by Brian Tomasik
First published: 7 Feb. 2017; last update: 15 Apr. 2017


Fischer (2016) discusses the question of whether ethical vegans should oppose entomophagy, the eating of insects. Fischer argues that vegans should ultimately favor entomophagy because the process of growing and harvesting plant-based foods causes great harm, to both vertebrates and invertebrates on crop fields.

This argument is flawed in many ways:

  1. Most insect farming relies heavily on cultivated crops as feed, which means that farming insects for human consumption usually causes more total crop cultivation than eating plants directly.
  2. Fischer overestimates the number of vertebrates killed on crop fields.
  3. Fischer's estimate of invertebrate deaths due to pesticides ignores the vast differences in size between different types of invertebrates.
  4. It's not clear if crop cultivation does in fact cause net harm to wild animals from a utilitarian standpoint.
  5. Fischer underestimates the suffering that insects endure when they're farmed and slaughtered.


Fischer (2016) admirably aims to consider the consequences of giving moral weight to invertebrates, a class of animals usually neglected in discussions of animal ethics. However, as I'll show in this piece, Fischer's argument relies on several questionable empirical assumptions. Even if Fischer's point could apply in a few edge cases for vegans with particular moral views, Fischer's argument is exactly wrong with respect to most commercial entomophagy.

Much of Fischer's article revolves around precautionary principles, which I won't discuss in detail. Personally, I'm not a fan of precautionary arguments. I would rather compute expected values and then use whatever qualitative heuristics may be necessary to temper those conclusions. For this reason, I was happy to see that Fischer proposed an expected-value calculation toward the end of his article. I'll discuss aspects of it in some sections of this piece.

Problem 1: Most human-edible-insect farming uses cultivated crops

Fischer argues in favor of eating insects over eating plants on the assumption that insect farming doesn't require mechanized crop cultivation. However, this assumption is usually wrong in practice: Most commercial insect farms that sell human-edible food in developed countries use primarily grains, and fruits/vegetables, as insect feed. I separated the evidence for this point out into a separate article: "Eating Insects Is Usually Less Efficient Than Eating Plants". This point completely reverses Fischer's argument. If Fischer is concerned about vertebrates harmed by crop cultivation, he should be encouraging vegans to eschew entomophagy.

Fischer explains in a single paragraph (p. 7) that his argument focuses on insects raised on food wastes rather than insects raised on newly cultivated plants. But he should have made this qualification more central to his discussion. As it stands now, Fischer's article risks causing people who don't read the fine print to eat grain-fed insects, thereby increasing harm not just to invertebrates but also to vertebrates, according to Fischer's own lights.

Problem 2: Fischer overestimates vertebrates killed by crop farming

Fischer's reconstruction of a hypothetical pro-vegan argument (p. 6) compares the harm of mechanized crop cultivation against the harm of killing insects for food. That argument includes this sentence: "For example, let’s estimate that 1000 clearly-sentient animals will die per acre as a result of standard agricultural practices." But 1000 deaths by clearly-sentient animals (which I take to mean "deaths by vertebrates") is probably too high. 1000 per acre is 2,471 per hectare. But Davis (2003), which Fischer cites when discussing harm to vertebrates on crop fields, estimates only 15 vertebrates killed per hectare per year by intensive cropping (p. 390). And Fischer also notes (p. 3) that "Lamey (2007) makes a good case that Davis overestimates the number of animals harmed in plant agriculture."

That said, Fraser (2012) cautions (p. 730) regarding the Davis/Lamey debate that "most of their calculations appear to have been based on data for Apodemus sylvaticus taken from Tew and Macdonald (1993) rather than more numerous species such as Microtus arvalis as studied, for example, by Jacob (2003)."

Archer (2011) argues that in Australia, due to mice plagues roughly every four years, "At least 100 mice are killed per hectare per year [...] to grow grain."

My best guess is that Fischer was overestimating vertebrates killed per hectare by crop farming because the argument he was making in that section was written from the perspective of a hypothetical vegan who was trying to show that even if vertebrate deaths on crop fields are extremely high, eating insects is still worse. However, it's important that readers don't mistake this overestimate as reasonable.

Following are some further data points to suggest that vertebrate deaths due to crop cultivation are much lower than 2,471 per hectare.


Table 1 of Gaston et al. (2003) estimates that breeding bird densities on cropland range between 1 and 5 per hectare. Gaston et al. (2003) caution (p. 1295): "These figures are based on breeding birds, and would be inflated by non-breeders during the breeding season, and by post-breeding individuals at other times." So maybe the actual numbers are several times more, but probably not vastly more.

And I would guess that most birds aren't killed by crop harvesting because they can fly away.


British mammals

This page estimates that there are ~285 million wild mammals in Britain in the spring pre-breeding season. Great Britain has a land area of 209,331 km2 ≈ 21 million hectares. So this suggests roughly (285 million) / (21 million) = 14 wild mammals per hectare. Unless mammals are vastly more densely packed on crop fields than elsewhere, this number gives a rough upper bound on annual mammal deaths due to crop farming.a

Pinkert et al. (2002)

Pinkert et al. (2002) sampled small mammals in two cornfields and neighboring regions in South Dakota using "mouse-sized snap traps" (p. 40). The authors used 7-by-7 trap grids (p. 40), which implies 49 traps in each location. The traps were separated 9.09 meters apart from one another (p. 40), which means each trap occupied about 9.092 = 83 m2. The authors also say they effectively sampled 0.4047 hectares = 4,047 m2 (p. 40), which confirms that the number of traps was (4,047 m2) / (83 m2 per trap) = 49 traps.

"[T]raps were set for three consecutive nights" (p. 40), over 8 different sampling areas: 4 pre-harvest and the same 4 post-harvest (p. 42). With 49 traps per sampling region, this implies (3 nights pre-harvest + 3 nights post-harvest) * (4 locations) * (49 traps per location) = 1176 "trap nights" in total, a number that the authors confirm (p. 41). It's unclear how to translate the number of mammals trapped per 3-day period into an estimate of total mammal density in the area.

