The Structure of Evolutionary Theory

by Stephen Jay Gould

  [Page 1060]

  vertex becomes the locus of canonical causes, while contributions from the other two vertices become constraints upon full determination by general laws of nature.

  Under this theory (Fig. 10-11b), adaptive form arises from the operation of general laws upon biological materials. But if, to understand any current ad­aptation, we need to invoke strictures based either upon: (1) passive inheri­tance within a specifically designated genealogical system (a constraint from the historical vertex, imposed by a unique and contingent biological particu­lar, thus detracting from a claim for full causation by general laws); or (2) upon the immediate construction of a particular adaptation by a biological process tied to specifics of adaptive pressures in one environment at one time (a constraint from the functional vertex of current natural selection) — then the full power of the purely physical model becomes compromised. Thus, as I show in Figure 10-11b, a purely physicalist theory for the direct generation of adaptive form by spatiotemporally invariant laws of nature, places its canoni­cal mechanism at the structural vertex, and regards inputs from both the historical vertex (strictures from past particulars) and the functional vertex (strictures imposed by the specifics of current biological situations) as con­straints.

  Finally (Fig. 10-11c), a pure (and caricatured) cladist, who believes that the reconstruction of genealogical pattern (without reference to modes of causa­tion) defines the goal and purpose of evolutionary biology, would locate his canonical mechanisms at the historical vertex, and view contributions from the other two vertices as constraints upon his ability to detect a pure genea­logical signal in the currently adaptive traits of organisms. Influences from the structural vertex must be counted as constraints because their timeless generality covers or distorts the desired signal of particular history with an unwanted contribution from causes with no specific genealogical content. And influences from the functional vertex impose a confusing immediate par­ticular — an autapomorphy offering no help at all in the reconstruction of lin­eages, and therefore conventionally omitted in cladistic analysis — degrading a phyletic signal that might otherwise map the organism's position in the ge­nealogical system of a more general lineage. (This insight about particular and immediately adaptive features — autapomorphies in cladistic terminology — has long been regarded as a truism in taxonomic practice. Darwin himself frequently emphasized (1859, chapter 13) that new and unique adapta­tions can only confound taxonomic relationships, and that systematists must privilege characters with broad homological residence in the taxa of larger genealogical groupings.)

  Thus, the cardboard Darwinian functionalist, the cardboard physical structuralist, and the cardboard genealogical cladist each chooses a different vertex for canonical causation, and must then define influences from the other two vertices as constraints upon the efficacy of his orthodox mecha­nism. My examples are purposefully cartoonish, but the principle thus illus­trated represents an important, and insufficiently appreciated, generality in [Page 1061] science — the theory-bound nature of terms, and particularly, in this case, the designation as “constraint” of all evolutionary causes lying outside the range of orthodox mechanisms, thereby compromising their power and generality. We should regard this terminological notion of constraint as positive in its ca­pacity to question accepted ways of thinking — the theme that shall now structure the remainder of this chapter.

  Deep Homology and Pervasive Parallelism: Historical

  Constraint as the Primary Gatekeeper and Guardian of

  Morphospace

  A HISTORICAL AND CONCEPTUAL ANALYSIS OF THE UNDERAPPRECIATED IMPORTANCE OF PARALLELISM FOR EVOLUTIONARY THEORY

  A context for excitement

  The last chapter of my first book, Ontogeny and Phylogeny, published in 1977, amounted to little more than a terminal exercise in frustration. I had written 500 pages on the history and evolutionary meaning of heterochrony, and had then been stymied, at the point of potential synthesis, by an inability to relate the well-documented (and reasonably well understood) subject of macroevolutionary changes wrought by shifts in developmental timing to any viable analysis (or even description) of the underlying genetic and embryological mechanisms.

