The Proximate–Ultimate Distinction and Evolutionary Developmental biology

What is the relationship between developmental and evolutionary biology?

What is the relationship between developmental and evolutionary biology? This apparently simple question (aren’t developmental systems the result of evolution?) has been controversial for more than half a century, which strongly hints at the possibility that the question isn’t just a scientific one (though it is that too), but also has inescapable philosophical dimensions.

My colleague Raphael Scholl and I published a paper on this very topic a couple of years ago in the journal Biology and Philosophy, and I think it is worth revisiting some of our arguments and conclusions.

The influential biologist Ernst Mayr argued back in 1961 that biology, despite its seeming unity, is made up of at least two fields that differ in their choice of, and approach to, research problems. On the one hand, he recognized a discipline of functional biology, of which he wrote:

“The functional biologist is vitally concerned with the operation and interaction of structural elements, from molecules up to organs and whole individuals. His ever-repeated question is ‘How?’”

Then he contrasted this with the interests of the evolutionary biologist:

“His basic question is ‘Why?’. […] To find the causes for the existing characteristics, and particularly adaptations, of organisms is the main preoccupation of the evolutionary biologist.”

Mayr could be read as distinguishing between ‘‘how’’ and ‘‘why’’ questions; between causes acting in the past and causes acting in the present; between different types of causes; or between scientific disciplines.

For Mayr, functional biology is concerned with ‘‘the operation and interaction of structural elements,’’ whereas evolutionary biology studies ‘‘the causes for the existing characteristics’’, ‘‘the reasons for this diversity [of the organic world] as well as the pathway by which it has been achieved’’ or ‘‘the forces that bring about changes in faunas and floras.’’ He then assigns the label ‘‘proximate’’ to the former and ‘‘ultimate’’ to the latter. He specifies the ultimate causes further:

“These are causes that have a history and that have been incorporated into the system through many thousands of generations of natural selection.”

Back in 1961, Mayr used to proximate-ultimate distinction to argue for the continued relevance of organismic biology in the face of molecular approaches. Later on, however, in the 1980s, he began to deploy the distinction against evolutionary developmental biology:

“The suggestion that it is the task of the Darwinians to explain development […] makes it evident that [critics of the Modern Synthesis] are unaware of the important difference between proximate and ultimate causations […]. Expressed in modern terminology, ultimate causations (largely natural selection) are those involved in the assembling of new genetic programmes, and proximate causations those that deal with the decoding of the genetic programme during ontogeny and subsequent life.”

The proximate-ultimate distinction, then, is claimed to give rise to an explanatory asymmetry: evolutionary causes explain how the genetic programs underlying functional causes have changed across generations. But there is no comparable reverse relationship — functional causes are of no relevance to evolutionary causes.

In recent years there have been interesting criticisms moved against Mayr’s neat distinction of causes in biology. The most salient critiques are by R. Amundson (2005) and by K.N. Leland and collaborators.

Since natural selection invariably operates on developmental processes, Amundsen has argued that any account of an evolutionary transition is causally incomplete so long as it does not include the relevant developmental processes: “In order to achieve a modification in adult form, evolution must modify the embryological processes responsible for that form. Therefore an understanding of evolution requires an understanding of development.”

Raphael and I, however, argue that Amundson’s critique fails to offer a reason why developmental processes in particular should be of interest to evolutionary biology. On the criterion of causal completeness, we might also ask for a biochemical-evolutionary synthesis (since biochemistry is both continuously operating in organisms and changed by selection) or indeed a physiological-evolutionary synthesis. Some might argue that these are indeed needed, but we think it is possible to give good reasons why development in particular is of interest to evolutionists.

More promising at first sight is the critique of Mayr by Laland and colleagues who argue that the proximate-ultimate distinction ignores certain prevalent types of reciprocal causation: “In reciprocal processes, ultimate explanations must include an account of the sources of selection (as these are modified by the evolutionary process) as well as the causes of the phenotypes subject to selection.”

