The Structure of Evolutionary Theory

by Stephen Jay Gould

  The attributes in the second major category of the exaptive pool are, by important contrast, actual entities, pieces of stuff, material things that have be­come parts of biological individuals for a variety of reasons (to be exemplified in the next section), but that have no current use (and also cause no substan­tial harm, thereby avoiding elimination by selection). Items in this second cat­egory of “available things” can originate in several ways — as nonadaptive spandrels (the most important subcategory, I shall argue), as previously useful structures that have become vestigial, or as neutral features fortuitously in­troduced “beneath” the notice of selection.

  I propose that we refer to these available but currently unused material or­gans and attributes as “miltons” to honor one of the most famous lines in the history of English poetry. John Milton ended his famous sonnet On His Blindness, written in 1652, by contrasting two styles of service to God: the frenetic activity of evangelists and conquerors and the internal righteousness of people with more limited access to worldly action:

  . . . thousands at his bidding speed,

  And post o'er land and ocean without rest;

  They also serve who only stand and wait.

  Miltons, in short, are actual things, presently without function, but holding within their inherent “goodness” the rich seeds of potential future utility. Now, they only stand and wait; tomorrow, they may be exapted as key inno­vations of great evolutionary lineages.

  Miltons constitute the radical counterpart to the conventionality of frank­lins within the exaptive pool. Miltons break the exclusivity of the adaptationist program by basing a large component of evolvability not upon the poten­tial of already functional (and adaptive) features to perform in other ways, but rather upon the existence of a substantial array of truly nonadaptive fea­tures — unused things in themselves rather than alternative potentials of fea­tures now functioning in other ways (and regulated by natural selection at all times). If features with truly nonadaptive origins occupy a substantial area in the full domain of evolvability, then we must grant this structuralist (and nonadaptationist) theme a generous and extensive space within the logic and mechanics of evolutionary theory. For this reason, I have argued that span­drels — already the most important category of miltons, but made far more significant by the inclusion, under their rubric, of cross-level effects of fea­tures originating at other levels (see pp. 1286–1294 of this section) — win [Page 1280] their central status in revising and expanding evolutionary theory because they represent the primary input of an overtly structuralist and nonadaptationist concept into the central logic of evolutionary causation.

  Choosing a Fundamentum Divisionis for taxonomy: an

  apparently arcane and linguistic matter that actually embodies a

  central scientific decision

  Table 11-2 presents my sketchy and preliminary proposal for taxonomy of subcategories in the exaptive pool. I shall retain franklins, or inherent poten­tials, as an integrity for now, not because I doubt that they could be usefully divided into subcategories, but simply because I wish to focus upon miltons, or available things, as the component of the exaptive pool that holds most re­formatory promise within evolutionary theory.

  I divide miltons into two major categories according to their different mo­dalities of generation: structurally, as automatic and nonadaptive sequelae or side consequences of changes in other features, or at other levels; and histori­cally, as nonadaptive features not linked by structural or mechanical necessity to another feature of a biological individual, but rather introduced sequen­tially in time by processes that can generate and tolerate such nonadaptive en­tities.

  I then divide each of these two subcategories into two further groups. Structural miltons, as necessarily and automatically consequential, are all, and collectively, spandrels. But spandrels come in two different “flavors,” with the second group of cross-level effects (newly so categorized here) repre­senting, in my judgment, the key addition that elevates spandrels to a posi­tion of central importance in evolutionary theory. (I will present my full ar­gument for considering cross-level effects as spandrels in the next section, pp. 1286–1294.) In any case, the first structural group of at-level, or architec­tural, spandrels includes the mechanical and automatic side consequences, deployed throughout the rest of the individual, of any primary change (usu­ally adaptive) evolved by other features of the same individual. When I orig­inally defined the biological concept of spandrels (Gould and Lewontin,

  Table 11-2. A Taxonomy for the Exaptive Pool

  A.

  Inherent potentials

  Franklins

  B.

  Available things

  Miltons

  1. As architectural consequences

  Spandrels

  Structurally

  i) At-level spandrels by geometry

  ii) Cross-level spandrels by injection

  2. As historical unemployments

  Manumissions

  Historically

  3. As invisible introductions

  Insinuations

  Historically

  [Page 1281]

  1979), with the pendentives of San Marco as my “holotype,” I described only this group — not for any good or principled reason, but simply because I had not recognized the status of cross-level effects in a hierarchical theory of selec­tion. Obviously, I already regarded this restricted set of at-level architectural byproducts as potentially significant in evolution. But the inclusion of cross-level effects as a second category moves the concept of spandrels from an edge of interest to a center of potential importance in evolutionary theory.