  • On the one hand, the number of mammals trapped might underestimate the actual mammal density if not all the mammals got trapped. As the authors note (p. 43): "Shrew population estimates may be underestimated because snap traps are not as effective in catching these species as are live or pitfall traps (Fowle and Edwards 1954, Mengak and Guynn 1987, Briese and Smith 1974, Wiener and Smith 1972)."
  • On the other hand, if mammals from outside the 7-by-7 grid entered the grid area and got trapped, the traps might sample more than just 0.4047 hectares.

If the number of mammals trapped roughly approximates the total number in the sampled area, we could divide the number of mammals caught by 0.4047 hectares to get rough densities per hectare. Numbers of mammals captured on cornfields in the below tables range from 4 to 12, which very roughly suggests densities around 4/0.4047 = 10 to 12/0.4047 = 30 per hectare.

Moreover, as you can see in the above tables, mammal numbers on cornfields were not reduced by harvesting, although I suppose it's possible that both (1) lots of mammals were killed on cornfields but (2) lots of other mammals came in to replace them. For example, maybe outside mammals were attracted to the grain on cornfields following harvest. Pinkert et al. (2002) mention (p. 43) that "Proportion of white-footed mice in the shelterbelt decreased after harvest, suggesting attraction to post-harvest cropland. This may be the result of increased waste grain and/or competition".

Fraser and MacRae (2011)

Fraser and MacRae (2011) present a nice summary of the literature regarding mammal deaths on crop fields (p. 583):

Agricultural practices such as harvesting and soil tillage can be extremely harmful to animals. Nass et al (1971) monitored radio-tagged Polynesian rats (Rattus exulans) during the mechanical harvesting of sugar cane. Of 33 animals studied in two trials, 14 were killed by the equipment, three suffocated in their burrows under compacted soil, and another three died from predation after the covering vegetation had been removed. Tew and Macdonald (1993) followed radio-tagged wood mice (Apodemus sylvaticus) during grain harvesting, also with a sample of 33 animals. Of these, one was killed by the combine harvester, 17 disappeared within a week likely because of predation when the covering vegetation was removed, and two were suffocated in their burrows when the stubble was burned. Jacob (2003) studied the effects of different farming practices on populations of common voles (Microtus arvalis); he found that mulching, mowing and the harvesting of wheat had little immediate effect on vole survival, whereas virtually all voles disappeared immediately after ploughing, and 44% disappeared after the harvesting of beans.b Disappearance in these cases was presumed to have been ‘mainly due to decreased survival’ (p 24) rather than emigration.c

These numbers suggest that a reasonably high fraction of animals on crop fields might be killed by ploughing or harvesting.

Fraser and MacRae (2011) then present the following table of mammal densities on cropland (p. 583). Note that while some of these numbers are in the hundreds of animals per hectare, none is in the thousands.


I don't have numbers for reptiles and amphibians, but I doubt mortality is vastly higher than for mammals (except maybe for crops like rice that can be more amenable to amphibians than dry fields are).

Also keep in mind that many reptiles, like snakes and frogs, are predators. This raises the question of whether killing them, tragic as it is, might prevent net long-run wild-animal deaths by sparing prey from being eaten alive. Of course, some prey may themselves be predators -- e.g., frogs can be prey for snakes. And it's not obvious whether releasing even herbivorous prey from predation reduces net suffering over the long run. These are complex questions, and they highlight the oversimplification involved in regarding collateral vertebrate deaths on crop fields as obviously bad overall. This theme is explored further in my "Problem 4" section below.

Deaths in rice production

There is one estimate of vertebrate deaths from crop cultivation that's greater than Fischer's 1000 per acre. Ward M. Clark, an opponent of animal rights, quotes "the following first-hand account on a Usenet forum" by "a rice farmer":

[A] conservative annualized estimate of vertebrate deaths in organic rice farming is ~20 pound. ... [T]his works out a bit less than two vertebrate deaths per square foot, and, again, is conservative. For conventionally grown rice, the gross body-count is at least several times that figure. ... [W]hen cutting the rice, there is a (visual) green waterfall of frogs and anoles moving in front of the combine. Sometimes the "waterfall" is just a gentle trickle (± 10,000 frogs per acre) crossing the header, total for both cuttings, other times it is a deluge (+50,000 acre).

It's a bit unclear to me how these numbers fit together. 2 vertebrates per square foot would be 87,000 per acre.

I'm a bit skeptical of such high reptile and amphibian densities, and the source isn't necessarily reliable. Still, it's certainly plausible that rice cultivation is particularly violent, and or this reason, I never eat rice except as someone else's leftovers that will otherwise go to waste. I'm very skeptical of comparable numbers of vertebrate deaths in the production of dry-land crops.

Problem 3: Fischer ignores size differences among invertebrates

Let's now ignore vertebrates and focus just on Fischer's calculations regarding invertebrate deaths. Fischer discounts insect suffering by a factor of 0.01, assuming a "very low" probability of 1% that insects are sentient. However, since I'm only comparing invertebrates in this section, I'll drop that factor.

Based on USDA statistics, Fischer reports (p. 6) that soybeans yield about 2868 pounds per acre = 3215 kg per hectare. This USDA document (p. 11) basically confirms that number, reporting that the "Yield per hectare" of "Soybeans for beans" was 3.23 metric tons per hectare in 2015 (and 3.53 metric tons per hectare in 2016).

Mealworm 01 PengoFischer says (p. 6): "A mealworm weighs about 100 milligrams, so it would take 13,009,000 mealworms to make up for that agricultural loss." This is because one acre of soybeans produces 2868 pounds = 1300.9 kg. We could adjust these numbers a bit to account for nutrition content per gram. For example, "Soybeans, green, raw" have 147 calories and 13 g of protein per 100 g, while mealworms have 223 calories and 20 g of protein per 100 g. But such an adjustment would probably change the calculation by less than a factor of 2, and Fischer doesn't do this adjustment, so I'll ignore it as well.

Fischer compares mealworm deaths from entomophagy to invertebrates killed on crop fields by pesticides. As mentioned in the "Problem 1" section, this comparison is often specious because mealworms are usually fed cultivated crops, but let's assume these mealworms are raised on food waste. In approximating invertebrate numbers on crop fields, Fischer mentions Pearse (1946), which estimates invertebrate densities in the Duke Forest. While forests are different ecosystems from crop fields, invertebrate densities seem unlikely to differ by many orders of magnitude between forests and (unsprayed) crop fields, given that net primary productivity doesn't differ dramatically between the two.

