Genetic drift at the traditional organismic level enjoys far more respect and currency today, but the basic argument of the Synthesis does have merit at this hierarchical level. Sexually-reproducing, multicellular organisms generally share two properties that greatly limit the efficacy of genetic drift: they live in populations far too large for random fixation in the face of nearly any measurable selection pressure; moreover, the style of individuality manifested by organisms, based on well-balanced functional integration among sub-parts, renders the traits of these interactors particularly subject to scrutiny by natural selection.
Do these good reasons for demoting random change at the organismic level doom this alternative style of evolution to weakness or impotency throughout the hierarchy? Clearly not, as the recent history of our profession proves; moreover, we may even invert the standard hope for extrapolation from the level we know best, and assert instead that the organismic level discourages random change as a peculiarity of individuality in this realm — and that analogs of genetic drift at other levels should expect healthy, if not dominant, relative frequencies.
All evolutionists also know that ideas of random change have enjoyed greatest success, based on inherent plausibility, at the genic level, where the so-called “neutral theory of molecular evolution,” most strongly associated with the great Japanese geneticist Motoo Kimura (1968, 1983, 1985, 1991a and b), but initiated and developed by others as well (Jukes, 1991), has [Page 686] often been hailed as the most interesting revision of evolutionary theory since Darwin.
When we consider the two properties of organisms that depress the frequency of fixation by drift at this level, we easily spot the difference that makes randomness so important at the lower genic level. Population size, also characteristically large for gene-individuals, cannot supply the reason. But the workings of DNA establish a strong supposition for absence of selective pressure from the organismal level at a high percentage of nucleotide sites, where alternative states do not influence the phenotypes of organisms — hence the designation of drift at this level as the neutral theory of molecular evolution.
Kimura's classical categories of evidence all depend upon the observation that maximal rates of nucleotide change occur at sites that do not influence the organismal phenotype — on the reasonable assumption that organismal selection usually acts in the stabilizing mode to preserve favorable sequences, and that sites under selective influence must therefore change at less than the maximal rate. The threefold confirmation of this prediction provides powerful evidence for the neutral theory — (1) for synonymous substitutions of the third nucleotide in a triplet; (2) for much higher rates of change in untranslated introns than in surrounding exons; and (3) for entirely untranslated pseudogenes, where rates at all three positions of triplets match the rapid third-position rate for translated DNA.
The move from mere plausibility to the important claim for high, or even dominant, relative frequency arises both by implication from the basic theory, and from observation. The three phenomena described above, after all, include a large percentage of all nucleotide changes — so neutralism must maintain a high relative frequency at this level if we have interpreted the rates of change correctly. At the broadest scale of geological time, the (admittedly approximate) ticking of the molecular clock in so many phylogenetic studies achieves its most plausible reading as a consequence of generally comparable rates for the high percentage of neutral substitutions. (The alternate explanation of averaging out for fluctuating selective control over sufficiently long periods of time cannot be dismissed a priori, but smacks of special pleading — whereas neutralism expects this result as the consequence of a central proposition.)
Kimura has always stressed the high frequency of neutral substitutions as his main challenge to Darwinian traditions. He writes, for example (1991a, p. 367), “in sharp contrast to the Darwinian theory of evolution by natural selection, the neutral theory claims that the overwhelming majority of evolutionary changes at the molecular level are caused by random fixation (due to random sampling drift in finite populations) of selectively neutral (i.e., selectively equivalent) mutants under continued inputs of mutations.” At the same time, Kimura also consistently insisted — and not, I think, merely for diplomacy's sake, or for any lack of resolve, but rather with genuine conviction (I discussed the matter several times with Kimura in person, so I will also stand as witness) — that the neutral theory did not contradict or dethrone [Page 687] Darwinism, but should rather be integrated with natural selection into a more complete and more generous account of evolution. Most neutral changes, after all, occur “below” the level of visibility to conventional Darwinian processes acting at the organismic level. Moreover, although most nucleotide changes may be neutral at their origin, the variability thus provided may then become indispensable for adaptive evolution of phenotypes if environmental change promotes formerly neutral substitutions to organismic visibility — an important style of cross-level exaptation (Vrba and Gould, 1986; Gould and Lloyd, 1999) that may serve as a chief prerequisite to the evolution of substantial phenotypic novelty. Kimura writes, for example (1985, p. 43): “Of course, Darwinian change is necessary to explain change at the phenotypic level — fish becoming man — but in terms of molecules, the vast majority of them are not like that. My view is that in every species, there is an enormous amount of molecular change. Eventually, some changes become phenotypically important; if the environment changes, some of the neutral molecules may be selected and this of course follows the Darwinian scheme.” Thus, Kimura's statement exemplifies the central principle that the various levels of evolution's hierarchy work in characteristically different ways — and that levels can interact fruitfully in these disparate modes.