  The second grouping of cross-level, or propagated, spandrels includes the expressed effects upon biological individuals of changes introduced for a definite reason (whether adaptational or not) at a different initiating level. I include such cross-level effects under the rubric of miltonic spandrels, rather than franklinian potentialities (as they have usually, if unconsciously, been re­garded, when conceptualized at all), because, like the architectural sequelae of my at-level group, they are actual, initially unused, nonadaptive things that also arise as side consequences — even though the side consequence in this sec­ond group of cross-level spandrels are propagated effects to other levels (of­ten with no expression at the level of origin for the primary change that gen­erated them), whereas the side consequences in the first group of at-level spandrels are immediate mechanical correlates of a primary change in the same individual, and therefore easier to spot and define.

  I then divide the second subcategory of historical miltons into two group­ings of markedly different status:

  1. Features that lose an original utility without gaining a new function. In a first group of “unemployments” or “manumissions,” previously utilized fea­tures become liberated from functional or selective control, and gain freedom to become exapted for other uses. Currently nonadaptive as a historical result of their altered status, they fall out of selective control and into the exaptive pool as actual items that must now “stand and wait,” but might serve again in an altered evolutionary future.

  We have generally granted little evolutionary potential to such vestigial unemployments because relatively quick reduction and loss — as in the stan­dard example of eyeless cavefishes — seems to follow as an inevitable injury of degeneration added to the initial insult of unemployment. But such manu­mitted miltons may be quite common, particularly at the gene level (where the process has achieved greater recognition). Many multiply repeated evolu­tionary phenomena — the deletion of larval forms in the evolution of direct development, or the shedding of adult stages in progenetic lineages, for exam­ple — must leave a substantial number of genes in such an “unemployed” state. Yet, as we also recognize, full unemployment may occur only rarely be­cause most genes function in more than one way. (This fact, however, should be regarded as salutary for future exaptive potentials in keeping partially “unemployed�
� genes in an active state of resistance to true operational discombobulation by accumulation of neutral mutations.)

  In the most fascinating confirmation of our literature (see p. 688 for more details in a different context), Hendriks et al. (1987) sequenced the alphaA crystallin gene in the blind mole rat Spalax ehrenbergi (which still grows a [Page 1282] lens in its vestigial eye, although the lens's irregular shape cannot focus an image, and the eye remains buried under skin and hair in any case). They found that the blind mole rat has accumulated mutational changes in alphaA crystallin at a rate far higher than that observed in nine other rodents — but still (to emphasize the intended point) at only 20 percent of the characteristic rate for truly neutral pseudogenes. The authors conclude that although the alphaA crystallin gene emplaces substantial gene product into a nonoperational lens (at least for vision), the gene must still serve some function, sup­ported by stabilizing selection, to resist the full neutral rate of change.

  In another context on p. 1245, I noted that, although the alphaA crystallin gene is more specialized for generating lens protein than its paralog alphaB crystallin, its product still appears in other organs in some mammalian spe­cies. Moreover, the nonseeing eye of blind mole rats may continue to function in other ways. Haim et al. (1983) show that blind mole rats still perceive light (and not by the obvious nonvisual route of correlates to temperature) in regu­lating their circadian clocks to photoperiod. Hendriks et al. (1987) argue that melatonin, secreted by the equally nonvisual retina, may act as a prime circa­dian regulator — and they therefore suggest that the developmental pathway leading to the eye and its lens, including the action of alphaA crystallin, may be adaptively preserved because the eye also performs essential nonvisual functions.

  2. Features introduced beneath selection's scrutiny. In the second group of “insinuations,” nonadaptive features may enter the exaptive pool by neutral drift. Just as we have unfairly discounted the role of manumitted miltons, we have probably underestimated the relative frequency of insinuated miltons, although for a different reason. We have regarded neutral changes as both pe­ripheral and rare (largely restricted to tiny populations on the verge of elimi­nation in any case) because we have granted too much power to selective control — thus permitting our orthodoxy to become self-fulfilling by circular argument. But the frequency of insinuated miltons may actually be quite high, especially given the inevitability of their substantial introduction via founder effects.

  In a fascinating recent example (Tsutsui et al., 2000; Queller, 2000), the Argentine ant Linepithema humile, “a superb invader of non-native habitats” (Queller, 2000, p. 519), has become firmly established in Mediterranean envi­ronments of California, much to the detriment of several native species and to the distress of humans. Queller notes (op. cit.): “We rarely understand why invading species succeed, although a common advantage is that they leave their predators, parasites and pathogens behind. Although this explanation could apply to Argentine ants, it seems that the most serious enemies left be­hind were the warring clans of its own species.”

