephalization quotient depends on which species is taken as the “standard” for the taxon. The 7.4-7.8 figure uses cats as the standard (EQ=1) animal for mammals. ↩︎

  16. Roth and Dcike 2005: 252. ↩︎

  17. Robert O. Deaner, Karin Isler, Judith Burkart, and Carel Van Schaik. 2007. “Overall brain size, and not encephalization quotient, best predicts cognitive ability across non-human primates.” Brain, Behavior and Evolution 70, no. 2: 115-124. ↩︎

  18. Ewan Smith and Gary Lewin. 2009. “Nociceptors: A Phylogenetic View.” Journal of Comparative Physiology A 195: 1096. ↩︎

  19. Lynne U. Sneddon. 2017. “Comparative Physiology of Nociception and Pain.” Physiology 33: 63-73. ↩︎

  20. Adrienne E. Dubin and Ardem Patapoutian. 2010. “Nociceptors: the Sensors of the Pain Pathway.” The Journal of Clinical Investigation 120: 3760-3772. ↩︎

  21. It is clear that nociception is not a sufficient condition for conscious suffering. Neither is it a necessary condition. Humans are able to suffer emotionally from stimuli that are not detected by nociceptors. ↩︎

  22. Reto Bisaz, Alessio Travaglia, and Cristina M. Alberini. 2014. “The neurobiological bases of memory formation: from physiological conditions to psychopathology.” Psychopathology 47, no. 6: 347-356. ↩︎

  23. Sneddon 2017: 67 ↩︎

  24. The term opiate refers to drugs derived from opium poppy. Opioid is broader, encompassing both naturally-occurring and synthetic substances. ↩︎

  25. Craig W. Stevens. 2009. “The evolution of vertebrate opioid receptors.” Frontiers in bioscience: a journal and virtual library 14 (2009): 1247-1269. ↩︎

  26. With the understanding that opioid receptors in vertebrates trivially qualify as opioid-like. ↩︎

  27. See Adam Shriver. 2006. “Minding Mammals.” Philosophical Psychology 19: 433-442 (especially §2) for an accessible overview. ↩︎

  28. Yuan Feng, Xiaozhou He, Yilin Yang, Dongman Chao, Lawrence H. Lazarus, and Ying Xia. 2012. “Current Research on Opioid Receptor Function.” Current Drug Targets 13 (2). ↩︎

  29. Here ‘understanding’ is not meant to be construed in a phenomenal, conscious sense. Non-conscious robots can understand their surroundings in this sense. ↩︎

  30. Peter Godfrey-Smith. 2016. “The Mind of an Octopus.” Scientific American Mind, 28: 62–69 ↩︎

  31. Graziano Fiorito et al. 2015. “Guidelines for the Care and Welfare of Cephalopods in Research–A consensus based on an initiative by CephRes, FELASA and the Boyd Group.” Laboratory Animals 49(S2): 1-90. ↩︎

  32. Todd E. Feinberg and Jon M. Mallatt. 2017. The Ancient Origins of Consciousness. MIT Press: 25. ↩︎

  33. Mirjam Appel and Robert W. Elwood. 2009. “Motivational Trade-Offs and Potential Pain Experience in Hermit Crabs.” Applied Animal Behaviour Science 119: 120–24. ↩︎

  34. Masafumi Oizumi, Larissa Albantakis, Giulio Tononi. 2014. “From the Phenomenology to the Mechanisms of Consciousness: Integrated Information Theory 3.0.” PLOS Computational Biology 10(5): e1003588. ↩︎

  35. More formally known as the mesencephalon ↩︎

  36. See, inter alia, §4 “The Capacity for Consciousness Depends on Functions of the Neocortex, a Brain Structure Unique to Mammals” in James D. Rose. 2002. “The Neurobehavioral Nature of Fishes and the Question of Awareness and Pain.” Reviews in Fisheries Science 10 (2). ↩︎

  37. Orrin Devinsky, Martha J. Morrell, and Brent A. Vogt. 1995. “Contributions of Anterior Cingulate Cortex to Behaviour.”Brain: A Journal of Neurology, 118(1), 279-306. ↩︎

