Generalization in Evo-Devo
How general are our findings concerning the evolution and development of life forms? I will discuss this question of generality and the related question concerning the existence of lawlike generalizations with specific reference to evolutionary developmental biology (evo-devo). On the one hand, evo-devo suggests that the evolution and development of life forms is more contingent than we have previously presumed, thus showing that we have accidental and non-generalizable results concerning evolutionary and developmental phenomena. On the other hand, evo-devo reveals the existence of generality-maintaining mechanisms, thus showing that certain evolutionary and developmental generalizations have stability and necessity. These two findings suggest a picture of evolution as being both stable and contingent, repeatable and unique, and generalizable and non-generalizable. This picture of evolution was already present during the Modern Synthesis. At the same time, evo-devo has provided us with new insights into what are generality-destroying, generality-maintaining, and generality-creating mechanisms behind the evolution and development of life forms.
KeywordsConstraints Contingency Natural kinds Generalizations Laws
How general are our findings concerning the evolution and development of extant and extinct life forms? What kinds of generalizations can be identified in evolutionary developmental biology (evo-devo)? This topic is related to many central epistemic, ontological, and methodological issues in both evolutionary and developmental studies. As a consequence, answering these questions is pertinent to ongoing scientific inquiry found in evo-devo.
A classic question surrounding generality in evolution is whether there is a direction to evolution or whether evolutionary outcomes can be predicted. Can we discern what evolutionary trajectories will be followed by traits? Can we predict when and where new traits or novelties will arise? Or, are we limited to only explain the origin of a novelty post facto? Are evolutionary processes capable of producing general and stable patterns, or do they yield idiosyncratic and unique results? If evolution is unique and non-repeatable, as evidence at least partly suggests, how and with what kind of evidence do we test phylogenetic hypotheses concerning the ancestral relationships between taxa, especially since extant and extinct life forms exhibit a considerable diversity of traits and body plans? Do evolutionary or developmental laws exist? What role do natural selection or various kinds of constraints have in creating, destroying, or maintaining generalities around which classical biologists formulated many bold “laws”? Can anything general be predicted about the evolution of possible life forms elsewhere based on our evolutionary and developmental understanding?
Answering all, or even many, of the above questions is beyond the scope of this chapter. The focus here will be on the question of how general results and findings concerning the evolution and development of life forms are and how evo-devo has helped answer this question vis-à-vis its predecessors.
In the “The Significance of Exceptions,” the traditional idea of biological generalizations being riddled with exceptions is discussed. Then, in “Accidental Products of Evolutionary History,” the picture of evolution being a contingent and historical process is portrayed as already (at least partly) included in the Modern Synthesis. The implication of these sections is that we lack universally true and stable biological generalizations and laws concerning evolutionary and developmental phenomena, since biological generalizations are both riddled with exceptions and accidental products of evolutionary history. However, in “From Contingency to Constraints and Generative Entrenchment,” the implications of evo-devo on traditional ideas about stable generalizations and accidental outcomes are discussed before treating the “Disappearance and Return of Laws and Natural Kinds” in biology. The final section concludes.
The Significance of Exceptions
Biologists have always sought generalizations that would allow predicting, explaining, and understanding phenomena. A common pattern is that a bold generalization is initially formulated based on empirical findings and data. Subsequent research, however, reveals an increasing number of exceptions to the original generalization.
