Evo-devo research is flourishing. However, I will argue that this high level of practical activity is not yet based on a fully coherent philosophical foundation. Although the general aims and key conceptual ingredients of evo-devo have been clearly articulated (e.g., Arthur2002; Hall2003), current practice in the field betrays a certain epistemological inconsistency. Because epistemological consistency is one of the fundamental characteristics of a fully mature science (cf. Ruse1996), evo-devo has not yet reached full maturity as a scientific discipline. The most important cause of this situation is the failure of many evo-devologists to properly appreciate the metaphysics of biology.

In this paper, I will suggest how evo-devo might be unburdened on the basis of an explication of the metaphysics of biology. Specifically, I develop my arguments as logical implications of the notion that species and other taxa, as well as their constituent parts or characters, are best understood as individuals rather then classes (hereafter the individuality thesis) (Ghiselin1997,2005a,b). First, I will illustrate the degree to which essentialist or typological thinking pervades evo-devo and how the individuality thesis suggests different interpretations. Second, I will illustrate how, in a context of typological thinking, evo-devo’s harmless preoccupation with distant ancestors has become transformed into a pernicious problem afflicting the choice of model organisms. I will expose the logical flaws underlying the almost universal assumption that basal taxa are the best choice for representing the clade of which they are a part, as well as for optimal extrapolation of results. It should be clear that my criticisms of published work are to be considered as arising from strictly adopting the individuality thesis, and, as such, refer only to selected aspects of some of the works.

I like to point out that I am a biologist, not a philosopher. It was not until I read Ghiselin’s masterful bookMetaphysics and the origin of species (1997) that I was rather surprised to discover how profoundly this exercise changed and clarified my views on many difficult issues, which heretofore I merely understood dimly. Conversations with colleagues gave me reason to think that many biologists share a general apathy (signifying lack of real study, not the perusal of an occasional abstract) toward the metaphysics of biology. Therefore, I hope that this paper will be an encouragement to others to study these important issues. The inevitable boon in understanding will be well worth the effort.

“Taxa, like other individuals, can change indefinitely, and the only thing that they must share is a common ancestor.” (Ghiselin2005b, p 91)

“Developmental systems are individuals: they can change indefinitely.” (Ghiselin2005b, p 99)

If you have never explicitly considered your metaphysical commitments, your agreement or disgreement with a few statements may help you gauge your general leanings: (1) I’m a member of the speciesHomo sapiens; (2) A type specimen of a species is generally a good example of that species; and (3) One of the defining features of Mammalia is hair.

When I offered these statements to the audience at theDevelopment and evolution of arthropods symposium, of which these are the proceedings, I received overwhelming votes of consent. This leaves no doubt that many researchers are intuitively comfortable with interpreting taxa as classes. If this manifestation of typological thinking is accepted as unproblematic, perhaps only implicitly, the door is wide open to serious misunderstandings of important issues in evolutionary biology. One member in the audience pejoratively dismissed the issue as “mere semantics.” As I hope to show in this paper, such a naive labeling of what must be considered to be both “the most difficult and most important subject” in evolutionary theory (Gould2002, p 599) can only lead to unfortunate, but preventable, errors in thinking.

First, the most revolutionary aspect of the Darwinian conception of evolution is a profound metaphysical shift from considering taxa as classes to considering them as individuals. It was not until Ghiselin (1974) proposed his “radical solution to the species problem” that the larger community of biologists started to take an explicit interest in this fundamental issue. Yet, the study of the recent evo-devo literature leaves little doubt that the profound implications of the individuality thesis are not widely appreciated.

For the purposes of this paper, it is sufficient to list the most important distinguishing features of classes and individuals. For an exhaustive and insightful defense of the individuality thesis, I refer the reader to Ghiselin (1997). At the outset, it should be realized that the distinction between classes and individuals is the most fundamental ontological dichotomy. Anything is either a class or an individual (Ghiselin1997). By accepting taxa as individuals, the correct responses to the above statements will become obvious. (1) Individuals are concrete and spatiotemporally restricted. Classes (including natural kinds) are abstract concepts outside of time and space. (2) Individuals, by virtue of being concrete, can partake in natural processes, while classes cannot. (3) The ontology of individuals is the part/whole relation, wherein high-level individuals are composed of low-level parts. Consequently, ontologically, the whole body of a multicellular organism vs its comprising parts (e.g., organs, cells) exhibits the same part/whole relation as a high-level monophyletic taxon vs its subclades or species. In contrast, the ontology of classes is the membership relation. As a result, and in contrast to individuals, classes have members, examples, and instances. Therefore, the first two statements posed above are wrong. (4) Individuals do not have defining properties. They can undergo an indefinite amount of change and still remain the same individual. A class has defining properties and, therefore, cannot change indefinitely without becoming a different class. (5) Natural laws do not refer to any particular individuals. Laws can only be formulated for classes, of which individuals may be members.

