Thirty years ago, Ted Garland (UC Riverside) published a seminal paper entitled “Why not to do two-species comparative studies: limitations for inferring adaptation” that was published in the journal Physiological Zoology (which then became Physiological and Biochemical Zoology, and is now Ecological and Evolutionary Physiology). This is an incisive and powerful paper, and it lays out the argument for why, whenever possible, comparative studies should incorporate multiple species and phylogenetic comparative analyses for inferring adaptation. This paper was published right at the advent of more broadly available phylogenies and increasing accessibility of phylogenetic comparative methods (such as independent contrasts). I recommend reading the paper- it is a clear and well-written paper with relevance for research today. However, I think the core message has been misremembered and misapplied in a way that can thwart the publication of solid research.
My interest in this matter comes from personal experience with publishing ecological and physiology research, often with only two species. Many a time the dreaded reviewer number two would trot out the first part of the title of Garland (1994) and tell us that we can’t do two species comparisons. But that paper is about how there are limitations for inferring adaptation with only two species, not that no study should ever be published with two species. There is no reason that you can’t include two species in a single paper to talk about variation within a group, or simply for convenience, as long as you careful about evolutionary inferences. Additionally, the philosophical issues that Garland (1994) was raising aren’t solved by simply adding (or subtracting) one species to your study- you need at least dozens of species for robust phylogenetic comparative analyses. In other words, there is no nadir of inference power that is centered at a sample size of two species. My close collaborator Michael Logan (at University Nevada Reno) and I were grousing about the kinds of reviews that we were getting, and so we decided to publish a perspective piece highlighting the contexts where two (or a few) species comparisons were appropriate (later published in Physiological and Biochemical Zoology). Given that this perspective piece was a sort of blog post, I thought it would be worth discussing on the blog.
To understand the context in which the original Garland (1994) paper was written, we need to go back to the type of comparative physiology work of the prior few decades. I think it is safe to say that many physiologists were not exceptionally well-versed in evolutionary biology, which at the time was still in initial development (this was in the modern synthesis era). There is no doubt that we learned a lot about physiology in this era, but some of these studies would not pass muster today. One of my favorite concepts (and literature) of comparative physiology of that era is the idea of symmorphosis. Symmorphosis is the idea that organismal systems should have quantitatively matched design for maximal efficiency, such that no component of the system is functioning at substantially greater capacity than the rest of the system. To put this in a real physiological context, symmorphosis would suggest that the digestive capacity of intestinal brush-border oligopeptidases that are the final step of protein digestion should be matched to the capacity of brush-border amino acid transporters that transport the resulting amino acids into the body. It is not really the point of this exercise, but symmorphosis as a context is probably true in general, but does not hold up in lots of situations (because of safety factors, because organisms are not optimized but the result of contingency and tradeoffs). Regardless, some of the foundational work underlying this field focused on respiratory physiology, and often compared such divergent species as dogs and goats. While we no doubt learned a lot from this work, inferring adaptation to differences between dogs and goats is fraught, because you only have two data points and they have been separated by many millions of years of evolution. In other words, it is hard to attribute any difference between the species to any single factor, because they differ in so many ways- diet (omnivore vs herbivore), locomotion (unguligrade rather than digitigrade), habitat type, and so on. All of this to say that the original Garland (1994) paper was a call to arms for rigor in comparative research, which was sorely needed at the time and I still think is an important message.
So what is the second reviewer getting wrong when they tell us that we cannot publish a two-species study? Well, I think they get several things wrong. First (and already addressed above), Garland (1994) is about inferring adaptation. If there are no inferences about adaptation, then bringing up Garland (1994) is irrelevant. Second, there is no criticism of a two-species comparison that does not apply to three-species studies, or four-species studies, or even one-species studies. But assembling data on the number of species necessary for phylogenetic comparative analyses might be difficult or impossible for complex physiological data, and it may take many years to accumulate enough research for comparative analyses. So what do you do if you want to think about evolution and adaptation, but cannot get data for enough species for comparative analyses? The final thing that the second reviewer can get wrong is that there are still ways to infer adaptation, even for only two or a few species, that are even highlighted in Garland (1994).