The chronological reaction of Darwinian hardliners to the neutral theory can be epitomized in a famous, if sardonic, observation about the fate of controversial theories. Tradition attributes this rueful observation to T. H. Huxley, but some form of the statement may well date to antiquity, the usual situation for such “universal” maxims. In any case, the earliest reference I know comes from the great embryologist von Baer, who attributed the line to Agassiz (von Baer, 1866, p. 63, my translation): “Agassiz says that when a new doctrine is presented, it must go through three stages. First, people say that it isn't true, then that it is against religion, and, in the third stage, that it has long been known.”
The first two stages unfolded in their conventional manner, with quizzical denial followed by principled refutation in theory (see p. 521 on Mayr's argument that neutralism cannot be true because we now know the ubiquity of selection). However, the third stage — still stubbornly occupied by some strict Darwinians — arose with an interesting twist, providing a cardinal illustration for this section's major theme: the dangers of parochialism, particularly the tendency to interpret all evolution from an organismal vantage point. Instead of simply stating that neutralism has long been known (so what's the big deal?), detractors now tend to say: “well, yes, it's true, and let's be generous and give Kimura and company due credit. But, after all, neutral substitutions only occur at sites without consequence for organismic phenotypes. So why focus upon such changes? Without any organismal effect, they can't be important in evolution. And no one can blame Darwin or Darwinian tradition for ignoring an invisible phenomenon.”
This exculpation of Darwin cannot be faulted in logic, but the rest of the argument reflects a narrow and discouraging attitude. Isn't the claim of unimportance absurd prima facie? How can anyone advance an argument for [Page 688] downgrading, as marginal, a process potentially responsible for more than half of all nucleotide substitutions — the supposed basis of evolution within a scientific ethos centered on reductionist preferences? Only a lingering prejudice for viewing organisms as a unique and intrinsic focal level could possibly generate such a claim.
Yes, an organism might view the world of its own compatriots as stable islands rising above an invi
sible sea — and choose to disregard random change within this swirling ocean of underlying, constant activity. But (if I may pursue this strained metaphor for a moment), any dynamic particle in the ocean could just as well, and perhaps with more merit, view the islands as rare and insignificant pedestals intruding into the truly fundamental substrate. May I just note the sterility of such a subjective argument, and state that any process with so strong an impact on change at any level cannot be unimportant in a world judged by relative frequencies.
As an illustration of the importance (and separability) of hierarchical levels, we may invoke balances produced by negative interaction among levels as a measure for the indispensability of molecular neutrality in full explanations of evolutionary phenomena. Just as a stable balance may arise by opposite forces of selection at adjacent levels, different processes — in this case neutrality at one level vs. selection at another — can also produce an intermediary result testifying to the importance of both styles of change. In such cases, moreover, neutrality enjoys a special heuristic advantage because random models yield general, quantitative predictions, while selectionist explanations usually require knowledge of particular circumstances that are much harder to decipher, and often impossible to quantify (for lack of requisite historical information).
For example, Spalax ehrenbergi, a blind Near Eastern mole rat, develops a rudimentary eye with an irregular lens that cannot focus an image. The eye is covered by thick skin and hair, and the animal shows no neurological response to powerful flashes of light (see p. 1282 for fuller discussion of this case in a different context). As expected under the neutral theory, the major lens protein, aA-crystallin, evolves much faster in S. ehrenbergi than in other murine rodents with normal vision (Hendricks et al., 1987) — nine amino acid replacements in a sequence of 173, over 40 million years of evolutionary separation, whereas the other nine rodents of this study show identical amino acid sequences, with no alterations at all from the ancestral state. But this rate of change for Spalax represents only 20 percent of the average for true pseudogenes, our best standard for a maximal and purely neutral pace of evolution. At a rate of alteration too fast for stabilizing selection, but too slow for pure neutrality, the results imply a dynamic balance between molecular drift and weakened selective control at the organismic level. (Suggestions for continued utility of a non-seeing lens include possible function in adjusting physiology to seasonal cues from changing day lengths, though we know no mechanism for perceiving such fluctuations without vision, see Haim et al., 1983; and developmental constraint based on formation of eyes as a necessary inducer of some later and fully functional feature in embryology.) [Page 689]
I close this woefully insufficient commentary by reemphasizing the point that our discomfort or disinterest in random change largely reflects the peculiarity of the individual and level that we know best — organisms — and does not record any rarity or impotence for stochastic forces as agents of phyletic change in evolution. Processes of drift probably exert least influence upon the organismic level, for the two reasons cited earlier: large population sizes, and a style of individuality that forges coherence by strict functional coordination of subparts, and therefore makes nearly every trait of the organism subject to selection strong enough to overwhelm drift.
