THE UBIQUITY OF SPANDRELS UNDER A HIERARCHICAL CONCEPTION OF EVOLUTIONARY LEVELS AND CAUSALITY. If Darwin's own view of natural selection as a single-level process operating on organisms had prevailed, spandrels would still be pervasive in nature and important in evolutionary theory. [Page 1267] But the scope of nonadaptive side consequences would then be limited to overt structures, physiologies and behaviors of bodies. That is, spandrels would be altered bits and pieces (often substantial) of organic “stuff,” molded as the propagated effects of primarily adaptational changes wrought by selection upon other parts of the body.
But under the revised and expanded hierarchical theory (Chapter 8), where selection works simultaneously on a nested hierarchy of biological individuals (genes, cell lineages, organisms, demes, species, clades), the domain of spandrels becomes much larger, and their importance to evolutionary theory expands accordingly, and for an interesting reason that has not been adequately addressed in the literature (see Gould and Lloyd, 1999), but will occupy most of the next and last section of this chapter. The expansion of spandrels under a hierarchical theory of selection establishes the most interesting and intricate union between the two central themes of this book — the defense of hierarchical selection (as an extension and alteration of Darwin's single-level organismal theory) on the first leg of the tripod of essential components in Darwinian logic; and the centrality of structural constraint (with non-adaptively originating spandrels as a primary constituent) for a rebalancing of relevant themes, and as a correction to the overly functionalist mechanics of selection on the second leg of the tripod (or branch of the tree — see Figure 1-4).
To epitomize the central argument: under a hierarchical theory of selection, any novelty introduced for any reason (usually adaptational) at any level, must propagate a series of effects to biological individuals at other levels of the hierarchy. Duplication of genetic elements by direct selection at the gene level, for example, propagates redundancies to the organismic level; any organismal adaptation at Darwin's level propagates changes to the encompassing species-individual, as expressed in such species traits as population size, geographic range, and coherence among subparts (organisms). These propagated effects must be defined and treated as spandrels. As “injections” from another level (where the initiating change probably had an adaptational, or physically automatic, basis), these propagated effects cannot be viewed as adaptations at the level under consideration. Moreover, because these effects exist as true properties at the level under consideration — that is, as actual “stuff” rather than unused potentials of features now operating in different ways — they must be treated as initially nonadaptive “things” or spandrels, rather than as mere potentialities of some hypothetical future utility. Therefore, under a hierarchical model, spandrels include both the architectural side consequences of adaptational changes at the level of their origin, and the large realm of effects propagated to other levels as nonadaptive consequences of changes wrought for directly causal reasons.
This concept of cross-level spandrels neatly explicates a variety of phenomena that have long been recognized as both true and essential, but that have remained puzzling or anomalous under the conventional Darwinian rubric of a functionalist, single-leveled theory of selection. For example, our canonical, [Page 1268] almost mantra-like, statement about the deleterious nature of most mutations achieves such an evident explanation that the resulting “aha” seems almost humorous in its suddenly obvious character. Each mutation arises for a perfectly good reason (usually chemical rather than adaptational in this case) at the gene level. But the effects then imposed upon organismal phenotypes must be designated as spandrels — that is, as nonadaptive side consequences expressed at another level. These effects will usually be deleterious, because the organism, as a highly complex, well integrated, and biochemically efficient object, will more often be hindered than helped by a change that arises as an “injection” from elsewhere, established by causes directly operating only in this elsewhere, and not subject to initial scrutiny at the level of injection. For the same reason, and in another mantra, we designate mutations as “random” — not in the mathematical sense of equally likely in all directions, but in the special evolutionary sense (see Eble, 1999) that such mutations originate without reference to the adaptive needs of the organism. When we recognize the phenotypic expression of mutations as cross-level spandrels, this property of “randomness” becomes entirely sensible, and no longer puzzling as a supposed sign of organic inefficiency.
Gene duplication, and other modes of origin for the repeated elements that constitute such a high percentage of the genomes of complex organisms (and that have been so puzzling under Darwinian assumptions about organisms as “lean and mean” machines honed to optimality by the relentless power of natural selection), represent the most important “playing field” yet identified for the evolutionary importance of cross-level spandrels. (I thank my colleague Jurgen Brosius for helping me to understand and work through the implications of this concept — see Brosius and Gould, 1992; Brosius, 1999; and Brosius and Tiedge, 1996.)
