The Structure of Evolutionary Theory

by Stephen Jay Gould

  The constraints of these rules have provided more flexibility in their fecund channels than limitations through their “forbidden places” — a theme rightly emphasized in the finest book yet written on the relationship of homologously shared and rule-bound developmental architecture to flexibility and evolvability in the phyletic richness of subsequent life (Gerhart and Kirschner, [Page 1273] 1997). But the importance of constraint and preexisting opportunity in chan­neling the pathways of change should not be underestimated — for the mes­sage of evo-devo does not proclaim that “all is possible under such flexible rules.” One might, for example, denigrate the importance of constraint in noting that a rod-like element of an ancestral agnathan gill arch exhibited sufficient malleability in form and function to become the tiny, disparately shaped and divergently functioning malleus of the mammalian middle ear. The gill-arch elements may therefore work as general building blocks of unconstraining Pharaonic bricks in the metaphor of my treatment of this sub­ject (pp. 1134–1142). But we must also remember that, absent some skeletal element of appropriate form and position (whatever its capacity for future modification to almost any other shape or function), vertebrates would prob­ably never have evolved jaw bones that could transmute to ear bones — and our lineage (if it had survived at all) might have remained an insignificant component of bottom-dwelling mud-sucking marine faunas, thus precluding, at least for this planet at this time, the evolution of any species with enough cognitive capacity to fret about such issues.

  2. From studies of genetic structures and sequences, the astonishingly high relative frequency of multiply repeated elements (with respect to previous assumptions about the nature of genomes), and the multiplicity of ways, from gene duplication to retrotransposition, for generating them, have docu­mented enormous redundancy and combinatorial flexibility within genomes — even if we have designated the objects of this discovery with the disrespect­ful name of “junk DNA” (but see p. 1269 for my favorable reevaluation of this term). Retrotransposons, for example, constitute about 40 percent of mammalian genomes (Kazazian, 2000). The human genome includes about 500,000 truncated versions and some 3000 to 5000 full copies of the LINE-1 long terminal repeat. Chromosome 2 of Arabidopsis includes 239 tandem du­plications involving 593 genes. A larger duplication of almost 2.5 million bases appears on two chromosomes in four large blocks. A long stretch of chromosome 4, including 37 genes, has been duplicated on chromosome 5. Chromosome 2 also contains a region with 75 percent of the mitochondrial genome, reflecting a recent transfer of a substantial block of DNA from an organelle to the nucleus (Meyerowitz, 1999) — quite a “gift” of “play” from the genic to the organismal level!

  3. From studies of the properties of populations, communities and interac­tions among evolving entities by mathematical modeling and computer simu­lation. Several researchers have used these methods in attempts to identify the abstract and general conditions that might confer flexibility, persistence, and capacity for change upon an evolving population or group of entities. Most notably, as discussed previously (p. 1210), Kauffman's (1993) claim that suc­cessful systems move towards “adaptation at the edge of chaos” rests upon attention to evolvability as a key ingredient of longterm success. Such systems must be adaptive, but too much (and too precise) a local fitting may freeze a system in transient optimality with insufficient capacity for future change. Too much chaos may prove fatal by excessive and unpredictable fluctuation, [Page 1274] both in external environments and internal states. But a capacity to adjust to chaotic situations also confers evolvability. Adaptation at the edge of chaos balances both desiderata of current functionality and potential for future change, or evolvability.

