The Structure of Evolutionary Theory

by Stephen Jay Gould

  The general conclusion may be stated in a simple manner, but I believe that the resulting implications for evolutionary theory are both profound and curiously underappreciated: If many features that operate as adaptations under [Page 1248] present regimes of natural selection were exapted from ancestral features with nonadaptive origins — and were not built as direct adaptations for their current use (or exapted from ancestral features with adaptive origins for different functions) — then we cannot explain all pathways of evolutionary change under functionalist mechanics of the theory of natural selection. In­stead, we must allow that many important (and currently adaptive) traits originated for nonadaptive reasons that cannot be attributed to the direct ac­tion of natural selection at all and, moreover, cannot be inferred from the exaptive utility of the trait in living species. Because the subject of evolution­ary biology must engage many crucial questions about the origins of features, and cannot be confined to the study of current utilities and selective regimes, nonadaptationist themes therefore assume an important role in any full ac­count of life's history and the mechanisms of evolutionary change.

  The key to this expansion of evolutionary theory therefore lies in the cate­gory of currently useful traits with nonadaptive origins — my rationale for prefacing this topic with two sections to develop the prerequisites of the argu­ment: one on Nietzsche's principle of general discordance between bases of current utility and reasons for historical origin, and another on the terminol­ogy of exaptation as a framework for describing and appreciating the impor­tance of currently useful structures coopted from a different ancestral status, rather than directly evolved for their present function. By introducing, in this final section, the theme of exaptations based upon features with nonadaptive origins, I complete the chart of Table 11-1 (see p. 1233) by recognizing two subcategories of exaptation: (1) Cooptations of features that originated for different adaptive reasons — the principle of “quirky functional shift” that en­riches (with structuralist “flavor”), but does not challenge, the functionalist control of evolution by natural selection, and that also establishes the ground of contingency within the Darwinian world view. (2) Cooptations of features with nonadaptive origins — the theoretically radical category that precludes any complete explanation of evolution in adaptationist terms, and that pro­vides a nonadaptationist alternative for evolutionary inquiry about the ori­gins of currently adaptive biological features.

  But if this argument is so simple to state, so airtight in logic, and at least interesting (I would say profound, but not with everyone's initial approbation!) in its implications for evolutionary theory, then why hasn't the category of currently useful features with nonadaptive origins been perceived as more troublesome by orthodox Darwinians? Why, to sharpen the paradox, did the concept not even receive a name in conventional theory? This complex ques­tion embraces many dimensions, including psychological and historical influ­ences lying well beyond my professional competence and the scope of this book. But, for the relevant dimension of the structure of Darwinian theory, two strong reasons — both invalid in my view, with their refutation as the pri­mary aim of this section — have permitted most defenders of strict natural se­lection to acknowledge the existence of nonadaptive features, but then to rel­egate them to a periphery of rarity and impotence where they can exert no effective role in setting pathways for the history of life, or in specifying principles [Page 1249] of evolutionary theory. Both arguments flow from the assumption (which I do not challenge) that most nonadaptive traits of organisms arise as struc­tural consequences of selection upon other features, or upon other aspects of the same trait. (I speak here of traits with enough stability and complexity to become exapted for meaningful utility in a descendant lineage. No one doubts, of course, that many truly trivial features of organisms have no bear­ing upon fitness, and lie well beneath the visibility of natural selection):

  Nooks and crannies. Darwinians have generally argued that most structural consequences of natural selection on other features can survive as true nonadaptations (and not be eliminated as inadaptive) only if they occupy little space or require minimal metabolic input. The mold marks on old bot­tles (made in two-piece molds) surely testify to mechanisms of manufacture, but serve no purpose. If we regarded them as ugly, or if they disrupted a bot­tle's utility, we could easily remove them. But we allow them to persist in their inconsequential triviality.

  Secondary and consequential status. This common argument commits the historical error that inspired Nietzsche's warning. Many people have assumed that nonadaptive origin as a consequence of selection upon an­other feature relegates a trait to permanent insignificance because it arose in a passive and sequential manner — a clear conflation of reasons for historical origin with potentials for subsequent utility. Perhaps, in an ancestral pelycosaur, the skull sutures remained unfused (at birth from an egg) as a trivial consequence of some developmental necessity in a forthcoming and free-living ontogeny. But this nonadaptive property may later become the prereq­uisite to the success of mammalian descendants, when live birth became the autapomorphic key to a new and highly successful mode of life.