In some cases, of course, gene amplication originates as an immediate adaptation at the organismal level, especially when the availability of more gene product provides a selective advantage to the organism. But, more commonly, amplification occurs for causal reasons at the genic level itself, often by the conventional Darwinian mechanism of increased reproductive success, in this case by generating more copies of oneself, and inserting them into various places in one's surrounding totality — that is, in the genome itself. (Such an argument about direct Darwinian selection at the gene level provides the rationale, as previously discussed (p. 693), for the important hypothesis of “selfish DNA” — see Orgel and Crick, 1980; and Doolittle and Sapienza, 1980 for the original publications.) Yet evolutionists have also recognized (see Ohno, 1970 for the classic statement) that these extra copies may strongly impact the evolutionary future of organisms by supplying flexibility for change through their redundancy. But this otherwise sensible argument also seems to raise a central dilemma in causality itself — since flexibility for future change cannot cause the current origin or maintenance of any feature! We can resolve this problem by recognizing augmented copies as nonadaptive (and cross-level) spandrels at the time of their initial expression at the organismic level. Later recruitment and utilization of spandrels represents a perfectly [Page 1269] sensible, indeed inevitable, concept under notions of constraint and hierarchical selection. I have, of course, and throughout this chapter, referred to such later utilization as exaptation — in this case by the cooptation of initially nonadaptive spandrels.
I have, in the past, objected to the usual terminology of such amplified elements as “junk DNA,” feeling that such a dismissive term could only record an adaptational bias towards viewing such currently “superfluous” stuff as an insult to Darwinian optimality. I wrote (in Brosius and Gould, 1992, p. 10706): “Genes duplicated or amplified by the tens to the thousands . . . have been named in an ambiguous or even derogatory manner (e.g., pseudo-gene or 'junk DNA'). Such names do not reflect the significance of retroposed sequences as large valuable assets for the future evolvability of species; and, as a result, it is more difficult to contemplate their significance, impact, and function.”
But I have changed my mind after reading an insightful commentary by Sydney Brenner (1999) on my 1997 paper about the meaning and significance of spandrels in evolution. Brenner begins by acknowledging the role of adaptational biases in our misunderstanding of the meaning and significance of amplified DNA:
There is a strong and widely held belief that all organisms are perfect and that everything within them is there for a function. Believers ascribe to the Darwinian natural selection process a fastidious prescience that it cannot possibly have and some go so far as to think that patently useless features of existing organisms are there as investments for the future ... Even today, long after the discovery of
repetitive sequences and introns, pointing out that 25% of our genome consists of millions of copies of one boring sequence, fails to move audiences. They are all convinced by the argument that if this DNA were totally useless, natural selection would already have removed it. Consequently, it must have a function that still remains to be discovered. Some think that it could even be there for evolution in the future — that is, to allow the creation of new genes. As this was done in the past, they argue, why not in the future?
But Brenner then defends the traditional terminology of junk DNA with an argument (based on the contrast of junk and garbage in vernacular English) that I had not considered, and that now strikes me as wise and useful:
Some years ago I noticed that there are two kinds or rubbish in the world and that most languages have different words to distinguish them. There is the rubbish we keep, which is junk, and the rubbish we throw away, which is garbage. The excess DNA in our genomes is junk, and it is there because it is harmless, as well as being useless, and because the molecular processes generating extra DNA outpace those getting rid of it. Were the extra DNA to become disadvantageous, it would become subject to selection, just as junk that takes up too much space, or is beginning to smell, is instantly converted to garbage. [Page 1270]
Brenner then ribs my literary and terminological pretensions (and I accept his criticism). But he also finds a resolution to the conceptual puzzles surrounding junk DNA in recognizing that such amplified sequences, when they arise causally at the gene level and then get propagated as effects to the organismal level, are nonadaptive spandrels with great potential for later exaptation to utility. Therefore, their designation as junk — that is, as currently useless, but harmless (as opposed to garbage), and replete with potential future value — seems entirely appropriate, and I belatedly embrace this term as a proper implication flowing from the definition and meaning of spandrels:
The paper [Gould, 1997e] has an important message and I strongly urge my readers at least to look at it, even if all the words in it can't be understood. I offer this brief summary as a guide.
The term spandrel originates in architecture and is used to describe spaces left over as a consequence of some other design decision, such as the triangles that remain behind when a rectangular wall is pierced by an arched opening. No self-respecting architect would simply leave such spaces, especially in a grand cathedral with a rich patron. Instead they would be decorated, as is the case of the four pendentives under the dome of San Marco in Venice, which are decorated with the four evangelists. This example is a good one, because the historical sequence of events is known. The spandrels are the consequence of a structural design decision; a by-product of placing a dome on rounded arches; three centuries later, mosaicists decorated these spaces. Thus spandrels are not primary adaptations but, because they can have later uses, they become in Gould's terminology, exaptations.
The Exaptive Pool: The Proper Conceptual Formula
and Ground of Evolvability
RESOLVING THE PARADOX OF EVOLVABILITY AND DEFINING
THE EXAPTIVE POOL
Conventional Darwinian organismal selection adapts creatures to their immediate local environments — a process of specialization, usually operating to produce particular contrivances that reduce organismal flexibility for future evolution to radically altered conditions, especially when adaptation leads to simplification and loss of structures, as in extreme, but common, cases of internal parasitism. In a subset of situations — especially emphasized by Darwin as the potential ground of general “progress” in the history of life (see p. 467) — local adaptation may be achieved by a generalized improvement in biomechanical design that might be construed as promoting future prospects and flexibilities, rather than restricting phyletic options by specialization. But few evolutionists would doubt that organismal selection leads far more often to diminution of future prospects by specializations and losses than to enlargement [Page 1271] by general biomechanical improvement. Natural selection in the organismal mode can only construct local adaptations in the here and now. We can all conjure up the conventional image of highly specialized and gorgeously adapted forms revelling in their successes of the moment, but then dying in the fullness of geological time, as marginal generalists parlay their staying power into phyletic persistence. (Natural selection, of course, may also favor such generalists in certain momentary environments, but not for their future prospects.)