How can you infer adaptation from two or a few species? When Mike and I started thinking about this, we were guided by the concept of robust inference from Platt and the severity principle from Mayo and colleagues, which I first learned about by reading the Dynamic Ecology blog. Basically, we think you can infer adaptation by 1) making appropriate comparisons of closely related species that are generally similar in most ways, 2) formulating hypotheses that can be subjected to integrative testing, 3) collecting lots of data to avoid statistical underdetermination that will allow severe tests of hypotheses, and 4) finally, systematically subjecting alternative hypotheses to severe tests. It is hard to actually do this, particularly in a single study, but I think my best work at least strives for this kind of robustness, either in a single study or across several studies. These basic ideas are already in Garland (1994), so our perspective piece really just expanded on those ideas and combined them with the concepts of robust inference and severity.
Our hope was that we can convince the integrative biology community (who reviews our manuscripts) to avoid the knee-jerk response of “Two-species = bad” based upon a superficial understanding of Garland (1994). Integrative organismal research is hard and time-consuming, but is important in the context of declining biodiversity and changing global climate. I do think comparative research with only a few species should be careful with inferences, and we don’t want a return to dog to goat comparisons. But we also don’t want to make the publication process more difficult or fraught than necessary, especially as it is only by accumulating single or few species studies that we will be able to deploy phylogenetic comparative analyses to test for adaptation in a macroevolutionary context.
References
Garland Jr, T., & Adolph, S. C. (1994). Why not to do two-species comparative studies: limitations on inferring adaptation. Physiological Zoology, 67(4), 797-828.
Cox, C. L., & Logan, M. L. (2021). Using integrative biology to infer adaptation from comparisons of two (or a few) species. Physiological and Biochemical Zoology, 94(3), 162-170.
Weibel, E. R., Taylor, C. R., Weber, J. M., Vock, R., Roberts, T. J., & Hoppeler, H. (1996). Design of the Oxygen and Substrate Pathways: VII. Different Structural Limits for Oxygen and Substrate Supply to Muscle Mitochondria. Journal of experimental biology, 199(8), 1699-1709.
Weibel, E. R., Marques, L. B., Constantinopol, M., Doffey, F., Gehr, P., & Taylor, C. R. (1987). Adaptive variation in the mammalian respiratory system in relation to nergic demand: VI. The pulmonary gas exchanger. Respiration physiology, 69(1), 81-100.
My interest in this matter comes from personal experience with publishing ecological and physiology research, often with only two species. Many a time the dreaded reviewer number two would trot out the first part of the title of Garland (1994) and tell us that we can’t do two species comparisons. But that paper is about how there are limitations for inferring adaptation with only two species, not that no study should ever be published with two species. There is no reason that you can’t include two species in a single paper to talk about variation within a group, or simply for convenience, as long as you careful about evolutionary inferences. Additionally, the philosophical issues that Garland (1994) was raising aren’t solved by simply adding (or subtracting) one species to your study- you need at least dozens of species for robust phylogenetic comparative analyses. In other words, there is no nadir of inference power that is centered at a sample size of two species. My close collaborator Michael Logan (at University Nevada Reno) and I were grousing about the kinds of reviews that we were getting, and so we decided to publish a perspective piece highlighting the contexts where two (or a few) species comparisons were appropriate (later published in Physiological and Biochemical Zoology). Given that this perspective piece was a sort of blog post, I thought it would be worth discussing on the blog.
