Cornelis – 

I’m glad to see such strong interest in the potential experiences of invertebrates. I’ll note that I agree with several major points: more research is needed, we don’t know what shrimp experience when shocked at low levels or when asphyxiated, and electrical shocks that don’t stun before slaughter seem worse than no stunning before slaughter. I elaborate on these points at the end. 

But first I’d like to weigh in on some of the neurobiological and physiological points you raise. I am an academic insect neurobiologist and thermal physiologist and have taken a strong interest in invertebrate nociception since 2021. I co-authored the Gibbons et al. 2022 pain review on insects and a separate article rebutting common arguments against pain in insects (Barrett and Fischer 2025). I currently research larval farmed insect nocifensive behaviors as well as slaughter and stunning methods and am thus generally very interested in how invertebrates may experience death. 

I spent only about two hours on these replies and so they may be imperfect as they rely mostly on my own background knowledge instead of new research (and I am not a shrimp nor human expert). Therefore, caution is advised and more reading into the peer-reviewed literature is warranted as not all of these areas are strong expert zones for me. Still, I think my replies will be useful. And I think there’s more that could be unpacked here with further research, which I did not have time to do. 

Finally, this post shouldn't be taken as an endorsement of electrical stunning for invertebrates or of SWP.

1. “As I understand it, there is no evidence that shrimp possess type C nociceptors … All you need is fast, sharp "first pain" -> the Aδ channel -> which shrimp likely possess…”

Here, your claim for the lack of type C nociceptors has a citation of the lack of fulfillment of criterion 3 in Crump et al. 2022 - i.e., the absence of identified connections between peripheral nociceptors and sensory integrative brain regions. However, this is strange evidence for your claim, as signal integration in the brain does not bear on nociceptive fiber type. In humans, signals from both types of fibers do reach the brain – so whether the signal reaches the brain does not actually bear on which type of nociceptor even a vertebrate has. But, a look at Crump et al. also shows we have not proved that the signal does not reach the brains of the shrimps – rather, it just hasn’t been studied whether any signals could reach the brain in enough detail to fulfill the criterion. As you note in your footnote, then, absence of evidence is far from evidence of absence. All we can take from this ‘very low’ evidence for Criterion 3 in Crump et al. is a statement about what scientists have cared to study in decapods.  

Second, comparing invertebrate nociceptors to vertebrate nociceptors based on “class” is always tricky business. For instance, shrimps have neither Aδ nor C type fibers in truth – these categorizations apply only to vertebrates, where nociceptors are classified based on “conduction velocity (fast or slow), diameter, amount of myelination, and the types of stimuli to which they respond” (Tracey, 2017). Aδ fibers, our ‘fast acting’ fibers, are myelinated and large, which allows for their rapid conduction velocities, and typically they only respond to one or two stimuli. However, shrimps (and other invertebrates) lack myelin in their nervous system, ruling out the possession of these fibers on the basis of myelin entirely. 

C type fibers, our ‘slow’ fibers, are unmyelinated and smaller in diameter in vertebrates, which actually makes them a better analog for invertebrate nociceptors than the Aδ on that basis alone. Further, they are polymodal, which makes them much more similar to insect class IV (see below) nociceptors, which are also polymodal.

I am using insect examples here because they are crustaceans’ closest relatives (forming the group ‘Pancrustacea’ together, as the ancestor of insects was a crustacean) and we know significantly more about their nociception thanks to the model system of the fruit fly. So, if you were going to try and make an analogy between fiber types in vertebrates and invertebrates like insects/shrimps (which I would not recommend!), C type fibers would be the better bet structurally and, in some specific ways, functionally. 

And, contra your claim that “since there is far better evidence that decapods show acute nociceptive-defensive responses than that they have a slow/tonic pain channel”, in fact, the correct prediction would be that shrimp nociceptors should be tonic, excitatory nociceptors. See, for instance, Puri and Faulkes (2010) hypothesizing that the shrimp nociceptors should be tonic: “Second, crustacean nociceptors are likely to be tonic excitatory neurons that will show a rapid and sustained increase in the rate of action potentials when exposed to potentially noxious stimuli.” As further evidence that invertebrate nociceptors can be both phasic and tonic, see the prolonged and non-accommodating phasic-tonic response of American cockroach nociceptive responses in Emanual and Libersat (2019).