But the organism is a unique and peculiar kind of individual — and these strictures upon drift do not apply so strongly at any other level. We have seen, in this section, how structural features of DNA impose neutrality or near-neutrality upon selection at a large percentage of sites, perhaps a majority. For this reason (and not by limitation of population size), randomness becomes a fundamental process of evolutionary change at the genic level, however weak such a force may be (or, indeed, may not be!) at the organismic level. We shall see that, at the highest levels of species and clades, randomness again attains a high relative frequency — but this time mostly as a result of low N for species in clades. If such different causes grant randomness a high relative frequency at several important levels of evolution's hierarchy — and if we can only assert low relative frequency at one level, and for reasons rooted in the peculiar character of individuality in this realm alone — then have we not committed a great conceptual error, and seriously narrowed our general view of evolution and the history of life, by giving short shrift to this most obvious of all alternatives to selection as a cause of change?
True genic selection. When future historians chronicle the interesting failure of exclusive gene selectionism (based largely on the confusion of bookkeeping with causality), and the growing acceptance of an opposite hierarchical model, I predict that they will identify a central irony in the embrace by gene selectionists of a special class of data, mistakenly read as crucial support, but actually providing strong evidence of their central error. Gene selectionists have always welcomed genuine cases of a phenomenon that they then falsely generalize to all evolution — that is, differential proliferation of genes within genomes for reasons acting at the genic level, and independent of effects introduced by downward causation from selection at any higher level.
Gene selectionists have naively embraced these examples as apparent confirmations of their belief that effectively all selection operates at this lowest level. If genes can work their magic even without a boost from the vehicles they usually employ as lumbering robots subject to their will, then our appreciation for their omnipotence can only increase. But such superficial admiration obscures a true distinction that actually illustrates the bankruptcy of exclusive gene selectionism. These examples do not showcase the maximal power of a ubiquitous phenomenon; rather, and quite to the contrary, they represent the only class of instances where pure and untrammeled gene selection can operate at all!
As argued previously in this chapter (pp. 613–644), when gene selectionists [Page 690] speak of genes using organisms as their vehicles, they commit a deep error by inverting causality and ascribing to genes (which only record the causal result, and therefore serve as good units of bookkeeping) the agency in natural selection that really belongs to the organism — for vehicles (or interactors) operate as units of selection, or causal agents of Darwinian evolution. But when genes do not use organisms as vehicles and engage in differential proliferation on their own accord, then the genes themselves do act as vehicles — and, consequently, can become units of selection. Gene selection only exists when genes can operate as vehicles (interactors); thus, these cases illustrate the restricted range of a process that gene selectionists naively regard as optimal illustrations of a ubiquitous phenomenon. The resulting irony deserves emphasis. Supposed best cases become only cases, and therefore disproofs of a generality when properly interpreted. Wilson and Sober (1994, p. 592) put the point well: “These examples have been received with great fanfare by gene-centered theorists as some sort of confirmation of their theory. However, they do not confirm the thesis that genes are replicators — all genes are replicators by definition and no documentation is needed. These examples are remarkable because they show that genes can sometimes be vehicles. They seem bizarre and disorienting because they violate our deeply rooted notion that individuals are organisms.”
Devotees of the genic level may eventually accept the defeat of their theory of exclusivity with good grace — for the supplanting hierarchical model provides more than enough room for true (and fascinating) examples of genuine genic selection, perhaps at quite high relative frequency once we acknowledge and learn to recognize the synergistic and orthogonal modes, as well as the better-documented examples of genic selection that harms organisms.
Moreover, when we recognize that many kinds and aggregations of genetic units can function in selection, the scope of this level becomes even wider. Selection may operate at the lowest unit of the nucleotide itself, if preferential substitution arises, for example, by differential production and consequently greater availability of one nucleotide vs. alternatives (the analog of natural
selection by birth biasing). Selection among entire genes and other DNA segments of comparable length may also hold great significance in evolution — as in Dover's important hypothesis of “molecular drive” (Dover, 1982).
In fact, we may be impeding a proper recognition of the substantial frequency of selection within genomes by naming the phenomenon for only one mode among many — “gene selection.” In the early days of Watson and Crick, biologists tended to conceptualize genomes as linear arrays of functional units (tightly strung beads with no spaces between in the usual metaphor). But we now know that most genes of eukaryotes, with their structure of exons separated by introns, do not maintain strict spatial continuity. Moreover, the functional genes of most complex metazoans represent, in any case, just a few percent of the full genome. All other kinds of genomic elements, forming an overwhelming majority of sites, can also evolve by processes of drift and selection. [Page 691]
For this reason, Brosius and Gould (1992) suggested that we use a more general term — “nuon” for nucleic acid sequence (DNA or RNA) — to recognize any stretch of nucleic acid, functional or not in organismic terms, that can evolve by differential origin or replication:
The Structure of Evolutionary Theory Page 110