Here's Table 1 from Pearse (1946)d:

The total is indeed around 100 million invertebrates per acree, which is the estimate that Fischer assumes for invertebrates on crop fields. But a big problem is that most of these invertebrates are orders of magnitude smaller than 100-mg mealworms. An invertebrate's size probably roughly correlates with its cognitive sophistication to some degree. In addition, those who care more about bigger brains because they contain more replication of neuronal processes will care more about bigger animals even if they have the same cognitive abilities. And even for those who think sentience is binary, similar considerations can still play a role in influencing an animal's probability of being sentient.

The three most common invertebrate taxa in the above table are (in descending order) mites, springtails, and thrips. Pearse (1946) notes in his literature-review section that mites and springtails are often extremely common (p. 129):

Frenzel [1936] found that mites constituted 42% and collembolans 30.9% of the geobionts, and many other investigators have obtained similar results. In dry soils, however, collembolans may exceed mites in numbers.

But mites, springtails, and thrips are very small:

  • This page says: "Mites of the suborder Oribatida [...] are the world's most numerous arthropods living in soil. [...] These slow moving mites are 0.2 - 1.0 mm in length". Note that Pearse (1946) reported that most of the mites in his study were oribatids (p. 135).
  • This page says: "Collembolans (common name 'springtails') are generally 0.2-5.0 mm in length."
  • This page says: "Thrips are generally tiny (1 mm long or less)".

Meanwhile, mealworm "Larvae typically measure about 2.5 cm or more". In other words, mealworms are at least ~10 times longer than these other invertebrates, which suggests that mealworms are, very roughly, ~103 times bigger in volume.f Perhaps we don't want to give a single mealworm 103 times as much moral weight as a single mite, but a trade ratio of 10 or 100 might not be implausible. And in fact, in order to tip the scales in Fischer's calculation, we need only give mealworms 2 times as much moral weight as mites, springtails, and thrips, as I'll show now.

The number of invertebrates other than mites, springtails, and thrips in Pearse (1946)'s Table 1 is 124,366,656 - 89,602,920 - 28,052,640 - 1,437,480 = 5,273,616. Let's assume (unrealistically) that all these invertebrates have the same sentience as a 100-mg mealworm. And let's say that the 119,093,040 mites, springtails, and thrips each matter half as much as a mealworm. Then the weighted number of invertebrates, measured in terms of mealworms, per acre is 5,273,616 + 0.5 * 119,093,040 = 64,820,136, which I'll round to 65 million.

Fischer supposes that 20% of these invertebrates are killed by pesticides. Fischer thinks 20% is probably too low. I'm uncertain, because I would guess that many of the invertebrates on crop fields are under litter or underground.g Brady (1974) says (p. 559):

Since the purpose of pesticides is to kill organisms, it is not surprising that some of them are toxic to specific soil organisms. At the same time, the diversity of the soil organism population is so great that except for a few fumigants, most pesticides do not kill a broad spectrum of soil organisms. It is perhaps surprising that the extensive use of pesticides in the United States has not provided more extensive evidence of damage to soil organism numbers. Even so, there is evidence that some commonly used pesticides adversely affect specific groups of organisms, some of which carry out important processes in the soil.

This quote from Brady (1974) doesn't tell us whether Fischer's 20% estimate is too high or too low. But if we stick with Fischer's 20% number, we get (20%) * (65 million) = 13 million mealworm-equivalent invertebrates killed by pesticides per acre. This contrasts with the 13.009 million mealworms that Fischer estimates are required to replace an acre of soybeans. Thus, by giving tiny invertebrates only slightly less moral weight per individual than mealworms, entomophagy is (very slightly) worse than eating soybeans directly, even for mealworms grown entirely on food scraps.

Obviously the above calculation is not robust; it can easily flip back and forth with tweaks to the assumptions. But given this fact, Fischer ought to have said "The analysis is very unclear, and we should do further research" rather than (p. 7) "Eating insects wins."

Personally, I think eating plants probably wins, both because of Problem 1 mentioned above and because I give much less weight (maybe at least ~10 times less?) to a single mite/springtail than to a single mealworm. Moreover, it's not clear how bad killing invertebrates with insecticides is, because the invertebrates would die naturally anyway. In contrast, farmed mealworms would not exist at all without entomophagy, and the suffering during their entire lives and deaths could be averted if they weren't farmed. (Of course, some of the food scraps given to mealworms might counterfactually have fed other invertebrates, and this also needs to be considered. However, some of that food also would have been eaten by bacteria/fungi/etc., not other bugs. In the ideal case, composting of food scraps can be done at least partly anaerobically, thereby preventing decomposer bugs from coming into existence at all, or at least reducing their numbers.)

The above calculation doesn't consider the number of pesticide applications per year, which would have to be incorporated into a more thorough analysis.

Fischer also doesn't consider organic soybeans, where synthetic pesticides would not be used. Of course, many organic farms still use pesticides of their own. For example, Bahlai et al. (2010) examined "four novel products to evaluate as potential reduced risk insecticides to include in integrated pest management programs for soybean aphid [...]. Two of these insecticides contained synthetic active ingredients, the other two are natural insecticides permitted for use in certified organic crops in Canada [17]." The authors "found that in addition to reduced efficacy against aphids compared to novel synthetic insecticides, organic approved insecticides had a similar or even greater negative impact on several natural enemy species in lab studies, were more detrimental to biological control organisms in field experiments, and had higher Environmental Impact Quotients at field use rates."h The authors also "included formulations of the two currently registered insecticides in the experiments as conventional controls." It was found that "generally, the two currently registered insecticides were most toxic to natural enemies under laboratory conditions."

And organic farmers may also use biological control, which involves painful predation or parasitism of many insects. It's an important open question whether organic foods entail more or less invertebrate suffering than conventional foods.

Of course, painful biological control also occurs in the natural ecosystems that would exist if land weren't used to grow soybeans. This fact underscores that "leaving nature alone" by recycling food scraps is not an option free from suffering and death either. This point is elaborated in the next section.