For anyone who attributes to taxa a real material existence in the world as particular and concrete things that are part of a unique genealogical nexus that is not a mere mental abstraction, the embrace of these metaphysical commitments is inevitable. The ontological status of taxa and their parts as individuals legitimates phylogenetics and systematic biology as the study of independently existing historical entities.

Although I think that the individuality thesis represents the majority view in biology however tacitly assumed, agreement about taxa as individuals is not unanimous. Discussion continues in both the biological and philosophical literature, notably in connection with the interminable debates about species concepts (Stamos2002; Reydon2003,2004; Crane2004), the use of phylogenetic nomenclature in taxonomy (the Phylocode) (Keller et al.2003; Sereno2005), and the conceptualization of characters (Wagner2001; Grant and Kluge2004; Rieppel2005a,b; Ghiselin2005b). However, to a biologist, some of this ongoing discussion seems ill-founded and confused, for example, when the philosophers Stamos (2002) and Crane (2004) attempt to deny or downplay the importance of a unique historical origin of individual species. In particular, there have been several recent attempts to conceptualize taxa and characters as classes in the form of natural kinds, including homeostatic property cluster kinds (Wagner2001; Keller et al.2003; Rieppel2005a,b). These efforts seem to be united in making classes appear to be more like individuals. For example, Keller et al. (2003) argue that because individuals lack defining characters, the natural kind concept needs to be changed to accommodate this as well. But so far, these workers have not been able to escape the ultimate choice between class and individual, however much they attempt to modify the former to emulate the latter.

Although I consider the natural kinds thesis to be flawed, readers may find it beneficial to try and grasp the relative strengths and weaknesses of these conflicting metaphysical commitments because, ultimately, one’s research will only acquire specific significance within the context of a certain metaphysical and ontological commitment, irrespective of whether this commitment is conscious or not.

Armed with this basic understanding of the distinction between classes and individuals and accepting the individuality thesis, we are ready to evaluate the current practice in evo-devo. The individuality thesis may especially shed light on the problems of homology assessment, the choice of model organisms, and their use in the formulation of generalizations and laws about development.

“As long as evo-devo involves developmental types, it is perniciously typological. From this perspective, the only way for evo-devo to form a synthesis with neo-Darwinism is for evo-devo to abandon its fascination with developmental types.” (Amundson2005, p 256)

Several evo-devo theoreticians are acutely aware of the presence of typological thinking in either the history of evo-devo or in current practice (Raff1996; Leroi1998; Hall1999; Richardson et al.1999,2001; Wilkins2002; Minelli2003). It should be noted that the concept of typology is related to the concepts of essentialism and idealism, and depending on the context, typological thinking may have a variety of different connotations in the history of biology (Amundson1998,2003,2005; Richards2002; Winsor2003). However, the interpretation of individuals as classes appears to be a common denominator of typological thinking, which is the meaning I adopt in this paper.

A prevalent manifestation of our intuitive appeal to typological thinking is the use of suggestive language. Some of this is a mere leftover from outdated philosophies of biology, for example, the continuing use of terms like “archetypes,” the Linnaean category “class,” and the identification of “members” of taxa. Just as my liver is not “a Ronald,” i.e., a member or example of me, so too is the title of this recent paper logically flawed: “Dinophilidae (Annelida) is most likely not a progenetic Eunicida...” (Struck et al.2005). Wilkins’ (2002, p 520) recent appeal to taxa as classes is even more obvious when he speaks of “phylogenetic sets” instead of clades.

Equally misleading is the definition of clades by synapomorphies rather than their mere diagnosis. For example, Slack et al. (1993, p 490) defined animals with the zootype, which represents a specific pattern of gene expression. However, individuals have no defining characters, merely diagnostic ones. Consequently, even if no trace whatsoever of the zootype can be traced in a particular animal, it does not mean that it is not an animal.