Classical evolutionary biologists, such as Mayr (1942) and Rensch (1960), devoted considerable time to finding generalities and elucidating generalizations for both evolutionary and developmental phenomena while trying to explain these as effects of natural selection. Many generalizations were dubbed as laws, at least initially. For instance, Cope’s law – a trend toward increased body size in many taxa through life’s history – was explained as a gradual pattern arising from selection for larger body sizes, which were thought to confer adaptive advantages on individuals (e.g., via reduced predation). Other examples of classical evolutionary and developmental generalizations include the biogenetic law, according to which ontogeny recapitulates phylogeny; Dollo’s law of phylogenetic irreversibility, which includes the idea that evolution cannot be reversed and that complex structures lost in evolution cannot be regained; Mendel’s laws, which include both segregation (during gamete formation, alleles dissociate so that each gamete carries only one allele for each genetic locus) and independent assortment (allele dissociation occurs independently at each genetic locus); the law of the generalist, according to which unspecialized species tend to avoid extinction longer than specialized species; Williston’s law (or the law of anisomerism), a phylogenetic trend in which serial, repetitive, similar, or unspecialized traits or parts in organisms evolve toward fewer numbers and more specialized functions; Mayr’s law (allopatric speciation), which holds that new species evolve when a population is separated by geographic isolation from its parent population; and the central dogma, according to which DNA is transcribed to RNA and RNA is translated to proteins unidirectionally.
Subsequent to the formulation of each of these classic “laws,” biologists have spent considerable time appraising the validity of these laws, questioning previous explanations given for them, and identifying exceptions to the purported generalizations. For example, Mendel’s laws have many exceptions (Crow 1979). Linkage is an exception to independent assortment because two (or more) alleles at different loci exhibit correlations in assortment due to proximity of location on a chromosome. Dollo’s law is considered by some to be relatively established, though there is plenty of discussion about how to define it, what would count as an exception, and what are proper methods for its testing (Gould 1970). There are clear exceptions to this law as well, such as the re-evolution of mandibular teeth in a frog genus after the disappearance of the trait in the lineage (Wiens 2011).
Cope’s law does not apply to flying animals in general; and, even in lineages that do exhibit this trend (e.g., equines), there is considerable stasis and reversals (i.e., decreases in body size) in lineages (MacFadden 1986). Some exceptions to Mayr’s law include polyploidy and sympatric speciation. Retroviruses provide a well-known exception to the classical central dogma of molecular genetics. Overall, similar considerations apply to other generalizations. They are riddled with exceptions, which seem to question the lawlike status of generalizations.
Some authors have suggested that the exception-riddled nature of generalizations is not damaging to their lawlike and general status, since the generalizations hold when ceteris is paribus, i.e., “when some other unknown interfering conditions remain the same or absent” (e.g., Carrier 1995). That is, when biological generalizations are qualified with “other things being equal” clauses, we have exceptionless generalizations because the exceptions are only apparent rather than genuine. Other things are not equal when generalizations are extended beyond their intended domains of application. For example, generalizations are not expected to apply to cases in which unknown interfering factors do not remain the same or are not absent. Moreover, in biology other things are rarely equal. Hence, according to these authors and contrary to appearances, biological generalizations are exceptionless in their relevant domains of application.
Let us presuppose that the ceteris paribus strategy is valid. There is another line of argumentation showing that even if generalizations were exceptionless and thus true (with or without ceteris paribus clauses), they might lack necessity, because the generalizations could be accidental rather than lawlike as generalizations, that is, true but only accidentally so.
The issue is thus not only whether evolutionary or developmental generalizations are true. This is only a necessary condition for lawlike and truly general generalizations. A generalization might be true due to prevailing conditions only, such as “all pure lumps of gold have a mass less than 1,000 kilograms.” Nothing guarantees that this generalization will hold in the future if some of the conditions change. The generalization “all pure lumps of uranium-235 have a mass less than 1,000 kilograms” is true as well. The difference is that the latter continues to be true even if various background conditions were changed. Nuclear physical facts guarantee the holding of the latter generalization in various background conditions, since 1,000 kilograms exceeds the critical mass of pure uranium-235. Another way to express the difference is that accidentally or contingently true, but non-lawlike generalizations, such as the gold generalization, lack the stability (Mitchell 1997, 2000) we associate with lawlike generalizations. Lawlike generalizations have necessity or stability that guarantees their holding even if background conditions changed in various different ways.
