The Structure of Evolutionary Theory

by Stephen Jay Gould

  Distinguishing and sharpening the two great questions

  THE STRUCTURAL VERTEX. I shall begin with the simple and more direct question posed from the structural vertex: does the existence of current adap­tations necessarily imply their functional origin at all (either as a direct re­sponse to current environments, or by inheritance of traits with a function­alist origin at their ancestral inception)? How, in other words, can good Darwinian design (aptations in my favored restriction, adaptations in the ver­nacular) arise by processes that do not involve functional adaptation?

  Since this category is defined negatively — to designate causes of functional characters not evolved by functionalist mechanisms — the structural vertex [Page 1054] becomes something of a miscellaneous repository, as Seilacher (1970) recog­nized for the same end member (that he called bautechnischer, or archi­tectural). In particular, the structural vertex includes two strikingly differ­ent subcategories, united in their common appeal to physical consequences rather than functional crafting, but otherwise strikingly disparate, even dia­metrically opposite, in their implications for a key question, admittedly sub­sidiary to our present inquiry about structural vs. functional origin, but of central importance to evolutionary study in general: namely, the role of his­tory and contingency in the interpretation of evolutionary lineages.

  I shall consider these two structural categories seriatim, and in depth, in Chapter 11, and will therefore only present the basic conceptual framework here. In the first category, some adaptive features of organisms may be di­rectly molded by, or may originate as immediate and deterministic conse­quences of, the physical properties of matter and the dynamical nature of forces — in other words, not by an accumulative process of functional honing through selection, and not (for that matter) by any uniquely biological pro­cess at all. When Williams (1966) famously, if a bit facetiously, remarked that we shouldn't consider a flying fish's capacity to fall back into the water as an adaptation because their descent represents a necessary consequence of physi­cal mass — even though this capacity may be vital to the continued life of the fish, and therefore strongly aptive — he invoked a direct physical property of matter (not subject to alteration by selection at all in this case!).

  D'Arcy Thompson's (1917, 1942) theory — that physical forces directly im­pose an optimal biomechanical form upon plastic organic material — marks the admittedly idiosyncratic locus classicus for this general attitude (see pp. 1179–1208). Stuart Kauffman's (1993) interesting concept of “order for free” (good design automatically generated by nature's laws, with no need for laborious construction by a particular biological process like natural selec­tion) provides the most fruitful current context for this approach to aptive or­ganic design.

  In the second category that Darwin designated as “correlations of growth,” and Gould and Lewontin (1979) called spandrels, features arise nonadaptively as physically necessary consequences of other changes that may (and, in all probability, usually do) have an adaptive basis, or as inevita­ble and unselected sequelae of general organic designs (that, again, generally arise for conventional functional reasons).

  Nonselected origin for structural reasons defines the common ground of the two categories — directly generated by physical forces in the first, indi­rectly developed as correlated consequences in the second. But the philosoph­ical implications of these two bases could not be more different in one crucial respect — hence the oppositional stance often adopted between champions of the two modes, despite their acknowledged common ground at the vertex of nonselected origin by physical necessity.

  Pure D'Arcy Thompsonians maintain little interest in history and phylogeny, and may even become overtly hostile to the commanding influence of these concepts in evolutionary biology. After all, if a trait arises by physical [Page 1055] necessity, why should we care about the specific contingencies that brought this or that lineage into the domain of the particular physical law under study. At any time, any lineage located in this domain must behave in the same way. This version of structuralism embraces the classical spatiotemporal invariance of natural law, and cares little (if at all) about historical pathways that happen to potentiate the law's operation in any particular case. Most evolu­tionists (including the author of this book) are historians at heart, and must view such derisory dismissal of phylogeny as anathema, however fascinating they find (as I do) the partial validity of this theme, and however much they may admire (and bravo again from this observer) the inimitable power of D'Arcy Thompson's prose style.

  Spandrelists, in strong contrast, generally share the evolutionary biologist's traditional fascination for contingent details of history in individual lineages under study. Spandrels do express general and predictable properties, but they originate as necessary consequences of particular triggers that can only be understood in a historical and phylogenetic context. If Julia Pastrana grew two rows of teeth as a correlated consequence of her abnormal hairiness (see p. 338 for a discussion of this example from Darwin's writing), then the forced correlation, set (in Darwin's view) by the constraining homology of hair and teeth, records and reflects the phyletic uniqueness of mammalian de­velopment (not the operation of invariant, universal laws), even if the extra teeth grew by enforced physical necessity. And even though the spandrels of San Marco must be built once the architects decide to mount hemisphaerical domes on four adjacently orthogonal rounded arches, we can only under­stand the basic blueprint that necessarily engendered the spandrels by study­ing the particular history of ecclesiastical architecture.

