(Edit 2022-10-20: I've revised this post to clarify a few things, add a summary and split the remainder of the text into sections. I've also made two corrections, more on which below.)
This post was written with a general audience in mind, but may still be of interest to EAs.
Great pain for little gain makes a man weary.
– Proverb
Summary
- Probably >100M vertebrates are used for research and ancillary purposes annually.
- Animal testing often seems to involve physical harm or psychological distress, and at minimum lifelong captivity and early death.
- Much animal testing isn't useful:
- Some animal testing is done for cosmetics, which doesn't seem very beneficial for humanity.
- One study found that only ¼ of animals used in research experiments ended up in published studies.
- Many published results either don't translate to human models or don't replicate.
More than 100M Vertebrates Every Year
You may have heard of the scandal where researchers from the University of Minnesota Twin Cities (UMN) intentionally tried to introduce bugs into the Linux kernel codebase as part of a research project. Quoting Ars Technica, the researchers "[emailed] their Trojan-horse patches to Linux kernel maintainers to see if the maintainers detected the more serious problem the researchers had introduced in the course of fixing a minor bug. Once the maintainers responded to the submitted patch, the UMN researchers pointed out the bug introduced by their patch and offered a 'proper' patch – one that did not introduce a newly exploitable condition – in its place."
This upset a lot of people – the Hacker News thread has received over 3000 points and around 2000 comments as I write this. But what was interesting to me was the way in which a lot of people expressed their anger. One user wrote, "Greg [Kroah-Hartman, one of the Linux kernel maintainers,] has all reasons to be unhappy since they were unknowingly experimented on and used as lab rats." "This is ridiculously unethical research," another wrote. "Despite the positive underlying reasons treating someone as a lab rat [...] feels almost sociopathic." The implication is that the UMN researchers have treated the Linux kernel maintainers badly and that, unlike lab rats, the maintainers have thereby been wronged.
But what about the lab rats? It's likely that well over 100 million vertebrates are used for research and ancillary purposes each year. This is in addition to the millions that are dissected for teaching purposes in the U.S. alone. I think the vast majority of the animals used in research are rodents, but it's hard to know for sure given the U.S.'s exclusion of rats and mice from reporting requirements.
(Edit 2022-10-20: The previous paragraph originally referred to the U.S.’s "baffling exclusion of rats and mice from research oversight". (I took this almost verbatim from the linked source.) I got some (legitimate) pushback on this, stating essentially that while rodents aren’t covered by the Animal Welfare Act, their use is now regulated by the NIH. It’s still the case (as the linked source and associated paper make clear) that they’re excluded from some reporting, which I suppose is a form of oversight, and I’ve updated the sentence to make that clear. Consider this a mistake.)
Most arguments against animal testing are based on utilitarian principles, but the view of animal rights that is most convincing to me is the one put forward by Christine M. Korsgaard. She argues that the other animals are what Immanuel Kant called ends in themselves. Because of this, Kant's Formula of Humanity, which says that we should treat people "always at the same time as an end, never merely as a means", also applies to the other animals, at least those that are conscious, like rats and other mammals. Using animals in scientific experiments seems like a clear case of using those animals as a mere means, just as the UMN researchers used those Linux kernel maintainers to further their own ends. The maintainers could not consent because they were given no opportunity to do so. Similarly, even if animals being tested upon could consent to it, which they cannot, it seems highly unlikely that they would choose to do so, if for no other reason than that they are often euthanised after the experiment is over.
If you are like me, when you think about ending animal testing, the first thing you think about is all the knowledge we humans would be giving up along with it. But if you are like me, you also don't think about all the knowledge we are leaving at the table through our refusal to subject humans to these sorts of experiments. Maybe the reasons that we feel differently about these two things are loss aversion and the endowment effect; the current state of things seems natural because we are used to it. To counteract those biases, Korsgaard invites us to a thought experiment:
[T]o check that your moral intuitions are really what you think they are, try to imagine that we live in a world where no one has ever used animals in research before, and someone proposes it for the very first time. "I know how we could find out if that stuff is toxic. We could drop it into a live rabbit's eye! I know how we can study withdrawal symptoms. We can get a bunch of dogs addicted, and then take away the drugs! I know how we can find out if that substance causes cancer. We can give it to a bunch of monkeys, and see if they get cancer! After all, these creatures are completely at our mercy, so why not?"