Problem 4: Crop cultivation isn't obviously net bad for utilitarians

Deontological vegans will probably always consider crop cultivation to be bad, because it entails some direct harm to wild animals. However, the question is far less clear for utilitarian vegans. The question partly depends on the net balance of happiness vs. suffering in nature. (I think that balance falls on the side of more suffering.) The question also depends on whether crop cultivation increases or decreases total wild-animal populations. This is again unclear; this piece outlines many factors to consider regarding the topic. Conventional wisdom often says that converting land to crop fields reduces wild-animal numbers, and if so, then utilitarians who are pessimistic about the net balance of welfare in the wild may favor it. I'm less certain, since in some cases, crop fields have higher net primary productivity than native vegetation.

From a utilitarian standpoint, it's also not obvious that insecticide use is bad. Indeed, it might be the case that insecticides prevent more insect deaths than they cause, if they keep insect populations on crop fields low enough for long enough.

The question is very complex, and sweeping conclusions are premature unless you're a deontologist. (But, as other authors have also pointed out, deontologist vegans should probably commit suicide given how much harm they inevitably cause to other creatures.)

Problem 5: Insect farming often involves significant suffering

Living conditions

Fischer quotes from p. 124 of Meyers (2013) in questioning how much farmed insects suffer:

Even if insects were capable of pain, the conditions that they would be raised in are conditions that would not cause them to suffer. Unlike cattle, pigs, or chickens -- and unlike even crabs, lobsters, or shrimp -- most insects actually prefer to live in crowded, hot, and filthy conditions.

I agree that living conditions for farmed insects are probably decent in many cases. However, farmed insects may still have significant pre-slaughter mortality. There may sometimes be problems with disease and cannibalism.

For example, this page reports:

Cricket paralysis virus can wipe out an entire farm’s supply in a matter of weeks, as it did at Top Hat Cricket Farms in Portage, Michigan, stopping all production and profit and forcing the company to lay off all of its employees. A cricket farm in Florida had to file bankruptcy as a result of the virus.

van Huis et al. (2013), p. 104:

The insect-rearing company Kreca used to sell more than 10 000 boxes of crickets (Acheta domesticus) each week. In 2000, 50 percent of the crickets reared by the company died within 8–12 hours, a population crash never previously experienced.

This article says:

The insects at Big Cricket Farms are housed in large troughs, each holding about 3,000 crickets, within a 5,000-square-foot warehouse. If a trough gets too full, the crickets get grumpy. "They’ll find a way to escape," says Bachhuber. "They’ll bite each other, they’ll eat each other. It’s not super common for them to eat each other, but it does happen a little bit."

This article says:

It turns out, captive-bred crickets aren't quite as hardy as their cousins who coat every inch of the American South five months out of the year. "We had this problem for a while ... our water was not drinking-water quality; it had too much chlorine. ... The water levels would spike and a couple hundred thousand or a million bugs would die. We'd installed water filters, but some of the stuff is really pernicious." Acceptable numbers of dead livestock in the cricket-farming world rival the casualty counts from battles on the Eastern Front. Thankfully, they're just friggin' crickets. [...]

Crickets have enough personality that Kevin can tell which ones are assholes: "Say you have 40k crickets in a bin. Eight to 12 of them might decide to be bullies, and they will block the other crickets from getting to the food tray. I don't know, man -- I just watch it sometimes and I'm like, 'What are you doing?'"

[...] people care more about the feelings of cows than of crickets. But all this cricket watching has made Kevin way more sympathetic than he ever was before[...].

This is a far cry from the absurd claim by Meyers (2013) "that there is no reason to worry about the suffering problem for insect farming" (p. 124).


Meyers (2013) says of farmed insects (p. 124): "As long as they could be slaughtered humanely, we would have an inexpensive and nutritious, karma free source of meat". Fischer likewise mentions (p. 2): "you might insist that even if insects are sentient, we wouldn’t harm them by killing them -- perhaps because they have no future-directed desires, and it would be easy to kill them painlessly." But is it easy to kill insects humanely, and is this done in practice?

Several insect farms kill their animals by freezing. For example:

At Big Cricket Farms, the “harvest,” as Bachhuber delicately calls it, happens on site. Farm workers carry the mature crickets’ troughs into a walk-in freezer, causing the bugs to go into suspended animation, an evolved response to cold temperatures. Once the bugs are unconscious, the temperature is lowered by another couple of degrees, which kills them.

Of the entomophagy advocates whom I've seen comment on slaughter, all claim that freezing amounts to "going to sleep" and is completely humane. But entomologists who don't have an agenda to push are less certain. The "BIAZA Recommendations for Ethical Euthanasia of Invertebrates (Version1.1-Apr2013)"i says of freezing (p. 3) that "research has suggested that this is probably one of the least ethical options (Bower et al, 1999; Pizzi et al, 2002; Pizzi 2012)." Freezing (p. 9) "Does not provide muscle relaxation or analgesia effects. Considered to be inhumane without prior anaesthesia." That said, there are many kinds of invertebrates, and maybe freezing is more humane for some than for others. Also, euthanasia of lab insects allows for more options than euthanasia of food insects, because those using lab euthanasia needn't worry about adding harmful chemicals to the insects. So the statement about "probably one of the least ethical options" should be read in the context of lab euthanasia, and I would guess that freezing is one of the least bad ways to kill farmed insects. The BIAZA paper says (p. 8):

Ideally, surplus invertebrates or invertebrates requiring euthanasia should be taken to a veterinary practice. However, this may not be practical or possible. Most members of the public do not have access to any chemical methods therefore physical crushing or freezing is still being advised by some institutions. Freezing is increasingly regarded to be inhumane without prior anaesthesia (Pizzi, 2012). However, if there is no alternative and euthanasia of an invertebrate MUST be carried out at home it is best to refrigerate first for a minimum of four hours then place the individual in the coldest freezer possible for at least two hours (Bushell, personal communication).

And unfortunately, some insects are killed in much worse ways than freezing. For instance, the Kreca company heats human-edible insects to death (p. 21): "When insects are reared for human consumption, more hygienic measures are taken: after sifting they sterilize them in hot water and then they place them in the fridge or freeze-dry them." This video shows that Enterra Feed Corporation's black soldier fly larvae are heated alive: "we cook them to about 108°C and then we bring it back down to room temperature." (Theoretically, these companies could freeze the bugs to death before heating them, which would be more humane. Perhaps animal-advocacy action alerts of the future will demand "freeze first" policies from companies like these.)