  Defining and defending spandrels: a revisit to San Marco

  When I chose the notorious spandrels of San Marco to illustrate these falla­cies with an architectural analogy that might not automatically raise the hackles of orthodox Darwinians (or thicken the scales over their eyes, to cite the other common biological metaphor for resistance to unfamiliar ideas), and also to introduce a term for the most common category of nonadaptive features with high potential for subsequent exaptive utility, I did not make a capricious selection for idiosyncratic reasons of personal interest (Gould and Lewontin, 1979). Rather, I chose this Venetian example because the historical record and current status of these particular four pendentives includes a suf­ficient richness and certainty of documentation to refute all the common objections raised against similar claims for the evolutionary insignificance of nonadaptive features in biological systems.* [Page 1250]

  When a hemispherical dome stands on a set of four rounded arches meet­ing at right angles to form a square — a very common design in ecclesiastical architecture and many other buildings — four tapering triangular spaces must appear directly under the dome, each formed by the space left over between the dome itself (above) and the pair of adjacent arches (at the sides) meeting at right angles (see Figure 11-8). These four spaces, called pendentives, must form as a structurally necessary side consequence of the architect's basic deci­sion to mount a dome on four rounded arches, so arrayed. Lewontin and I never claimed that these spaces did nothing useful (for obvious and trivial starters, their roofing keeps out the birds and the rain). But we did argue that their ineluctable size and shape — their number and triangular form — arose as side consequences of a previous architectural decision, and could not be viewed as adaptations in themselves. (The four pendentives, in other words and by analogy with my previous example, hold the same status as mold marks on an old bottle — necessary side consequences of an architectural deci­sion, not functional features in themselves.)

  The general architectural term for such “spaces left over” is spandrel — a lovely name derived from the primordial human tool of measurement, the span of our own hand (or of the corresponding feature in an anthrop­omorphized divine architect — as in Isaiah's God (40:11-12) “who hath measured the waters in the hollow of his hand, and meted out heaven with the span”). Classical spandrels are two-dimensional spaces left over (Figure 11-9) including the vertical boards between steps of a staircase, the triangu­lar spaces between arches arrayed in a linear row, and the flat horizontal stretches (called “spandrel courses”) on large office buildings, located between the tops of windows on the floor below and the bottoms of windows in the next story just above. Apparently — see documentation in Gould, 1997e — some architects restrict the term “spandrel” to two-dimensional spaces left over, whereas others, particularly in European usag
e, extend the term to any space that arises as a side-consequence of a prior decision, and not as an ex­plicitly designed feature in itself, thus including the three-dimensional pen­dentives of San Marco and thousands of other buildings. In any case, I con­sciously decided to apply this remarkably appropriate term to San Marco's pendentives because their shared property with classical two-dimensional spandrels — their status as architectural byproducts (at least for their forms, numbers, and placements) — cannot be denied. I also felt that biology needed a term for such architectural sequelae of “adaptive” decisions, and that this well known term from a related discipline could serve admirably.

  [Page 1251]

  11-8. Four spandrels with brilliant mosaics of the evangelists under a circular dome of the Cathedral of San Marco in Venice. From Demus, 1984.

  [Page 1252]

  11-9. A three-dimensional spandrel, or pen-dentive, contrasted with a more conventional two dimensional spandrel between arches in a linear row (below). From Gould, 1997e.

  In any case, although spandrels must originate as necessary side-consequences of an architectural decision, and not as forms explicitly chosen to serve a purpose, they still exist in undeniable abundance, and can therefore be secondarily used in important and interesting ways by clever architects, artists and patrons of buildings. (I grew up in New York City, and have al­ways admired the lovely ornamentations on the spandrel courses of many of our finest art-deco skyscrapers, particularly the zoological motifs on the Chanin Building just opposite Grand Central Station.) In my “holotype” of the central dome of San Marco's cathedral in Venice, where the entire interior space has been covered in glorious mosaics, the four pendentives of the central dome have been ornamented in a complex way, stunningly well “fit” both for the space occupied, and for the symbolic meanings portrayed by the mo­saic designs. The four evangelists (including St. Mark, the cathedral's patron and most celebrated resident interred therein) occupy the widest top sec­tions of the spandrels, under a motto in Latin doggerel: Sic actus Christi, describuunt quatuor isti — thus did these four describe the acts of Christ. Be­neath the evangelists, in the narrowing triangular spaces, personifications of Eden's four rivers (see Genesis 2) hold amphoras (Greek pitchers) over their shoulders, each pouring water onto a single flower wedged into the narrow and elongated space at the base of each pendentive.

  The design — beautiful, complex, and particularly appropriate both to Christian symbolic representation and to the surrounding space — exudes util­ity. But no one would make the mistake of arguing that the spandrels exist to house the evangelists. The spandrels originated as a nonadaptive side-consequence of a prior architectural decision. These originally nonadaptive spaces were then coopted (several centuries later, in this case) as “canvasses” for wonderfully appropriate designs. In biological terms, the mosaic designs are secondary adaptations, and the spandrels themselves then become exaptations [Page 1253] for the residency of these designs. Again, I chose this architectural analog because I felt that a similarly unambiguous case in organisms might become conceptually muddled — for biologists have been trained to regard anything that “works well” as an adaptation, and might therefore not “see” the originally nonadaptive nature of the spaces. But I felt — rightly, in retro­spect — that this quite precise architectural analog would not generate enough emotional salience to act as a barrier against understanding the intended point.