  I could only wave my hands and write a few vague paragraphs about the putative importance of “regulatory” genes — then an almost purely abstract concept (at least for eukaryotic development), based on no direct documen­tation of any worth, and supported only by three inferential forms of argu­ment: analogies to rudimentary knowledge about the different systems of prokaryotic regulation (primarily the work of Jacob and Monod); general models suggesting the necessity of a regulatory hierarchy, with some genes operating as primary controls on rates and placements of structural genes and their products (Britten and Davidson, 1971); and such conclusions by negative inference as King and Wilson's (1975) famous calculation of more than 99 percent identity between human and chimp polypeptides, im­plying that the considerable phenotypic differences between the two species must therefore reside in the action of a small class of unknown regulatory genes.

  Of course, this frustration only recorded a technological inability to specify these regulators, not any failure to grasp the centrality of the subject. I wrote (1977b, p. 406): “The most important event in evolutionary biology during the past decade has been the development of electrophoretic techniques for the routine measurement of genetic variation in natural populations. Yet this imposing edifice of new data and interpretation rests upon the shaky founda­tion of its concentration on structural genes alone (faute de mieux, to be sure; [Page 1062] it is notoriously difficult to measure differences in genes that vary only in the timing and amount of their products in ontogeny, while genes that code for stable proteins are easily assessed).”

  Barely 20 years later, this statement reads like a quaint conceptual fossil from an “ancient” time of crossbows and arquebuses, when we could only re­construct the anatomy of genes from their protein products (and could not recognize regulatory genes that did not deposit such results in explicit flesh and blood). I therefore succumbed to the necessity of technical limits and ended a long book with the weakest of conclusions — a future hope, however heartfelt and (in retrospect) accurately surmised: “I believe that an under­standing of regulation must lie at the center of any rapprochement between molecular and evolutionary biology; for a synthesis of the two biologies will surely take place, if it occurs at all, on the common field of development” (Gould, 1977b, p. 408).

  Now, as I begin this chapter in the summer of 1999, I can only express both my joy and astonishment at a subsequent speed of resolution and discovery that has sustained my predictions, but also made my earlier book effectively obsolete, not only within my own lifetime, but during my active mid-career. The field of evolutionary developmental biology (known as “evo-devo” to practitioners), while still in its infancy, has invented the tools — and already cashed out a host of stunning and unexpected examples — for decoding the basic genetic structure of regulation, and for tracing the locations and timings of regulatory networks in the early development of complex multicellular creatures.

  But this very pace of growth and excitement presents a problem for a book like this, with a “lead time” measured in months to years, rather than the professional journal's weeks to months or the popular press's days. The dis­coveries of deep homology and pervasive parallelism among phyla separated for more than 500 million years continue to accumulate at an accelerating pace, based on methodological refinements and extensions, in both speed and accuracy that could hardly have been conceptualized even a decade ago.

  This situation places me in a quandary (although I could hardly imagine a happier form of puzzlement). The data of evo-devo constitute the largest and most exciting body of novel empirics to support this book's general thesis. Since I have tried to provide thorough overviews of empirical documentation for other central elements of my overall theory, I should now be tabulating and evaluating th
ese cases of deep genetic homology in extenso. But I am hoist by my own petard of emphasis on appropriate scales. The data of evo-devo accumulate and improve at such a pace that any thought of a “review article” written more than two years before anticipated publication can only be regarded as absurd. In other words, this book's timescale of production must be labeled as geological compared with a pace of discovery that can only be measured in ecological time.

  I will therefore adopt the following strategy as appropriate to the circum­stance. I will exemplify the best and most informative of current empirical cases, but I ask readers to heed the following label of warning: “I wrote this [Page 1063] section in the closing months of the last millennium. The cases discussed herein represented a 'state of the art' at this historical moment. This 'state' will be obsolete and superseded by the book's publication, but I am confident that the general themes and directions will hold and grow. Please consider the empirical discussion as exemplification, not as fulfillment.”