Of course, one might argue that these manifestations of typologically tainted language are scarcely convincing evidence of real distorted thinking, but the use of suggestive language could quite easily predispose one to think typological thoughts. In the following sections, I will show that typological thinking has indeed penetrated deeper conceptual layers of evo-devo.

“Let two forms have not a single character in common, yet if these extreme forms are connected together by a chain of intermediate groups, we may at once infer their community of descent.” (Darwin1859, p 409)

Ever since Darwin, we have understood evolution as descent with modification. Consequently, no degree of modification can be used as evidence against common descent. In the chapter on “Mutual affinities of organic beings: morphology: embryology: rudimentary organs,” Darwin made it very clear that no amount of difference between organisms due to various degrees of modification could impact our decisions about genealogy (Darwin1859). Unfortunately, this fact has not been internalized by all biologists. A failure to take the individuality thesis to heart has ensured a continuous thread of typological thinking throughout the history of biology.

The potential for indefinite change of individuals is at the very core of Darwinian metaphysics. It underscores the fundamental distinction between classes and individuals. By definition, the members in a class need to share some properties on the basis of which membership in the class is defined. On this view, the members of a particular class can exhibit no indefinite change without losing the characteristics necessary for membership in the class.

In contrast, individuals can change indefinitely while remaining the same individual as long as their spatiotemporal continuity or persistence remains unbroken. I am still the same individual as the zygote in my mother’s womb 33 years ago, and this would even be so if every single nucleotide position in my genome had accumulated at least one mutation during my lifetime (admittedly, I might then be an exceedingly unhealthy individual). Similarly, a particular individual lineage of nucleotides may accumulate unlimited mutational changes without becoming a different lineage. Would this not be the case, then molecular phylogenetics would instantly become impossible because we would loose the ability to assign homology to different nucleotides occupying the same position in aligned sequences.

In the context of the individuality thesis, it is crucial to be aware of an asymmetry in the ease of recognition or diagnosis of potential homologies. Similarities are used to diagnose potential homology, but differences cannot be used to diagnose nonhomology. Differences are simply impotent with respect to the establishment of homology and phylogeny. Homology is a correspondence relation between homologues based on shared common ancestry (Ghiselin1997,2005b). As Scholtz (2005, p 130) writes, it is perfectly justified to use similarity to help identify possible homologues because “inherited similarity is a function of the correspondence of parts.” However, as time since common ancestry of two homologues increases, their similarity (indeed, initial identity) may deteriorate. This does not make the features less homologous. It just makes homology less easy to diagnose. Similarity is no more than a helpful operational criterion to identify possible homologues, but it should not be part of the definition of homology. After all, individuals, including characters and taxa, can change indefinitely without negating their unique historical origin.

Apparently not all workers are comfortable with the implications of the individuality thesis. The idea that at least something immutable, however small, must characterize the same individual, can easily be discerned in the literature of evo-devo and appears to be one of the most tenacious strongholds of typology. For example, although an ardent advocate of the individuality thesis, Gould (2002, p 602) wrote that “an individual may undergo some, even substantial, change during its lifetime, but not so much either to become unrecognisable or to encourage redefinition as a different thing.” However, this application of the vernacular sense of individuality, which attaches a new name to an individual after a certain amount of change, is not supported by the philosophical understanding of what an individual is. Rather, it treats individuals as classes with an intensional definition, which necessitates the designation of a new class when the old definition no longer applies.

Another clear expression of lingering typological thinking in evo-devo is prominently displayed in the homology concepts propounded by various evo-devologists. For example, Wilkins (2002, p 167) expresses his unease thus: “the idea that the ‘same’ trait in two different organisms may actually exhibit more points of visible difference than of discernable identity seems counterintuitive, to put it mildly.”

To codify the logical necessity of the conservation of at least some identity or observable similarity in homologues, these and several other evo-devologists have conceived homology concepts alternative to the widely accepted evolutionary homology concept. See for example, the biological homology concept propounded by Günther Wagner (see discussion in Amundson2005), the partial and combinatorial homology concepts advocated by Minelli (1998,2003), Wilkins (2002, pp 164–165) and Sanetra et al. (2005), and Hall’s (2003) recent attempt to conceive of a continuum of homology by incorporating notions of similarity in developmental processes in the concept. All these conceptualizations conflict with the central tenet of the individuality thesis that individuals can change indefinitely. According to these authors, homology cannot profitably be seen as an “either, or” question, and the amount of accumulated differences must somehow be incorporated into homology statements.

