The Structure of Evolutionary Theory

by Stephen Jay Gould

  Any wider hold of homology would have to inspire suspicions that the cen­tral tenet of orthodox Darwinism can no longer be sustained: the control of rates and directions of evolutionary change by the functional force of natural selection. In a particularly revealing quote within the greatest summary docu­ment of the Modern Synthesis, for example, Mayr (1963, p. 609) formulated the issue in a forthright manner (see p. 539 for previous discussion of this statement). After all, he argued, more than 500 million years of independent evolution must erase any extensive genetic homology among phyla if natural selection holds such power to generate favorable change. Adaptive evolution, over these long intervals, must have crafted and recrafted every genetic locus, indeed every nucleotide position, time and time again to meet the constantly changing selective requirements of continually varying environments. At this degree of cladistic separation, any independently evolved phenotypic similar­ity in basic adaptive architecture must represent the selective power of sepa­rate shaping by convergence, and cannot record the conserved influence of re­tained genetic sequences, or common generation by parallelism: “In the early days of Mendelism there was much search for homologous genes that would account for such similarities. Much that has been learned about gene physiol­ogy makes it evident that the search for homologous genes is quite futile ex­cept in very close relatives.”

  But we now know that extensive genetic homology for fundamental fea­tures of development does hold across the most disparate animal phyla. For an orthodox Darwinian functionalist, only one fallback position remains via­ble in this new and undeniable light (and Ernst Mayr, vigorous as ever at age 95 as I write these words, would be the first to welcome this illumination wholeheartedly, and to laugh at his old cloudy crystal ball on this single is­sue). One can admit the high frequency and great importance of such genetic [Page 1067] constraints (and also designate their discovery as stunningly unexpected), while continuing to claim that natural selection holds exclusive sway over evolutionary change because deep homology only imposes limits upon styles and ranges of developmental pathways, but cannot power any particular phyletic alteration. Natural selection can still reign supreme as the pool cue of actual evolutionary motion.

  But a formalist defender of positive constraint will reply that such unantici­pated deep homology also channels change in positive ways — and that the key to this central argument resides in an old distinction that, unfortunately, cannot be matched for both conceptual and terminological confusion, and for consequent failure of most evolutionists to engage the issue seriously: namely, the differences in causal meaning (not just in geometric pattern) be­tween parallelism and convergence. The next section shall treat the history and logic of this issue in detail, but I shall first present the following basic for­mulation in relevant terms of balances between constraint and selection:

  Even the most committed adaptationist would not deny that the indepen­dent evolution of similar phenotypic features (in both form and function) in two closely related lineages may be facilitated by the presence, in both ances­tors, of the same genes and developmental pathways inherited from a recent common ancestor. (The independently evolved features of these two lineages cannot be called homologous on basic definitional grounds, but the features may still be built by homologous genes and along homologous developmental pathways.)

  For example, no adaptationist would be fazed by the suggestion that rela­tive increase in antler size within two separate cervid lineages undergoing phyletic increase in body size occurred because both lineages retained an an­cestral allometry that may well be homologically pervasive within the Cervidae (Huxley, 1932, for this classic case of positive allometry). Inherited constraint may set a preferred channel, but selection must still guide any lin­eage into such an internally biased path. So a functionalist may view such un­deniably positive constraints as, at most, helpmeets or facilitators of natural selection, while continuing to regard selection as a necessary instigator, and therefore as the primary cause of change.

  But — and now we come to the nub of the issue, and to the central role of positive developmental constraint as a major challenge to selectionist ortho­doxy — the attribution of similar evolutionary changes in independent lin­eages to internal constraint of homologous genes and developmental path­ways, and not only to an external impetus of common selective pressures, must be limited to very close relatives still capable of maintaining substantial genetic identity as a consequence of recent common ancestry. Mayr's charac­terization of selectionist orthodoxy comes again to mind: distantly related lineages cannot be subject to such internal limitation or channeling because the pervasive scrutiny and ruthless efficiency of natural selection, operating on every feature over countless generations in geological immensity, must have fractured any homological hold by underlying genes and developmental pathways over the freedom of phenotypes to follow wherever selection leads. [Page 1068] Darwin's famous words, so often quoted, haunt the background of this dis­cussion (1859, p. 84): “It may be said that natural selection is daily and hourly scrutinising, throughout the world, every variation, even the slightest; rejecting that which is bad, preserving and adding up all that is good; silently and insensibly working, whenever and wherever opportunity offers, at the improvement of each organic being in relation to its organic and inorganic conditions of life.”

  Therefore, any uncannily detailed phenotypic similarity evolved between distantly related groups must arise by convergence from substrates of nonhomologous genotypes — thus affirming our usual view of selection's over­arching power, especially if common function for the two similar forms can validate the hypothesis of generation within a comparable adaptational ma­trix. (Note the logical danger of circularity that intrudes upon the argument at this point, for this extent of detailed similarity — the very datum that, in an unbiased approach, would lead one to entertain parallelism based upon com­mon internal constraint as a viable alternative to convergence based on simi­lar adaptive needs — now becomes an a priori affirmation of selection's power, the hypothesis supposedly under test.)