In truth, though, it is better to leave behind entirely the categorizations of C and Aδ fibers when moving from mammals to invertebrates. In insects, we refer to their nociceptors as “class IV multidendritic arborization neurons” (for heat, mechanical and chemical responsiveness) and “class III multidendritic arborization neurons” (for cold). These neurons are unmyelinated and, as I described, polymodal with (at least in some cases) both phasic and tonic capabilities. The stimuli applied can also change how the nociceptor responds (re: phasic or tonic, as a single nociceptor can do both sometimes), at least in some insects (Caron et al. 2020) – and how this works likely changes by species.

2. “Ischemic and hypoxic tissue accumulates lactate and protons, and that acidosis activates acid-sensing type C nociceptors and this seems to directly lead to the experience of long-term suffering in humans.”

Zheng et al. (2021) indicates that in decapods that experience chronic hypoxia, cells will switch to anaerobic respiration to generate energy, with the resultant excessive accumulation of lactic acid occurring if metabolic demand outstrips removal capacity. So, it’s true that hypoxic tissue may accumulate lactic acid to toxic levels in decapods under hypoxia as well.

Importantly, in insects where these channels are better studied, class IV nociceptors are sensitive to some acids, including in insect larvae (some of which have pretty simple nervous systems with fewer neurons than shrimps; Lopez-Bellido et al. 2019) – leaving open the possibility of sensing tissue acidosis at the periphery with the nociceptors that these animals may plausibly have. 

Regarding decapods specifically and their capacity to sense tissue acidosis as noxious via the nociceptors, it is hard to know what to make of the results highlighted in Puri and Faulkes (2010). They show an apparent lack of 1 mol/L and 6 mol/L (pH 0 and pH -0.78) hydrochloric acid sensing when applied externally on the antennae of multiple decapod species. 

A few possible notes, however: acid nociception is not consistent across all acids in insects. For instance, in fruit fly larvae 9% citric acid (pH = 1.71) never elicits nocifensive behaviors while both lower pH 9% sulfuric and hydrochloric acid (pH ~ -0.4) and higher pH 9% acetic acid (pH = 2.23) do elicit responses (Lopez-Bellido et al. 2019). Puri and Faulkes (2010) do not detect responses when using hydrochloric acid; but Barr et al. (2008) and Elwood et al. (2017) both used higher pH acetic acid and found nocifensive responses on the antennae, eyestalk, or mouth. So, this could reflect differences in the kinds of chemicals the nociceptive channels can sense. 

Alternately, variation in swabbing protocol across individuals could have impacted the results, as Puri and Faulkes note no consistent effects of the acid across all individuals, not no effects of the acid at all: “Although some neurons in some individuals increased their activity in the noxious condition, the variation from individual to individual indicates that these were spontaneous variations in neural activity rather than responses evoked by the noxious stimuli.” Importantly, it’s strange that they interpret this variation as spontaneous – as the traces shown in Figure 5 mostly seem to show this variation in the noxious treatment and not the saline controls (but, they may not be representative traces or I may be misinterpreting the result with such a quick review as I’m doing here). Lastly, single studies are never great evidence on their own - in general, we do want science to be built on a replicable and repeated body of work with molecular mechanisms also understood. The paucity of evidence here is extreme and it makes conclusions difficult to draw.

However, given that invertebrates have entirely different nociceptors than vertebrates, how much should we even port over from our understanding of human asphyxiation at all at this juncture? Further, given that asphyxiation may be experienced very differently for aquatic animals v. terrestrial animals, it seems at least plausible that we should be extraordinarily cautious with claims that attempt to generalize across such wide taxonomic and environmental/physiological boundaries. This definitely seems like an avenue where further investigation could be beneficial to better understand the arthropod experience of asphyxiation (which I have argued for insects is also unclear, as they may also be asphyxiated during slaughter; Barrett et al. 2022).