And yet, flexibility for future change manifestly exists in differential degrees among organisms. This flexibility contributes mightily to the longterm macroevolutionary success of lineages, but cannot be directly built or maintained by ordinary natural selection in the organismal mode. We designate this differential capacity for success and extent of future change by the vague and loosely defined name of “evolvability” — a concept that, until recently, remained unpopular among Darwinian biologists by evoking feelings of discomfort and confusion. The reasons for this usual distaste flow from the failure of conventional theory to provide a context that could make such a concept intelligible rather than paradoxical. After all, if “evolvability” seems contrary to the general workings of natural selection, and if natural selection represents the fundamental mechanism of evolutionary change in populations and lineages, then how could “evolvability” be defined or characterized as anything other than an accident or a passive residuum? Phenomena without direct mechanisms generally do not win much interest or approbation from working scientists.
For example, in their important 1998 article entitled “Evolvability,” Kirschner and Gerhart express both apparent paradoxes attending this crucial concept within a strict Darwinian context: the seemingly logical need to impute selective advantages to supraorganismic levels (within a theory committed to the primacy, or even exclusivity, of organismal selection), and the almost unavoidable “feeling” that benefits of evolvability can only accrue to future states (which, in any standard account of causality itself, cannot be influencing the present evolution of beneficial features). Kirschner and Gerhart write (1998, p. 8426): “The proposal that evolvability has been selected in metazoan evolution raises difficulties because it seems to be a trait of lineages or clades rather than individuals. Clade selection is often considered an 'explanation of last resort.' Also, evolvability seems to confer future rather than present benefit to the individual.”
But if we follow the expanded Darwinian logic of this book, the paradoxes become only apparent because the theoretical revisions developed herein validate both apparent peculiarities of “evolvability,” thus bringing this crucial concept within the rubric of a revised evolutionary theory. First, selection at higher levels is an important force in evolution — and evolvability can therefore originate directly at the level of its evident advantage. Second, the structuralist validation of exaptation establishes, as a central aspect of evolutionary theory, the future cooptation of features initially evolved for other reasons. Thus, hierarchy and positive constraint — the two primary revisions [Page 1272] to the first two components of Darwinian central logic — show their mettle in providing theoretical resolutions for each of the superficially paradoxical aspects of “evolvability,” a concept that evolutionary biologists have lately recognized as vital, but treated so gingerly, or even apologetically, in the absence of a proper theoretical framework for admitting something so evidently important among accepted modes of causality and explanation.
Indeed, a remarkable change has been brewing for the last decade or so in evolutionary studies. “Evolvability” has suddenly become a hot topic, even among the most orthodox of modern Darwinians (Dawkins, 1996, for example). This change has occurred for at least three good reasons listed below, each reflecting one of the major topics of these two chapters on the biology of evolutionary constraint. But I feel, as stated just above, that the subject still languishes for want of a proper theoretical context in revisions and expansions of a Darwinian world view t
hat had become too narrow in its focus on organismal adaptation and the sufficiency of known microevolutionary mechanisms to explain all scales of evolutionary change. In this final section, I therefore try to provide a context for evolvability by combining the two central theoretical reformulations of this book: (1) hierarchical models of selection; and (2) the importance of structuralist approaches to biological form and function, as expressed in concepts of constraint and, especially for elucidating this particular topic, in the importance of spandrels as nonadaptively originating side consequences, then available for later cooptation to utility as exaptations.
1. From studies of evo-devo, the discovery of extensive genetic and developmental homology among distantly related phyla, especially the common presence, spatial orientation, and mode of action of Hox genes in bilaterian phyla, has focussed attention upon the flexibility inherent in the great range of interesting, workable, and often realized permutations that can be generated from developmental rules shared by all complex animal phyla. In particular, and as discussed previously, the disproof and subsequent reversal of Lewis's original “bottoms up” hypothesis of sequential addition and differentiation of Hox genes in causal concert with the complexification and differentiation of arthropod phenotypes, has emphasized the enormous flexibility inherent in broad rules emplaced at the outset, and plesiomorphic among all bilaterian phyla — for the common ancestor of protostomes and deuterostomes already possessed a full complement of Hox genes, as do the most homonomous of living groups, the Onychophora and the Myriapoda (see pp. 1147–1150). Thus, the realized diversity of bilaterians evolved in a “top down” fashion (at least for features regulated by the Hox series) from a common ancestor with a full set of basic components and their rules of action already in place.
The Structure of Evolutionary Theory Page 202