People cooking bugs at home aren't necessarily more humane. This how-to page on cricket farming instructs readers to "Harvest your crop, then boil it in rolling water and sauté with some salt, basil, and olive oil". No mention is made of freezing the insects first.

This article gives an account of cooking with an entomologist at her home:

She handed me a container full of bran and beetle larvae -- skinny, crusty, yellowish -- commonly known as mealworms. I shook the mixture through a sieve; as I rinsed off the last of the bran, the worms clung to the side like sailors on a capsized ship. Dunkel dumped them in a buttery frying pan, where they hissed and squirmed before going suddenly still.

Heat is one of the stimuli that has been most studied as a cause of pain in invertebrates; to see this, search "heat" and "temperature" on Wikipedia's "Pain in invertebrates" article. Some examples from the order Orthoptera:

  • Forman (1984) trained locusts "to maintain a particular range of joint angle in order to turn off an aversive stimulus. The aversive stimulus was provided by a focussed illuminator which heated the head of the locust" (p. 130). Forman (1984) also mentions by way of background (p. 129): "Locusts exhibit characteristic thermoregulatory postures and behaviors (Uvarov, 1977; Fraenkel and Gunn, 1961)."
  • Zabala and Gómez (1991) found (p. 887) that crickets given morphine took longer to escape from a heated box (an effect that could be blocked by naloxone, a morphine antagonist). Zabala and Gómez (1991) conclude (p. 890): "The cricket [escape reaction time] ERT at different temperatures had characteristics that are similar to that of the escape response shown by mice in the D'Amour hot plate (4)."

Unfortunately, farmed crickets who are heated to death have no escape.

Choosing less harmful plant foods

If, like Fischer, you're concerned about killing animals via crop cultivation, you can probably have more effect on this issue by choosing which types of plant foods you eat than by causing additional harm by eating insects.

For example, it's plausible to me that harvesting cereal grasses (corn, rice, wheat, oats, etc.) kills more vertebrates per hectare than harvesting fruits, vegetables, and legumes because cereal fields tend to be thick, dense, and good hiding places for small mammals. This page says: "Keeping grass neatly trimmed ranks as the easiest method of preventing field mouse infestations. Regularly mowing the lawn eliminates the tall grass which field mice often feed on and seek shelter within." Jacob (2003) (mentioned earlier in this piece) also explains (p. 322): "Reduction in vegetation height has been used as a management tool to decrease the abundance of rodents (Singleton and Brown, 1999)." That said, Jacob (2003) found similar percent reductions in M. arvalis population densities two months after harvesting for both wheat (a cereal grass) and beans (a legume) (Fig. 1(a), p. 323). So my suggestion about legumes causing less collateral damage to small mammals than cereal grasses may not be correct.

Another possibility could be to buy foods produced using no-till farming. Jacob (2003) found that "Common voles disappeared after ploughing" (p. 324), i.e., the common-vole population was basically eliminated. In contrast, harvesting beans only reduced common-vole populations by about ~1/2 (Fig. 1(a), p. 323). Of course, no-till farming has other impacts on wild-animal suffering as well that would need to be part of a full analysis. (Sadly, I fear that no-till farming may increase the world's total wild-invertebrate population.)

Meanwhile, if one is concerned about insect deaths, you could research crops that require less pest control, perhaps because the plants are more naturally pest-resistant, because the farm uses preventative cultural-control methods like crop rotation and sanitation, etc.


This section discusses how mushrooms, like insects and other farmed animals, are capable of growing on organic wastes. But how do mushrooms score relative to Fischer's concerns about (1) vertebrate deaths due to mechanical crop-cultivation equipment and (2) invertebrate deaths due to insecticides?

Concern #1 seemingly doesn't apply to indoor mushroom farming, at least when mushrooms are grown on organic wastes rather than newly harvested biomass.

Concern #2 does apply somewhat. This page says:

The sciarid fly or phorid fly may lay eggs in the growth medium, which hatch into maggots and damage developing mushrooms during all growth stages. [...] Pesticides [...] are available to use against these infestations.[16][23] Biological controls for insect sciarid and phorid flies have also been proposed.[24]

The Environmental Working Group ranks mushrooms as 35th out of 50 on a list from most to fewest pesticide residues among common foods -- i.e., mushrooms have fewer pesticide residues than 34 other foods on the list.

Unfortunately, the early stages of mushroom compost preparation may create lots of invertebrates. Beyer et al. (1999):

The organic matter on the composting wharf has the potential for supporting very large populations of house flies (Musca domestica) and stable flies (Stomxys calcitrans). Together, growers call these nuisance flies. Nuisance flies around composting areas are mostly managed with cultural controls to make the habitat unsuitable for breeding, such as eliminating sites with standing water, and by releasing Pteromalid parasitoids against the pupal stage of nuisance flies, or by encouraging barn swallows to roost close to the composting site since barn swallows feed on nuisance flies.

That said, if the decomposing material would be composted in a way that supports bug populations regardless of mushroom farming (would it?j), then mushroom farming would not cause net harm relative to the counterfactual along this dimension.

Borrell (2009) reports that pasteurization of mushroom growing substrate can be hot and partly anaerobic, which suggests to me that few bugs are created by this form of waste decomposition:

The fragrant process that brings fungi to the dinner table begins when Needham's crew salvages straw bedding from horse tracks and stables, mixes in fresh hay and manure, and spreads it out in long rows for seven to 14 days to naturally pasteurize it. As bacteria living in the manure break down its nutrients, their metabolism raises the temperature to more than 120 degrees F (50 degrees C), leaving only the hardiest organisms alive. [...] after a few days, oxygen levels in the pile decline, encouraging growth of anaerobic bacteria[...].

Beyer et al. (1999) mention "steam pasteurization that eliminates most insects and nematodes from the substrate." Presumably this heating process is very painful to those bugs already in the substrate, which is pretty bad. That said, these bugs would have died naturally somehow or other, whereas the bugs killed in insect farming needn't have been born in the first place and hence needn't have died at all.

What about grass-fed dairy?

Davis (2003) mentions no-till farming as a way to reduce vertebrate deaths (p. 390):

"The predominant feeling among wildlife ecologists is that no-till agriculture will have broadly positive effects on mammalian wildlife" populations (Wooley et al., 1984). Pasture-forage production, with herbivores harvesting the forage, would be the ultimate in "no-till" agriculture.