  38. Bjorn Merker. 2007. “Consciousness without a Cerebral Cortex: A challenge for Neuroscience and Medicine.” Behavioral and Brain Sciences, 30(1), 63-81. It should be noted that this claim only applies to children born without a neocortex. Adults with damaged neocortices remain completely vegetative. ↩︎

  39. Erich D. Jarvis et al. 2005. "Avian brains and a new understanding of vertebrate brain evolution." Nature Reviews. Neuroscience. 6 (2): 151–9. ↩︎

  40. Andrew B. Barron and Colin Klein. 2016. “What Insects Can Tell Us about the Origins of Consciousness.” Proceedings of the National Academy of Sciences of the United States of America 113 (18): 4900-4908. ↩︎

  41. We used TimeTree to determine the dates for last common ancestor. TimeTree is a free interactive tool from Temple University’s Center for Biodiversity. It relies on the published data from over 3,000 studies, covering nearly 100,000 species. ↩︎

  42. See Peter Godfrey-Smith. 2016. Other Minds: The Octopus, The Sea, and the Deep Origins of Consciousness. New York: Farrar, Straus and Giroux: 87-97 for an accessible overview. ↩︎

  43. See, inter alia, Stanislas Dehaene. 2014. Consciousness and the Brain: Deciphering How the Brain Codes Our Thoughts. Penguin Books. ↩︎

  44. See, inter alia, Michael Trestman. 2013. “The Cambrian Explosion and the Origins of Embodied Cognition.” Biological Theory 8:80-92. ↩︎

  45. C. H. Eisemann, W. K. Jorgensen, D. J. Merritt, M. J. Rice, B. W. Cribb, P. D. Webb and M. P. Zalucki. 1984. “Do Insects Feel Pain? - A Biological View.” Experentia 40: 164-167. ↩︎

  46. There are literally only three exceptions: killer whales, short-finned pilot whales, and humans. Darren P. Croft, Lauren JN Brent, Daniel W. Franks, and Michael A. Cant. 2015. “The evolution of prolonged life after reproduction.” Trends in Ecology & Evolution 30, no. 7: 407-416. ↩︎

  47. Thanks to Kim Cuddington for raising this point. ↩︎

  48. Ferran Sayol, Joan Maspons, Oriol Lapiedra, Andrew N. Iwaniuk, Tamás Székely, and Daniel Sol. 2016. “Environmental variation and the evolution of large brains in birds.” Nature communications 7: 13971. ↩︎

  49. Note also that animals with faster metabolisms and smaller body sizes tend, according to some metrics, to process information faster. Thus, there is some reason to think that smaller animals have, in general, faster subjective experiences. These considerations could mean that small, short-lived creatures have proportionally greater need for conscious pain and this would weigh against the considerations described in this section. ↩︎

  50. ‘Motility’ should not be confused with ‘mobility.’ ‘Mobility’ is merely the ability of an object to be moved. A basketball is mobile, but it is not motile. ↩︎

  51. Colin Klein and Andrew B Barron. 2016. “Insects Have the Capacity for Subjective Experience.” Animal Sentience 9: 1–19. ↩︎

  52. Michael Trestman. 2013. “The Cambrian Explosion and the Origins of Embodied Cognition.” Biological Theory 8:80-92. ↩︎

  53. Jerome M Siegel. 2011. “Sleep in Animals: A State of Adaptive Inactivity.” Principles and Practice of Sleep Medicine. ↩︎

  54. Stefania Piscopo. 2009. “Sleep and Its Possible Role in Learning: A Phylogenetic View.” Frontiers in Bioscience S1: 437-447. ↩︎

  55. Gümüştekín K, Seven B, Karabulut N, Aktaş O, Gürsan N, Aslan S, Keleş M, Varoglu E, Dane S. 2004. “Effects of Sleep Deprivation, Nicotine, and Selenium on Wound Healing in Rats.” International Journal of Neuroscience. 114: 1433–1442. ↩︎

  56. Siegel 2011. ↩︎

  57. These changes describe non-REM sleep. REM sleep is only found in mammals and some juvenile birds. D. Purves, G. J. Augustine, D. Fitzpatrick, L. C. Katz, A. S. LaMantia, J. O. McNamara, and S. M. Williams. 2001. “Physiological changes in sleep states.” Neuroscience. 2nd edn. Sunderland, MA: Sinauer Associates. ↩︎