3. “A living being with type-C nociceptors slowly dies from suffocation:

• chemoreceptors -> chemoreflex -> maybe this isn't experienced negatively?
• acidosis -> type C -> more acidosis -> sustained C-fibre firing -> "long pain" that accumulates over minutes”

I know very little about human asphyxiation, so I cede a lack of peer-reviewed evidence in what briefly follows. It seems like asphyxiation plausibly contains more negative experiences than only pain, even for human beings that have type-C nociceptors. I asked Gemini about asphyxiation to death for humans; Gemini replied: “Suffocating triggers a violent, terrifying panic known as "air hunger". It feels like a crushing weight on the chest, a burning fire in the lungs, and a desperate, inescapable urge to breathe that no gasp can satisfy. Physical Sensations: Air hunger: A fierce panic reflex in the brain screaming for oxygen. Chest pressure: A heavy, tight weight crushing the ribs. Burning pain: A hot, stinging ache in the throat and lungs. Head rush: Throbbing temples, dizziness, and a hot, swelling pressure in the head”. This all seems like a pretty negative experience to me and much of it is happening before lactate has built up to the point that external tissue acidosis is the major problem.

My Cleveland Clinic also notes: “Your body will try to start breathing again, which may look frightening. Your eyes may bulge, your skin may change color, your hands may grab at your throat and you may weakly cough. Your blood pressure and heart rate increase sharply (spike), your blood pH drops and your body releases catecholamines. 

Eventually, your blood pressure drops, and your heart slows down.” 

[As we’ll see later with shrimps, a heart rate increase followed by a heart rate slow down is also exhibited in decapods subject to air exposure that are not adapted to emersion (point 5).] It seems to me, then, that this experience plausibly causes intense suffering even without necessarily inducing peripheral nociceptor activation, at least in the human case.

Further, it seems plausible to contest that the primary experience of suffering in humans is not actually due to acidosis in acute hypoxia of the kind that causes rapid death (perhaps it is the case in chronic hypoxia, where people have a disease that leads to a much slower accumulation of hypoxic tissue?) - my very minimal non-peer-reviewed reading suggests that tissue acidosis generally happens in minute 2-5 of suffocation while loss of consciousness often happens faster in average individuals under oxygen-store depletion models (1-3 minutes). So, it seems like most of the suffering may be unrelated to nociceptor activation through acidosis, which may be occurring most acutely after loss of consciousness. Also, the chemoreflex is apparently only one of several mechanisms regulating hypoxia response in humans and is not likely to be sufficient for explaining how we experience air hunger given what’s written in that article.

Finally, I suspect that we should take care in generalizing across even all experiences of ‘air hunger’. For instance, in humans I think this experience would be fundamentally different based on the way my air was short or cut off – for instance, if I were choked, I expect the suffering might be extreme; if I were a trained free diver, I might (?) have a different experience prior to passing out; and if my air were steadily replaced, the gas used and amount could do a lot of work in determining my suffering level (carbon monoxide v. carbon dioxide, for instance). 

4. “But patients in a persistent vegetative state breathe spontaneously and mount chemoreflex responses to hypercapnia and hypoxia. The brainstem loop is intact; the experience is absent. So we have direct proof that the ventilatory chemoreflex and the felt state of air hunger are dissociable.

Ergo, asphyxia may plausibly produce two dissociable bad things: air hunger (interoceptive, non-nociceptive) and acidosis-driven nociceptive pain. And it may well be the latter which is important for establishing whether a being actually suffers.”