  However, the timescale of this book also permits a luxury not afforded to authors of journal articles. For I can balance this guaranteed empirical super­annuation against a discussion of general significance that, if properly situ­ated within this book's broader subject of the history and structure of macro-evolutionary theory, may succeed in exemplifying the signal importance of evo-devo in changing and expanding our basic conception of evolutionary causality (even while I must fail to capture what the favored cliche of the mo­ment calls the “cutting edge” of actual discovery). I will therefore focus my treatment of evo-devo upon some crucial issues in the structure of evolu­tionary theory — all rooted in the concept of “constraint” in relationship to natural selection — that have frequently been overlooked, bypassed, or short­changed in the midst of immediate excitement generated by the novel data of this burgeoning field.

  What features generally lead scientists to strongly shared feelings about the unusual importance of a set of discoveries? We might nominate sheer novelty as an initial, base level property — especially when enhanced by a conquest over nature's previous taunt to scientists: you know where to look in theory, but you haven't developed the proper tools for perception. (The canonical ex­ample of such rare triumphs, Galileo's Sidereus nuncius of 1610, comes inevi­tably to mind — a mere “pamphlet” that packed more oomph per paragraph than any other document in the history of printing. After all, the first tele­scopic look at a previously invisible cosmos necessarily “skimmed off” a set of magnificent and unexpected novelties, including the composition of the Milky Way as a sea of stars, the satellites of Jupiter, the phases of Venus, and the topography of the Moon.)

  Much of our fascination with the data of evo-devo arises from the sheer novelty of discovery in biological domains that had been previously and to­tally inaccessible. These empirical gems also illustrate, even in these early days, the integrating power of scientific conclusions to translate a previous descriptive chaos into explanatory sensibility. As an example, consider the name given to the truly elegant theory of floral genesis, as developed by stu­dents of Arabidopsis, the “Drosophila” of angiosperm biology — the ABC Model (Coen and Meyerowitz, 1991; Weigel and Meyerowitz, 1994; Jurgens, 1997; Busch, Bomblies, and Weigel, 1999; Wagner, Sablowski, and Meyero­witz, 1999).

  In this elegantly simple model (see Fig. 10-12), based on genes with homeotic effects upon serially repeated structures arranged in systematic order (with repetition in concentric whorls rather than linearly along a body axis), A genes operating alone determine the form of the outermost whorl of leaf-like sepals; A plus B genes regulate petals in the next whorl within; B plus C genes mark the male stamens, while C genes working alone determine the [Page 1064] most interior female carpels. Moreover, leafy, a “higher control” gene previ­ously recognized as an initiator or suppressor of floral growth and placement in general (Weigel and Nilsson, 1995), apparently also regulates the more specific operation of the ABC series. (Busch et al., 1999, demonstrate that a protein produced by leafy bonds directly to a particular DNA segment of a C gene responsible for the generation of carpels.)

  This model enjoys obvious significance for the full gamut of evolutionary issues, ranging from the most theoretical (in “updating” Goethe's formalist theory (see pp. 281–291) that all parts regulated by the ABC series conform to a generalized “leaf” archetype), to the most practical (hopes of florists to enhance AB interactions and grow flowers with larger and more numerous petals). But in the present context, I merely wish to highlight a linguistic point: the selected terminology of ABC surely encapsulates the accurate im­pressions (and the excitement) of researchers who recognize their role as pio­neers engaged in the construction of a basic alphabet for a new understanding of nature.

  The pure discoveries of evo-devo may fit the heroic image of science as con­queror of previous ignorance (the tabula rasa model of light upon previous darkness or, literally, the first writing on a blank slate). But the most stunning of scientific novelties surely gain their status by virtue of their unexpected or surprising character — that is, their failure to match, or even their power to mock, the anticipated constitution of a part of the natural world previously

  10-12. The elegant simplicity of the ABC model, from Weigel and Meyerowitz, 1994. In this model, based upon the genetics and development of Arabidopsis, the four circlets of sepals, petals, stamens, and carpels achieve their distinctive forms under the following influences, as shown in the diagram. A genes deter­mine the development of sepals; A and B petals; B and C stamens; and C alone, carpels.