  58. Jerome M.Siegel. “Do All Animals Sleep?.” Trends in Neurosciences 31: 208-213. ↩︎

  59. David Papineau and Howard Selina. 2000. Introducing Consciousness. New York: Totem. ↩︎

  60. Jennifer Mather. 2008. “Cephalopod Consciousness: Behavioural Evidence.” Consciousness and Cognition 17: 37-48. ↩︎

  61. Of course, phylogenetic distance is an imperfect proxy for physiological similarity. Physiologies across very distantly related taxa can converge due to the fitness landscape for a given morphological space, and relatively close taxa can have important differences in physiology. Thanks to Tegan McCaslin for bringing this point to my attention. ↩︎

  62. Robert H. Ressler, Robert B. Cialdini, Mitchell L. Ghoca, and Suzanne M. Kleist. 1968. “Alarm Pheromone in the Earthworm Lumbricus terrestris.” Science 161: 597-599. ↩︎

  63. Kamal Kaur Sekhon, Samantha R. Fashler, Judith Versloot, Spencer Lee, and Kenneth D. Craig. 2017. “Children’s behavioral pain cues: Implicit automaticity and control dimensions in observational measures.” Pain Research and Management 2017: 3017837. ↩︎

  64. C. H. Eisemann, W. K. Jorgensen, D. J. Merritt, M. J. Rice, B. W. Cribb, P. D. Webb and M. P. Zalucki. 1984. “Do Insects Feel Pain? - A Biological View.” Experentia 40: 164-167. ↩︎

  65. Michael Tye. 2016. “Are Insects Sentient?”. Animal Sentience 9 (5): 3. ↩︎

  66. Victoria Hurst, Philip C. Stevenson, and Geraldine A. Wright. 2014. “Toxins nduce 'malaise’ behaviour in the honeybee (Apis mellifera).” Journal of Comparative Physiology A 200: 881-890. ↩︎

  67. Marianna I. Zhukovskaya. 2014. “Grooming Behavior in American Cockroach is Affected by Novelty and Odor.” The Scientific World Journal, article ID 329514. ↩︎

  68. Timothy Murphy et al. 2015. “A Behavioral Assay for Mechanosensation of MARCM-based Clones in Drosophila melanogaster.” Journal of Visualized Experiments 106: 53537. ↩︎

  69. Many invertebrates, such as crustaceans and spiders, engage in autotomy (self-amputation under threat, e.g., a lizard casting off its tail when grasped by a predator), but we have classified autotomy as a defensive behavior rather than a protective behavior. Autotomy is a response to danger, not a response to injury. ↩︎

  70. Opaline is a whitish fluid secreted by sea hares that becomes viscous upon contact with water. ↩︎

  71. Cynthia E. Kicklighter, Markus Germann, Michiya Kamio, and Charles D. Derby. 2007. “Molecular identification of alarm cues in the defensive secretions of the sea hare Aplysia californica.” Animal Behaviour 74, no. 5: 1481-1492. ↩︎

  72. Michael H. Robinson, Lawrence G. Abele, Barbara Robinson. 1970. “Attack Autotomy: A Defense against Predators.” Science 169: 300-301. ↩︎

  73. Robert J.Blanchard, D. Caroline Blanchard, Guy Griebel, David Nutt. 2008. “Introduction to the Handbook on Fear and Anxiety.” Handbook of Behavioral Neuroscience 17: 3-7. ↩︎

  74. Christoph P. Wiedenmayer. 2009. “Plasticity of Defensive Behavior and Fear in Early Development.” Neuroscience and Biobehavioral Reviews 33: 432-441. ↩︎

  75. Here we use "vocalization" in a broad and not exclusively phonetic sense to mean more than sounds that literally emanate from vocal chords. ↩︎

  76. Gerhard Manteuffel, Birger Puppe, and Peter C. Schön. 2004. “Vocalization of farm animals as a measure of welfare.” Applied Animal Behaviour Science 88, no. 1-2: 163-182. ↩︎

  77. Bo Karlsson, Gunilla Burell, Ulla-Maria Anderberg, and Kurt Svärdsudd. 2015. “Cognitive behaviour therapy in women with fibromyalgia: A randomized clinical trial.” Scandinavian Journal of Pain 9: 11-21. ↩︎