Accidental Products of Evolutionary History
Beatty (1995) has argued that evolution is a contingent process that can lead to different outcomes despite the same or similar starting points and even with the same or similar selection pressures (see also Gould 1989). That is, given the same or similar selection pressures, the same or similar adaptations do not necessarily follow. This implies that our evolutionary and developmental generalizations are contingent, accidental, and unique outcomes since their evolution can be or could have been easily switched to another track or be disturbed by minimal changes in their initial or past historical conditions.
Some classical reasons why our generalizations concerning evolutionary and developmental phenomena are contingent are that, in addition to natural selection, stochastic and random forces, such as mutations, founder effect, and genetic drift, affect the outcomes of evolution. Moreover, even natural selection typically has multiple trait variants from which to choose, which are similar in their fitness, but differ in their realization (cf. the argument from the multiple realizability of biological properties to the non-existence of biological laws in Rosenberg 1985: 59–65). That is, even natural selection leads to contingency in biology.
Had there been no mitosis or some equivalently fit or fitter alternative to mitosis in the past, then meiosis would not have evolved, since meiosis evolved from mitosis. Consequently, Mendel’s laws would also not have evolved on this planet, since the validity of these generalizations depends on the operation of meiosis and mitosis. In this manner, the generalizations behind both Mendel’s “laws” are contingent and accidental products of history. In fact, non-Mendelian mechanisms of inheritance occur on our planet. If the future environment changes so that non-Mendelian mechanisms become as fit or fitter than Mendelian ones, then they could become as omnipresent as Mendelian mechanisms are presently.
Similar considerations apply to other previously discussed generalizations. Evolutionary and developmental generalizations are accidental products of evolutionary history, even if they are true and lack exceptions. The generalizations lack the stability traditionally associated with laws. Different regularities and generalizations could have evolved and held on this planet had different initial or past historical conditions prevailed. Similarly, a slight change in our current conditions would amount to changes in the holding of our current developmental and evolutionary generalizations.
From Contingency to Constraints and Generative Entrenchment
Some central findings of evo-devo suggest that generalizations concerning evolutionary and developmental phenomena are more contingent and accidental than already suggested.
Developmental plasticity and, more generally, eco-evo-devo suggest that the way phenotypes of organisms are determined does not solely, and in some cases perhaps not even mainly, depend on their genotypes, but on the environmental conditions (see the chapters “Developmental Plasticity and Evolution” and “Eco-Evo-Devo”). For instance, the final developmental form of an insect species may depend on its ecological environment, such as the presence of predator species (cf. Abouheif et al. 2014). Environmentally induced phenotypic variation, such as polyphenisms, facilitates the adaptations of phenotypes and allows for rapid changes in phenotypes vis-à-vis environmental changes. In other words, ontogeny as a process is more complex and contingent on environmental factors, including the syn-ecological or community context of a species, than what was previously presumed.
The idea of genetic material being the only way to carry inheritance information from parents to offspring is being questioned by findings about epigenetic factors, such as DNA methylation and histone modifications. Epigenetic factors, heritable or not, are often essential for normal development. Again, the implication is that our generalizations concerning developmental phenomena are contingent, since non-genetic factors, such as epigenetic and environmental or ecological factors, not only broaden our understanding of heredity and development but also provide new sources of variation for ontogeny.
These findings concerning ontogeny could have major phylogenetic consequences. As phenotypic plasticity, polyphenism, and epigenetic factors provide new sources of variation for development and selection to act on, variation in these factors may lead to evolutionary novelties, reproductive isolation, and speciation events (see the chapter “Developmental Innovation and Phenotypic Novelty”). In other words, our generalizations concerning developmental and evolutionary phenomena may be true, but they are contingently true. Perhaps even more so than suggested by the Modern Synthesis, there are many other contingent factors and sources of variation than genetic mutations and drift upon which the truth of these generalizations depend.