Matheny (2003) argues (p. 507) that grazing may cause more animal deaths than crop cultivation based on a mathematical error in Davis (2003):

Davis suggests the number of wild animals killed per hectare in crop production (15) is twice that killed in ruminant-pasture (7.5). If this is true, then as long as crop production uses less than half as many hectares as ruminant-pasture to deliver the same amount of food, a vegetarian will kill fewer animals than an omnivore. [...] In one year, 1,000 kilograms of protein can be produced on as few as 1.0 hectares planted with soy and corn, 2.6 hectares used as pasture for grass-fed dairy cows, or 10 hectares used as pasture for grass-fed beef cattle (Vandehaar, 1998; UNFAO, 1996).

Southan (2011) contends in reply to Matheny (2003) that grazing may actually kill fewer vertebrates after all, because Davis (2003) was plausibly too generous toward crop cultivation in his assumptions:

Davis scuttled himself by being over-confident and allowing for so many wild animal deaths in pasture-raised animals. Because Davis was (by his own admission) making the numbers up, he didn’t want to skew them too blatantly in his own favor. Especially since he didn’t feel that he needed to, because he didn’t predict Matheny’s counter-argument about differences in yield.

That said, Southan (2011) doesn't seem to take a strong stance on this particular issue, given uncertainty and variation in farming practices. Rather, Southan (2011) mainly highlights "how simplistic it is to generalize that pasture-raising animals kills exactly half the number of wild animals per hectare that raising plants does."

I'm don't know where I stand on this debate.

While I'm not an expert, I would guess that two of the main sources of vertebrate mortality from cattle grazing are planting forage crops (if this is done) and making hay.

I'm uncertain what fraction of pastures are planted. Moreover: "Hayfields do not have to be reseeded each year in the way that grain crops are, but regular fertilizing is usually desirable, and overseeding a field every few years helps increase yield."

Whereas harvesting crops typically involves one pass through a field, making hay requires at least two: one to cut the grass ("cutting") and another to harvest it a few days later after it has dried ("baling"). Hay farmers may also turn over the hay in between: "During the drying period, which can take several days, the process is usually sped up by turning the cut hay over with a hay rake or spreading it out with a tedder."

My family harvested hay every summer when I was a child. We did three passes through the field: mowing, raking to turn the hay over, and then baling. There were a few times during baling when my family found a snake that got trapped in a hay bale (i.e., presumably the snake was killed by baling, unless it had already died and was merely lying on the ground to be scooped up in the baler). That said, we baled about ~8 hectares per year, so I would guess the number of snakes visibly killed by baling was probably less than ~0.5 per hectare per year. Of course, there might have been snakes that we didn't see, perhaps crushed by tractor tires aboveground or in their burrows. Basically no birds were killed, but this was partly because my family harvested hay later in the season than normal in order to reduce bird mortality.

This page explains:

There is also a risk that hay bales may [...] contain decaying carcasses of small creatures that were accidentally killed by baling equipment and swept up into the bale, which can produce toxins such as botulism. Both can be deadly to non-ruminant herbivores, such as horses, and when this occurs, the entire contaminated bale generally is thrown out [...].

My impression is that most cattle farmers use hay or other harvested feed over winter, although some farmers even manage to avoid this:

  • Southan (2011) quotes one South African farmer: "I didn’t make hay, having no equipment or money for the operation. Instead, I practiced rotational grazing, shutting off some paddocks in early summer to produce winter forage to be rotationally grazed in due time."
  • This podcast episode discusses grazing cattle year-round without hay, even in cold regions like Canada. Cattle can eat through snow, so with the right management techniques, they can overwinter without using hay, except perhaps in cases of emergency. That said, some of the techniques discussed in the episode involve planting certain forages, so even this approach might not be completely free of mechanical passes through fields.k The episode mentions the book Kick the Hay Habit : A Practical Guide to Year-Around Grazing.

It's also worth remembering that hay is only used for part of the year -- let's say half the year for simplicity. So if, as Matheny (2003) says, dairy cows need 2.6 hectares to produce the same amount of protein as 1 hectare of soybeans, they may only need ~2.6/2 = 1.3 hectares of hay. And if hay production involves fewer vertebrate deaths than soybean farming (such as due to less need for plowing, pesticides, etc.), then Davis (2003) might be vindicated in the case of grass-fed dairy (if not beef).

And in the case of those rare farmers who don't use hay at all, perhaps even grass-fed beef would cause fewer vertebrate deaths than soybeans. While we shouldn't base generalized conclusions about grass-fed cattle protein on rare production methods, neither should we evaluate human-edible insects in industrialized countries on those rare cases where farmed human-edible insects are raised on organic wastes. A charitable interpretation of Fischer's article is that he's encouraging more investment into waste-based entomophagy, rather than claiming that it's widespread. But in the same way, one could encourage more investment into cattle grazing methods that kill fewer vertebrates.

How about invertebrates? I assume that most pastures don't apply a lot of insecticides, unlike crop fields, although insecticides are sometimes used on pastures. Bugs may also be crushed by cattle hooves and farm equipment. That said, cattle grazing also often reduces invertebrate populations, preventing lots of bugs from being forced to be born. For this reason alone, I suspect that cattle grazing may be net beneficial for wild-animal welfare, both for negative utilitarians and for those classical utilitarians who think invertebrates endure more suffering than happiness during their brief lives and painful deaths.

See also

A Reddit discussion of Fischer's paper makes some of the same points as I did above.

Fischer's reply

On 8 Feb. 2017, Bob Fischer wrote a reply to this piece, which I'm sharing with his permission:

First off, many thanks to Brian for engaging in such detail with my work. I'll take his points in turn.

First, Brian is correct that it matters what insects are fed. I was thinking about a case where a woman in Denver is raising crickets using spent grain that, she claims, she gets for free from a local brewery. If that spent grain really would be wasted otherwise (and I admit that I have no way of verifying whether that's the case), then it might still be possible to run the argument that I had in mind, assuming the other objections to be addressed. But if not, then not.