The main utilitarian argument for animal testing (though I don't mean to imply that utilitarians as a group are for animal testing; I don't know) is exactly this, that though it does consign the animals to short, mostly joyless and sometimes painful lives, it adds more than enough disability-adjusted life years (DALYs) for humans to compensate. Korsgaard makes the point that this is a difficult calculation to make. Even if you can find one particular invention – insulin is often brought up – and estimate how many DALYs it has netted us, it is nearly impossible to figure out how many animals were used in developing it, simply due to the incremental, cumulative nature of research. It's also hard to know if and when it would've been invented without animal studies. Maybe it is possible to come up with some sensible approximate analyses of this sort; I leave that to utilitarians to figure out. I follow Korsgaard's Kantian reasoning that animal testing is categorically wrong (though for practical reasons I think it should be ended incrementally).
(Edit 2022-10-20: The previous paragraph originally referred to lab animals' "short lives full of agony". I now think "full of agony" was an exaggeration and have changed it to "short, mostly joyless and sometimes painful lives".)
The Lives and Origin of Lab Rats
At this point, I ought to admit that this is personal for me, not because I am involved in this kind of research or because I am a Linux kernel maintainer or anything of that sort, but because my wife recently got two pet rats, two sweet if somewhat shy girls given by their previous owner the names Jim and Beam. They are gentle, curious, playful and unexpectedly intelligent creatures. Jim is a little bit bolder and wilder; Beam plans for her future by hoarding food. I have known them a mere week and still I cannot stand the thought of someone subjecting them to invasive experiments. They deserve better.
Like today's laboratory rats, Jim and Beam are domesticated Norway rats, also known as fancy rats, descendants of the brown rat. Brown rats were scuttling around the Asian steppes until the 17th century, when they began to live alongside humans and alongside us spread over the world. They were then domesticated in Europe, not for research purposes, but chiefly for use in rat-baiting, a blood sport in which a promoter would put a terrier in a pit full of rats and time it to see how long it took until it had killed every single one of them. This was a common gambling sport well into the 19th century. The Wikipedia page quotes an article from 1822 that gives a colourful description of one event:
Thursday night, Oct. 24, at a quarter before eight o'clock, the lovers of rat killing enjoyed a feast of delight in a prodigious raticide at the Cockpit, Westminster. The place was crowded. The famous dog Billy, of rat-killing notoriety, 26 lb. weight, was wagered, for 20 sovereigns, to kill 100 rats in 12 minutes. The rats were turned out loose at once in a 12-foot square, and the floor whitened, so that the rats might be visible to all. [...] At four minutes and three-quarters, as the hero's head was covered with gore, he was removed from the pit, and his chaps being washed, he lapped some water to cool his throat. Again he entered the arena, and in vain did the unfortunate victims labour to obtain security by climbing against the sides of the pit, or by crouching beneath the hero. By twos and threes, they were caught, and soon their mangled corpses proved the valour of the victor. [...] At seven minutes and a quarter, [...] the victor relinquished the glorious pursuit, for all his foes lay slaughtered on the ensanguined plain.
In the mid-19th century, humans started breeding these rats specifically for use in laboratories. I suspect that accounts of those early animal experiments are a horror show, whereas contemporary research involves a more humane treatment of the animals. What follows are brief descriptions of four recent rat studies that I found on Google Scholar. The sole inclusion criterion was my being able to understand the paper.
- Hobson et al. injected rats with diisopropyl fluorophosphate, a pesticide that causes brain damage, followed by injections of substances that reduce mortality. This would cause strong seizures, tremors and uncontrolled urination, salivation and excretion among other things.
- Magalhães et al. gave their rats a "chronic unpredictable stress protocol". If you are curious, as I was, about what that really means, have a look at the supplementary documents, where you can read that it involves the researchers' daily restraining the rats for an hour, putting them in cold water for an hour, exposing them to hot air, shackling them, overcrowding them or reverting the light/dark cycle for a full day.
