9-11. From Lenski and Travisano, 1994. At too broad a scale, the size increases within Lenski's bacterial lineages seem gradual. But a finer scale analysis shows a stepwise punctuational pattern of clone selection with stasis recording the waiting time between favorable mutations, and punctuation caused by rapid sweep of these rarely favorable mutants through the population.
[Page 811]
non-causally correlated with coincident evolutionary expressions at other scales; and (2) in principle, genetic expressions of a common causal structure do not rank as intrinsically more “deep,” “real,” “fundamental,” or “basic” than other manifestations in different forms and at different levels; causal relevance depends upon the questions we ask and the processes that organisms undergo.)
Second, the “allometric” effects of scaling either render the same process in a very different manner at various scales, or (perhaps more frequently, and primarily in this case at least) generate the distinctive patterns of different scales by independent processes, acting simultaneously, but with each process primarily responsible for results at its own appropriate level.
If I could affirm, as may well often be the case, that punctuated equilibrium regulated the phenotypic pattern of evolution in a given clade, while genotypic distances conformed closely to a “molecular clock,” I would not conclude that punctuated equilibrium had therefore been downgraded, or exposed as incorrect, superficial, or illusory — with genetic continuity as a physically underlying (and conceptually overarching) reality. Rather, I would regard each result as true and appropriate for its own scale and realm — with the full pattern of legitimate difference standing as an intriguing example of resolvable complexity in evolutionary scaling and causality. Moreover, this particular pattern might easily result from a highly plausible scenario of complex and multileveled causation — namely, that neutral substitutions at the nucleotide level impart a signal sufficiently like a genomic metronome to dominate the molecular results, while ordinary speciation both regulates the phenotypic history of populations, and works by the expected pattern of punctuated equilibrium. The genomic results, in principle, need not extrapolate to encompass the pattern of speciational (macroevolutionary) change. After all, we do understand that gene trees do not entirely match organism trees in phylogeny!
In this way — as in the foregoing example of predictable differences between asexual unicells and sexually reproducing metazoa — punctuated equilibrium proves its value primarily by hypothesizing sensible distinctions: that is, by operating at scales and biological conditions where cladogenetic speciation plausibly sets evolutionary pattern. Punctuated equilibrium should not prevail where species cannot exist as Darwinian individuals, or where continuously occurring, and largely nonadaptive, substitution of nucleotides probably regulates the bulk of genomic change. In this crucial sense, punctuated equilibrium becomes a valuable hypothesis by delineating such testable distinctions, rather than allowing evolution to be conceptualized monistically as a single style of alteration, or a single kind of process either flowing from, or applicable to, all scales of change.
The question of consistency between observed genetic patterns in living species, and the relative frequency of punctuated equilibrium in their phylogeny, shall be treated in the next section on the correspondence of punctuated equilibrium with predictions of evolutionary modeling. But one genetic issue has been widely discussed in the literature, and should be included in this section [Page 812] on empirical results. Several researchers have noted that punctuated equilibrium implies a primary prediction about patterns of genetic differences among species: if most change accumulates at ruptures of stasis during events of speciation, and not continuously along the anagenetic history of a population, then overall genetic differences between pairs of species should correlate more closely with the estimated number of speciation events separating them, than with chronological time since divergence from common ancestry. (This prediction might be clouded by several factors, including the foregoing discussion on attributing the bulk of genomic change to continuity at a lower level, and a number of potential reasons for discordance between phenotypic effect and extent of responsible genetic change. But I certainly will not quibble, and I do allow that punctuated equilibrium suggests the broad generality of such a result.)
In the early days of debate about punctuated equilibrium, Avise (1977) performed an interesting and widely discussed test. In comparing genetic and morphological differences among species in two fish clades of apparently equal age but markedly different frequencies of speciation, Avise found a higher correlation of distances with age than with frequency of branching, and therefore favored gradualism over punctuated equilibrium as an explanation of his results. But Mayden (1986) then showed that Avise's test did not apply well to his chosen case (primarily because we cannot be sure of roughly equal antiquity for the two clades). He then argued, as several supporters of cladistic methodology had urged, that such tests should be applied only to well-confirmed cladistic sister groups — for, in such cases, even if paleontological data permit no certainty about the actual time of joint origin from common ancestry, at least we can be confident that the two clades are equally old! Mindel et al. (1989) then performed such a properly constituted test on the reptilian genus Sceloporus, and more loosely on allozymic data in general, and found a positive correlation between evolutionary distance and frequency of speciation — thus validating the primary prediction of punctuated equilibrium.
Empirical tests of conformity with models
Limitations of the fossil record restrict prospects for testing punctuated equilibrium by inductive enumeration of individual species and lineages. Cases with sufficient resolution may not be common enough to establish a robust relative frequency; or systematic biases based on imperfections in the fossil record may lead to artifactual preferences for punctuated equilibrium — thus making the data unusable as a fair test for a minimal frequency. (I do not regard these problems as particularly serious, and I will provide several examples of adequate resolution in the next section of this chapter. But we should, in the light of these difficulties, also be exploring other ways of testing punctuated equilibrium, as considered below.)
