How much should we be concerned about suffering in the wild?

Venkateshan K
Jul 13 · 14 min read
Source: Avel Chuklanov on Unsplash

Animals in the wild suffer in a variety of ways — attacks by predators, exposure to diseases, physical injury, indefinite starvation and thirst among others.

Despite increasing attention to the enormity of suffering inflicted by humans in factory farms , animal testing laboratories, etc., there remains the more ambiguous issue that most animal rights advocates would have considered at one point or the other — what should be done about wild animal suffering?

This is difficult to answer and there is a great deal of uncertainty about what, if anything, can be done to minimize the harm to wild animals. Whatever the position one may choose to take on this issue, there is one aspect that is no doubt paramount: the importance of establishing the relative magnitude of wild animal suffering.

Eliminating all suffering would be the ideal, but effective animal advocacy involves identifying areas where there is the greatest scope for improvement in welfare. From that perspective, it makes a lot of difference whether there is only a tiny amount of suffering in the wild or it accounts for the predominant misery of all sentient species.

A natural benchmark for comparison is the totality of pain inflicted on domesticated animals from factory farms to animal experimentation. However, we encounter two problems with this approach: first, how do we compare the suffering of two different species when science isn’t developed enough to provide straightforward metrics, and second, how many of each species are there in the wild?

My focus here is to address both of these issues. Part of the reason why we need more focus on these questions is that currently we rely on estimates that could be wildly off and are using assumptions that may be very flawed.

A scenario where wild animal suffering is 0.001 that of domesticated animals is very different — from a consequential perspective — from one that is 1000 times.

Understanding capacity to experience pain

The modes of physiological, chemical and behavioral response to stimuli varies a lot across species. While only the most solipsistic would argue that chimpanzees are merely exhibiting a reflexive programmed response to negative stimuli instead of experiencing pain, the situation become more uncertain as we go down the species hierarchy. To an extent, the presence of a central nervous system serves as a good proxy for ability to experience pain. But almost all organisms that show bilateral symmetry (bilateria) have a nervous system with some rudimentary form of a brain. Bilateria includes not just vertebrates but also crustaceans, insects and worms.

One way around this difficulty has been to use the number of neurons as a proxy for the degree of sentience of the animals. Not just the raw number of neurons (because they range over orders of magnitude) that would perhaps place too small a weight to species like insects. For example in this discussion 0.5*log[#Neurons] is used (among others) as a weighting factor when considering the relative importance of a given species.

Even if one accepts this as a crude representation of the capacity to suffer, there is something wrong in defining the utility function by merely multiplying this factor by the population size of that species. We can all agree for example that 10 units of suffering by one individual is much worse than 10 individuals suffering one unit of pain. Going further, a universe with 1 unit of suffering by 10 individual is much worse than an alternate one where a million individuals each suffering 10^(-5) units even though the total amount of pain is the same in both cases (assuming simple additivity of pain). In fact, one can set a threshold such that when an individual’s pain is less than that threshold, it is so negligible that we can ignore it.

If we accept the above argument, then how much importance should we accord to the suffering of 10^(18) (billion trillion!) insects each of whose capacity to experience pain may only be a small fraction of mammal’s? If, for example, an insect were to suffer 10^(-3) of a mammal, then a naive estimation of the total suffering would be 10^(-3)*10^(18) = 10^(15) in mammalian terms! But maybe that 10^(-3) factor of suffering that we have assigned to the insect is perhaps too little and that, as a species, it cannot be compared to mammals or birds.

Obviously one will rightly object that all this is very speculative. The point here however is not the so much accuracy of the estimate, but the fact that we will arrive at radically different conclusions depending on distinct methodological approaches — (1) we simply sum a certain negligible suffering by a large enough population, then the result can be arbitrarily large that it can outweigh all mammals and birds and (2) recognize that negligible suffering does not count and we place no moral significance regardless of the population size.

To be sure, the two cases represent only extremes, and one can consider other intermediate ways to define the total negative utility. It may be somewhat tempting to consider using a concave function on the population size itself — say square root of the population, for example- but that is problematic (the importance of an individual depends on the size of the population of the species it belongs to!) . We could use instead a second weighting factor that explains the relative importance of that amount of suffering. For instance, in the above example where insect capacity to suffer is assumed to be 10^(-3) that of a mammal, we can introduce a second factor, square of the first, 10^(-6) , that quantifies the significance of that extent of suffering to the individual. This effectively gives a weight of 10^(-9) for each insect. Even so, if we now multiply by the total insect population of 10^(18) we get an amount equivalent to a billion mammals. If the insect populations were larger than that estimate, we will soon be attributing most suffering to insects than all mammals and birds.

Given the highly divergent scenarios that different choices lead it, my personal inclination would be to focus first on those animals whose suffering is far less ambiguous. And this naturally brings us to the top of the hierarchy and the second issue where uncertainty reigns — the estimation of wild population of birds and mammals.