- Atalay et al. exposed the backs of naked rats to UVA radiation regularly for a period of four weeks, after which the rats were anaesthetised and had their hearts cut out. (Some of the rats also received protective treatment.) UVA radiation causes skin damage, including cancer.
- Ouyang et al. subjected rats to high-pressure blasts in order to give them tinnitus, which it did. They then tested the rats' hearing by startling them with sudden sounds. Five weeks later, 8/13 rats had developed chronic tinnitus.
Much Animal Testing Isn't Useful
Those four studies are probably among the best of what we can wish for in animal testing. To begin with, they aim to gain knowledge that would be useful in treating human health problems. By contrast, many animal tests carried out today are done to validate the safety of cosmetic products. To me, the case for ending the use of animal tests in developing mascara, deodorants, shampoos and other beauty products looks very strong due to the limited benefits these products have. This practice has already been phased out in the United Kingdom, the European Union, Israel, Norway, India, New Zealand, Taiwan, Switzerland, Guatemala, California, and Colombia. A positive recent development is China's exempting some beauty products from animal testing requirements; previously, such products had to be tested on animals in order to be sold there. But I see no good reason that animal testing of cosmetics should not be banned totally and everywhere.
Even if we narrow our focus to animal tests that are done in science, for purposes like developing medicines or studying diseases, we see that a lot of the research basically isn't useful. For example, van der Naald et al. looked at seven years' biomedical research using animals at the University Medical Center Utrecht and found that only 60% of the studies led to one or more publications, and that of the 5590 animals used in the studies they looked at, only 26% ended up in published research. (For small animals, which made up the majority of those used, the number was 23%.) The main reasons for not publishing were "lack of statistical significance, technical problems and objections from supervisors and peer reviewers".
Here is Gwern Branwen summarising the research:
On the general topic of animal model external validity & translation to humans, a number of op-eds, reviews, and meta-analyses have been done; reading through some of the literature up to March 2013, I would summarize them as indicating that the animal research literature in general is of considerably lower quality than human research, and that for those and intrinsic biological reasons, the probability of meaningful transfer from animal to human can be astoundingly low, far below 50% and in some categories of results, 0%.
The primary reasons identified for this poor performance are generally: small samples (much smaller than the already underpowered norms in human research), lack of blinding in taking measurements, pseudo-replication due to animals being correlated by genetic relatedness/living in same cage/same room/same lab, extensive non-normality in data, large differences between labs due to local differences in reagents/procedures/personnel illustrating the importance of "tacit knowledge", publication bias (small cheap samples + little perceived ethical need to publish + no preregistration norms), unnatural & unnaturally easy lab environments (more naturalistic environments both offer more realistic measurements & challenge animals), large genetic differences due to inbreeding/engineering/drift of lab strains mean the same treatment can produce dramatically different results in different strains (or sexes) of the same species, different species can have different responses, and none of them may be like humans in the relevant biological way in the first place.
An interesting review here is that by Gardner. He notes that very few drugs that make it through animal trials also succeed in human trials. That is usually not because they end up being unsafe for humans, but because they end up being ineffective. As an example, Garner cites a review of over 200 interventions for Alzheimer's disease; all of them were effective in mice, zero in humans. (I believe these kinds of drugs are tested using rodents specially bred or engineered to have early onset Alzheimer's, a cruel thing in its own right.) Gardner argues that humans are so different from the other animals that "an animal experiment (as currently conducted) cannot reasonably predict the outcome of a human trial".
One of the reasons all of this matters is that, if we want to reduce animal suffering, we could likely have a larger impact eliminating some of the experiments that wouldn't produce any useful results than marginally improving the conditions of test animals in general. In addition to being a waste of animals, bad research is also a waste of time and money. But another reason is its showing us that the loss involved in ending animal testing is probably smaller than we think.