Simultaneously, evo-devo has revealed that there are also generality-maintaining mechanisms in evolution and development (see the chapter “Mechanisms in Evo-Devo”). These findings suggest that our generalizations concerning evolutionary and developmental phenomena might be more stable, necessary, and less prone to exceptions than formerly presumed. Relevant phenomena include generative entrenchment and constraints (see the chapter “A Macroevolutionary Perspective on Developmental Constraints in Animals”). The idea of generative entrenchment is simple: even though, for instance, a trait or a gene was originally a highly contingent result of an evolutionary history, it can become a functional necessity that is essential and cannot be changed. A constraint is here defined as any factor that limits evolutionary change.
Although on this planet it is a true generalization that hereditary information – with the possible exception of epigenetic factors – is carried by nucleic acids, this is a conditional and contingent fact of our planet’s and life’s history. Had the past conditions or other initial background conditions been different, then other materials could have evolved to do the same thing. Hence, the ubiquity of the genetic code is an accident whose evolution could have been disturbed or switched to another track. Note that contingency and accident do not imply that the code is arbitrary or without adaptive advantages.
Yet throughout the history of life, this code has become so generatively entrenched as to be nearly impossible to change because so many other things depend on it. In other words, the code and the way it is realized represent a functional necessity. The same is true of many basic biological mechanisms, the presence and functioning of many ancestral or primitive characters or traits: Mendel’s laws and the mechanisms of mitosis and meiosis, diplobiontic life cycle, cell respiration, the Krebs cycle, and the use of ATP in metabolic processes, eukaryoticity, homeobox genes (see below), the mechanism of photosynthesis, bilateral symmetry, dorsal-ventral polarity, initiator and terminator codons, DNA ligase and DNA methylation, morphogenesis and organogenesis in general, and the specific mechanisms of apoptosis in ontogeny. It is not just genetic factors that might become generatively entrenched, but also epigenetic factors.
Generatively entrenched traits (Wimsatt 2001) are functional necessities for organisms’ development, survival, and reproduction, which resist evolutionary change, even if we suppose that evolution proceeds via forces other than selection, such as mutation and drift. For a trait to become generatively entrenched, it does not matter how or under what evolutionary forces it has evolved or will evolve. The number of other traits that come to depend on the functioning or development of the trait in question is what matters (see also the chapter on “The Concept of Burden in Evo-Devo”). A change in a deeply generatively entrenched trait has serious, deleterious effects on other traits in the development or functioning of an organism. Therefore, deeply generatively entrenched traits are preserved and resist evolutionary change, regardless of whether the trait was a consequence of drift, mutation, or selection.
Generative entrenchment thus provides us with exception- and contingency buffering and stability- and generality-maintaining mechanisms, thereby allowing for the discovery of repeatable, projectable, and stable generalizations concerning evolutionary and developmental phenomena. Generative entrenchment, however, also restricts the emergence of evolutionary novelties by forcing evolution to be locked into certain directions, since it is more likely that conserved traits are preserved rather than changed – the functioning of so many other things depend on their preservation. This again implies that evolution might not be as contingent in its results, nor so easily switched to another track, as was suggested earlier. Our generalizations concerning evolutionary and developmental phenomena might be associated with stability and necessity after all.
When a generatively entrenched trait is successfully changed, however, this can produce novel traits and even new forms. The same generality-begetting mechanisms can produce novel generalizations in the context of macroevolution.
Regulatory genes, such as homeobox genes, switch other genes on or off during the development of an organism. Changes in regulatory genes (or changes in their regulatory networks) are also quite likely responsible for major evolutionary and developmental changes. Different organismal traits, shapes, and forms follow not from differences between genotypes per se, but from how and when different genes are (in)activated during ontogeny. Many of these regulatory genes are not only conserved, but almost all animals use the same or similar regulatory genes, such as homeobox genes, during development. There is thus a puzzle. On the one hand, many regulatory genes are highly conserved and generatively entrenched within diverse taxa that, on the other hand, differ greatly in development. One way to both preserve regulatory genes and allow diverse phenotypes to develop is through the duplication of conserved genes, which has happened in the case of many different regulatory genes.