Second, Brian is correct that I did not mean to suggest that there are 1000 non-insect deaths per acre in plant agriculture. That was supposed to be a gift on behalf of the strict vegan to her critic, and one that turns out not matter (or so I argued). (I've got another paper where I look at all the sources of animal death due to plant agriculture, and try to provide the best estimate I can of the total. When it's ready, I'll be happy to share my number, but I have every reason to think but it will be much lower than 1000.) In any case, there is no substantive disagreement here; only the perfectly reasonable point that I should've been clearer about how I pitched that section.

Third, Brian is right that a whole lot depends on the specific values that we assign to the likelihood of consciousness in various invertebrate species. However, I'm a bit surprised that he was willing to run with my low estimate, and then discount certain species – such as springtails — even further. The rhetorical effect of that section was supposed to be something like this: "Of course, it's implausible to assign such low numbers, but the argument works even when we do; so, we should expect the argument to work when we assign much more reasonable numbers." That said, I acknowledge that there is plenty of room for reasonable disagreement about what numbers we should assign, so I should retract the claim – or implication — that sub 1% numbers aren't implausible. Now, however, the dispute is an empirical one about the likelihood of particular mental state in particular species, and we're all in a hard position there. To my mind, that's a reason to be worried about the value of expected utility arguments, except as rough heuristics, but I know that Brian and I disagree about that.

As for Brian's fourth point, the disagreement here is complex, and I'm not sure that I have space to do it justice. But for what it's worth, I don't think that things are any simpler re: insecticide use given a deontological framework. Also, I'm wary of interpreting utilitarianism in a way where it says that it isn't bad to use insecticides. It seems to be best to say that it is bad, though perhaps morally obligatory because the badness is outweighed. We shouldn't make utilitarianism anymore heartless than it already is. :-)

Brian's fifth point is basically irrelevant if we use one of the precautionary principle versions of the argument. But if we don't, it's still worth noting that insect deaths are likely to be much faster than mammalian or avian deaths. So, although insect deaths might not be humane, that doesn't show that the calculation cuts against me. We would need to do much more detailed calculations to figure out how things go. But all that aside, Brian is absolutely right that such calculations are necessary, and I really regret that I didn't think to include more material on that issue.

My replies to points #3 to #5:

  • #3: I think insects are (much) more than 1% likely to be sentient, but the moral weight I give to a single insect even if it is sentient is many times less than the moral weight I give to a single mammal (Tomasik 2013). In any case, as long as we give individual mites, springtails, and thrips at most half the moral weight or probability of sentience of individual mealworms, then my "Problem 3" objection remains—whether mealworms have a 1% or 100% probability of sentience.
  • #4: As far as "It seems to be best to say that it is bad, though perhaps morally obligatory because the badness is outweighed", this is what I meant, but my wording wasn't clear. I changed "bad" to "net bad" in my statement of "Problem 4" in an effort to address this.
  • #5: As far as "insect deaths are likely to be much faster than mammalian or avian deaths", this might or might not be true in terms of how long it takes to die, since my impression is that insects may take hours or more(??) to die when frozen. But the spirit of this point is plausibly often true: frozen insects probably have better deaths than many mammals and birds killed by crop-farming equipment. That said, insects who are heated alive rather than frozen may have worse deaths than many mammals and birds killed on crop fields (relative to their respective levels of sentience). Of course, these are mainly my speculations based on anthropomorphic imagination of the situations, combined with a little bit of empirical knowledge about how insects react to cold vs. heat.


  1. Presumably the number of mammal deaths due to farming is less than the total number of small mammals on crop fields, since usually not all mammals present are killed. On the other hand, fields are typically subject to mechanical disturbance several times per year, such as from plowing and harvesting, so theoretically the total deaths per year could be higher than the instantaneous population density on a crop field, if, for example, most of the animals present were killed by plowing in the spring, then populations rebounded, and then most of the animals present were again killed by harvesting in the fall. This seems to me unlikely, though.  (back)
  2. This wording by Fraser and MacRae (2011) is true but might be misunderstood. It sounds from the Fraser and MacRae (2011) quote like wheat harvesting didn't have much impact. In fact, Jacob (2003) found (p. 324): "Mowing and harvesting wheat were followed by an initial increase in density but density decreased 2 months later." If you look at Jacob (2003)'s Fig. 1(a) (p. 323), you'll see that the percent decrease in population density of M. arvalis after 2 months is approximately the same for harvesting wheat as for harvesting beans -- roughly 50% in both cases.

    Jacob (2003) notes (p. 324): "spilled grain after harvesting wheat may have provided food to sustain a moderate density of common voles (112 individuals/ha)."

    I haven't read Nass et al. (1971) or Tew and Macdonald (1993) and so can't comment on whether Fraser and MacRae (2011) omitted some fine print from those studies.  (back)

  3. There's a typo in Fraser and MacRae (2011). The page number for this statement by Jacob (2003) is actually 324, not "p 24". The full quote from Jacob (2003) is (p. 324): "Radio-tracked common voles exposed to farming practices show no emigration (Jacob and Hempel, 2002), and decreases in density after farming practices may be mainly due to decreased survival."  (back)
  4. [This footnote is uninteresting to most readers.]

    The "Summary" of Pearse (1946) includes a puzzling statement when compared against Table 1. I'm uncertain whether it's an error by Pearse or whether I misunderstand his results.

    Point #2 in Pearse's summary begins (p. 144): "From the 576 samples collected and placed on Berlese funnels 353,026 animals were recovered". So far so good. This matches with Table 1.

    Pearse continues by saying (p. 144) this was "an average of 1719 per square foot each week (Table 4)". Pearse's Table 4 shows an average of 6,874 animals collected per week per station. Since Pearse said "Twelve collections were made each week" (p. 130), and there were four collecting stations, I would have thought there were 12/4 = 3 samples collected at each station in a week. Each sample was 1 square foot in size (p. 130). But 6,874 animals per station per week divided by 3 square feet per station gives 2,291 animals per square foot each week, not 1719. It turns out we can get 1719 animals per square feet if we divide 6,874 by 4 rather than 3. That's odd; maybe there were 4 samples per station each week?