For example, according to Wilkins (2002, pp 164–165; Sanetra et al.2005present a similar example), the gene circuitry that underpins the development of a particular phenotypic character may change its constituent genetic elements over time, leading to a gradual diminishing of homology as increasingly more elements are replaced or changed. Wilkins concludes that “in terms of the whole genetic circuitry the homology is partial.” The logical equivalent would be to conclude that two molecular sequences are partially homologous because one sequence may have acquired a single nucleotide insertion. This logic confuses homology with similarity, which is the most common logical error in comparative biology (Ghiselin1997,2005b).

At best, the designation of partial homologues is the result of a failure to specify in detail which parts of a complex whole are in fact homologous to parts in another complex whole and which are not. Consequently, the heuristic value of partial or combinatorial homology concepts is unclear to me and seems to represent, at best, a coarse-grained summary view of complex characters and their underpinning developmental mechanisms that are composed of both homologous and nonhomologous parts. At worst, partial or combinatorial homology is merely a disguised version of the long repudiated claims of molecular phylogeneticists that homology of sequences comes in percentages.

The key question in considering partial or combinatorial homology is whether it is accepted that homology of a whole can gradually decrease with time as its constituent parts change, via modification, replacement, loss, or addition of new parts. On the basis of the individuality thesis this can clearly not be accepted. As Darwin himself stressed, no amount of modification can erase history. Homology can only decrease in one step: when phylogenetic continuity is lost through loss of a homologue.

Svensson (2004) attempted to defend the logical foundation for the evo-devo strategy of using the expression patterns of developmental regulatory genes in “testing hypotheses of organ homology.” The supposed test is based on the premise that nonhomologous phenotypic characters will be underpinned by differences in gene expression patterns. Hrycaj and Popadic (2005, p 179) advocate a similar logic when they write that nonhomologous characters should display “an appreciable degree of regulatory differences.” However, accepting the ontology of the part/whole relation of individuals, a logical justification for this approach is lacking. Parts of any given individual can change without negating the identity of the higher-level individual. This reasoning can be extended to expression patterns of developmental genes if these are regarded as parts of a developmental process or phenotypic character they help to form. If gene expression patterns are not accepted as parts of a developmental process or the character they help to form, there is even less reason why they should be relevant. Only when the character is the expressed mRNA or protein itself will there be a direct relationship between gene expression and homology of its product. Similar (conserved) developmental (genetic) processes may underpin the development of complex homologous phenotypic characters, but this is not stipulated by necessity.

This theoretical argument is amply born out by the empirics of evo-devo (Scholtz2005). The dissociation between developmental process and pattern led Scholtz to the profound conclusion that “there is no ontogenetic homology criterion.” This implies that observed differences in the ontogenetic processes underlying the development of hypothesized homologous characters do not imply nonhomology of these characters. Strikingly and perhaps upsettingly, this implies that evo-devo studies of developmental gene expression patterns are virtually impotent in shedding light on the homology of their phenotypic products.

It is not too surprising to note that differences between taxa are incorrectly interpreted as potential evidence against homology by evo-devologists. Such reasoning is still deeply ingrained in systematic biology. For example, in his new textbook on phylogenetics, Wägele2005, pp 121, 209) asserts that nonhomologous characters will only be superficially similar. The problem is that homologous characters may be equally superficially similar as a result of modification.

The current literature on invertebrate zoology yields many additional examples of the incorrect use of differences as evidence against homology and monophyly. I present a few. Nezlin (2000) compared hemichordate tornaria larvae with echinoderm larvae and concluded that differences in their nervous system are evidence against the widely accepted homology of these larvae. Pilato et al. (2005) considered the existence of morphological and developmental differences between different ecdysozoans as evidence against ecdysozoan monophyly. Schram and Koenemann (2001) concluded that the fossilsRehbachiella andLepidocaris could not be considered crown-group branchiopods, because their inferred limb developmental pattern was deemed incompatible with the inferred ground pattern of crown-group branchiopods. The logically flawed use of differences as evidence against monophyly and homology was taken to an extreme in Willmer (1990). Not surprisingly, Willmer’s summary “phylogeny” of the Metazoa was massively polyphyletic.