  For this reason, such detailed functional and structural similarities, evolved independently in distantly related lineages, have become “poster boy” exam­ples of convergence — itself the “poster boy” phenomenon and general con­cept for showcasing selection's dominant sway — precisely because similarities evolved in this mode cannot, by Mayr's argument, be ascribed to parallelism based on positive constraint imposed by homologous genetic and develop­mental pathways. With internal channeling thus theoretically barred as a po­tential source of impressive similarity, convergence becomes the favored ex­planation by default. The argument, surely “tight” in logic and principle, seems incontrovertible.

  Since I do not wish to dwell on the previous errors that we all committed on this issue, let me simply illustrate the older view, and the magnitude of cur­rent reversal, with one of my own mistakes — from a 1976 paper extolling convergence as evolutionary biology's closest natural analog to replication in the experimental sciences (Gould, 1976, p. 177):

  The convergent evolution of similar structures fulfills, at least imper­fectly, the criterion of independent replication that any experiment re­quires. An adequate theory of functional morphology must explain adaptive design by studying how different organisms react to the same selective regime. If we want to know whether plate tectonics is a true, universal physics of large bodies or only a descriptive account of this planet's history, other planets must be studied. If we want to know whether the biochemical unities of all life on earth have general import as optimal designs (given the nature of universal chemistry) or merely re­flect the monophyletic origin of life on earth (and the homologous status of ATP and left-handed amino acids), then we shall have to hope for life on Mars. If, to retreat to something more immediate, one wishes to assess [Page 1069] the functional limits or mechanical constraint upon the human eye, one would do well — as J. Z. Young and others have done — to study the oct
opus.

  But, as I shall discuss on pp. 1123–1132, one of the major discoveries of evo-devo has revealed a deep genetic homology underlying and promot­ing the separate evolution of lens eyes in cephalopods and vertebrates. The overt phenotypes do record substantial convergence (for different body tis­sues build corresponding structures in the two groups), but both phyla share key underlying genes and developmental pathways as homologies, and the example has lost its former status as the principal textbook case of natural se­lection's power to craft stunning similarities from utterly disparate raw mate­rials. Eyes of such strikingly similar design owe their independent origin as much to genetic and developmental parallelism, based on internal constraints of homologous genes and developmental pathways, as to selection's capacity for iterating nearly identical adaptations from scratch by convergence.

  With this “one liner” of maximal force — evo-devo has reinterpreted several textbook examples of convergence as consequences of substantial parallel­ism — we can encapsulate the depth of theoretical disturbance introduced by this subject into the heart of Darwinian theory. Our former best examples of full efficacy for the functional force of natural selection only exist because in­ternal constraints of homologous genes and developmental pathways have kept fruitful channels of change open and parallel, even in the most disparate and most genealogically distant bilaterian phyla. The homological hold of historical constraint channels change at all levels, even for the broadest pat­terning of morphospace, and not only for details of parallel evolution in very closely related groups.

  A terminological excursis on the meaning of parallelism

  THE NINE FATEFUL LITTLE WORDS OF E. RAY LANKESTER. The transforming power of this discovery upon evolutionary theory would stand out more clearly if the key terms and concepts had not become so muddled in our literature, and therefore so widely misunderstood or disregarded by modern researchers. (This situation cannot validate the graybeard's perennial lament: “them young fellers just don't keep up with the views of the older guys, like we did when we wuz gettin' started.” The concepts and terminology sur­rounding the origin and status of similar structures in different lineages have inspired particular difficulty and unclear thinking ever since Darwin, and even before. In their classic paper on the subject, still the best treatment ever published, Haas and Simpson (1946) devoted the bulk of their long text to the history of confusion over differences between parallelism and convergence — with the two authors finally agreeing to disagree about the most fruitful definitions, even as they resolved the conceptual confusions.)

  We should begin by recalling a central distinction that we all know, and probably all regard as refreshingly free from conceptual ambiguity: the differ­ence between homology and homoplasy. Homologous structures are similar [Page 1070] by inheritance from a common ancestor. Homoplastic structures are simi­lar by independent evolution, for we can infer that the common ancestor did not possess the structure. In other words, the dichotomy of these two terms captures the essential difference between common ancestry and independent origin. At a first level of interpretation (but here we immediately plunge a toe into troubled waters, as we shall soon see), the dichotomy also marks a conceptual distinction between the hold of history and the power of adap­tation.

  So far so good — and I will not challenge the accepted and codified current definition of these two terms for describing an important logical distinction in evolutionary biology. But we often gain better understanding — and do not merely indulge an antiquarian passion for trivial and superannuated detail — when we explore the historical origin of a word, and then discover a marked discrepancy between initial and current usage. I intend no criticism of current usage in making such an observation. Words change their meanings, just as organisms evolve. We would impose an enormous burden upon our economy if we insisted upon payment in cattle every time we identified a bonus as a pe­cuniary advantage (from the Latin pecus, or cattle, a verbal fossil from a for­mer commercial reality).