  [Page 1065]

  inaccessible to investigation. (The most stunning property of Jupiter's four large moons, when first seen by Galileo, lay not in their mere existence, but in the recognition that their revolution about the planet would fracture the crys­talline sphere that, in the “certain” knowledge of previous views, marked Ju­piter's domain in a geocentric universe — and that such a sphere, therefore, could not exist.) In the same manner, the central significance of our dawning understanding of the genetics of development lies not in the simple discovery of something utterly unknown (the ABC floral model, or the specification of anteroposterior differentiation by arthropod Hox genes), but in the explic­itly unexpected character of these findings, and in the revisions and exten­sions thus required of evolutionary theory.

  The discovery that has so discombobulated the confident expectations of orthodox theory can be stated briefly and baldly: the extensive “deep homology” now documented in both the genetic structure and developmental archi­tecture of phyla separated at least since the Cambrian explosion (ca. 530 mil­lion years ago) should not, and cannot, exist under conventional concepts of natural selection as the dominant cause of evolutionary change. Natural se­lection must therefore operate in a context of far greater constraint (in both the “negative” sense of limits upon freedom to craft particular adaptive solu­tions, and in the “positive” sense of synergism in the specification of preexist­ing or preferred internal channels) than the usual functionalist characteriza­tions of Darwinian theory envisage.

  I am not trying to construct straw men or cardboard images for easy demo­lition. Of course, no good Darwinian naturalist ever conceptualized organic matter as pure putty molded by natural selection to local optimality. The hold of phenotypic homology has always fascinated evolutionary biologists and served as the basis for classification and phylogenetic reconstruction. Even the most orthodox Darwinian systematists have always recognized that “putty-like” characters — maximally labile and malleable by natural selec­tion in an unconstrained way — must be shunned in phyletic reconstruction (as sources of autapomorphic traits and manifestors of convergence), while taxonomists must base their hierarchical orderings on nested levels of homo-logical retention among related taxa.

  But two classical views about homology have traditionally served to inte­grate this cardinal principle of historical constraint with a functionalist the­ory of evolutionary mechanisms.
First, as previously discussed in more de­tail (pp. 251 and 1058), Darwinian biology attributes the origin of shared homologous characters to ordinary adaptation by natural selection in a com­mon ancestor. Moreover, homologous characters not only continue to ex­press their adaptive origin, but also remain fully subject to further adaptive change — even to the point of losing their ready identity as homologies — if they become inadaptive in the environment of any descendant lineage. Homological similarity in related taxa living in different environments there­fore indicates a lack of selective pressure for alteration, not a limitation upon the power of selection to generate such changes. (At the Chicago Macroevolution meeting of 1980, for example, Maynard Smith acknowledged the allometric [Page 1066] basis of many homologies, but stated that the attribution of such simi­larity to “developmental constraint” would represent what he proposed to christen as the “Gould-Lewontin fallacy” — for natural selection can unlock any inherited developmental correlation if adaptation to immediate environ­ments favors such an alteration.)

  Second, homological holds must be limited in taxonomic and structural ex­tent to close relatives of similar Bauplan and functional design. The basic ar­chitectural building blocks of life — the DNA code, or the biomolecular struc­ture of fundamental organic compounds, for example — may be widely shared by homology among phyla. But the particular blueprints of actual designs and the pathways of their construction — the form of the Gothic cathedral rather than the chemical formula of calcite in the facing stone (see pp. 1134–1142 for extensive treatment of this point) — must be limited to clades of closer relationship. The identical topology of bones in mammalian forearms of markedly different utility (the whale's flipper, the horse's leg, the bat's wing, and my typing, or literally manipulating, digits) can be homologous, but we expect no comparable hold of history upon the more generally similar segmentation of arthropod metameres and vertebrate somites (not to mention the non-homology of bones in mammalian forearms and teleost forefins).