With respect to arthropods, several of the most influential workers in the 20th century provide exemplary illustrations of the misguided attribution of phylogenetic significance to differences (Manton and Harding1964; Anderson1973; Fryer1996; Schram1978). The central methodological premise adopted by these workers is the assignment of phylogenetic significance to autapomorphies, the unique specializations of individual taxa. If taxa are too different from each other, this is taken as evidence that they cannot share common ancestry. In other words, although evolution may be understood as common descent with modification, these workers impose a limit on the amount of modification that is compatible with an interpretation of common descent.

As a final and most extreme example of the abuse of differences to argue against common descent, consider the proponents of the bogus sciences intelligent design and baraminology (study of the taxonomy of created kinds). A central goal of these versions of phylogenetics from the Dark Side is to label differences between taxa as unbridgeable gaps, which are evidence of separate acts of creation. What is extremely worrying is that proponents of these ideas have already managed to parasitize legitimate scientific outlets, from scientific journals to meetings of professional societies, including the Geological Society of America and the Society of Developmental Biology (Jenner2005). For biologists to have a proper grasp of the metaphysics of their discipline is a necessary first step toward preventing such alien intrusions.

A few remarks about laws in developmental biology are apposite here. Richardson et al. (1999,2001, p 279) are very clear in dismissing the search for laws or universal statements in evo-devo as “idiosyncratic and archaic.” They argue that statements about “laws” or “universals” in biology carry tacit assumptions about phylogenetic relationships and homology. Richardson et al. (1999, p 6) write “we define ‘laws’ (or universal principles) in this context as hypotheses about the generality of developmental mechanisms in a major taxon.” This reveals a fundamental metaphysical confusion.

Laws are by definition ahistorical. They refer to spatially unrestricted and timeless classes that are intensionally (by providing defining criteria) or extensionally (by enumerating members) defined. This does not mean that biology cannot have real laws (Ghiselin1997). By logical necessity, a developmental law or universal principle does not refer to any particular taxon or part of a phylogeny, which are individuals. Individuals can be members of a class if they possess those defining characters that are singly or jointly necessary and sufficient for the definition of that class. Most biological laws have exceptions and, thus, are not laws, either because they refer to an individual (a particular taxon) or a class of which the assumed membership is not properly established. However, if a class is intensionally defined in such a way that its members show no variation with respect to its defining criteria, legitimate laws can be formulated for those members with reference to the defining characters that lend them membership to the class. That is why Richardson et al.’s (1999,2001) criticism of attempts to formulate proper laws for classes such as the “universal positional field” is misguided.

For example, Davidson’s (1991,2001) definition of three types of metazoan embryogenesis in principle allows lawful regularities to be formulated for each type, which are predicated on the defining properties of these types. Type I embryos develop bodies of a small number of differentiated cell types through direct cell type specification mechanisms. That led to the prediction that, for the development of such relatively simple bilaterian body plans, no spatial pattern formation mechanisms such as the use ofHox genes would be necessary (Davidson et al.1995). This could then be tested by studying the development of primary larvae of invertebrates (Davidson2001), providing tentative support for the hypothesis. And if these types of development happen to coincide more or less with a particular taxon, that would not disprove laws formulated for that type of development. It would simply mean that taxa that lack the defining characteristics of that type of development are not members of the class defined by that type of development.

Evo-devologists are intensely fascinated with ancestors. Ancestors originally played a central epistemological role in evolutionary biology. Phylogenetics was originally defined as “a backward looking endeavor, the search for and study of common ancestors. The starting point in such an analysis is a particular taxon and the student of phylogeny attempts to infer the properties of its ancestors” (Mayr and Bock2002, p 175). In the mid-20th century, however, ancestors were marginalized in the epistemology of evolutionary biology as a direct result of the widespread adoption of Hennig’s phylogenetic philosophy, which left no place for ancestors as Hennig assumed they ceased to exist during speciation, as two new sister species formed. With the sole focus on sister group relationships, cladistic analysis marginalized ancestors as a mere inferential by-product relegated to the internal nodes of cladograms.