  A desire for clarity in illustration served as my primary motive, but our original article (Gould and Lewontin, 1979) does not sufficiently emphasize my other major reason for choosing this example as a “holotype” to illustrate the important category of nonadaptive features originating as architectural side consequences. I chose the San Marco spandrels because they so evidently refute, in terms of the architectural analogy, the two standard arguments raised against a similar importance for nonadaptive structures in biological morphology (as outlined above on p. 1249):

  Nooks and crannies. One cannot brand the spandrels as trivial in occupied space or peripheral position — as one might legitimately hold for the mold marks on a bottle. The four spandrels under any dome occupy a sub­stantial area, surely equal to a large percentage (if not the totality) of the area of the dome above. As we shall see, this generous size and central place­ment also refutes a major aspect of the second dismissal based on consequentiality.

  Consequentiality. The false inference of evolutionary insignificance from secondary original status as a side-consequence of a primary adaptation includes two arguments of different logical standing, but each equally invalid:

  1. The empirical claim. For biological examples, many scientists have assumed that the temporally sequential status of any exapted utility (as im­posed upon a primary nonadaptation) must relegate any subsequent use to marginal importance. Thus, the evangelists and rivers are adaptive in their purposeful and lovely fit within a preexisting (and initially nonadaptive) space, and in their important message as conveyed within the larger aim of the building's role as a Christian church. But these mosaic designs are just as surely secondary and sequential — as later adaptations restricted by prior con­straints upon the number and form of an initially nonadaptive housing. The mosaicists made an adaptive choice, but preexisting constraints strongly lim­ited their options. The four spandrels could not house, in any easy or adap­tive way, the three children saved from the fiery furnace or the five books of Moses (not to mention the inelegance of setting 2.5 Commandments in each spandrel). Biologists often conflate a genuine limitation upon options with a false inference that constrained solutions, however adaptive, cannot generate structures of major importance either to the current working of organisms, or to their future evolutionary potential.

  But, as Nietzsche argued, a secondary and constrained origin implies noth­ing about potential for either present or future importance — and the designs in San Marco's spandrels clearly expose this fallacy. Extensive feedback from [Page 1254] the spandrels to the mosaics of the dome proves that secondary features can exert pervasive influence upon the basic design of a totality. The domes of San Marco are radially symmetrical and therefore provide, in se and considered alone, no reason for favoring a quadripartite mosaic design. Yet all but one of San Marco's five domes contain mosaics arranged in four-part symmetry — clearly, in each case, to harmonize with the iconography in the four triangular spandrels below. For example, in the mosaic design of our “holotype” central dome, three circles of figures radiate out from a central image of Christ: an­gels, disciples, and virtues. Each circle is divided into quadrants, even though the dome itself is radially symmetrical in structure. Each quadrant meets one of the four spandrels in the arches below the dome.

  Another dome contains angels in the spandrels and the twelve apostles in the dome, arranged in four groups of three, with each group centered on one of the four spandrels below. Yet another dome presents four male saints in the dome and four female saints in the spandrels, with each male perfectly cen­tered between two of the females. Thus, an ineluctable architectural byprod­uct can, nonetheless, determine the fundamental design of a totality that or­dained its consequential origin. The natural world abounds in recursions and feedbacks of this sort. Mustn't the ever-cascading spandrels of the human brain be more weighty than the putative primary adaptations of ancient Afri­can hunter-gatherer ancestors in setting the outlines of what we now call “hu­man nature”?

  2. The methodological claim based on operationality. As dis­cussed throughout this section, biologists have often been reluctant to base terminological distinctions on differing historical pathways to a similar cur­rent result — for the good reason that a poverty of historical records often de­nies us the data needed to reach a firm conclusion, whereas current situations can always be directly observed or experimentally manipulated. Thus, a biol­ogist might argue that the distinction between exaptation and adaptation, al­though logically sound and conceptually interesting, cannot be “cashed out” in a suffic
iently high percentage of cases because we so often lack enough his­torical data to determine whether a currently useful structure originated by natural selection for its present function (adaptation) or got coopted to its current role from an initial status either as an adaptation for some other func­tion, or as a nonadaptive spandrel (exaptation).

  I accept this point, of course, and have given my response in a previous sec­tion (p. 1233): when we cannot resolve the historical antecedents of a cur­rently useful feature, we need not apply the terminology of adaptation vs. exaptation at all. The feature remains an aptation (in its current status), and may be so named, whatever its unresolved historical origin. But I also admit that if the distinction between exaptation and adaptation can be drawn in only a small percentage of cases, and only under unusually favorable circum­stances, then the concepts enjoy little practical or operational use, and the terms might as well be abandoned in the actual practice of science.