In another strategy that has been pursued by some researchers, but could (and should) be exploited to a much wider and more varied extent, we might characterize, in quantitative fashion, broader patterns in the deployment of [Page 813] diversity through time and space in major taxonomic groups — and then devise tests to distinguish among contrasting causes: anagenetic vs. cladogenetic; gradual vs. punctuational. If certain well-defined patterns can only be generated, say, by branching speciation rather than by anagenetic transformation (or vice versa, of course), then we can use the fit of broad results with distinctive models, rather than minute documentation on a case-by-base basis, to establish the relative frequency of punctuated equilibrium.
In an important study, for example, Lemen and Freeman (1989) investigated “the properties of cladistic data sets from small monophyletic groups (6-12 species) ... using computer simulations of macroevolution” (p. 1538). They contrasted the differing outcomes of data generated under anagenetic gradualism vs. punctuated equilibrium, and then examined cladograms of extant monophyletic groups for consistency with these “abstract, end-member” alternatives. They claimed better support for gradualism, but several flaws in their logic and data render their conclusions moot, much as their pioneering approach may be applauded and recommended for further study.
Lemen and Freeman tested actual data against three modelled differences between cladograms generated by gradualism vs. punctuated equilibrium.
1. Punctuated equilibrium should produce a strongly positive correlation between number of branching points and apomorphies of species — because change occurs at speciation and does not further correlate with passage of time per se. Lemen and Freeman's models did affirm this expected result, and real data, somewhat ir
onically, revealed “higher correlations of apomorphies and branch points than could be explained by either mode of macroevolution” (p. 1549). But the authors rejected an interpretation of this information as favorable to punctuated equilibrium because anagenesis, under certain conditions, also yields positive correlations, and because high correlations can also arise artificially by errors in establishing cladograms: “a consistent error in polarity can profoundly affect the correlation of total apomorphies and branch points” (p. 1551). Fair enough, but Lemen and Freeman are not nearly so circumspect when equally flawed data seem to favor their preferred alternative of gradualism.
2. The modal (but not the mean or median) number of autapomorphies will always be zero under strict punctuated equilibrium. This odd-sounding situation arises because, in the cladograms, an event of branching produces a daughter species with some autapomorphies and a persisting parental species remaining in stasis with none. With no change except at branching points, the value of zero autapomorphies must remain most common across all species on the cladogram. Under gradualism, autapomorphies simply accumulate through time, whatever the pattern of branching, so zero should not mark a preferred or particularly common value.
Lemen and Freeman never found a mode of zero autapomorphies in real data, and therefore rejected punctuated equilibrium as a predominant style of evolution. But had they pursued explanations based on artifacts (as they did so assiduously when the data seemed to favor punctuated equilibrium), they would have realized that taxonomic practice precludes the definition (or even [Page 814] the recognition) of species without autapomorphies. Such species arise frequently in the modelled system as a necessary consequence of the chosen rules of generation and the general logic of cladistic analysis. But, in neontological practices of naming, a species without autapomorphies represents an oxymoronic concept, and such taxa could never be designated at all. Lemen and Freeman recognize this point in writing about their various forms of gradualistic modelling (p. 1551): “When distinctness of species is demanded the lack of autapomorphies may not be the most expected condition.”
3. Under punctuated equilibrium, “as the number of characters used in the analysis increases, the distribution of the number of autapomorphies per species becomes bimodal. Under gradualism, the distribution of autapomorphies remains unimodal under all conditions” (1538). This situation, a spinoff from their second criterion, arises because each branch, in an event of punctuated equilibrium, produces one changed descendant and one persisting ancestor — and the more characters you measure, the more you pick up the differences between stasis on one branch and change on the other. Under gradualism, total change correlates only with elapsed time; so accumulating autapomorphies should form a unimodal distribution so long as species duration remains unimodal as well.
Lemen and Freeman found no bimodal distributions in real data, and therefore concluded again in favor of gradualism. But, once more, the differences between idealized modeling and data from real organisms scuttles this conclusion. In the models, we know for sure that long arms without branching are truly so constituted, for we have perfect information of all simulated events. These unbranched arms, under punctuated equilibrium, should accumulate no autapomorphies — and the low mode of the bimodal distribution arises thereby. But, in real data of cladograms based on living organisms, long unbranched arms usually (I would say, virtually always) record our ignorance of numerous and transient speciational branchings that quickly became extinct and left no fossil record. (Moreover, since Lemen and Freeman's cladograms only include living organisms, even if successful and well-represented fossil species existed, they would not be included.) When we note a long arm without branches on a modern cladogram, and then assume (as Lemen and Freeman did) that accumulated autapomorphies between node and terminus must have arisen gradually and anagenetically, we commit a major blunder. We have no idea how many unrecorded speciation events separate node and terminus, and we cannot assert that recorded autapomorphies did not occur at these (probably frequent) branchings. In other words, Lemen and Freeman's bimodality test assumes that unbranched arms of their cladograms truly feature no speciation events along their routes, whereas numerous transient and extinct species must populate effectively all of these pathways.
