The Structure of Evolutionary Theory

by Stephen Jay Gould

• Consequences (miltons arising as spandrels)

  • Manumissions (miltons arising by unemployment)

  • Insinuations (miltons arising by random drift)

  All taxonomies — thus embodying the richness of fascination of systematics as a scientific subject (Gould, 2000c) — mix aspects of nature's objective order with human preferences for utility or intelligibility. Even if we allow that these four categories exist “out there” in nature — and even I, although I developed this scheme, would not go so far in trying to craft a naturalistic defense, for I recognize that the objective items of the exaptive pool could be parsed in other ways — our decisions about their ranking and secondary ordering require a choice among several logically legitimate alternatives. All taxonomies base such choices on the designation of a fundamentum divisionis, or basis of primary ordering. The differences among alternative fundamenta reflect the theories we favor as most useful in understanding and [Page 1285] explaining the phenomena encompassed by the taxonomy. (Thus, as Ernst Mayr has so forcefully argued throughout his career, taxonomies are theories about the basis of order, not boring and neutral hatracks, pigeonholes or stamp albums for accommodating the objective and uniquely arranged items of nature.)

  For the exaptive pool, I can imagine three fairly obvious candidates for fundamenta (and I am sure that other possibilities have eluded me). My choice of franklins vs. miltons reflects my convictions and arguments about the most appropriate theoretical context for understanding the evolutionary meaning and importance of the exaptive pool.

  First, suppose I were working within a more orthodox context that as­sumed effective control over evolution by natural selection. Suppose further that, within this context, I still appreciated the importance of specifying the currently unused attributes of organisms that might contribute to future suc­cess. From such a standpoint, I would probably be tempted to choose a fundamentum that drew a primary distinction between the ineluctable un­used attributes that arise even when natural selection works in its “leanest and meanest” mode of optimization, and a second groups of unused attrib­utes that arise either from the cessation or the weakness (or at least the non-exclusivity) of natural selection. In such a taxonomic system, I would place all franklins together with all miltonic spandrels in my first category of in­eluctable attributes of pure selection (inherent potentials of features opti­mized by selection for other functions, and inevitable side consequences of similarly optimizing selection). I would then devote my second category to the other two groupings of miltons — manumissions and insinuations — aris­ing as historical results of selection's relinquished control: features falling out of its purview for the first group, and features insinuated beneath its notice in the second group.

  In a second fundamentum that might also follow from stronger selectionist commitments, I might decide to make a primary division between attributes that originated either as direct adaptations or as inherencies of direct adapta­tions for the first group, and attributes with truly nonadaptive origins for a second group. In this case, I would place miltonic manumissions and frank­lins in the first category (as adaptations that become unemployed, and as ad­ditional potentials of features that arose as direct adaptations). I would then unite miltonic spandrels and insinuations into a second category of attributes originating as nonadaptations — spandrels as features that are nonadaptive in their own isolated selves (whatever the adaptive status of other features that generated them as side consequences), and insinuations as nonadaptive fea­tures that originated beneath the notice and malleability of selection.

  I have chosen a quite different third alternative — inherent potentials (franklins) vs. available things (miltons) — because I wish to emphasize the structuralist and nonadaptionist components of evolutionary theory that have, in Darwinian traditions, been downplayed or ignored. This funda­mentum stresses the primary difference between consequences of purely adaptationist mechanisms applied to directly adaptive features (the franklins, [Page 1286] or inherent potentials, that, ever since Darwin's excellent arguments against Mivart, have been recognized as the basis of quirky functional shift in unpre­dictable evolutionary sequences remaining entirely under selectionist con­trol); and the extensive and influential set of material features that actually “stand and wait” by themselves as nonadaptive attributes (not as additional but now only potential uses of present adaptations). In other words, I choose this fundamentum because miltons (and not franklins) pose a genuine chal­lenge to the exclusivity of adaptationist mechanisms. Franklins enlarge the scope and sophistication of selectionist argument, adding a genuine flavor of formalist limitation and potentiation to an otherwise naively functionalist theory based only upon organic accommodation to selective pressures of an external environment. But miltons emplace a genuinely nonadaptionist com­ponent into the heart of evolutionary explanation — for if many features origi­nate as nonadaptations, and if nonadaptations, as material items of miltonic “stuff,” stand and wait while occupying a substantial percentage of the exaptive pool, then evolutionary explanations for both the origin of novelties, and for the differential capacity of lineages to enjoy future phyletic expansion and success, will require a revised and expanded version of Darwinism, enriched by nonselectionist themes of a formalist and structuralist research program. I therefore choose my fundamentum as the best taxonomic device for explor­ing the role of nonadaptation and structural constraint in the exaptive pool of evolvability.