Consequently, the deep fascination with ancestors and Urbilaterian fantasies that are prominently displayed in many works on evo-devo may come across as somewhat anachronistic to modern systematists. Yet, ancestors have a real existence, which is sufficient justification for our fascination with them.

Unfortunately, within the context of typological thinking, these harmless ancestral attractions are frequently transformed into a distortion of proper evolutionary reasoning during the selection of model organisms for comparative evo-devo research. Model organisms are typically chosen based on how closely they approximate an assumed ancestor or ancestral condition of a character of interest. The result is the almost universal focus on basal taxa. Unfortunately, accepting the individuality thesis, the logical foundation of an uncritical focus on ancestral taxa or on basal taxa is extremely problematic. Amundson (2005, p 3) states a common opinion when he writes about “the reliance of evo-devo on a relatively small number of model organisms” as a “practical barrier” between evolutionary biology and evo-devo. As I will argue, it is a rather more fundamental theoretical barrier to a full synthesis.

In her often-cited paper on model systems in developmental biology, Bolker (1995, p 451) wrote that the model systems approach is “clearly an extraordinarily powerful way to analyze animal development.” However, the extent to which this is true depends on the validity of two assumptions (Bolker1995, p 451): (1) “we can extrapolate what we learn from a few model species to many other organisms,” and (2) “the models themselves are a representative sample of extant diversity.” She then concludes that in reality, these assumptions may be violated by biased sampling (as opposed to merely insufficient sampling). Let us evaluate these statements in detail.

First, to say that the model system approach is an extraordinarily powerful way to study animal development or any other biological process is merely a truism in one sense. If one just wants to study a particular biological problem without concern for comparisons, one just picks any suitable organism and that is it. Model organism choice for comparative or evolutionary research is more complex.

I will consider Bolker’s two assumptions in conjunction because they are intimately related. According to Bolker (1995, pp 451, 453) the goal of the proper choice of a model organism is to select one that represents or exemplifies the larger clade that it is a part of, so that one can extrapolate or generalize findings from the model to other organisms. These assumptions would be perfectly valid if taxa are interpreted as classes. These would have real examples that would fully represent other members in the same class with respect to the defining characters of the class. This would allow smooth extrapolation of results between members of the same class for the shared defining features.

However, if one accepts the individuality thesis, it follows that no individual is a real example or representation of any other. One individual can only be said to exemplify or represent another individual in so far as they share something. The extent to which a chosen model organism is informative about homologous characters shared with other taxa is entirely dependent upon how much evolutionary change has affected the characters of interest. As will be argued below, there is no simple criterion for choosing an optimally representative model organism.

Bolker (1995, pp 451–452) worried that frequently used criteria for choosing model organisms, such as rapid developmental rate and short generation time, create a biased and “less random” sample of model organisms than we would like. I do not understand why nonrandom sampling is problematic. Model organisms are chosen to study a particular aspect of their biology. By definition, this prevents a random or unbiased sample. Any choice of a model organism, irrespective of its phylogenetic position (see section below) creates an inherent bias. It is inescapable. If it happens that model organisms, which are independently chosen for the study of particular problems, share certain characteristics, then this may be an aid rather than a hindrance as Bolker thinks. Their shared characteristics might be used to define classes for which universal statements or laws may be formulated, as argued above.

For most comparative research, the ultimate goal of a model organism is to capture the nature of a distant ancestor. Hence, the focus is on taxa that possess primitive characters. Some researchers prominently display the importance of choosing such a model organism by going to great lengths trying to justify the primitiveness of their models. Arendt and colleagues provide a nice example. On their website and particularly in Tessmar-Raible and Arendt (2003), they have argued fervently for the “ancestrality” of their chosen model. Not wanting the reader to have any doubts, they write that “the nereid polychaetePlatynereis dumerilii—and its close relativeNereis virens—have been chosen primarily for their ancestrality,” and “Platynereis has been chosen for ancestral development, body plan and genes” (Tessmar-Raible and Arendt2003, pp 335, 338). It does not become clear relative to what these polychaetes are ancestral, because certainly their phylogenetic position deep within the polychaetes doesn’t support any unambiguous “ancestrality.” No real evidence in support of their conclusions is presented.