  Yet, however much biologists can document and articulate these compo­nents of evolvability, the general concept itself has remained uncomfortable, even paradoxical, because the clear existence of “flexibility for future bene­fit,” and the equally obvious importance of evolvability to the long-term per­sistence and success of lineages, cannot be rendered as a directly causal and explicit outcome of the Darwinian mechanisms that we have viewed as fully sufficient for understanding the causes of evolutionary change. Organismal selection for traits that confer differential reproductive success in the ecologi­cal moment simply cannot generate, in any active or direct manner, a set of features that achieves evolutionary significance only by imparting flexi­bility for change in distant futures. We cannot deny that these features of evolvability deeply “matter” in the history of lineages; but how can benefits for futures arise by any causal process in the here and now? Moreover, and adding insult to anomaly, the major components of evolvability apparently reside in “superfluous” genetic elements, and in supernumerary items of anat­omy, that almost seem to mock our usual concepts of the stark efficiency of selection as a natural arbiter between the immediately useful and the discardable junk. As an example of this discomfort and puzzlement, just con­sider the language of Kazazian's excellent review of mammalian retrotransposons (2000, pp. 1152-1153), as he struggles to grasp and communicate the apparent discordance between current irrelevancy (at the level of neces­sary generation) and future benefits: “Although retrotransposons have been viewed as selfish DNAs that provide no benefit to their host cell, we now know that over evolutionary time they have increased the diversity of the ge­nome through a variety of mechanisms providing it with considerable 'added value.' ... It is clear that LI elements are the master mammalian retro­transposons. Although L1s may be selfish, they are clearly not junk, for they have played a major part in our evolution and the evolution of our genomes.” But the genuine junk of today can be exapted for the triumphs of tomorrow. The spores of penicillin didn't do us much good (and even imposed substan­tial harm in spoiling our foods) until Dr. Fleming made his fortuitous discov­ery of a previously unrecognized, and now eminently lifesaving, property.

  As its central theme and purpose, this book proposes a set of expansions and revisions to Darwinian theory that, among other salutary features, can resolve the paradox of evolvability by exposing the issue as a Scheinproblem, or problem of appearance — that is, a spurious appearance within an overly restricted theory presently lacking the language and concepts for granting causal intelligibility to this evidently vital theme in evolutionary studies. The key revisions proposed on each leg of the logical structure that I have called the tripod of essential components in Darwinian theory provide, in their en­semble, an explanation of evolvability in hierarchical and structuralist terms.

  1. The expansion of selection to a hierarchical theory of simultaneous operation [Page 1275] on Darwinian individuals at several nested levels: Future flexibilities cannot be targets of conventional organismal selection; nor can such an at­tribute enter any calculation of the differential organismal fitness that fuels Darwin's mechanism. But this conclusion does not imply that attributes of evolvability must remain uninvolved as agents in any form of selection — for the most commonly cited components of evolvability can act as positive traits to enhance the fitness of species-individuals in selection at the species level. Just consider the two most widely recognized boosts to evolvability at the species level: the propensity of some species to generate relatively more daughter species, and the resistance to extinction conferred upon some spe­cies by such organismal features as generalized anatomy and wide ecological tolerance. These two properties represent the primary analogs, at the species level, of the two cardinal attributes — birth rates and death rates — that virtu­ally define the fitness of organisms in the calculus of Darwinian benefits for traits that must ultimately express their selective advantages by correlation with enhanced birth or retarded death of organisms. Organismal selection cannot craft propensities for speciation or resistances to extinction in any di­rect way, but these properties act as primary features in the higher-level pro­cess of species selection, and therefore figure prominently and directly in the calculus of selective advantage under hierarchical models.

  But a pote
ntial problem still remains. Species selection can certainly utilize and maintain these important traits of species-individuals, but species selec­tion cannot always fashion such traits, especially when they emerge into this higher level as spandrels of selection upon organisms — as must occur when emergent fitness at the species level resides in the higher-level expression of organismal traits that can only originate by conventional organismal selec­tion (see Lloyd and Gould, 1993; Gould and Lloyd, 1999 for an analysis of the logic of this issue). For example, in the case cited above, the enhanced re­sistance of a species to extinction emerges as a consequence of such organ­ismal traits as generalized anatomy and broad physiological tolerance.