  E. Ray Lankester, T. H. Huxley's protege and the finest evolutionary morphologist in the generation just after Darwin, proposed the concept of homoplasy in 1870 (see Lester and Bowler, 1995, and Gould, 1999a, on Lankester's life and general views). I suspect that most evolutionary biologists could cite Lankester as a source, but I will wager a substantial sum that very few colleagues could identify (to their pecuniary benefit) a supreme irony in Lankester's original paper, entitled “On the use of the term homology in modern zoology, and the distinction between homogenetic and homoplas­tic agreements” — namely, that he defined homoplasy as a subcategory of homology, in apparent defiance of current usage (which, I repeat, I do not challenge) of homology and homoplasy as dichotomous opposites. The rea­sons for his distinctions, and for subsequent changes and refinements of meaning, tell an interesting story that can unlock the essential distinction between parallelism and convergence, and also explain the significance of evo-devo in unleashing the capacity of parallelism to rebalance formalist and functionalist causes within evolutionary theory.

  Richard Owen enjoyed the height of his influence when Lankester wrote his paper, and the younger morphologist properly went to the source of all later usage (Owen, 1848 and 1849) in defining his terms. Lankester, as a Dar­winian acolyte, also correctly noted the philosophical difficulty facing anyone who sought to translate such vital terms as homology into the new evolution­ary context. Owen, as Lankester notes, defined homology within a Platonic theory of archetypal form (see discussion of Owen's concepts on pp. 312–329). How can the term be carried over into Darwin's world? Lankester (1870, p. 34) began his paper by stating this problem:

  Whilst the adoption of the theory of evolution has broken down the no­tions at one time held by zoologists and botanists as to the existence of [Page 1071] more or less symmetrical classes and groups in the organic world, estab­lished by some inherent law of Nature which limited her productive powers to arbitrary special plans or types of structure, and has taught us to see, in the variously isolated and variously connected kinds of animals and plants, simply the parts of one great genealogical tree, which have become detached and separated from one another in a thousand differ­ent degrees, through the operation of the great destroyer Time, yet certain terms and ideas are still in use which belonged to the old Platonic school, and have not been defined afresh in accordance with the doctrine of descent.

  In particular, Owen had specified three categories of homology: special, general, and serial. (His classical and definitive 1848 treatise, “On the arche­type and homologies of the vertebrate skeleton,” comprises three chapters, ti­tled “special homology,” “general homology,” and “serial homology” respec­tively.)

  Owen's famously vague and broad definition of “homologue” as “the same organ in different animals under every variety of form and function” (1848, p. 7, repeated from 1843, p. 374) invokes a Platonic notion of sameness as “proceeding from a common archetype.” Lankester had the good sense and vision to recognize (and we continue to assent today) that this concept did enjoy philosophical coherence, and could be translated into evolutionary terms — but that the Darwinian version implied different distinctions, requir­ing a subdivision of meanings and significances within a general notion that remained usefully unitary.

  Owen's three categories share tighter bonding in the idea that parts can be called homologous so long as they can be construed as expressions or em­bodiments of the same idealized archetype (a key pre-evolutionary notion of formalist, as opposed to functionalist, thinking, and therefore particularly difficult to translate into a functionalist theory of evolution like natural selec­tion). Obviously, as a first pass for evolutionary translation, we should rede­fine the Owenian archetype as the Darwinian common ancestor — thus substi­tuting the real flesh and blood of physical continuity for a Platonic notion of formal identity. We can then proceed, as Lankester notes, with evolutionary versions of Owen's three categories.

  For Owen (1848, p. 7)
, special homology refers to “the correspondence of a part or organ, determined by its relative position and connections, with a part or organ in a different animal.” In evolutionary terms, we regard these two parts (in two different organisms) as homologous because they descend from the same feature in a common ancestor. Lankester (1870, p. 36, first paragraph) recognized this criterion of common ancestry as paramount — the definition that “without doubt the majority of evolutionists” would assign to the concept of homology. Lankester proposed — although his name never took hold — that this aspect of Owen's broader concept (“special homology”) be called homogeny (or homology sensu stricto.) What then becomes of Owen's other two categories?

  Owen defined general homology as “a higher relation ... in which a part [Page 1072] or series of parts stands to the fundamental or general type, and its enuncia­tion involves and implies a knowledge of the type on which a natural group of animals, the vertebrate for example, is constructed” (1848, p. 7). This idea that we can assert a form of homology between two parts (in two organ­isms) because both express the same general archetype, rather than because one part can be designated as the “same” as the other part (as in special homology), does not translate by a similar criterion of descent from common ancestry because, as Owen noted, general homology records “a higher rela­tion.” For example, Owen regarded arms, legs, and heads for that matter, as derivations from a common vertebral archetype. Thus, the forearm of an aardvark is a general homolog of my leg (not to mention the shrew's head and the whale's flipper).