  Cross-level effects as Miltonic spandrels, not Franklinian

  potentials: the nub of integration and radical importance

  As one of its most interesting and potentially reformatory implications, the hierarchical expansion of selectionist theory introduces an extensive array of features into the exaptive pool (and upon the consciousness of evolutionary biologists) as effects propagated to other levels by features that arise for directly causal reasons at a focal level. (One need not challenge the conven­tional view that most direct reasons at focal levels will be selectionist and adaptational — for the propagated effects may still assume a different and non-adaptive status.) What shall we call such effects that only become manifest at other levels and may be truly invisible at the focal level of their generation as consequences? Shall we interpret them as franklins, or inherent and presently unexploited utilities of features originating for other functions — with their only difference from more conventional franklins (flight capacities of thermoregulatory feathers) resting upon the fact that they happen to reside in biological individuals at levels other than the level of the feature that generated them?

  I do not think that such cross-level effects can be interpreted as franklins, or mere potentials. Rather, cross-level effects are available things, albeit avail­able only to biological individuals at other levels. Cross-level effects are there­fore miltons, not franklins. Franklins are potentials, not things. As such, franklins can only be recruited sequentially in time from the primary adapta­tions in which they inhere as alternative utilities (feathers for flight following [Page 1287] feathers for thermoregulation to cite the canonical example once again). (If feathers performed both functions from the start — a perfectly plausible sce­nario, of course — then we would never have designated their aerodynamic role as a franklin, for this function would have been part of their adaptive ex­pression ab initio.) But cross-level effects, as miltonic things standing and waiting in the exaptive pool, become available as separately cooptable attrib­utes right from their origin. They can therefore be utilized (by exaptation at their different level) simultaneously with the continuing primary adaptive function of the generating feature at its focal level of origin.

  For example, the duplicated gene that arose by gene level selection, achiev­ing an adaptive advantage thereby at the genic level (by definition, and through its plurifaction), may be simultaneously exapted at the organismal level by undergoing a mutational change (only now of potential benefit to the organism as a consequence of the gene's redundancy),
and contributing thereby to a new organismal function. Similarly, the organismal form that adapted to its immediate environment by evolving a lecithotrophic bottom-dwelling larva from a planktotrophic ancestor, may simultaneously impart an exaptive effect to its species by enhancing the speciation rate via the altered demic structure of isolated subpopulations that no longer experience the gene flow previously potentiated by floating planktonic larvae. I am not, by the way, inventing these cases as personal speculations. Each represents the most widely discussed potential example of cross-level exaptation for its pair of levels.

  A recent example (Podos, 2001; Ryan, 2001) illustrates the range and probable ubiquity of simultaneous emplacement of spandrels to other levels as consequences of primary adaptations at a focal level. In the best known case of Darwin's finches, Podos (2001) shows that ordinary adaptation of bill sizes and shapes in response to climatic changes and competition with other species (as so superbly documented in the continuing work of the Grants and others — 1986 for an early summary, for example) imposes automatic conse­quences upon the form, style, and range of the resulting song, because “two functional systems — that used for feeding and that used for singing — share a common morphology, the beak” (Podos, 2001, p. 186). Sharper and nar­rower beaks permit wider ranges and precision of song, whereas heavier and blunter beaks impose greater “constraint” (Podos's term) upon potentials and specificities of resulting vocal production. Since songs function as powerful premating isolating mechanisms, the automatic divergence of song, arising as a side-consequence of ordinary adaptation of bills in feeding, and the differ­ent degrees of distinctiveness attached to specific forms of the bill, may have profound consequences in a resulting (and ultimately highly exaptive) differ­ential capacity for speciation among different subclades of this classic group (based upon varying capacities of the resulting song to act as an effective sig­nal for mate recognition). Ryan's commentary acknowledges Podos's infer­ence about exaptive effects on speciation rate as necessarily conjectural for now, but as the most interesting larger implication of this important study.

  This property of simultaneous utilization carries two important implications [Page 1288] for the status of cross-level effects. First, cross-level effects are thereby identified as things rather than potentials — that is, as miltons rather than franklins. (Differential variation in range of song among subclades is a “thing” that may impose emergent fitness at the species level at the same time as its generating adaptations in the form of the bill operate in their ordinary Darwinian manner at the organismal level.) Second, cross-level effects, as things, generate a potentially radical challenge to conventional Darwinian concepts. Unused potentials, as argued several times previously, remain fully within the adaptationist program as possible future uses of features that arose as adaptations and will always be adaptations (albeit for a different function in the future). But unused things begin as nonadaptations right at their origin (whatever their future importance as exaptations) — and the demonstration of their high relative frequency in the exaptive pool, and of their importance to the evolutionary history of many lineages, would introduce a significant nonadaptationist element into evolutionary theory.

  So if we agree that cross-level effects are miltonic things and not franklinian potentials, into what category of miltons (see Table 11-2) do they fall? If the previous components of this argument prove acceptable, then this final question enjoys a simple and unambiguous resolution: they are spandrels. The key property of spandrels lies in their automatically consequential char­acter as things necessarily enjoined by other changes. Cross-level effects fit this definition in all ways, for they arise in concert with the primary change, and as a necessary consequence thereof. But they do manifest an interestingly different property from such conventional at-level spandrels as the pendentives of San Marco or the cylindrical space that can be exapted as a brooding chamber at the center of a gastropod shell: they express themselves in a bio­logical individual at a different level from the individual bearing the feature that generated the effect.