  But when we probe further, and draw the appropriate analogy to the level we understand best, we recognize that ability to utilize, combined with inabil­ity to fashion, represents a norm within Darwinian theory, not a distinctive anomaly provoked by evolvability at the species level. Ordinary natural selec­tion doesn't manufacture its variational raw material either. Darwinians have always understood that their theory's most quirky, most original, and most brilliant intellectual “move” explains how a process that creates nothing di­rectly can, nonetheless, operate on raw material of different origin to become a “creative” force in the construction of novel and useful features. Indeed, I devoted much of Chapter 2 to illustrating how Darwin's contemporaries rarely grasped this intensely paradoxical, but defining, property of natural se­lection's genuine creativity, based only upon its power to enhance or eliminate, but not to craft variation in a direct and dedicated manner. And I pointed out that the isotropy (or undirectedness) of variational raw material acts as a prerequisite for granting natural selection this potentially creative role. [Page 1276]

  Thus, Darwinians have always argued that mutational raw material must be generated by a process other than organismal selection, and must be “ran­dom” (in the crucial sense of undirected towards adaptive states) with respect to realized pathways of evolutionary change. Traits that confer evolvability upon species-individuals, but arise by selection upon organisms, provide a precise analog at the species level to the classical role of mutation at the organismal level. Because these traits of species evolvability arise by a differ­ent process (organismal selection), unrelated to the selective needs of species, they may emerge at the species level as “random” raw material, potentially utilizable as traits for species selection.

  The phenotypic effects of mutations are, in exactly the same manner, span­drels at the organismal level — that is, nonadaptive and automatic manifesta­tions at a higher level of different kinds of causes acting directly at a lower level. The exaptation of a small and beneficial subset of these spandrels virtu­ally defines the process of natural selection. Why else do we so commonly re­fer to the theory of natural selection as an interplay of “chance” (for the spandrels of raw material in mutational variation) and “necessity” (for the locally predictable directions of selection towards adaptation). Similarly, spe­cies selection operates by exapting emergent spandrels from causal processes acting upon organisms.

  2. The acknowledgment of structural components as joint causes and spe­cifiers, along with natural selection, for the directions of evolutionary change: (In the most radical version, these structural inputs operate as positive con­straints generated as consequences of features with nonadaptive origins, thus precluding a purely functionalist or adaptationist account for both the ori­gins, and the channels of subsequent change, of organismal traits.) To sum­marize this major claim of the preceding section: if important components of evolvability must emerge as spandrels of natural selection on other features (for they cannot be fashioned as direct products of a process that cannot ex­plicitly make “things” for potential future benefits), and if these spandrels serve as exaptable raw material for higher-level processes of change, then we will need to describe macroevolution at least partly in the language of channeling by historical and structural constraint (often based upon features with nonadaptive origins), and not entirely in conventional functionalist terms of selective modelling to current environments, potentiated by an effectively un­fettered capacity of mutational raw material to provide the wherewithal for evolutionary movement in any immediately adaptive direction.

  3. The conventional mechanisms of microevolution cannot, in their extrapolation through geological immensity, fully explain the causes and patterns of evolutionary change at larger scales: I have argued throughout this book that the biological components of nonextrapolation lie within critiques of the first two legs of the tripod — and I have therefore invoked the external “disobedi­ence” of the geological stage to represent this theme as a surrogate from the domain of another relevant subject (particularly for paleontologists like my­self). Similarly, for this case of evolvability, species level selection on the first leg, and nonadaptive spandrels on the second leg, identify the major barriers [Page 1277] to explaining evolvability by extrapolated microevolutionary Darwinian pro­cesses.

  Above all, and to move forward towards a clear and operational definition of concepts, the crucial subject of evolvability requires a taxonomy for its nu­merous modes, and their strengths and distributions. The key to an inclusive accounting lies in the general notion of usable features (for promoting long-term diversification and success) now unused. What kinds of attributes thus contribute to evolvability? What encourages their production or augments their number (especially since natural selection in the organismal mode can­not craft such features directly)? Who has more of them and why?

  Clearly, organisms and populations maintain what we might call a “fund” or “pool” of potential utilities now doing something else, or at least doing no harm. I propose that we designate this ground of evolvability as “The Exaptive Pool,”* and that we try to establish a logical, interesting, and em­pirically workable taxonomy for the various attributes in this fund of exaptable potential. The exaptive pool represents the structural basis of evolvability, the potential vouchsafed to future episodes of selection (at all levels) in a world of strongly polyhedral objects always and ineluctably built with interesting corners and facets that facilitate and channel the directions of evolutionary movement.