This claim seems potentially philosophically confused. First, nociception is not the only thing that produces suffering (and, pain can also happen without nociceptive input – see Tracey 2017), as already stated. Second, we also have direct proof that nociception and pain are dissociable. In humans, there is a nocifensive reflex that withdraws your hand from a hot stove that is dissociable from the felt state of pain experienced in your brain. An anesthetized person, without paralytics, can still respond reflexively to noxious stimulation, presumably without the experience of pain. Yet, the fact that there is a reflex involved does not dissuade us from being concerned about the felt state that may also be involved in humans. So, it does not follow any more than for nociception-pain that, just because the ventilatory chemoreflex and the felt state of air hunger can be dissociated, they are in fact totally dissociated in the living animal and thus it’s all reflex and no negative experience (or, at least, it doesn’t follow any more than this claim follows for nociceptive reflexes).  

5. “Green crabs held in air didn't ramp up: gill pumping dropped within minutes and stayed low for an hour. Weak evidence, since pumping rate can't distinguish "no drive" from "drive that can't be expressed." But it's the only decapod measurement I can find, and it points away from air hunger being a thing in shrimps rather than toward it.”

Green crabs, according to the abstract, are a tidal species that naturally and regularly experiences daily in-air emersion as a part of their life cycle (“in nature, the green crab exhibits emersion and terrestrial activity at low tide”) and thereby have significant emersion adaptations. They will even preferentially go into air to avoid unfavorable positions in the water and are, explicitly, amphibious (living in both water and air by design). It is also described in the introduction of your cited study as a species with unusual hypoxia tolerance and unique gill structures that prevent collapse in air. This study looked at the effects of exercise on land for a species well-adapted to emersion. These data do not bear on marine shrimp that are not hypoxia tolerant, are not amphibious, and consequently have no emersion adaptations. 

Better evidence for the effects of air exposure stress on shrimps not adapted to emersion might come from Wang et al. (2020), who found that air exposure activates the hypoxia response pathway and induced cell death in gill and hepatopancreas cells in one non-emersion-adapted shrimp species. Heart rate initially increased in response to air exposure, before beginning to decrease right around the time mortality picks up in the population and right after lactate levels peak (Figure 4a). This meta-analysis suggests that emersion is one of the most potent stressors for decapods not adapted to living in air (e.g., non-amphibious decapods), only preceded by trawling. I suspect more research on shrimp with variable hypoxia tolerance may be out there, but these studies were the first to come up in my search. 

Do we have any other evidence that hypoxia seems to cause stress - potentially, a kind of suffering - for decapods behaviorally or physiologically? A review of the literature by Zheng et al. 2021 shows that, across diverse species, acute and chronic hypoxia can damage gill tissue, affect energy metabolism, impair intestinal function, diminish survival, cause cellular self-instructed death, induce oxidative damage (including increasing MDA levels, which may be worthy of investigation as a cross-taxa indicator of negative welfare: Beaulieu 2024), and impair reproductive tissue. Evidence points to the animals detecting hypoxic environments: they jump, reduce feeding frequency, rise to the surface, preferentially go to areas that are shallow with aquatic plants, and attempt escape (species that can go on land, such as crabs, apparently will favor emersion over hypoxic water). From a fitness perspective, it makes sense for these animals to be able to detect and attempt to avoid this fate - and welfare states are expected to be linked to fitness-affecting stimuli, encouraging animals to avoid what is bad for them by making it feel bad.

According to the review, acute hypoxic stress triggers the following physiological strategies, dependent on the species and context: higher ventilation and circulation rates to attempt to enhance oxygen levels in smaller-bodied species like shrimps; increasing oxygen-binding affinity of hemolymph proteins; and reducing movement to attempt to decrease metabolic demand. In entirely marine crustacean species without the right gill structure (unlike their amphibious or terrestrial counterparts), air emersion can eventually trigger collapse of the gills and this strongly correlates with death (Johnson and Uglow 1985).

6. “Acute pain like that from electric stunning does not require type C  nociceptor activation. All you need is fast, sharp "first pain" -> the Aδ channel -> which shrimp likely possess[3],”

Electrical stunning causes pain by directly depolarizing neurons, and any and every neuron is sensitive to this stimulation regardless of their nociceptive potential or ‘fast’ or ‘slow’ conduction velocity. All these neurons are depolarized because they contain voltage-gated ion channels as part of their signaling mechanisms and so all that is necessary to trigger them is sufficient voltage. 