Estimation wild population of birds and mammals

There seems to be a general consensus that the number of wild mammals and birds is far greater than their domesticated counterparts.

I will argue against this and more specifically, demonstrate the numbers used in most analysis is very likely to be systematically overestimated.

Biomass Comparison

To begin with, let’s state a fact that is beyond any doubt — if it is the total biomass that we are interested in, then the contribution of domesticated species is most certainly higher. In their work on estimating the global biomass distribution, Bar-On et al have determined that the domesticated mammals have a biomass of 0.1 Gigaton Carbon (roughly 10–15% of body mass is in the form of carbon compounds) as compared to a mere 0.007 GtC for the wild (see figure below). Likewise the numbers for domesticated and wild birds is 0.005 and 0.002 Gt C respectively (see Supplementary material).

Source: Bar-On et al,PNAS Jun 2018

The question then is, given the disproportionately larger number of smaller animals in the wild, are their total numbers overwhelmingly large?

Current Estimates

In his article on wild animal populations, Brian Tomasik refers to the analysis found in Gaston et al (2003) where they estimate the typical number of birds globally to be 86 billion (with lower and higher estimates 40 and 135 billion, respectively). The authors arrive at these numbers by using the estimates for density of birds in different biomes/land-use classes from multiple sources. Despite identifying and using only the most reliable of studies, they acknowledge, and Brian Tomasik notes, that these estimates could be biased because the studies are usually carried out in high density areas. It is a well-known problem that population density estimates are negatively correlated with study area. The primary reason for this appears to be that the location chosen for smaller study areas tends to have a higher abundance and this selection bias is significantly weaker when the study areas are larger (something alluded to earlier as well). In fact, this issue is so important that it explains most of the variation in the estimated population densities of species. If this trend were to be extrapolated, it would suggest that the overall population density of a given species is probably less than what most estimates obtained from study areas.

Also, note that the areas of the land-classes (temperate forest, boreal forest, tropical woodland, etc) over which we estimate species population (using density estimate in a given study area) extend to millions of square kilometers. For reference, the areas of India, California and France are about 3, 0.42 and 0.64 million sq. kilometers respectively. Given the variations in density across these regions, one can imagine how a biased estimate much can affect the final number.

In the same article, Brian Tomasik reports that the total wild mammal population is about 100 billion, based on different types of estimates from various sources. Again, none of the estimation methods described in the sources is very reliable, even up to an order of magnitude.

More doubts on these numbers are raised indirectly from further examination of the Bar-On et al (2018) paper on estimating the overall global biomass distribution. The authors use the data in Novosolov et al (2017) to estimate the biomass of birds and mammals. One of their methods involved fitting a linear model to this data to explain the population numbers of a given species (in the particular study) in terms of the reported study area and the body mass of the species (see Supplementary material and their Jupyter notebook on this analysis).


I re-analyzed the estimates for mammals obtained from the linear model using the original data in Novosolov et al(2017)(paper is under a subscription firewall but I reached out to Marie Novosolov, the lead author, who kindly shared all the raw data ). There is some ambiguity in the Bar-On et al’s work because they don’t specify the units in which the variables are meant to be entered in the linear model. Through trial and error, I settled on those units that led to the best fit of the data. The resulting equation is given below (the coefficients are identical to what they have used):

Shown below is the scatter plot of the log of the actual population of mammals in the study area and the estimate from the model (the red diagonal line represents the curve corresponding to equality between the two) .

Scatter plot of the log of actual population estimated in study area (x-axis) and that obtained from the linear model (y-axis) for mammals in the Novosolov et al (2017) dataset. The red line is the 45 degree diagonal.

If we take that model fit to estimate the total population of a given species by replacing the study area by the geographic range size (as Bar-On et al. have done, without much justification, to estimate the biomass of mammals) then we can estimate the total population of the nearly exhaustive list of mammals in the PanTHERIA data-base (over 5000 extant or recently extinct mammals).

The total number of mammals in the wild based on that calculation is about 25.4 billion.

Again, we should certainly consider that estimate with a grain of salt, but it is likely to be more accurate than most of the other, even less well-justified methods described.

(This calculation applies only to terrestrial mammals but the number of marine mammals is much smaller. Bar-On et al estimate the total biomass of marine mammals is about 0.004 GtC, which is approximately 30 billion kg of wet mass. The smallest of marine mammals weigh on an average at least 10kg and even if we assume that they contribute the all the biomass, we get the upper bound of 3 billion.)


I repeated this analysis for birds for which Bar On et al have provided a similar model but the fit in that case turned out to be quite bad. And this was true despite trying different units for body mass and study area.

I then independently determined the best fit by running the regression for log of population on the log of study area and log of body mass and obtained the following:

(In order to focus more on the region corresponding to higher population and avoid underestimation , I fit it only on the data points where population was greater than some threshold. The R² on the full data-set was around 0.4 although despite the correction one can see below that it is somewhat underestimating the higher higher population numbers).