When it comes to developing vaccines and drugs specifically, the debate on animal testing often assumes that the only alternative to it is testing on humans. People think that animal testing is good because it is worse to subject a human to experiments than it is to subject other animals to them. In a way that is true, because most humans have richer lives with more integrated viewpoints than the other animals. But it is also false, because animals cannot meaningfully consent to being experimented upon, whereas humans can. We alone can understand the risks and dangers (and benefits) involved and make an informed choice.
But more importantly, I think that framing is a mistake for the reason that there are alternatives to animal testing that don't involve testing on humans. Take for example vaccine development. There have so far been something like 7-13 million excess deaths worldwide during the COVID-19 pandemic. The best way we have of responding to these sorts of crises is by developing vaccines quickly. Animal testing, though it is time-consuming and has a low success rate, is used extensively in developing human vaccines. A recent paper by Busquet et al. reviews alternative, animal-free methods for developing human vaccines. These new methods have, they write, "consistently proved to be human-relevant and effective, allowing safe progression to clinical testing in a shorter amount of time as compared to traditional animal testing". As Siobhan Ballan writes:
Instead of using animals to test vaccines, newer models involve the use of tissue cultures or computer software. Healthy or diseased humans are also used to complement findings from other methods. [...] There have been two key developments in animal-free testing in Europe. These are the establishment of the European Center for the Validation of Alternative Methods and the area having received substantial funding of around €50 million per year over the last 15 years. Animal testing for cosmetic purposes is banned within the E.U., which has incentivised a shift towards animal-free testing for medical purposes. [...] Today, around 90% of research methods in academia and industry don’t involve animal testing. Instead, cell-culture based methods play a major role.
Conclusion
In short, much of animal-testing research is carried out in the service of comparatively trivial goals like producing cosmetics. Much of the rest either produces no useful knowledge at all, or knowledge that doesn't translate to humans. And much of what does seemingly translate to humans does not replicate. No one should be surprised that there is a lot of bad and uninformative research out there, but everyone should be horrified when such research involves the mistreatment and killing of innocent animals.
Now, it is true that animals used in research are both far fewer than those used in food production and also produce more benefits, and that we therefore should focus our resources on reducing harms from factory farming. One effective way of doing that is reducing your meat and animal product consumption. Another, more effective way is to donate to charities like the Animal Welfare Fund or any of those recommended by Animal Charity Evaluators. That said, there are a number of things you can do that would help reduce the suffering involved in animal testing on the margin.
If you are a researcher or have influence over animal model research, you may want to follow the recommendations of van der Naald et al., chiefly preregistering your study and sharing your data. Preregistration increases the probability of quality research (and decreases the probability that the research is useless). Data sharing (including of null findings) makes sure the research body isn't skewed towards positive results, and may also reduce study duplication.
If you are not a researcher and have no influence over animal model research, your options are more limited, and it probably makes sense to focus on factory farming instead. But you can try to buy cosmetics and over-the-counter drugs that are explicitly vegan (anecdotally, it seems to be getting more common for cosmetics manufacturers to label their products as vegan when that is the case, at least in Germany). Another option might be to donate to Faunalytics which distills research on animal testing. A final possibility – though I am not sure about the impact here – might be, if you live in a country or U.S. state where it is still legal to use animals for testing cosmetics, to advocate locally for a ban on such testing.
How many studies that use non-animal models end up in published research? Lack of statistical significance leading to non-publication is an issue that is not specific to research on animals, but fairly widespread.
Again the counterfactual is really unclear. If we didn't use animal models even fewer drugs would succeed in human trials.
those are both good questions. i tried to find base rates with a cursory search but came up empty-handed. maybe i just didn't use the right search terms, though. but even if the numbers here are the same as the base rate, i would argue that's still pretty bad, because the costs involved in animal testing are higher. i think it makes sense to judge animal-testing research more stringently than other research. though base rates would be useful to see e.g. how difficult it would be to improve methodologies and reduce the amount of unproductive research.
one thing i didn't make clear in the post but which i now realise i should've is that an experiment not getting published due to lack of statistical significance (or more precisely rejection of the null hypothesis) doesn't mean that research wasn't valuable -- it could have been rejected due to publication bias.