Thus, both human A and C channels will be depolarized by electricity, as well as sensory interneurons that carry nociceptive signals onward and a host of other neurons – such as motor neurons – causing spasming as those cells activate, even though they are non-nociceptive. There may be different activation thresholds – for instance, larger and myelinated A-fibers seem to have lower electrical activation thresholds than smaller, unmyelinated C-fibers (from informal reading). But, at the thresholds used to stun an animal, you’re past the activation threshold of all these fiber types – in this zone, electricity is the great neuronal activation equalizer. Similarly, in invertebrates, all nociceptors depolarize with voltage-gated ion channels and are thus sensitive to electrical stimulation. In theory, electricity could also cause pain or suffering by activating more central neurons – not just peripheral nociceptors – as pain can occur without nociceptor activation in theory (Tracey 2017).

7. “Reading the preprint, matthes found that at lower shock voltage/duration, neural activity sometimes increased, only 3/6 animals showed a 90% drop within 30 min, and that group had worse outcomes than 0°C ice slurry alone.”

Electricity characteristics will matter to the experience of the animal undergoing stunning. For instance, pulse width, duration, and amplitude, voltage/current, how the electricity routes through the body, and even stunning context, like wet or dry, can affect the animals’ experiences. Significant increases in neuronal firing after electrical stun is actually expected in most animals: this is typically described as a general, epileptiform seizure in animals like pigs, characterized by sudden, abnormal surges of activity in the brain. In humans, pain is a recognized but rare part of epileptiform seizures – and pain more often accompanies the seizure ending and the return of normal brain activity. So, increases in neural activity are not inherently problematic following stunning, as consciousness can still be offline and pain thereby minimized during seizures that are tonic-clonic and generalized (in the way electrical stuns are expected to be). Still, even the Somerville study acknowledges that further research is needed to confirm if this increase in neural activity actually represents an epileptic insult and I think from their behavioral data that is an important gap to fill.

That said, Matthes is right to surmise that we should expect an incorrect stunning protocol that does not result in loss of consciousness followed by cold shock to be worse than cold shock alone. But, to me this is not surprising a priori? The same is true for chickens that are inadequately shocked before being blasted with hot water on the line. This does not, on its face, however indicate that correctly-done stunning at appropriate voltages and durations is worse for chickens – or for shrimps. It just means we should be very careful to dial in the stunning parameters to maximize successful, <1 second stunning that achieves loss of consciousness while minimizing the risk of accidentally shocking-but-not-stunning animals. 

In your later sections on proposed research, you write about the value of looking at stunning that does not induce unconsciousness v. asphyxiation for shrimps – but actually, I think this question misses the point: we do not need to be investigating low-grade stunning because we never intend to be delivering ineffective stuns in the first place. Instead, we should be making stunning protocols that are definitively consciousness-losing in less than one second (and generate a loss of consciousness long-lasting enough to sustain that state during cold shock). That’s the only target worth aiming for. And, if we can successfully do that, then the animal’s response to lower voltages is a non-issue (barring the expected error rate of all stunning methods). My point is that we shouldn’t conduct research that may be extremely painful for the animals involved if we think it also doesn’t bear on the real-world application of the method we actually want to employ.

I have so many further thoughts on this – but perhaps I’ll save them for SWP, as I think I’m about out of the time I allotted myself here. 😉

So, where do agree?: “I think experiments here are genuinely missing?”

Yes. Absolutely. But I will tell you – as an animal researcher working on behavioral paradigms in poorly studied, non-model taxa with no baseline cognition research and little descriptions of even their sensory abilities, the experiments you propose are not easy to do. In insects, motivational trade-offs and conditioned place avoidance work have barely ever been done, even in bees and fruit flies, where thousands of labs work on these animals every year with millions of research dollars. Despite all that funding, effort, talent, and ingenuity aimed at these animals for nearly 60 years, these experimental paradigms have still proven somewhat challenging to test in even model invertebrates. 