Scatter plot of the log of actual population estimated in study area (x-axis) and that obtained from the linear model (y-axis) for birds in the Novosolov et al (2017) dataset. The red line is the 45 degree diagonal.

When extrapolated by substituting study area by geographic range size (as in the case of mammals), I obtained the total population of the 830 species given in Novosolov (2017) database to be less than 150 million.

A further extrapolation to the nearly 9000 extant species of birds identified (assuming similar distribution of populations for those species), we get the total abundance of all birds in the wild is less than 2 billion.

Admittedly, this is quite likely to an underestimate but I don’t see any reason why it is less accurate than estimates of about a 100 -400 billion.

Population Density-Body Mass

As I have shown, if it is the biomass that if of interest, then there is no question that domesticated animals are the dominant contributors. Therefore the claims of larger numbers of animals in the wild rests on the accuracy of estimating animals that are relatively small.

One common approach to extrapolating population densities is allometric scaling (referred in mammalian population estimation). Allometric scaling is the hypothesis in biology and population ecology that different quantitative characteristics of a species (such as its metabolic rate, embryonic growth, life span etc. ) scales as an exponent of the body mass.

The most relevant instance of allometric scaling that is of importance to us is the negative scaling between population density and body. In other words, population size and the body mass of a species are believed to be inversely related, i.e., smaller the body mass, the greater the abundance of the species, and the relation is of the approximate form

To understand why it is important to test the validity of this hypothesis, consider a hypothetical selection of species whose body masses are 0.01kg, 0.1 kg, 1kg , 10kg and 100kg. Assuming that the exponential factor (gamma) in the above equation is unity then the relative population of the different species might look like something in the following table (not real data)

Observe that the abundance of the species whose body mass is 0.01kg is nearly a factor of ten more than all the other species put together!

Given the remarkable significance of the scaling relation, let’s examine how accurate it is. To do that, I once again use the data in Novosolov et al. and consider the scatter plot of the log of population density and the body mass of a given species for birds.

Plot of log of population density (y-axis) against log of body mass. The red line is the fit obtained by the model.

There is a clear negative slope relating the two quantities, but the slope of the line (gamma) in the equation is quite small -0.19.

A closer examination of the above plot reveals more. Observe that for small body masses (log values less than 1.5) the negative dependence is even weaker (if it exists at all). We’ll narrow our focus on that region and replot the dependence

Scatter plot of log ode population density and log of body mass of birds (where body mass is less than 100 grams)

There is almost no perceptible relation (a linear model confirms that (not shown)) between the two quantities whenever the body mass is less than 10² =100g! This fact has a considerable bearing on our calculation. Indeed, as we pointed out earlier, if allometric scaling were to hold across the entire spectrum of body mass sizes, then the greatest contribution to the total number of individuals would come from species with the smallest body masses. However it is precisely this small body mass range where there is little to no scaling dependence and hence, the estimated numbers when this correction is taken into account will be reduced by orders of magnitude!

A similar analysis with mammalian masses and population densities reveal the same property. In fact, the issues of allometric scaling has been reported in literature (see Silva et al (1994)).


I have shown here that claims that the total numbers of wild birds and mammals being 100–200 and 100–400 billion respectively is most likely an exaggeration with the true numbers closer to 20 billion or less. At least for these two vertebrate classes, presumably the most important in terms of our confidence in their ability to experience pain, the numbers in the wild are not far greater than their domestic counterparts.

As for the numerous species that science is unable to conclusively decide on the sentience, our estimates of their overall contribution to suffering is highly unreliable. More to the point, there is a virtually infinite variation in our estimation of total suffering when it comes to animals in the wild.

There is no doubt that wild animal suffering is an important issue and is often neglected in the animal advocacy community. Nonetheless, before we commit resources towards developing strategies to minimize it, we should recognize the great uncertainty in our estimates of the relevant quantities in this area.

The case for liberation of domesticated animals is very strong and can be summarized by the followed key arguments: (a) mammals and birds most certainly experience pain and constitute the 70 billion (land animals) are slaughtered globally every year for food ( b) humans can live perfectly healthy lives without depending on animal products (c) the lives of vast majority of farmed animals is horrible to say the least, with scant regard for their welfare from birth to death ( d) the treatment of animals is antithetical to other progressive values humans claim to uphold (e) there is a clear potential and scope to minimize if not completely eliminate animal agriculture.

Most of these points do not apply to wild animals, unfortunately. Given that we are still very, very far from ending human exploitation of animals, one way to strengthen the case for wild animals is by improving our methods to estimate the severity of the problem.


Post Darwinian Speciesist

Towards ending abuse and exploitation of sentient animals

Venkateshan K

Written by

Physics PhD.; seeking an informed perspective

Post Darwinian Speciesist

Towards ending abuse and exploitation of sentient animals

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade