Obviously, these pairings do not represent homology by direct descent from common ancestry. And yet, we would not deny that some legitimate evolutionary commonality links my leg and aardvark's forearm (forget the shrew's head, although I would not be shocked if the old vertebral theory for the origin of the skull reemerges some day in a renewed form of validity). Moreover, we can be quite confident that the similarity marks a genuine historical hold, not a fortuity, or a convergence separately evolved from different archetypal bases. Still, the hold cannot be equated with true common ancestry, and must arise instead as a constraint based on common genesis from a source that imposes limitations or sets preferred channels of change from within. In his Platonic perspective, Owen called this common source an archetype. We would identify such a generating source as a developmental constraint from the historical vertex of the aptive triangle — perhaps arising from homologous genes or homologous developmental pathways in the two separated lineages.
Finally, Owen defined serial homology as the iteration of an archetypal form within the same organism in a set of repeated parts, perhaps each specialized for a particular function, but still bearing signs of the common architectural plan — as in the biramous appendages of arthropods, whether specialized as antennae, mouth parts, walking legs or genital claspers; and in the arms and legs of tetrapods.
When asked how he could square serial homology with his basic definition of “the same organ in different animals...” Owen would reply that the criterion of sameness trumped the requirement for different organisms, and that different places within the same organism would suffice. And we do not judge this response today as flippant or invalid because we share Owen's feeling that some common principle — although not common ancestry — validates a legitimate comparison of my leg and the aardvark's forearm (general homology) and the aardvark's forearm with the aardvark's own leg (serial homology). We would also identify this common principle as developmental constraint based on homologous genes and embryological pathways (whether expressed as arms and legs in the same animal, or as similar structures in two animals — although we would call such structures nonhomologous because they do not descend from a common ancestor, even though they [Page 1073] owe their structural similarity, in large part, to construction by homologous genes and developmental pathways. After all, if no common genetics or development influenced the ontogeny of the aardvark's arm and leg, we would view the two limbs as purely convergent and unaffected by internal constraint, while Owen, in such circumstances, would not have defined them as serially homologous).
We may now return to Lankester's problem: Owen's special homology translates easily into evolutionary language as descent from common ancestry. This phenomenon presents no conceptual problem, and Lankester therefore chose to separate this subcategory as the unambiguous “best case” of homogeny. But how shall general and special homology be represented in evolutionary language? On the one hand, Owen applied these terms to separate structures that do not descend from the same structure in a common ancestor. They should therefore be distinguished from special homology (Lankester's homogeny) because we must be able to identify true and unbroken continuity in physical descent as a basis for phylogenetic reconstruction.
On the other hand, general and serial homologies do record a hold of history over descent within clades. Some common property, present in a clade as a consequence of phylogenetic history, does generate the similarities in these two other Owenian categories. But this property can only be identified as a common generating pattern, a common constraint, a common pathway of development, or a common set of hereditary tendencies — and explicitly not as an overt common ancestral structure retained by descent in subsequent branches of the clade.
What then shall we call these products of common phylogenetic patterns in organic architecture — these separately evolved results of common developmental constraints, we would say today — but not of overt and expressed common ancestral phenotypes? Lankester proposed that we call them homoplasts in contrast with the homogens of common structural origin, and that we designate the process of their production as homoplasy, as distinguished from the homogeny of strict descent from common ancestral structures. But he considered both processes as subdivisions of a larger and coherent concept of homology — homogeny for Owen's special homology, and homoplasy for Owen's general and serial homology.
In defending his placement of homoplasy within a broad but coherent concept of homology, Lankester asks (1870, p. 38): “What is the other quantity covered by the term homology over and above homogeny?” Lankester answers that many similarities not due to inheritance of common ancestral structures nonetheless arise as consequences of the inheritance of unique phylogenetically constrained building patterns — and therefore deserve inclusion within a broader category of similarity based upon descent (as opposed to similarity derived purely by independent adaptation, with no contribution by constraint from an organism's past history). These independently evolved, but historically constrained, similarities — we would now call them parallelisms — define Lankester's original concept of homoplasy. Lankester does acknowledge that homoplastic similarities must be evoked by similar environmental [Page 1074] pressures (the pool cue of natural selection, in Darwinian terms), but he stresses the internal basis of inherited common building patterns and materials (1870, p. 42): “Under the term 'homology,' belonging to another philosophy, evolutionists have described and do describe two kinds of agreement — the one, now proposed to be called 'homogeny,' depending simply on the inheritance of a common part, the other, proposed to be called 'homoplasy,' depending on a common action of evoking causes or moulding environment on such homogenous parts, or on parts which for other reasons offer a likeness of material to begin with.”
The foregoing exegesis raises an obvious question: if Lankester restricted homoplasy to independent origin of similar features based on common and phyletically distinctive internal constraints (though not common ancestral structures) in two or more lineages — thus drawing the phenomenon close enough to the essential and defining theme of homology (the “hold of history”) to rank, in Lankester's system, as a subcategory of homology (broadly defined) — then how did the term migrate to the opposite meaning now universally and unambiguously understood today? In other words, how did homoplasy move from a subcategory of homology to become the diametric opposite of homology, with the domain of homology then shrinking to encompass only Lankester's narrower category of homogeny, and the domain of homoplasy expanding to include all similarities evolved independently and not directly inherited from a common ancestral structure?
What looks like an enormous difference — the expulsion of homoplasy as a subcategory of homology (sensu lato), and its establishment as a phenomenon directly contrary to homology (sensu stricto) — actually rests upon a small point: the migration of convergence into the category of homoplasy as now defined. If we decide that the crucial distinction between homology and homoplasy should rest upon common ancestry vs. independent origin, then one important phenomenon, necessarily included within homoplasy by the defining criterion of independent origin for similar structures, shares too much conceptual overlap with homology to permit a clear and comfortable theoretical separation (however firm the descriptive division): independent origin channeled by common internal constraints of homologous genes or developmental pathways — in other words, the phenomenon known as parallelism.
But Lankester originally defined homoplasy exclusively on the basis of phenomena that we would attribute to parallelism (Owen's general and serial homologies). Therefore, for him, homoplasy could legitimately count as a subcategory of homology (sensu lato), even though he recognized that he had to separate homoplasy from homogeny (homology sensu stricto) by the genealogical criterion of common ancestry vs. independent origin. But, if the scope of homoplasy ever expanded to embrace convergences as well — a defendable move because convergences also recor
d an independent origin of similarities — then the combination of parallelisms plus convergences into one category would destroy the conceptual linkage of homoplasy with homology. With the addition of convergence (based on explicit denial of common [Page 1075] internal constraints, and exclusive focus on a common external context of adaptation), homoplasy loses its former common ground with homology (in positing historical hold — whether of structures in homogeny, or of genes, pathways and potential in homoplasy — as the shared causal basis of similar structures in two lineages). Moreover, this expanded category of homoplasy now includes only one universal feature to define its own coherence: independent origin, in both parallelism and convergence. This feature also places homoplasy into antithesis with homology (common ancestry vs. independent origin).
Convergence did join parallelism to build an expanded category of homoplasy, thus setting the opposition that continues to define these terms as an exhaustive and dichotomous division today. Ironically, as a final point, Lankester himself — in a logical inconsistency within his own paper — spawned this dramatic shift in his own intended parsings by adding those “nine fateful little words” (of my title to this subsection) to the end of one statement in his original article. On page 41, as he tries to distinguish his newly formulated concept of homoplasy from the older notion of analogy, he presents (in this single passage) such a broad definition of homoplasy (in the midst of a “generous” attempt to show that analogy must be construed as broader still, and therefore not synonymous with homoplasy no matter how far we extend the concept) that he actually includes independent evolution by convergence — not a subcategory of homology by any stretch of the imagination! — in those nine words at the end (presented in italics below, but not in Lankester's original): “Homoplasy includes all cases of close resemblances of form which are not traceable to homogeny, all details of agreement not homogenous, in structures which are broadly homogenous, as well as in structures having no genetic affinity.”
Once pure convergence had been added to homoplasy via these nine fateful little words, the linkage of homoplasy to homology could no longer be defended — and the two concepts moved from their initial union to their current antithesis.
I recount this story at some length because I know no better way to illustrate the central tension and conceptual confusion within the concept of homoplasy. Parallelism and convergence do share the common descriptive feature of defining an independent origin for similar structures in two lineages. But in causal terms, particularly for assessing the relative weights of formal vs. functional factors in evolutionary change, the conceptual difference could not be more important — for parallelism marks the formal influence of internal constraint, while convergence reflects the functional operation of natural selection upon two substrates different enough to exclude internal factors as influences upon the resulting similarity.
This recognition of internal channeling as the root cause of parallelism — the principal basis for ascribing evolutionary change, and not only limitation, to historical constraint — lies at the heart of evo-devo's theoretical novelty and importance to the Darwinian worldview. This context behooves us to formulate, and to clarify, the causal distinction between parallelism and convergence — [Page 1076] and not just to lump these two principles together by their single common property of specifying an independent origin for similar features in separate lineages. I believe that the history and logic of debate about the meaning of parallelism provide our best path to understanding this important revision in evolutionary theory.
THE TERMINOLOGICAL ORIGIN AND DEBATE ABOUT THE MEANING AND UTILITY OF PARALLELISM. After struggling through dense paragraphs and conceptual thickets, G. G. Simpson (in Haas and Simpson, 1946, p. 325) finally conceded that traditional confusion about the evolutionary meaning of similarity rested upon a logical dilemma, not an absence of empirical data for resolution of a factual issue. Why, to state the dilemma succinctly, does parallelism resist easy fitting into a coherent conceptual structure for the terminology of evolutionary similarity? In particular, why, when parallelism comfortably joins convergence to establish a coherent larger category (called homoplasy) for similar structures evolved independently, have so many good biologists, from the first formulation of the concept until today, continued to “feel in their bones” that something about parallelism veers off towards the supposedly opposite category of homology? Why, to use a vernacular expression often invoked in this discussion (as by Patterson, 1988), does parallelism seem to occupy a “gray zone” between the clear homology of evident retention by common descent, and the clear homoplasy of convergence by selective production of strikingly similar structures (in both form and function) from entirely different points of origin (the “cup coral” shape of rudistid bivalves, prorichthofeniid brachiopods and rugosan corals, for example)?