Other applications of this method — modeling of alternative outcomes and testing of contrasting predictions against patterns of real data — have yielded results favorable to punctuated equilibrium. In a path breaking paper, Stanley (1975 — see elaboration in Stanley, 1979 and 1982) first proposed this style of testing and developed four putative criteria, all affirming punctuated equilibrium. [Page 815] (Stanley's tests may be reduced to three, as his second “test of the Pontian cockles” represents a particular instance of his first “test of adaptive radiations.” Stanley argued:
Test of adaptive radiation. After calculating average species durations from the fossil record, one can affirm that pure anagenetic gradualism (or temporal stacking of species end-to-end) cannot account for the magnitude of recorded adaptive radiations in the time available — so rapid cladogenesis must be invoked.
Test of living fossils. Punctuated equilibrium associates realized amounts of change primarily with frequency of speciation, anagenetic gradualism primarily with elapsed time. If so-called “living fossils” — ancient groups with little recorded change — also show unvarying low diversity through time, then we can affirm the primarily prediction of punctuated equilibrium, and refute the corresponding expectation of gradualism (for these groups are ancient). Stanley then documented such a correlation between clades identified as “living fossils” and persistently low diversity in these clades.
Test of generation time. Under gradualism, amounts of realized evolution should correlate strongly with generation time — for the time that should mark accumulated evolutionary change does not tick by an abstract Newtonian clock, but by number of elapsed generations, representing the number of opportunities for natural selection to operate. But, under punctuated equilibrium, amount of change correlates primarily with frequency of speciation — a property with no known relationship to generation time. Stanley then cited the well-documented lack of correlation between evolutionary rate and generation time as evidence for the prevalence of punctuated equilibrium (fast-evolving elephants vs. stable invertebrates with short generations).
Much as I regard Stanley's arguments as suggestive, I cannot accept them as conclusive for two basic reasons. First, other plausible explanations exist for the patterns noted. For example, many reasons other than the prevalence of punctuated equilibrium might explain a lack of correlation between realized evolution and generation time, even in a world of anagenetic gradualism. The correlation might simply be weak or too easily overwhelmed (and therefore rendered invisible) by such other systematic factors as variation in the intensity of selection. (Maybe elephants, on average, experience selection pressures higher by an order of magnitude than those affecting short-lived invertebrates; maybe population size overwhelms the factor of generation time.)
Second, most of Stanley's tests (particularly his key claim about adaptive radiation) don't really oppose punctuated equilibrium to gradualism, but rather contrast a more general claim about the speciational basis of change (whatever the mode of speciation) with anagenesis. Moreover, the tests employ a somewhat unfairly caricatured concept of gradualism. I doubt that the most committed gradualist ever tried to encompass the maximal change between ancestor and any descendant in an adaptive radiation by stacking species end to end, and then calculating whether the full effect could arise in the allotted time. A committed gradualist might fairly say of an adaptive radiation: [Page 816] “of course the magnitude of change in both form and diversity correlates with number of branching events (what else could a 'radiation' mean). But adaptive radiations only accelerate the frequency of branching in response to ecological opportunity ('open' environments just invaded or just cleared out by ex
tinction); they do not affect the modality of change. I will allow that, in adaptive radiations, most new species arise in less time than usual, but still gradualistically. If full speciation takes half the average time (one million rather than a modal two million years, for example), but still occurs imperceptibly and still occupies a large percentage of an average species's lifetime, then gradualism encounters no threat in adaptive radiation.”
However, in another crucial sense, at least one of Stanley's tests does illustrate the most salutary potential role for punctuated equilibrium: its capacity to act as a prod for expansive thought and new hypotheses, whatever the outcome of the empirical debate about relative frequency. Paleontologists had been truly stymied in their thinking about the important and contentious topic of “living fossils.” Neither of the two conventional explanations could claim any real plausibility. Every textbook that I ever consulted as a student dutifully repeated the old saw that living fossils had probably achieved optimal adaptation to their environment. Therefore, no alternative construction could selectively replace an ideal form achieved so long ago. But no one ever presented any even vaguely plausible evidence for such a confident assertion. Why should horseshoe crabs lie closer to optimality than any other arthropod? What works so well in the design of lingulid vs. other brachiopods? What superiority can a lungfish assert over a marlin or tuna? In fact, since living fossils also (by traditional depiction) present such a “primitive” or “archaic” look, the claim for optimality seemed specially puzzling.
The Structure of Evolutionary Theory Page 130