Reappraising adaptive radiation

Abstract

Books with a point of view are more interesting than those without one. The editors of Molecular Evolution and Adaptive Radiation, Tom Givnish and Ken Sytsma, have mixed fascinating case studies of adaptive radiations in a diversity of taxa with a very particular point of view regarding this ill-defined evolutionary process. Their notion of adaptive radiation is provocative, and, to the editors’ credit, they do not just implore the community to reappraise the concept adaptive radiation; they take a stab at it themselves. In a preface, two introductory conceptual chapters and a third empirical chapter, they argue for one definition of adaptive radiation, outline procedures for studying it, and criticize current methodologies along the way. According to the editors, one of the primary motivations for the book (and the symposium that prompted it) is to take advantage of robust molecular trees to study adaptive radiation from a phylogenetic perspective. The irony is that much of their critique is aimed at approaches that are explicitly phylogenetic: the “comparative method,” the phylogenetic analysis of adaptation, and the phylogenetic analysis of lineage diversification rates. A perusal of the book's contributed chapters does not immediately resolve this apparent contradiction. A diversity of opinion is expressed about what an adaptive radiation is and how it should be studied, and many contributors disagree with at least some elements of the general framework proposed by the editors. Regardless, the book represents an outstanding compendium of case studies of morphological and ecological diversification analyzed in a phylogenetic framework. Of the 19 empirical chapters, eight are studies of plant taxa (all angiosperms). Four of the angiosperm studies are island radiations in the strictest sense: Hawaiian silverswords (Baldwin), Hawaiian Caryophyllaceae (Sakai et al.), and Macaronesian Argyranthemum (Francisco-Ortega et al.), or involve island-like scenarios, as in bromeliads on Guyanan tepuis (Givnish et al.). Three others involve putative transitions to new adaptive zones, as in the aquatic family Pontederiaceae (Barrett and Graham), or orchid twig epiphytes (Chase and Palmer), and one explicitly considers the role of so-called “key innovations” in radiations (nectar spurs in columbines and other angiosperms—Hodges). The last paper, on the orchid genus Platanthera, is less easily categorized. Among other things, it addresses one of the book's recurring themes—that morphological data are too homoplastic to be useful for inferring phylogeny (more on that later). Strangely missing from the plant studies are cases of adaptive radiations at higher taxonomic levels. Not a single plant study considers adaptive radiation at even the level of angiosperm families, let alone, say, angiosperms as whole. The animal studies are less biased in this respect. They include the radiation of marsupials (Springer et al.), echinoids (Smith and Littlewood), Old World fruitbats (Kirsch and Lapointe), and New World Monkeys (Horowitz and Meyer). One wonders if modes of adaptive radiation at higher levels differ in plants and animals, or whether this is merely a sampling artifact in what botanists are studying, or the choice of contributors to the book. Classic island-type scenarios in animal taxa are also included, including Hawaiian Drosophilids (Kamysellis and Craddock), East African rift lake cichlids (Reinthal and Meyer), and Caribbean Anolis lizards (Jackman et al.). A few papers are difficult to pigeonhole, including a review of rates of speciation in fishes (McCune), one on phenotypic evolution and plasticity in Daphnia (Colbourne et al.), and one on experimental ecological approaches to diversification in stickleback fishes (Taylor et al.). The papers are characterized by an unusually uniform and careful attention to methodological issues in phylogeny reconstruction. Contributors (and perhaps we have the editors to thank) were zealous in their attention to the details. Nearly all the trees have support values (either bootstrap or decay indices or both); tree statistics are reported in figure captions and trees are labeled with the source of data, which makes wading through multiple lines of evidence in many papers much easier. Even more delightful is the large number of trees that have characters mapped onto them, often with illustrations of morphology or ecological traits attached. This makes the book a valuable source of studies of character evolution in its own right. These studies exemplify the potential for phylogenetic history to inform us about the chronicle of evolution, to paraphrase Springer et al.'s chapter title. In an ambitious historical and conceptual overview, one of the editors, Tom Givnish, argues for a view of adaptive radiation that is a subset of conventional usage. Minimally, modern views of adaptive radiation tend to consider three issues: lineage diversification (speciation and extinction), character (usually morphological) diversification and ecological diversification. The relationship between characters and ecology is potentially quite direct—via adaptation—if not easily documented. The relationship between lineage diversification and the other two is less obvious, but ever since Simpson's work, the potential link between rapid lineage diversification and adaptive divergence of traits has formed part of the conceptual framework surrounding adaptive radiation. Simpson's definition as cited by Givnish is “More or less simultaneous divergence of numerous lines from much the same ancestral adaptive type.” Note that this explicitly mentions the number of lineages (“numerous”) and the rapidity of their divergence (“simultaneous”), but Givnish rejects this definition in favor of the older and, I think, vaguer view of Osborne's: “Differentiation in habit in several directions from a primitive type.” Osborne does not mention rate, and it is likely that he was not thinking about elevated rates of speciation to the same extent that Simpson was. Givnish make no bones about that point in his own definition: “The origin of a diversity of ecological roles and attendant adaptations in