  THE HISTORICAL VERTEX. The structural vertex poses a direct question about the origin of currently adaptive features themselves: what percentage of items in this category did not originate by a process of adaptation, but were coopted for present utility from non-adaptive beginnings? If we can de­termine a high relative frequency in general, or even if we could only specify a subset of crucial evolutionary situations for such nonadaptive origins, then an exclusively adaptationist theory for the genesis of aptive structures will no longer suffice, and evolutionary theory will require enrichment from struc­turalist alternatives promoted to a more than marginal or peripheral status.

  The historical vertex, on the other hand, poses a more indirect challenge that might better be designated as a metaquestion: Given a functional origin for presently adaptive features (either by immediate construction for a cur­rent role, or by adaptive origin in an ancestor, with subsequent maintenance by homology in descent), may we also regard the markedly inhomogeneous distributions of organisms across the potential morphospace of good organic design as a best set of solutions to functional problems, or do we need to in­voke internal constraints and channels to explain substantial aspects of this decidedly “clumped,” and decisively non-random, occupation of a theoreti­cal “design space”? [Page 1056]

  In other words — and I label the inquiry as a “metaquestion” for this rea­son — in what ways does the skewed and partial occupancy of the attainable morphospace of adaptive design record the operation of internal constraints (both negative limitations and positive channels), and not only the simple failure of a limited number of unconstrained lineages to reach all possible po­sitions in the allotted time? (Geological history may be long, and the number of evolutionary lineages immense, but even these substantial quantities must be risibly small compared with the number of spatiotemporal positions in po­tentially “colonizable” morphospace.)

  In attempting to explain such non-random clumping in adaptive morpho­space, Darwinians have traditionally emphasized the contingency of limited time and numbers — rather than any failure to populate accessible regions as a consequence of active constraint — because their functionalist theory presup­poses the power of natural selection to break such constraints (whose exis­tence, needless to say, they cannot and do not wish to deny), and the conse­quent accessibility (at different levels of effort and probability, to be sure) of all physically possible adaptive designs.


  If the influence of historical constraints must be integrated with the con­ventional mechanism of unfettered adaptive exploration (limited by accidents of historical opportunity) to explain the markedly non-random clumping of actual organisms in the potential morphospace of adaptive and theoretically accessible organic form, then this metaquestion about nonfunctional causes for the distribution of adaptive features poses a different kind of challenge to our usual views about the power and range of natural selection in the expla­nation of functional design.

  In my presentation thus far, I have epitomized this crucial issue in an ab­stract way, but I shall, in the final section of this chapter, present the empirical results — primarily from the burgeoning study of genetic bases for the major developmental patterns of organic Bauplan (“evo-devo,” or evolution of de­velopment to its devotees) — that have so surprised the biological sciences in recent years, bringing this new study of ancient themes to the forefront of our science.

  I shall argue that two prominent discoveries have magnified the importance of historical constraints vs. the free operation of natural selection to a point where this historical aspect of constraint can no longer be denied prominence in designating the causes of evolutionary change: First, “deep homology” or the discovery that major phyla, separated by more than 500 million years of independent evolutionary history, still share substantial (if not predominant) channels of development based on levels of genetic retention that proponents of the Modern Synthesis had specifically declared inconceivable, given the presumed power of natural selection to modify any independent line in its own uniquely adaptive direction. Second, the importance of parallelism (a concept rooted in internal constraint) for explaining independently evolved features of distant phyla; traits long touted as textbook examples of convergence (a concept rooted in externally conditioned adaptation).

  I shall also argue that deep homology often embodies the “negative” empirical [Page 1057] theme of constraints as limitation, while parallelism features the “positive” empirical theme of constraints as enabling channels. Both themes, however, have forced evolutionary biologists to reassess the importance of constraints at the historical vertex for explaining the actual distribution of adaptive form within potential organic morphospace. In this sense, both themes count as “positive” in my second (or conceptual) sense that any pow­erful argument challenging a stale and limiting consensus must be treasured in science.

  No good or experienced naturalist could ever fully espouse the reductionistic belief that all problems of organic form might be answered by dis­solving organisms into separate features, each with a specified function, and each optimized independently by natural selection. But theories do drive, or at least nudge, adherents towards their extreme formulations — and even such sophisticated versions of Darwinism as the Modern Synthesis (see Chapter 7) biased the perspectives of biologists in this direction by advocating natural se­lection as, effectively, the sole cause of evolutionary change. Various pleas, heard with increasing frequency during the past generation (Goodwin, 1994, for example), to “put the organism back into evolution,” or to “reestablish a meaningful science of morphology,” should be understood as expressions of a growing conviction that theories of part-by-part functionalism cannot ex­plain the major patterns of life's history and current morphological distri­bution.