Most other researchers, however, almost unanimously focus on basal taxa as a proxy for ancestrality. To justify this, some workers claim that basal taxa are the closest relatives of the last common ancestor of a group (e.g., Muller and Muller2003 and Tudge2000 for sponges and cephalocarids, respectively). This is incorrect because all descendants are equally related to their last common ancestor. The urge to designate basal taxa is so great that Krell and Cranston (2004) recently published a poignant editorial inSystematic Entomology to stop the flood of papers that falsely labeled taxa as basal. Very frequently, the basal clade was simply the sister group of the one considered to be less basal. This exceedingly common error is based on what O’Hara (1992) called “differential resolution.” By judiciously collapsing subtaxa of selected clades, an asymmetry is produced that many workers then intuitively reify as a time axis. This allows the linearizing of taxa, akin to imposing ascala naturae on a bush of diversity. This becomes especially clear when intermediate taxa are specified as well. For example, Peterson et al. (1998, p 547) falsely label a thysanuran insect “basal,” a hemipteran “intermediate,” and hymenopterans and dipterans “more derived.” Species-poor taxa are often incorrectly labeled as basal with respect to more species-rich sister groups. This common strategy significantly distorts thinking (Crisp and Cook2005).

For example, consider the focus on spiders in recent evo-devo studies (Damen et al.2000; Schoppmeier and Damen2001). Sanetra et al. (2005) write: “the more basal phylogenetic position of spiders in relation toDrosophila suggests a derived mode of segmentation in the latter.” Of course, the impression of a basal spider position is solely the result of focusing on a phylogeny depicting only the major arthropod groups. Chelicerates branch off first, then myriapods, and finally a clade of hexapods and crustaceans. When you expand these clades to show individual species, spiders suddenly appear in a comparably derived position as insects. A far greater problem is Sanetra et al.’s equation of basal phylogenetic position with the possession of primitive character states.

This problem may represent one of the least appreciated problems of comparative biology. Taxa can be labeled as more or less basal with respect to a given ancestral node, by measuring their crown-ward distance in terms of the number of internodes between taxon and ancestral node. This conceptualization corresponds to common practice. Assuming such a designation of basal, Bolker (1995, p 454) writes that “more basal lineages within a clade have (by definition) fewer derived characters, and more characters retaining ancestral states that may be generalized to other groups.” These arguments appear to be accepted by many evo-devologists and systematists alike. Wilkins (2002, p 520) writes that when one goes “more basal” “one inevitably learns something about probable basal states of the character in question.” Damen et al. (2000) write “the basal phylogenetic position of chelicerates within the arthropod clade allows us to draw more general conclusions on the degree of conservation of this part of the segmentation–gene hierarchy.” Ladurner et al. (2005) consider their favored model platyhelminth to be especially valuable for studying Urbilateria because it is “more basal than any of the model organisms currently used.”

Are these statements correct? I think not. Basal taxa do not by definition have less derived characters. Of course, we all know that already, which makes the enormous attraction to basal taxa even more surprising. Is there, nonetheless, reason for assuming that basal equates primitive in many cases? This problem can (and should eventually) be tackled both empirically and theoretically. However, within the constraints of this paper, a logical approach is more concise.

There is absolutely nothing in the logic of phylogenetics that stipulates that basal taxa are less derived in their characters than less basal taxa. In a clade of extant species, each species should be equally informative about the last common ancestor of the clade, because they are all equally related to it and equally distant in time from it. The only potential distinction between these species is the number of speciation events or internodes that separates them from the common ancestor. However, this information becomes relevant only if we know that character change occurs solely or predominantly in association with speciation events. Advocates of punctuated equilibrium, such as Stanley (1998) and Gould (2002), argue that this is indeed the case for morphological characters in certain taxa under certain circumstances. However, evolutionary change in unbranching lineages is equally real (Levinton2001; Carroll1997), and the relative importance of these two modes of change are not reliably established.

Although it seems reasonable to assume that character evolution temporarily speeds up during speciation, perhaps relating to population size effects or character displacement in a diverging population, it is not clear that for the phenotype as a whole, speciation will speed up change or that change accrued as a result of speciation will always disproportionally contribute to total change compared to change accumulated during nonbranching times. For molecular data, there is some evidence that speciation events may be associated with episodes of elevated genetic change (Webster et al.2003), but we do not know whether an organism’s phenotype as a whole may exhibit faster rates of change during speciation. In this context, it is noteworthy that rates of molecular and morphological evolution need not to be linked (Bromham et al.2002).