  In Argentina, ants of different colonies fight, and these antagonisms, and the resulting elimination of many colonies, keep the entire species at relatively low population densities, thus leading to an ecosystem that includes many other equally successful ant species. But, in California, Linepithema humile does not fragment into mutually antagonistic colonies, but lives as a single [Page 1283] aggregate. Queller states (op. cit.): “They form a vast supercolony within which there is little aggression and extensive interchange of both workers and queens. Such ants are called unicolonial because a whole population effec­tively becomes one colony.” The Californian success of Linepithema humile does not arise from their prowess as fighting individuals (for these relatively small ants do not seem especially gifted as gladiators in combat by the myrmecine equivalent of mano a mano), but from the sheer power of their num­bers — “because of their ability to rapidly recruit legions of troops from their network of nests. So peace with flanking nests generates advantages in com­petition with other species” (Queller, 2000, p. 519).

  Fighting among different colonies in native Argentina depends upon the ants's ability to recognize degrees of genetic relationship as assessed by differ­ences at seven microsatellite loci of nuclear DNA. (Interestingly, in another form of exaptation — a cross-level spandrel in this instance — the usefulness of these markers presumably arises from the rapid and substantial variability thus imparted among populations, and flowing upwards as an effect from the neutral status of most mutations in these microsatellites.) Thus, in Argentina, ants tolerate conspecifics of closely related colonies and fight with genealogi­cally more distant conspecifics in colonies of greater genetic disparity.

  However, and unsurprisingly, the ancestors of the California invaders passed through a genetic bottleneck in their initially small population, proba­bly derived from a single native colony. The California ants lost about two-thirds of the genetic variability at microsatellite loci. “So all of the ants are ge­netically alike, and those applying the old similarity-tolerance rule could be fooled into accepting everyone as kin” (Queller, 2000, p. 519). Thus, as an example of an insinuated milton (a feature introduced nonadaptatively by neutral drift) that has been exapted for marked, if transient, utility, the suc­cess of Linepithema humile in California seems to rest upon a loss of genetic variability that induced the ants to form a single supercolony operating as a military phalanx (to continue the dubious tradition of anthropomorphic description for behaviors of social insects) of remarkable power and efficiency. Queller notes (p. 5190): “Paradoxically, the ecological success of the intro­duced populations stems not from adaptation but from the loss of an adapta­tion — colony recognition — due to genetic drift.” But, then, if we recognized the hierarchical nature of selection, and repressed our adaptationist biases, this situation might not appear so paradoxical, and we might even search ex­plicitly, encouraged by a conceptual reform and expectation, for phenomena that may be very common in nature, but that elude our notice because we find them anomalous, and do not recognize their existence until they stare us in the face.

  Finally, this case also suggests an interesting flip-side of potential longterm disadvantages for such transient success — a near necessity in our hierarchical world of potentially conflicting levels, lest victory at one position propagate throughout the system to full and permanent triumph. (The scientifically as­tute Tennyson may have experienced a more universal insight than we usually acknowledge in penning his famous line in Morte d'Arthur (1842): “The old [Page 1284] order changeth, yielding place to new . . . lest one good custom should cor­rupt the world.”) If reduced conflict builds supercolonies of such competitive success, why does the same species live so differently, and so much more mod­estly, in its native Argentinian abode?

  Queller (2000, p. 520) presents the interesting and cogent explanation that, by losing their capacity to identify degrees of kinship, the ants “can­not improve or even sustain their cooperative behavior. Without relatedness, adaptive modifications of cooperative worker behavior cannot be favored and maladaptive ones cannot be disfavored.” Queller (op. cit.) therefore con­cludes that the randomly established feature (the miltonic insinuation) be­hind transient triumph must also spell eventual doom: “Random drift will become important again, this time because of the absence of any opposing force rather than a small population size. The ants are like a casino gam­bler who is lucky once but cannot quit: chance got them their stake, but over the long run it can only lead to ruin.” Thus, either the California ants will eventually restore their genetic variability, split into fighting colonies, and become restricted to smaller overall populations with greater staying power; or they will suffer “the lingering death of decay by drift” (p. 520). Tsutsui et al. (2000) propose what, to most people, would sound like an absurd and counterintuitive strategy for control, although the suggestion clearly makes theoretical (but perhaps not practical) sense
to anyone schooled in the intrica­cies of evolutionary theory: introduce more ants with substantial variation at those microsatellite loci, thus encouraging self-regulation by the reinitiation of conflict!

  Thus, in summarizing the categories in my proposed taxonomy of the exaptive pool, the currently unused but eminently usable features that build the ground of differential evolvability fall into two groups of inherent poten­tials (franklins) and available things (miltons). Miltons, in turn, include three distinct groupings ordered by different sources of origin. The exaptive pool therefore contains:

  • Potentials (franklins)