  We do need to reformulate, in modern and operational ways, the old no­tions of organic integrity, and structural determination from the “inside” of genetics and development, thus balancing our former functionalist faith in the full efficacy of adaptationism with positive concepts of internal and struc­tural constraint. Only in this way can we forge a unified science of form to in­tegrate the architecture and history of organisms with their daily struggles to survive, prosper, and propagate in a complex ecological surround — a world that Western culture once perceived as “the face of nature bright with glad­ness” (Darwin, 1859, p. 62), but that we now recognize as the material do­main of natural selection, a process carrying no moral implications for human life (thus permitting us to throw aside the crutch of comforting imagery that Darwin so rightly rejected), but operating with relentless (though not ex­clusive) force throughout living nature.

  An epitome for the theory-bound nature of constraint terminology

  We may use the model of the aptive triangle to illustrate how my second, or conceptually “positive,” meaning of constraint (including both the “positive” and “negative” empirical modes of channels and limits) rests upon the the­ory-bound nature of all scientific terms and definitions. As I argued previ­ously (see pp. 1032–1037), if we designate a set of causes as canonical within an orthodox theory, then vernacular usage designates other causes lying out­side the theory, but nonetheless influencing phenomena that should fall under the aegis of orthodoxy, as “constraints” upon the power or validity of the standard theory. (My designation of such constraints as “positive” then follows [Page 1058] the usual ethos of science in placing special value on new ideas that chal­lenge complacent conventionalities.) In this crucial sense, the identification of certain causes as constraints depends upon the claims and nature of a stan­dard theory.

  The historical and structural vertices of the aptive triangle become sources of constraint when this standard theory emanates from the third, or func­tional, vertex — as classical Darwinism does, and as Darwin himself so clearly specified in his own analysis of the three vertices (without explicitly identify­ing such a model) in closing Chapter 6, entitled “Difficulties on Theory,” in the Origin of Species — again, see pp. 251-260 of this book for my analysis of this key Darwinian argument). In making this “pure end-member” analysis of canonical causes vs. constraints, I am consciously presenting extreme, or cardboard, versions of central theories in order to clarify a logical (and termi­nological) point about the naming of expectations and exceptions. (Just as I argued previously that no actual empirical case would fall precisely at a vertex of total determination by one factor alone, I also acknowledge that no subtle thinker's theory will fall right on a vertex either. Nonetheless, discus­sion in terms of pure end-members may be defended as a conceptual device for clarifying the central content and primary commitment of more complex theories.)

  Figure 10-11 illustrates the changing terminology of constraint and con­vention under three pure end-member theories of causation at vertices of the aptive triangle. For the pure adaptationist, committed to natural selection as the controlling and functionalist mechanism of evolutionary change, all causes of currently adaptive form that cannot be attributed to direct selec­tion for immediate utility must count as constraints. (Fig. 10-1 la depicts this version, with the canonical cause placed at the functional vertex, and with other vertices making contributions that then be called constraints upon the full and free operation of current natural selection to forge immediate utility.)

  As another virtue of these simplified representations, we can also grasp how any pure end-member theorist must treat the exceptions (“constraints”) that cannot be denied as causal contributors to currently adaptive form. In this case (Fig. 10-lla), I have already noted Darwin's own excellent strategy (see pp. 251-260): admit the historical inputs, but attribute their cause to natural selection in the past; then admit the structural inputs as genuine ex­ceptions, but relegate them to a low and insignificant relative frequency. Thus, all “constraints” either record the operation of the canonical mechanism in the past, or stand as genuine exceptions rendered impotent by their rarity.

  But if I were committed to a view that the direct action of physical forces (as expressed in the spatiotemporal invariance of natural law) builds the adaptive forms of organisms directly, and without any appeal to functionalist or distinctively biological principles like natural selection — D'Arcy Thomp­son, in fact, advocated this general view in as pure a form as any 20th cen­tury biologist dared to espouse (see pp. 1179–1208) — then the structural

  [Page 1059]
/>   10-11a. What we label as primary causes and what we consider as constraints follow from the logic of our preferred theory. Under strictly functional views of natural selection, the functional vertex represents the orthodox cause of current action by natural selection, whereas the historical vertex imposes constraints based on past adaptations and the structural vertex imposes architectural constraints based on limitation in the nature of building materials.

  10-11b. For a structuralist thinker like D'Arcy Thompson, optimal forms built directly by physical laws become canonical, and influences from other vertices become constraints — either of ecological particulars from the functional vertex, or of contingency from the historical vertex.

  10-11c. For a cladist interested in historical reconstruction of branching sequences, the historical vertex becomes primary, and influences from the other two vertices represent constraints — ahistorical constraints of universal physics from the structural vertex and autapomorphies of particular and immediate adaptation from the functional vertex.