  THE TAXONOMY OF THE EXAPTIVE POOL

  Franklins and Miltons, or inherent potentials vs. available things

  The American dime, the smallest and thinnest (but not the least valuable) of our coins, can't purchase much these days, but still functions as legal tender, the primary cause of its manufacture and useful persistence. American dimes have also, starting a few years after his death near the end of World War II, borne a representation of Franklin D. Roosevelt on their recto (“heads”) side. As an unintended consequence of their thinness, American dimes (and no other indigenous coins) also happen to fit snugly into the operative groove on the head of a standard screw — and dimes therefore work very well as adventi­tious screwdrivers. (I strongly suspect that virtually every adult American has used a dime in this well-known supplementary way, just as many of us know [Page 1278] how to jimmy a door latch with a credit card.) American dimes are therefore adaptations as money, and exaptations as screwdrivers.

  This potential for supplementary use as a screwdriver represents an inher­ent capacity of the dime's size and shape, not a separate and unused entity arising as a side consequence of some other change. The inherent potentials of any object (for uses other than their intended purpose of manufacture) es­tablish a large and important category of attributes in the exaptive pool of any biological individual. I propose that we refer to these inherent (but cur­rently, or at least usually, unexploited) potential functions as “franklins” to honor the most famous exaptation of the American dime. (Indeed, if our cur­rency inflates much further, dimes may become virtually useless as money, and their exaptive role as screwdrivers may achieve a primary status in imme­diate function. When Russia's currency hyper inflated after the collapse of the Soviet Union, I witnessed a remarkable example of this pheno
menon. Five-kopeck coins had become truly worthless as currency, but only objects of their size, shape and weight could operate public telephones. Sharp entrepre­neurs therefore stood at telephone booths, offering to sell old five kopeck coins for 100 rubles each — 2000 times their official monetary value, but when ya gotta call, ya gotta call.)

  Franklins are not actual but unemployed “things out there.” Franklins are alternative potential functions of objects now being used in another way. In the words of my subtitle, franklins are inherent potentials, not available things. This distinction may seem trivial, inordinately fine as an exercise in logic chopping, or even parodic to hard-nosed scientists committed to the professional ethos of “just give me the facts, and leave linguistic and theoreti­cal niceties to effete humanists who lack the luxury of an objectively empiri­cal subject matter.” But I will attempt to show, in the next two sections, that the distinction between inherent potentials and available things, as the two fundamental categories of the exaptive pool, provides the conceptual key to understanding the importance of spandrels, and for recognizing the strong weight that must be applied to structuralist, and particularly to nonadaptationist, elements in the exaptive pool — thus defining the revisionary power of this concept in evolutionary theory, and exposing the depth of different ex­planations required to understand evolvability vs. ordinary adaptation by natural selection.

  I will just mention, for now, that franklins constitute the theoretically untroubling category embodied in the Darwinian notion of quirky functional shift (as discussed on pp. 1218–1229 of this chapter). Franklins capture the important concept so poorly expressed in the old term “preadaptation” — that is, suitability for another function not presently exploited because the feature has been adapted by natural selection for a different utility. When feathers function for enhanced thermoregulation on the arm of a small run­ning dinosaur, their potential aerodynamic benefits are franklins, or inherent but unused potentials. When Michelin's rubber works as an automobile tire, its suitability for manufacturing cheap and durable sandals for poor children in developing countries is a franklin. In short, franklins represent future potentials [Page 1279] within structures now adapted to a different utility. When evolution then coopts the structure for actual service in this formerly potential role, the franklin becomes an exaptation. The entire process remains under the control of Darwinian selection and adaptation throughout. Franklins do not vie with current utilities, and they cannot be construed as “things” (whether adaptive or nonadaptive) waiting for a potential place in the adaptive sun. Franklins are the inherent potentials that permit Darwinian pathways, under the con­trol of natural selection leading to adaptation at all times, to undergo quirky and unpredictable shifts from one function to a qualitatively different utility.