This is, in some ways, surprising. Different species of bees used as models for cognition research have been shown to count; to have abstract concepts of same and different; to remember human faces (see my 80K episode for more); and to independently solve the chimp-box-banana problem. Yet there has been only one attempt to study conditioned place/feeder avoidance in the species (and, done poorly, we contend in Gibbons et al. 2022). It’s the only such study I’m currently aware of in bees. So, should we expect this to be easy to do in decapods? Well, we have already demonstrated in this post that we barely have working, replicable acid nociception demonstration in any decapod… so, it would take a lot of work to go from the body of research you currently have to the kind of work you currently want. 

Note: This isn’t to dissuade you! I am excited to spend my next ~forty years dedicated to the cognitive and nociceptive abilities of poorly studied animals. But you should expect the data you want to be methodologically difficult to obtain, expensive, and full of false negatives that have more to do with your assay design than your species’ capabilities. This work can be very, very hard to do well despite, on its face, being relatively easy to imagine. 

I have some specific thoughts on the exact conduct of the experiments you proposed and the value of the data they’d provide. If you find someone interested in performing this work (and qualified), and a funder who cares to take it on, then I’d be happy to work with them on these questions and the question of whether the data are worth the money required to get them, given their expected impacts on our credences in the current empirical landscape. 

I’m also interested in the question of cold nociception in shrimps, as you are; I’m deep in the tail flip literature, trying to understand how we can parse what a tail flip means when it comes to the nervous system’s response to see if that might give us some clues. This, it turns out, is complicated. 

Finally, we also agree that we should do a good job validating our stunning protocols to ensure they are actually rendering animals rapidly unconscious and not simply shocking them before they go through slaughter. But, as expected, this work too is very challenging. The empirical literature published to date reflects the extraordinary difficulty, as does my own experience trying to get this to work in invertebrates. But, as Matthes suggests, it may be possible. And, if it is, it may indeed save the animals from the many plausible kinds of suffering that could be generated from asphyxiation. 

Last thought: Of course, there is much more to learn on all fronts and I would encourage significant empirical work in this and many other areas of invertebrate welfare science. I returned to academia precisely for this reason, and believe strongly in the value empirical research can provide to the EA movement and the animals it aims to help. I live this belief through my own animal research program every day. 

But a word of caution to all of us invertebrate-excited folks here on the Forum: I regularly see many inaccuracies when describing invertebrate nervous and physiological systems. Even with over a decade of research experience in this area, the vast array of adaptations and their extraordinary differences from mammals (and one another!) can lead me astray in making hasty assumptions about what invertebrates can do or how their bodies are structured. Please – please, please! – be careful with what you write on the Forum about these animals without getting expert peer-review. It is easy to make empirical mistakes with long-lasting impacts on others’ beliefs; I’m sure it’s possible I’ve even made some myself in my hasty and under-researched reply today. 

To do better by these animals, we must start by bettering our understanding of these animals. We can only do that by providing each other with the best available, most accurate biological information we have. This rigor is a critical part of the EA community – and the invertebrates may need that from us more than most.

I wanted to note that I really appreciate the robust discussion about, and interest in, this post.  As an insect neuroecologist and physiologist, it's gratifying to see a community of folks engaging with insect neuroscience and behavior in such a meaningful way. 

I really do encourage folks to read the full paper - it addresses many of the things you might be interested in (it's hard to capture 75 pages of research in only 1000 words, so we necessarily lost some detail). If for some reason you can't access it, I'm happy to share a pdf copy or you can email Dr. Chittka, the corresponding author for the paper.

I've reached my time limit for replying to comments here, so please don't take any future silence as indicative that your questions are in any way bad/uninteresting/I don't care about them. I've just gone back to being busy doing more research on bugs.

I replied to your comment before you edited it and added the following, so I will make quick replies to these new questions throughout.

To the extent that we found research on these orders O and criteria C, each of the orders satisfies each of the criteria.

- I would rephrase: For the OxC combinations that we found enough research to make a determination about whether each order O satisfies or fails each criterion C, we found that each order satisfied each criterion.