Consequently, current evidence provides no convincing theoretical or empirical justification for assuming that phylogenetic position is a reliable proxy for the degree of character modification. This area clearly needs much more research before confident conclusions can be drawn.

I think that the widely held intuition that basal taxa may posses less derived characters may, in a large part, be self-fulfilling circular reasoning. Since phylogenies can only pick up change “between” rather than “within” taxa, a positive correlation can be expected between the number of internodes along a given evolutionary lineage from an ancestor to a descendant species and the probability of inferring character change along that lineage. This phenomenon, known as the node density effect, has long been known in molecular phylogenetics (Fitch and Beintema1990), but its significance for phenotypic data is much less understood. If the node density effect is also relevant for phenotypic data and developmental data, then this may have serious implications for general practice in comparative biology as this could mean that the amount of character change inferred in more basal taxa is systematically underestimated. This argument is more fully developed in a separate publication.

There is also no logical reason for the common claim that findings from basal taxa can be more generally extrapolated than those of more derived taxa. More generally, the degree to which findings can be extrapolated from a model organism to other taxa is not logically related to the phylogenetic position of the model against the recommendations of Hughes and Kaufman (2000). We should not be surprised, as Wilkins (2002, p 520) seems to be, that “model systems have, paradoxically, tended to constrict and distort thinking to a degree.” Even if model systems are primitive for their respective taxa, they will always constrain and distort thinking for they can never represent all diversity, and even an actual ancestor of a clade may not best represent the clade’s characters. A model with a more derived phenotype or less basal phylogenetic position may actually allow results to be extrapolated more broadly. A focus on hagfish does not necessarily generate results that can be more generally extrapolated to other vertebrates than a focus on zebra fish. In view of the large number of extant teleost species probably it is quite the reverse.

It is important to note here the inverse relationship between explanatory force and explanatory range. This is rooted in the hierarchical nature of phylogenetic relationships and the accumulation of evolutionary change over time. Increasingly distant relatives may allow generalizations across an increasingly large number of taxa, but this comes at the cost of losing explanatory force (Rieppel2005a). For example, the study of a bird would allow forceful generalizations across the range of birds. The study of a theropod dinosaur allows a less forceful generalization across birds, but the explanatory range now also includes other theropod dinosaurs. This is important to keep in mind when a model organism is chosen with the express purpose of testing the evolutionary conservation of a developmental character.

In conclusion, the model organism approach as currently practiced as a shortcut for comparative research shows important logical flaws. If taxa are individuals, and individuals can change indefinitely, no single taxon can represent another without specifying what characters are focused on and without taking the amount of character change into account. Phylogenetic position by itself is no reliable indication of how well a taxon represents other taxa or how broadly the results can be extrapolated. Basal taxa are not special, and their place in the spotlight may at least partly be an unrecognized consequence of not taking the node density effect into account for phenotypic characters. This issue needs to be studied in depth if support for the widespread assumption that basal equals more primitive is to transcend mere anecdotal evidence. Future grant proposals may be required to more fully substantiate the choice of a model organism than by simply stating it is basal.

I think that much current thinking about model systems in comparative research is a relapse into typological thinking. No single taxon can simply be abstracted as a representative of the larger clade of which it is a part. All taxa are independent and may change indefinitely. Yet, it is clear that model organisms are widely seen as what Ghiselin (1997, p 293) calls “domesticated essences.”

Achieving a mature evo-devo synthesis is dependent upon removing metaphysical confusion, which hinders the development of a fully consistent epistemological foundation. I argue that the best way to achieve this is to consider taxa and their parts ontological individuals, not classes, and to purge all lingering typological thought from evo-devo. Typological thinking is manifested from suggestive language to the misuse of differences in phylogenetic reasoning and the elaboration of homology concepts that allow degrees of homology. Typological thinking also seems to be the context that has distorted thinking about model systems. In contrast to common claims, model systems cannot represent other taxa in any straightforward manner. There is no reason why taxa with ancestral character states or basal taxa are the best representatives of other taxa, or that they allow optimal extrapolation of results to other taxa. Choice of model systems should be less constrained by considering the phylogenetic position and should take more account of the characters and their modifications that are the focus of comparative study.