Wimsatt’s (2001) notion of generative entrenchment is deliberately broader and somewhat different than that of traditional constraints (on constraints, see Richardson and Chipman 2003 and Shanahan 2008). Many kinds of constraints are discussed in the literature, such as physical, chemical, phylogenetic, architectural, and developmental. Generative entrenchment as a notion encompasses the meaning of many of these. A functional necessity begets stability to the development and evolution of the system in which it operates regardless of the nature or origin of the constraint. Moreover, generative entrenchment not only restricts variation and produces inertia to developmental and evolutionary change – functioning similarly to traditional constraints – but can work as an engine for evolutionary and developmental change and function as a source of novelties as well (see Nuño de la Rosa and Villegas 2019 for a similar function of constraints). The importance of generative entrenchment is thus that it is capable of explaining the stability and generality of micro and macro evolutionary trends while at the same time allowing for evolutionary changes or origins of novelties when generative entrenched traits are successfully changed (cf. homeobox genes, above).
Disappearance and Return of Laws and Natural Kinds
Traditional philosophers of biology, such as Rosenberg (1985), were interested in the issue of whether there exists lawlike biological generalizations. The putative lack of laws would not only imply that biology is inferior and perhaps reducible to the physical sciences; it would also indicate that there exist no autonomous and distinctive biological explanations and predictions. Laws were deemed necessary for scientific explanations and prediction (see Hempel 1965).
For many philosophers, laws were expressed as true, empirically testable, and universally quantified statements that included predicate terms making reference to natural kinds or classes (see again Hempel 1965). Natural kinds or classes refer to such terms as “mass,” “gold,” “regulatory gene,” or “predator,” which make sure that laws make no reference to particulars. Why no reference to particulars? Because laws needed to be distinguished from accidentally true, i.e., non-lawlike, generalizations that typically refer to particulars or hold because of particular background conditions but not outside of them (e.g., “all coins in my pocket today are fifty cents”).
In addition to exceptions and contingency, another issue concerning the lawlikeness in biology was that species and many other taxa, such as Mammalia, and perhaps many central causal developmental factors, such as the pax-6 and hox genes that evolutionary and developmental biologists focused on, were not natural kinds or classes, but something else, such as individuals (Ghiselin 1974; Mayr 1976), i.e., particulars. Thus, no laws were to be found concerning developmental or evolutionary phenomena, according to some philosophers, since biological generalizations were riddled with references made to particulars rather than being expressed as universally true generalizations concerning nature’s classes (see Lange 1995 for further discussion and references).
Philosophers have more recently questioned the notion of lawlikeness as an important property of generalizations that scientists, including biologists, should be and in fact are looking for to ground their predictions and explanations of nature. For instance, according to Woodward (2001) the central property of causal generalizations is their invariance under manipulations rather than their lawlikeness (see also Waters 1998; Raerinne 2013). Under this framework, the issue of whether generalizations refer to natural kinds is obsolete. What matters is whether generalizations give us accurate recipes for successful manipulations of natural phenomena. This aligns with the ideas of some central defenders of the Modern Synthesis as well, who rejected typological or kind and class thinking (cf. Sober 1980 and Lewens 2009 for discussion and references).
Despite this, some issues that were discussed in the traditional philosophical context of laws are being reintroduced to evo-devo literature, such as the existence of natural kinds in evo-devo (Rieppel 2005a, b; see also the chapter “Typology and Natural Kinds in Evo-Devo”). Examples of kinds are, for instance, body plans and developmental types that distinct taxa share. Natural kindness of other concepts in evo-devo has also been discussed, such as “modularity.” The existence and stability of kinds and that distinct taxa share kinds pose no problems for evo-devo. Generative entrenchment and constraints can be used as explanations (but see Wagner 1996 for an alternative explanation concerning modularity as a kind).