    Finally, Pearse says (p. 144) that 1719 per square foot each week implies "2,590,082 per acre (Table 1)." But that's not right. 1 acre is 43,560 square feet, so the number per acre should be 1719 * 43,560 = 74,879,640, which is neither 2,590,082 nor the 124,366,656 per acre total in Table 1. Pearse does say (p. 131) that "one square foot of litter was collected and the metal rings that were driven into the soil to collect the 0-2 in. and the 2-5 in. samples collected only 0.66 of a square foot." So maybe there's wiggle room to slightly increase the numbers to account for some of the sampling being done only on 2/3 of a square foot, but this fudge factor doesn't seem to help us here. A fudge factor that does solve our problem is 48: 2,590,082 times 48 approximately gives the 124,366,656 in the bottom of Table 1. This looks like Pearse is reporting 2,590,082 animals per week times 4 weeks per month times 12 months per year. Does this mean that 124,366,656 is not an instantaneous density of animals but rather is 48 times the average instantaneous density? If so, then the population of invertebrates here is much less than I thought.

    Anyway, I remain confused by these numbers, and I hope I'm not misinterpreting Pearse's Table 1 as implying higher densities per acre than were actually found. If I am misinterpreting Table 1, so is Fischer (2016), not to mention Sabrosky (1952), who also cites Pearse's paper and says (p. 5): "A. S. Pearse calculated that there were approximately 124 million animals per acre."  (back)

  5. The actual number is probably somewhat higher because not all invertebrates in soil are collected using a Berlese funnel. Berlese sampling "gives a biased sample of soil fauna, because it is based on specific avoidance behavior triggered by dryness and thus best captures animals that are mobile and do not desiccate easily. Immobile larvae, endophagous nymphs and soft-bodied invertebrates such as nematodes are not extracted by a Berlese funnel."

    Pearse (1946) himself notes this point (p. 131):

    Some of the numbers in Table 1 doubtless do not indicate the comparative number of animals that occurred in the sample collected. For examples, only thirteen nematode worms were secured and they probably were present by thousands but did not go from the soil, through the screen, and into the bottle below the Berlese funnels. Some animals (some earthworms, insects, myriapods, snails, etc.) were too large to pass through the apertures (1/16 in.) in the screen. Others (some spiders) ran over the edge of the screen and escaped. However the figures in the table as far down as earthworms probably give a fairly accurate estimate of the comparative numbers of animals present.


  6. Mealworms are plausibly somewhat narrower than mites/springtails/thrips, so a 10:1 ratio of lengths may not scale up to a full 103:1 ratio of volumes. However, the ratio of lengths is plausibly more than 10 times, especially for the smallest and most numerous of mites and springtails, in which case a volume ratio on the order of ~103 still seems plausible. In any case, the exact ratio isn't essential for my point here.  (back)
  7. This claim is based on my experience examining invertebrates with a microscope camera in a small garden and in surrounding grassy areas. The most numerous arthropods I find are usually mites and springtails, which live mostly in the soil or within litter.

    This textbook explains (p. 436) regarding herbivory: "On average, about 13% of terrestrial [net primary productivity] NPP is consumed (range 0.1%-75%)[...] (Cebrian and Lartigue 2004)." The rest ends up as detrital biomass.  (back)

  8. It's unclear to me whether and how much the authors of this paper were pro-industry biased. The paper says: "The funders had no role in data collection and analysis, decision to publish, or preparation of the manuscript. [Agriculture and Agri-Food Canada] AAFC suggested the insecticide list for testing, and this is the only role any funders played in study design." However, that list might have been important to the outcomes. More importantly, the researchers themselves may have had unconscious or conscious biases, in part because "In support of other projects unrelated to this study, the authors' research group has received competitive research grants from grower organizations and government bodies and contracts and/or in-kind contributions from manufacturers of both organic and synthetic pesticides."

    The "Acknowledgments" section says: "We would also like to thank Syngenta, Bayer, FMC, UAP, and Laverlam for providing insecticides for our experiments". These companies provided 5 out of 6 of the insecticides tested (p. 2, Table 1), both synthetic and organic, so it's not obvious there's a bias to favor one side or the other here.  (back)

  9. This document is sadly not available online. I got a copy by contacting BIAZA using the email address found on its website.  (back)
  10. I'm not very informed on the details, and probably the answer to this question varies a lot from one mushroom operation to another and one counterfactual waste-disposal method to another. For example, suppose that organic waste, if not used for mushroom farming, would counterfactually be composted at high temperatures in massive industrial-composting operations; in this case, I expect it wouldn't contain lots of invertebrate animals per kilogram. Likewise if the compost were landfilled. On the other hand, if compost, instead of being used for mushroom farming, were to decompose in smaller, more air-exposed piles, it might give rise to large numbers of invertebrates -- perhaps more invertebrates per kilogram than compost prepared for use as mushroom substrate gives rise to?

    Beyer (n.d.) says: "The industry of composting urban waste for disposal differs from our mushroom composting goals. Urban composters promote slow temperature composting (110-130 °F) in small piles that support very active thermophilic fungi, bacteria, and actinomycetes." In contrast, composting for mushroom farms is a "high temperature (160-180 °F) or thermogenic composting process". My guess is that either type of composting has temperatures that don't allow for much invertebrate growth, except maybe around the cooler edges? Here are examples of the upper bounds of insect temperature tolerance:

    • Hornets are killed above 115°F.
    • This page reports: "Lethal temperatures for bed bugs range from 117 degrees Fahrenheit to 122 F."

    I would guess that even a 110°F compost pile is too hot for most insects to grow in.

    It's also relevant to consider what happens to "post mushroom substrate", the mushroom-growing substrate left over after mushrooms are grown on it. Beyer (n.d.) mentions several ways post-mushroom substrate can be disposed of. One is just to let it sit in heaps: "Piles of spent substrate in all parts of the landscape are no longer ignored by neighbors or environmental regulatory agencies. The piles may become anaerobic and give off offensive odors." Presumably anaerobic piles contain few invertebrates except on the aerobic exterior?  (back)

  11. As one example, Melancon (2013) discusses planting brassicas to "extend the grazing period well into late winter." It's recommended to plant "with a seed drill for best results." How much collteral harm does a seed drill cause?

    Based on videos like this and this, my guess is that aboveground vertebrates (if any) would notice the oncoming machinery and flee in time. If so, then maybe the main source of mortality would be crushing/suffocating animals in burrows??  (back)