I have used Simpson's insight to construct the enlarged chart presented as Table 10-1. In an incisive footnote, explicating his differences with coauthor Otto Haas, Simpson makes the logical point that, although homology and homoplasy do cohere as dichotomous opposites encompassing all cases, they bear to each other the odd relationship of a positive claim (A) contrasted with an absence thereof (not-A). Nothing in logic forbids such a taxonomy, but scientists, maintaining a deeply engrained (if unconscious) preference for classification by causes, feel discomfited, in a way that they may not even be able to articulate, about a scheme that contrasts a positive assertion (homology as descent from common ancestry) with its descriptive absence (homoplasy as similarity not by descent from the same structure in a common ancestor). Simpson wrote (in Haas and Simpson, 1946, p. 325):
Homology, as we agree, is best defined as similarity interpreted as due to common ancestry. Homoplasy, as we also agree, is best defined as similarity (or as including any process leading to similarity) that is not explicitly interpreted as due to common ancestry. Both terms rather than being purely descriptive . . . express an opinion, one positive and one negative. Homology expresses an opinion as to how the similarity arose. Homoplasy [Page 1077] expresses an opinion as to how the similarity did not arise, i.e., that it did not arise by homology, but it does not express an opinion as to how the similarity did arise. I do not... see these as alternatives at the same categorical level. The set is not positive, “a” and “b” as mutually exclusive categories, but is a dichotomy of “a” and “not-a.” Under “not-a” it is still possible to have a sequence of alternatives, such as “b,” “c,” etc., that are positive categories on the same level as “a.”
The standard literature does include a venerable term — analogy — that might establish a contrast with homology in the causal sense that wins our almost visceral assent as more satisfactory, with homology as positive A (similarity due to common descent, with no need to invoke direct selective molding), and analogy as oppositely positive B (similarity due to common pressures of natural selection upon backgrounds of no common descent).
We now encounter the logical dilemma that underlies nearly all our extensive and lamentable confusion on this issue. Homoplasy and analogy might strike us, at first, as fully synonymous, for both invoke natural selection as the source of separate evolution for similar structures in two lineages. This synonymy certainly applies for convergence. But homoplasy comes in two flavors: parallelism and convergence — with parallelism as the historical root (in Lankester's original definition of homoplasy), but only convergence carrying the full flavor of synonymy. That is, convergence stands opposite to homology by both criteria — the negative not-A of origin not by common descent, and the positive B of origin by natural selection working in a similar way upon two unrelated substrates.
Unfortunately, a common error of human thinking leads us to define broad and variable categories by their clearest extreme cases. Thus, many scientists have assumed that all homoplasy, whether by parallelism or by convergence, must originate entirely for functional reasons, and not at all by const
raint (the B category of exclusively dichotomous logic); whereas, the “not-A” of independent origin identifies the only property truly required for inclusion within the broad definition of homoplasy. Simpson continues (in Simpson and Haas, 1946, ibid.):
Moreover, the implication is usually present to some degree and it has sometimes been explicitly stated that the structural similarity here in question is not due to homology but is correlated with community of function as opposed to community of ancestry. It is in this sense that analogy is a true alternative (but not the only alternative) to homology as a positive category on the same level, a “b” category rather than a “not-a” category or something on a different level altogether. That is, analogy, when used in this way, expresses a positive opinion, or theory, that a structural resemblance is correlated with function, just as homology expresses the view that it is correlated with common ancestry. Unlike homoplasy, analogy offers an alternative theory as to the basis of the resemblance in question. [Page 1078]
The flavor of parallelism, however, lies in the gray zone of the A-B type classification that our traditions favor (see Table 10-1). In a descriptive sense, parallelism surely ranks within not-A by definition. But when we assess parallelism's relationship to the B category of our alternative scheme, we run right into the aforementioned logical conundrum. Parallelism partakes of B in its invocation of similar selective regimes to produce homoplastic structures from two separate starting points lacking the structure. But parallelism also includes too much A-ness (that is, claims for genuine homology of some sort) to rank as pure B.
The logical solution, had the issue been properly formulated, is not, and never has been, particularly arcane or difficult. Parallelism lies in a gray zone as a consequence of its different status within two conceptual schemes that do not parse nature in exactly the same way, but that we tend to conflate (because both capture important properties of phyletic change) when we consider the meaning of similarity in evolution. Parallelism includes aspects of both constraint and independent selection — not as a wishy-washy mixture in one grand pluralistic glop of all-things-for-all-people, but in rigorously different parsings for different levels of consideration (again, as evident in Table 10-1. Several authors have stressed the dependency of these terms on the hierarchical
The Structure of Evolutionary Theory Page 171