  78. Sessile organisms such as plants have necessarily evolved different strategies to protect themselves from noxious stimuli. ↩︎

  79. Gunjan Pandey and Rakesh K. Jain. 2002. “Bacterial Chemotaxis toward Environmental Pollutants: Role in Bioremediation.” Applied and Environmental Microbiology 68 (12): 5789–5795. ↩︎

Sorted by Click to highlight new comments since:

Great post, I really like this series of posts from RP and look forward to the rest. I have a few comments about this one:

-Should a distinction be made between operant and classical (associative) conditioning in the requirement for valence to facility learning? I agree that learning to associate positive or negative experiences with an environmental state (such as probcius or sting extension reflexes in honeybees) require a valence cue.

However, the role of valence is less clear during operant condition which is often used to tune how a reflexive sensorimotor action is executed. For example, fixation paradigms have been used to study insect flight behaviour (an insect tries to position a visual object frontally) - if the coupling between their turning behaviour is artificially reversed (such that by turning right the object also appears to move right) an insect can learn to reverse the usual direction of their motor output to regain control visual position of the object. This is definitely (sensorimotor) learning, but doesn't require an extrinsic valence cue to achieve (although the insect has an internal prefered world state it is comparing its sensory experience to, I'm not sure that the state error is analogous to an internal valence cue).

-The length of a life span correlating to the potential for learning are not entirely clear, as I think that most relatively long lived insects still have a lot of reflexive behavioral cues they rely on (and perhaps tune slightly with operant conditioning as above). Eusocial central place foragers a clear exception in that they are well known to have excellent capability for navigation and associative learning.

But in the example of long distance insect migrations (e.g. Monarchs butterflies or Bogong moths), most seem to simply follow genetically programmed instincts about where to go. Intuitively one might think that these insects will need to learn which flowers to forage on at different stages of their migration, although it could be that the have genetically programmed innate preferences for flowers that work well along the entire migration route. There is a bit of work on associative learning for Monarch butterflies - it seems they are capable of associative learning over a time scale of days (very slow compared to bees which can manage single trial learning), but they also have strong innate preferences for flower colours.

-How would you class an animal releasing a warning pheromone as a reaction to noxious stimuli? Lots of eusocial species do this to summon extra soldiers to attack a threat so I would probably call that case a defensive behaviour (in addition to being a physiological response).

Yet, in other cases warning pheromones are released to warn other conspecifics (for instance, aphid alarm pheromone causes dispersal) - for the insect being attacked this isn't really defensive as it doesn't benefit from the other aphids avoiding the threat (which are themselves moving away from the signal of a noxious stimulus); maybe it is almost analogous to a chemical vocalization?

Aside, even some plants issue chemicals that warn other plants or summon protective insects when they are attacked by herbivores.

Hi Gavin,

Thanks for the compliment and especially for the thoughtful reply. I’ll take your comments in turn.

  1. In the third part of the mini-series on features potentially relevant to invertebrate sentience, we discuss a number of learning indicators, including both classical and operant conditioning. That post is going up June 12. I would be interested to hear your take on the relevant sections.

  2. There is certainly not going to be a perfect correlation between lifespan and potential for learning. (Indeed, there might not be any correlation at all.) The claim that we’re defending is that, in general, longer-lived organisms would benefit more from learning abilities than shorter-live organisms. We expect there to be exceptions both ways (i.e., relatively short-lived organisms that would benefit from learning abilities and relatively long-lived organisms that wouldn't). Much depends on context of various kinds. Your point about the learning abilities of monarchs vs. bees is well-taken. In future work (to be published mid-July), we take an especially close look at eusocial insects, which are pretty amazing.

  3. Your question about warning pheromones is a great example of a difficulty that has hounded us for the length of the project. Classifying and assessing complex behaviors is context-sensitive. I think you’re right that warning pheromones could fall into at least three categories. (Or maybe different pheromones fall into different categories?) Assessing the evidential force of these features is often even more context-sensitive. A behavior that looks like good evidence for sentience in one context doesn’t always look like good evidence for sentience in a different context. (e.g., a human reporting “I am in pain” is normally great evidence of painful experience. A very simple robot programmed to utter the same sounds is not great evidence of painful experience.)

Curated and popular this week
Relevant opportunities