different species within a lineage.” The number of species is presumably irrelevant as long as it is more than one. He then criticizes recent efforts to quantify rates of lineage diversification—the production of new species—in phylogenies and interprets these efforts as an attempt to restrict the definition of adaptive radiation to lineage diversification (p. 10): Defining adaptive radiation in this way is inappropriate and, I believe, ultimately unproductive. It ignores the traditional usage and conceptual framework established for adaptive radiation, dating back more than a century, at least to Osborne (1893) if not Darwin (1859). It focuses on species number (perhaps a preoccupation of some systematists) to the exclusion of all other evolutionary phenomena. Most importantly it conflates “adaptive radiation” with mere speciation. As Huxley (1942) and Mayr (1942, 1963, 1970) argued so convincingly as a central evolutionary thesis, speciation often occurs simply because gene flow is interrupted by geographic barriers or intrinsic biological factors… Clearly, such speciation may have nothing to do with adaptive radiation. He cites six papers published in the last five years (including three of mine) as representing this view. Such a view would be overly restrictive, but, inexplicably, five of the six papers cited (Slowinski and Guyer, 1993 [cited as 1994 in the book]; Sanderson and Donoghue 1994, 1996; Barraclough, Harvey, and Nee, 1996; Sanderson and Wojciechowski, 1996) never even use the term “adaptive radiation” and hence it is unlikely they were attempting to redefine it. Instead, those papers examined lineage diversification rates, either in their own right, or in conjunction with so-called “key innovations”, another slippery concept that sometimes is discussed in relation to adaptive radiation, and sometimes is not (Nitecki, 1990). Granted, another paper by Guyer and Slowinski (1993) did discuss lineage diversification rates in the context of adaptive radiation, but in several places in that paper they specifically referred to rapid lineage diversification per se as “one type of adaptive radiation,” i.e., the kind that is associated with increases in species richness. Givnish may be reacting to a movement in the literature that doesn't exist—a mythical attempt to restrict the meaning of adaptive radiation to “mere speciation.” This conception of adaptive radiation seems to be gaining a foothold. Several other contributors to the book question the importance, in principle, and in their own case studies, of lineage diversification in adaptive radiation. Barrett and Graham suggest that “…the number of lineages arising from an adaptive radiation is of secondary importance to the pattern of character diversification among lineages” (p. 225). They hint that, if anything, the 50-odd invasions of aquatic habitats by angiosperm taxa have often been accompanied by reduced lineage diversification rates. Jackman et al. state that “Contrary to some recent discussions of adaptive radiation, the relevant criterion is not the number of species, but the adaptive disparity among these species; even clades with relatively few species can constitute an adaptive radiation if the species demonstrate considerable ecological and morphological disparity” (p. 535). They argue that different traits may actually be responsible for different components of diversification. The dewlap, an extensible fold of skin on the throat used for behavioral display, may have promoted speciation rates in anoles, but a separate innovation, toepads, led to arboreality, the trait involved in the main adaptive diversification of morphology and ecology. However, other contributors take a less dismissive view of the importance of lineage diversification. Three papers go so far as to estimate lineage diversification rates. McCune reviews a vast body of literature on species-richness in clades of fishes, arguing that the East African cichlids are undergoing net rates of speciation several orders of magnitude higher than other taxa. Hodges argues that nectar spurs in columbines have increased rates of lineage diversification, a pattern that he sees in seven of eight other comparisons of angiosperm taxa with floral spurs. Baldwin points out that the Hawaiian silversword alliance is more species-rich than its continental relatives, although sample size is too small for significance using conventional null models of diversification. Several other authors use the phrase “explosive speciation” as if it were an issue of direct relevance to adaptive radiation. But the idea that lineage diversification can and should be decoupled from other aspects of diversification is a tantalizing one. One way to summarize the possible modes (and definitions) of adaptive radiation is by reference to a table such as Fig. 1, which portrays various combinations of lineage and “adaptive” (character/ecological) diversification. Givnish (and others) argue that bona fide adaptive radiations are those that fall into the high adaptive/low lineage diversity cell of the matrix, whereas other “nonadaptive radiations” fall into the low adaptive/high lineage diversity cell. Givnish et al. (Chapter 8) contrast two closely related bromeliad genera, Brocchinia (∼20 spp.), which exhibits tremendous variety in mechanisms of nutrient capture, and Navia (76 spp.), which is much less variable ecologically. The former they regard as an adaptive radiation, the latter a nonadaptive radiation. If this framework is useful, it raises the exciting prospect that many large angiosperm genera will have to be evaluated in the context of competing hypotheses of adaptive vs. nonadaptive radiation. Astragalus, Senecio, Euphorbia, Carex, and Psychotria all have more than 1000 species. Are any of them adaptive radiations? Adaptively diverse clades that are also species-rich fall into the fourth cell of the matrix of Fig. 1 but are relatively neglected by this book (except Chapters 12 and 21, which deal with East African cichlids). Angiosperms and birds are obvious examples. However, a consideration of this cell brings up a lurking and troublesome issue of scale. How should adaptive radiations, large or small, be