Typological themes in evo-devo have already been discussed by others. For instance, Lewens (2009) views typological thinking in evo-devo as a complementary explanatory strategy to the population thinking of the Modern Synthesis. Nevertheless, there are aspects of it that have received less attention. What is it that makes kinds explanatory? It seems that certain authors think that kinds are explanatory due to their generality and unifying power. This was also the basic premise of a covering law account of scientific explanation by Hempel (1965). The more diverse and distinct the set of phenomena a law covers as its instances, the more general and unifying the law, and the better the explanation or prediction by the law. If this is so, then typologists might be understood as arguing for the return of laws and law-based accounts of scientific explanation in evo-devo.
Natural kinds are postulated not only as explanantia, but as explananda or targets of explanations as well. For instance, “modularity” and “developmental types” are both used in explanations and as targets of explanations. In general, not much discussion exists on different explanatory, heuristic, evidential, or classificatory roles of natural kinds in evo-devo (but see Love 2009 for discussion and references) and by virtue of which properties the kinds function in these roles (e.g., causality vs. unification; fruitfulness vs. carving nature at its joints; and so on). In evo-devo, natural kinds seem to have diverse, if not elusive, functions and roles.
How has evo-devo changed our understanding of evolutionary and developmental phenomena and especially what has it contributed to the question of how general our evolutionary and developmental findings are? The same general picture of evolution currently reigns as during the days of Modern Synthesis. Rensch summarized the central findings and tenets of the Modern Synthesis as follows: “Now these new results [of evolutionary research] allow two kinds of conclusions, which seem to be very contradictory. On the one hand, evolution may be looked at as an undirected unique historical process; on the other hand, it seems to be determined by a great number of laws and rules” (Rensch 1960: 95).
Our understanding of the causes and mechanisms of why evolutionary and developmental phenomena appear to be both stable and contingent, repeatable and unique, and generalizable and non-generalizable have changed, however. Evo-devo has provided us with new insights about the sources of variation in the development and evolution of life forms, such as developmental plasticity and epigenetics (see also the chapter on “Evo-Devo’s Contributions to the Extended Evolutionary Synthesis”). Evo-devo has contributed a better understanding of how evolutionary novelties could arise through changes in highly conserved traits of organisms. Simultaneously, it has provided us with insights of what constraints on variation and causes of stability and generality in evolution and development are, such as generative entrenchment.
It is an open question to what extent evo-devo implies changes in epistemic practices. Many authors believe that natural kinds have important and even central roles in evo-devo. But it is still unclear what diverse and distinct roles kinds have in evo-devo and what it is about kinds that furnishes them with such roles. Most of the discussion on kinds in evo-devo has focused on ontic issues (e.g., homeostatic property clusters or traditional essences), whereas the diverse epistemic roles of kinds deserve equal attention.
- Beatty J (1995) The evolutionary contingency thesis. In: Wolters G, Lennox J (eds) Concepts, theories, and rationality in the biological sciences. Universitätsverlag Konstanz, Konstanz, pp 45–81Google Scholar
- Carrier M (1995) Evolutionary change and lawlikeness. Beatty of biological generalizations. In: Wolters G, Lennox J (eds) Concepts, theories, and rationality in the biological sciences. Universitätsverlag Konstanz, Konstanz, pp 83–97Google Scholar
- Gould S (1989) Wonderful life. Norton, LondonGoogle Scholar
- Hempel C (1965) Aspects of scientific explanation. Free Press, New YorkGoogle Scholar
- Mayr E (1942) Systematics and the origin of species. Columbia University Press, New YorkGoogle Scholar
- Nuño de la Rosa L, Villegas C (2019) Chances and propensities in evo-devo. Brit J Phil Sci. https://doi.org/10.1093/bjps/axz048
- Rensch B (1960) The laws of evolution. In: Tax S (ed) Evolution after Darwin, vol 1. University of Chicago, Chicago, pp 95–116Google Scholar
- Rieppel O (2005b) Type’ in morphology and phylogeny. J Morphol 267:525–535Google Scholar
- Wimsatt W (2001) Generative entrenchment and the developmental approach to evolutionary processes. In: Oyama S, Griffiths P, Gray R (eds) Cycles of contingency. MIT, Cambridge, pp 219–237Google Scholar