  As stated before, and as the heart of my argument, I believe that the desig­nation of cross-level effects as a second class of spandrels greatly increases the range and importance of this concept — for cross-level spandrels are probably far more common, and of far more frequent importance in potentiating the future direction of evolutionary lineages, than the at-level spandrels that pro­voked the initial formulation of this concept. As an almost naively evident de­fense of this claim, I would point out that more cross-levels exist than focal levels (obviously, as focal levels must be singular!), so cross-level spandrels have more “places to go” than at-level spandrels. Moreover, any cross-level attribute holds greater potential, prima facie, for manifestation as a quirky and oddly nonadaptive feature — for anything arising at one level and injected into another must enter its new domain adventitiously and without reference to the norms and needs of its adopted “home”; whereas an at-level spandrel can only be tolerated if it meshes reasonably well with a design already estab­lished for its kind of entity.

  These attributes of cross-level spandrels embody their importance in revis­ing our usual understanding of evolution, particularly at the macroevolutionary level of the species-individual. As argued above, at-level spandrels [Page 1289] must be nearly neutral (or at least not too burdensome) because they “come with the territory” and must immediately integrate themselves as parts of the larger structure built at their level — for at-level spandrels share the same kind of form, and are made of the same sorts of materials, as the primary adapta­tions themselves. If pendentives made buildings collapse, or doubled their cost without compromising their mechanical function, then architects would not choose the designs that generate them as necessary consequences of oth­erwise favorable properties.

  But cross-level spandrels, especially when injected upwards into more slowly cycling biological individuals at higher levels, can establish themselves more easily, and beyond any screening power of the higher-level individ­ual, by pressure of numerous introductions within single generations of the slowly-cycling higher-level individual. I have been emphasizing the future exaptive potential of nonadaptive spandrels, but nonadaptive features can also work in an opposite manner, becoming detrimental (truly inadaptive) to their bearers — a fate that probably befalls at-level spandrels only rarely (be­cause inadaptive effects will generally preclude their introduction in the first place), but that may represent a common outcome of cross-level spandrels in­jected into higher-level individuals, and not readily suppressed, at least ini­tially, because they can become rooted before any episode of generational cycling reveals their disutility to the higher-level individual.

  The claims of the last paragraph may seem arcane and distant (in my ab­stract formulation) from empirical reality. But this phenomenon has long been recognized at the species level, even though evolutionary biology previ­ously lacked the conceptual apparatus to offer a general explanation. We all acknowledge that many organismal adaptations impose strongly negative consequences upon the geological longevity of their lineage. Any highly complex, metabolically expensive, and intensely specialized adaptation (the peacock's tail, or virtually any elaborate contrivance of runaway sexual selec­tion); any alteration, especially involving the loss and simplification of com­plex ancestral structures, that adapts an organism to a transient and highly specialized environment (the “degenerate” parasite utterly dependent on a unique and unusual host); must strongly compromise the geological potential of a subclade bearing its autapomorphy, relative to a sister subclade retaining an ancestral and generalized morphology and ecology. In fact, any feature — and they must be legion — that provides adaptive benefits at the organismal level, but that simultaneously injects such “negative” spandrels into the en­compassing species-individual (either suppressing its rate of speciation or de­creasing its geological longevity thereby), will be inadaptive at the higher level, but unpreventable by ins
ertion before the species-individual can “no­tice” and reject the feature.

  This pairing of organismal adaptation with injected spandrels that prove inadaptive to the encompassing species-individual sets the proper conceptual context, under the hierarchical theory of selection, for what our literature has long called, in an ambiguous and merely descriptive way, the “opportunism” of evolution. “Opportunism,” like “preadaptation,” should be recognized as [Page 1290] an odd and inappropriately anthropocentric term, designating one subtype in the large and general category of features established for good causal reasons at one level that then impose effects at other times, or upon other levels, either positively (as in “preadaptation”) or negatively (as in “opportunism” for im­mediate advantages, leading to extinction in the long run). Moreover, al­though the term has been fading from use of late, all textbooks of evolution used to include an explicit section on a phenomenon called “overspecialization” — another nonsense phrase devised to treat, when theory lacked the proper concepts, the important observation that many organismal adapta­tions impose inadaptive effects as spandrels upon the encompassing species-individual. What in this pairing could possibly be called “overspecialization” with any justification? The organism becomes adaptively specialized to its own immediate benefit, and the species suffers as a nonadaptive (ultimately inadaptive) side consequence of spandrels representing the adaptation's ex­pression at the species level.*

  But I don't want to leave the impression that upwardly-injected cross-level spandrels always spell dissolution, or even doom, at the higher level — and that the species-level, in particular, suffers from this phenomenon, even to the point of becoming a weakened or ineffective locus of evolution thereby (for such a conclusion would greatly compromise the hierarchical theory itself by an argument akin to Fisher's rejected claim (see pp. 644–652) for the logical ineluctability, but practical insignificance, of species selection). Two related arguments reinforce the evolutionary importance of cross-level spandrels, while also reaffirming the power of the species-individual as a biological agent in evolution.