We are not saying anything about the degree to which a particular O satisfies a particular C. [Uhm, I am not sure why. Are the criteria extremely binary, even if you measure them statistically? Or were you looking at the degrees, and every O satisfied every C to a high enough degree that you just decided not to talk about it in the post?]

- These criteria, like many in science, are actually particularly binary. E.g., you either have nociceptors or you do not have nociceptors (you either are an insect, or you are not an insect!). So, in this way, we are assessing satisfaction/failure in a binary sense. 

- But, of course, there are relevant degrees that emerge after determining satisfaction or failure. For example, you might have more types or fewer types of nociceptors. They might be expressed in greater or fewer numbers. Ion channels could be expressed in different types of cells or sequestered on the interior of the cell following expression for different amounts of time/due to different physiological causes. In all cases, these degrees would not change our determination about whether the animal group possesses nociceptors (e.g., satisfies/fails the criterion). But, of course, these degrees might have some relevant effects on our eventual credence for pain! We do spend some time on this in the paper (which is 75 pages and I could not replicate here! But see criterion 7 for some of this light discussion of degrees).

- To my mind, the point of this framework, and determining pass/fail for the criterion, is to 1) determine whether it is worth taking the idea of pain seriously in an animal group (e.g., providing evidence for or against applying some version of a precautionary principle); and 2) determining where we should direct research effort by identifying areas where we don’t yet have high-quality evidence for or against the satisfaction of criteria that might be relevant to insect pain.  

To recap: you don't talk about the degrees-of-satisfying-criteria, and any research that existed pointed towards sufficient-degree-of-C, for any O and C. Given this, the tables in this post essentially just depict "How much quality-adjusted research we found on this."

- We talk about this briefly, for criterion 7, but to be clear, there was relatively little ‘degrees-of-satisfying’ evidence to be found in insects at this time. In most OxC cases, as the table demonstrates, there wasn’t even sufficient evidence to demonstrate with high/very high confidence that the order met or failed to meet the binary condition of the criterion – much less the degrees of meeting/failure, after having met or failed it.

In particular, the tables do not depict anything like "Do we think these insects can feel pain, according to this measure?". Actually, you believe that probably once there is enough high-quality research, the research will conclude that all insects will satisfy all of the criteria. (Or all orders of insects sufficiently similar to the ones you studied.)

[Here, I mean "believe" in the Bayesian sense where if you had to bet, this is what you would bet on. Not in the sense of you being confident that all the research will come up this way. In particular, no offense meant by this :-) .]

- I don’t really understand the first part of this. But I guess I would say that the table itself doesn’t represent any particular quantifiable credence that insects feel pain. However, the summation of these lines of evidence can provide some traction for thinking about how seriously to take the idea of insect pain at all - even if it again doesn't give us any particular credence level.

- Re: point 2, I strongly disagree that this is my belief. I believe that once there is enough high quality research for each criterion and order, we will be able to conclude whether or not insect orders satisfy or fail all criteria. There are a select few criteria where available evidence suggests that we might bet on more research coming up satisfactorily – e.g., criterion 3 in adults, where evidence for integrated nociception is distributed across the phylogeny and criterions 1 + 2 (the preconditions) are also robustly met, plus we know that both 1 + 2 are robustly conserved across all the insect orders. However, in most O x C cases we have very, very little data – or even no real data (criterion 8, particularly!) – that can lead us to make any specific or generalizable conclusions about the likelihood of any or all insects meeting that criterion.

This comment is representative only of MRB's opinions and expertise, and not the other post/publication authors.

These are great questions. I want to say at the outset here: we explicitly chose to stick closely and without major adjustments to the Birch et al. 2021 framework for this review, such that our results would be directly comparable to their study of decapods and cephalopods that led to the protection of those groups in the Animal Welfare (Sentience) Act 2022 in the UK. 