To understand the context in which the original Garland (1994) paper was written, we need to go back to the type of comparative physiology work of the prior few decades. I think it is safe to say that many physiologists were not exceptionally well-versed in evolutionary biology, which at the time was still in initial development (this was in the modern synthesis era). There is no doubt that we learned a lot about physiology in this era, but some of these studies would not pass muster today. One of my favorite concepts (and literature) of comparative physiology of that era is the idea of symmorphosis. Symmorphosis is the idea that organismal systems should have quantitatively matched design for maximal efficiency, such that no component of the system is functioning at substantially greater capacity than the rest of the system. To put this in a real physiological context, symmorphosis would suggest that the digestive capacity of intestinal brush-border oligopeptidases that are the final step of protein digestion should be matched to the capacity of brush-border amino acid transporters that transport the resulting amino acids into the body. It is not really the point of this exercise, but symmorphosis as a context is probably true in general, but does not hold up in lots of situations (because of safety factors, because organisms are not optimized but the result of contingency and tradeoffs). Regardless, some of the foundational work underlying this field focused on respiratory physiology, and often compared such divergent species as dogs and goats. While we no doubt learned a lot from this work, inferring adaptation to differences between dogs and goats is fraught, because you only have two data points and they have been separated by many millions of years of evolution. In other words, it is hard to attribute any difference between the species to any single factor, because they differ in so many ways- diet (omnivore vs herbivore), locomotion (unguligrade rather than digitigrade), habitat type, and so on. All of this to say that the original Garland (1994) paper was a call to arms for rigor in comparative research, which was sorely needed at the time and I still think is an important message.
So what is the second reviewer getting wrong when they tell us that we cannot publish a two-species study? Well, I think they get several things wrong. First (and already addressed above), Garland (1994) is about inferring adaptation. If there are no inferences about adaptation, then bringing up Garland (1994) is irrelevant. Second, there is no criticism of a two-species comparison that does not apply to three-species studies, or four-species studies, or even one-species studies. But assembling data on the number of species necessary for phylogenetic comparative analyses might be difficult or impossible for complex physiological data, and it may take many years to accumulate enough research for comparative analyses. So what do you do if you want to think about evolution and adaptation, but cannot get data for enough species for comparative analyses? The final thing that the second reviewer can get wrong is that there are still ways to infer adaptation, even for only two or a few species, that are even highlighted in Garland (1994).
How can you infer adaptation from two or a few species? When Mike and I started thinking about this, we were guided by the concept of robust inference from Platt and the severity principle from Mayo and colleagues, which I first learned about by reading the Dynamic Ecology blog. Basically, we think you can infer adaptation by 1) making appropriate comparisons of closely related species that are generally similar in most ways, 2) formulating hypotheses that can be subjected to integrative testing, 3) collecting lots of data to avoid statistical underdetermination that will allow severe tests of hypotheses, and 4) finally, systematically subjecting alternative hypotheses to severe tests. It is hard to actually do this, particularly in a single study, but I think my best work at least strives for this kind of robustness, either in a single study or across several studies. These basic ideas are already in Garland (1994), so our perspective piece really just expanded on those ideas and combined them with the concepts of robust inference and severity.
Our hope was that we can convince the integrative biology community (who reviews our manuscripts) to avoid the knee-jerk response of “Two-species = bad” based upon a superficial understanding of Garland (1994). Integrative organismal research is hard and time-consuming, but is important in the context of declining biodiversity and changing global climate. I do think comparative research with only a few species should be careful with inferences, and we don’t want a return to dog to goat comparisons. But we also don’t want to make the publication process more difficult or fraught than necessary, especially as it is only by accumulating single or few species studies that we will be able to deploy phylogenetic comparative analyses to test for adaptation in a macroevolutionary context.
References
Garland Jr, T., & Adolph, S. C. (1994). Why not to do two-species comparative studies: limitations on inferring adaptation. Physiological Zoology, 67(4), 797-828.
Cox, C. L., & Logan, M. L. (2021). Using integrative biology to infer adaptation from comparisons of two (or a few) species. Physiological and Biochemical Zoology, 94(3), 162-170.
Weibel, E. R., Taylor, C. R., Weber, J. M., Vock, R., Roberts, T. J., & Hoppeler, H. (1996). Design of the Oxygen and Substrate Pathways: VII. Different Structural Limits for Oxygen and Substrate Supply to Muscle Mitochondria. Journal of experimental biology, 199(8), 1699-1709.
Weibel, E. R., Marques, L. B., Constantinopol, M., Doffey, F., Gehr, P., & Taylor, C. R. (1987). Adaptive variation in the mammalian respiratory system in relation to nergic demand: VI. The pulmonary gas exchanger. Respiration physiology, 69(1), 81-100.
RSS Feed