These authors’ paradigmatic case is niche construction, where selection creates organisms which alter their environment and thus, in turn, alter selection pressures. Laland et al. regard Mayr’s view as one of linear causation, where a series of genotypes are successively adapted to static selection pressures. In niche construction, however, there is a phenotype-mediated alteration of the selective environment, and thus reciprocal causation.

However, the criterion of reciprocal causation fails to pick out all the cases of interest, and only those. On the one hand, reciprocal causation is again too inclusive since it is ever present. On the other hand, reciprocal causation is not sufficiently inclusive: there are phenomena that are of great theoretical interest to developmental evolutionists but that are not neatly captured as instances of reciprocal causation.

Our thesis, which we defend in detail in the paper, is that it is a mistake to think about the proximate-ultimate distinction in terms of allegedly omitted biological causes. It is much more fruitful to think in terms of abstraction in causal explanations: the issue is not whether certain types of causation (e.g., between genotype and phenotype, or between phenotype and selective environment) exist, but whether these causal paths carry much weight in the explanations we give. We should not ask: Is there causation between genotype and phenotype? Of course there is. Or: Is there causation between phenotype and environment? Again, of course there is. We should ask instead what motivates the foregrounding or backgrounding of some parts of a complete causal account of any given evolutionary transition. The key concern is not causal relevance, in other words, but rather explanatory salience.

Mayr considered natural selection to be the main force in evolutionary explanations. Thus, he foregrounded selection and abstracted development, although both are uncontroversially part of a complete causal account of evolutionary transitions. The key question, then, becomes: Under what circumstances can selection be taken to be the main explanatory force in evolutionary explanations?

Raphael and I maintain that two issues have become conflated in this debate. The first issue is the proximate-ultimate distinction, which does not necessarily imply an explanatory asymmetry between evolution and development. We think a ‘‘lean version’’ of the distinction should be maintained, since it separates research agendas that are indeed concerned with different aspects of causation in nature: proximate questions about biological mechanisms are different from ultimate questions about the evolutionary processes that have produced these same mechanisms. Proximate and ultimate causes answer different contrastive questions. The proximate questions asks why this bird flies south in contrast to another, otherwise identical bird that lacks the same neural mechanism. The ultimate question asks why these birds fly south in contrast to another population of birds with a different history of natural selection.

The second issue is the claim that developmental mechanisms carry no force in evolutionary explanations. This is only true in cases where biological variation is isotropic (i.e., does not have preferential direction, no “developmental constraints,” to use Stephen Jay Gould’s phrase), and this empirical assumption is independent of the lean version of the distinction. On the assumption of isotropic variation, developmental mechanisms are indeed nothing but gory mechanistic detail in evolutionary explanations. When variation is not isotropic (and, it turns out empirically, it often isn’t), however, developmental causes do carry explanatory force in evolutionary transitions — and our lean version of the distinction is entirely compatible with this possibility. The task of proponents of the “evo-devo” (evolution of development) approach, then, has always been to argue not against the proximate-ultimate distinction, but against the generality of the isotropy assumption — to argue that some evolutionary explanations require us to foreground developmental mechanisms.

In summary, Raphael and I argue that the critiques of ‘‘causal completeness’’ and ‘‘reciprocal causation’’ of the proximate-ultimate distinction mentioned above fail because they do not reliably pick out those aspects of individual evolutionary episodes which are of interest to developmental evolutionists or experts in niche construction. Instead, the proper framing of the question concerns explanatory abstractions in causal models: under what circumstances is it appropriate for explanatory purposes to foreground or background certain aspects of the complete causal account of any given evolutionary transition? Ernst Mayr abstracted from developmental causes because he assumed isotropy of variation. In cases where this assumption is not warranted, however, developmental causes carry explanatory force and must be included. In other cases, interactions between the phenotype and the environment may carry particular explanatory force. This approach permits us a charitable reading of Mayr, whose focus on development-less natural selection can be seen as a special case where variation is isotropic. In the half-century since Mayr’s discussion of the proximate–ultimate distinction, it has become apparent that the isotropy assumption does not hold generally. Thus, developmental causes can carry explanatory force in both the proximate and the ultimate domain.