Here is what the paper says about the framework, and the confidence levels:

The five possible confidence levels are:

(1) “Very high confidence”, when the weight of scientific evidence leaves no scope for reasonable doubt;

(2) “High confidence”, when we are convinced that the animal satisfies or fails the criterion, although scope for reasonable doubt remains;

(3) “Medium confidence”, when concerns about the evidence's reliability and quality prevent us from having high confidence;

(4) “Low confidence”, when there is little or flawed evidence;

(5) “Very low confidence”, when the evidence is seriously inadequate or when there is no evidence whatsoever to make a determination (e.g., no confidence).

So, to be clear the colors are not ratings of 'very high confidence that they satisfy the criteria' but rather 'very high confidence that we can determine whether they satisfy or fail the criteria'. This is why it's important that we clarify that there is no evidence that insects fail any criteria.  Green happens to mean 'satisfies' and not 'fails' in all cases for our study, but that's a result of how the evidence shakes out - and not specified by the framework itself, which does allow for the collection of 'evidence against' the criterion.

So, in Rockoptera - let's imagine we had really, really high quality evidence that they did not meet any criteria. We would find a row of green boxes - we are very highly confident in our determination that they fail 8/8 criteria. Then, when we do our final summation of the evidence, as in Section 4 of the paper: satisfying 0/8 criteria is classified as "capacity for pain unknown or unlikely" in the group. They explicitly state "If remaining indicators are uncertain [e.g., white/red] rather than shown absent, sentience (or pain) is simply unknown. However, if high-quality scientific work shows the other indicators to be absent [e.g., green], pain is unlikely" (emphasis and brackets mine). So, we would say, according to Birch et al. 2021 that we have very high confidence that pain is unlikely in Rockoptera. Whew! :)
 

I do think there's the potential here for color = "pain satisfaction" to be a common misunderstanding; so it seems like future iterations of this work (as you note) might be presented more clearly by adding some kind of symbology that interacts with the color, such that you know whether green = satisfy vs. fail upon immediately looking at it.

Last thought - re: 1000 high quality papers, split down the middle and with no way to resolve the data by considering the biological context. I would classify this as 'Medium confidence' - we do not have the evidence to be convinced of satisfaction or failure (e.g., high) but the evidence is apparently neither little nor flawed (e.g., low).

This comment is representative only of MRB's opinions and expertise, and not the other post/publication authors.

Vasco - thanks so much! Sajedeh did a wonderful job with the tables (as well as many other great figures in the publication itself).

Rethink Priorities has an interdisciplinary team working on the many complexities of moral weight (including how important to consider neuron numbers). I know more work will be coming out from that team in the next two weeks on this topic, which I'm sure they can speak about more elegantly and critically than I can.

You can follow that project here: https://forum.effectivealtruism.org/s/y5n47MfgrKvTLE3pw

This comment is representative only of MRB's opinions and expertise, and not the other post/publication authors.

Multiple groups of animals do fail certain, or even likely all, criteria. For example, sponges do not have neurons and therefore fail criteria 1 - 3. Although I'm not aware of tests of 4-8 in the sponges, it seems reasonable to suppose that the lack of a nervous system would also preclude endogenous neurotransmitter systems and many of the behavioral criteria.

Importantly, it is not just the number of criteria but also which criteria are fulfilled when considering the likelihood of pain or sentience. Some criteria provide more important evidence for sentience than others. In particular, criteria 2/3 and 5 have been considered important evidence by many philosophers and scientists. The roundworm, C. elegans, fails criterion 2 according to Irvine (2022); we can have reasonably high confidence that this failure is not due to an absence of evidence, but rather evidence of absence, given that we have a complete connectome of the C. elegans nervous system (302 neurons per animal). The fact that all adult insects fulfill criterion 2 with very high confidence, and that we find fulfilment of criterion 3 across the Holo/Hemimetabola split (suggesting the potential for broad taxonomic conservation of this criterion), is a meaningful distinction between these animal groups (at least to me, in the context of this framework).

Irvine E (2022) Independence, weight, and priority of evidence for sentience. Animal Sentience 32 (10). DOI: 10.51291/2377-7478.1724.

This comment is representative only of MRB's opinions and expertise, and not the other post/publication authors.