circumscribed? Consider that every famous, well-studied island adaptive radiation is nested within some larger, more species-rich clade, so it is always possible to turn a species-poor/adaptively diverse group into a species rich/adaptively diverse group just be moving deeper in its phylogenetic history. One answer is that phylogenies come with ready-made comparisons in the form of sister groups. The equivalence of sister group ages has been exploited to compare relative rates of molecular evolution, relative rates of lineage diversification, and to test hypotheses of rate constancy. Surprisingly, however, adaptive divergence has rarely been examined by sister group comparisons. One way to test whether a clade exhibits a pattern of unusually high adaptive divergence would be to compare it to its sister group using measures of morphological disparity, for example (Foote, 1996). By this method, rank-based and other taxonomic artifacts are eliminated. A test of whether or not Brocchinia is really an adaptive radiation would involve not a comparison with some other genus of close but uncertain relationship, but rather with Brocchinia's sister taxon (rank irrelevant). Alternatively, knowledge of the timing of key events in the diversification of a clade would permit an estimate of absolute rates of morphological, ecological, and lineage diversification, which in turn could be compared from taxon to taxon. This, in fact, is the only way in which we can meaningfully compare one taxon to another one that is not its sister group. Thus to compare Brocchinia to Navia (assuming they are not sister taxa), we need to know the age of the most recent common ancestor of each, from which we can infer the average rate of diversification based on extant diversities. However, in a somewhat radical view, some contributors argue that it is not the rate of adaptive divergence that matters but the amount. Barrett and Graham say that, “a radiation in slow motion is still a radiation” (p. 225). Givnish goes further: “. . . the Australian marsupials would still be an adaptive radiation whether they diverged in ten thousand or ten million years” (p. 8). If it is true that differences in rate of divergence of three orders of magnitude are unimportant to the definition of adaptive radiation, then the task of distinguishing adaptive from nonadaptive radiations may be very difficult indeed. The only way this view makes any sense is if there is some threshold amount of divergence required for any clade to be considered an adaptive radiation. Above that threshold (and hence over many orders of magnitude above it) is an adaptive radiation, but below it is not. However, if one moves deeper in the tree, that threshold will surely be achieved. Perhaps the reductionist scheme of Fig. 1 is just too simple. Either explicitly or implicitly, most concepts of adaptive radiation have species diversity lurking within them. I suspect that the authors who question the relevance of species-richness are mainly concerned with causality: character diversification may promote, allow, or even drive lineage diversification, but not the other way around. If true, this view makes character diversification the key ingredient in an adaptive radiation. But at a basic level it cannot be true. A clade of one species cannot show high character diversity—nor can a clade of two species. The number of combinations of morphological/ ecological features that can be found across a clade increases monotonically with the number of species in that clade. Character and species rates can interact in many different ways. Rates of speciation may sometimes be the rate-limiting component in adaptive radiations, in cases in which adaptive divergence is extremely rapid, or, alternatively, rapid high lineage turnover may permit at least some taxa to adapt to appropriate novel opportunities. Even in the narrower definition of adaptive radiation advocated by the editors, the “adaptive” part is important, and the methodology for studying adaptation is a key issue in this book. The study of adaptation has changed dramatically in the last 20 years in what has been termed the “historical approach,” largely spurred by Gould and Vrba's macroevolutionary redefinitions (aptation, exaptation, adaptation, etc.), and subsequent phylogenetic refinements of those concepts (see Rose and Lauder, 1996). These tools permit the falsification of certain hypotheses about adaptations if, for example, it is discovered that the putative functional advantage of a trait actually evolved prior to or long after the supposedly adaptive trait evolved. Contemporaneously, the development of the statistical “comparative method,” which permits corrections for phylogenetic nonindependence, had a dramatic impact on studies of the relationship of a trait and hypothesized ecological correlates (Harvey and Pagel, 1991). In purely phylogenetic/historical analyses of adaptation, in which experimental studies of fitness are not possible or practical, these two historical approaches represent important refinements of mere descriptive or correlational studies. However, they are not completely uncontroversial. For example, Lauder (1996) scrutinized the phylogenetic study of adaptation and identified a number of weaknesses. Givnish (p. 17) is also skeptical of its role in the study of adaptive radiation: “…while heuristic, the aptation/exaptation terminology does not seem particularly useful.” The comparative method has recently come under even stronger attack, primarily from ecologists (e.g., Ricklefs, 1996) and Givnish strongly echoes this critique in listing four shortcomings of this phylogenetic approach to adaptation (pp. 22–23). Jackman et al.'s paper on the multiple radiations of Anolis lizards takes the opposing view. Knowledge of the phylogenetic relationships of anoles and their relatives reveals that toepads evolved slightly after arboreality (life in trees), and hence it is a phylogenetically sensible hypothesis that they represent an adaptation to enhance arboreal capabilities (had they evolved prior to arboreality they would be an exaptation). This seems to be a “useful” distinction after all. This same paper also contains one of only two comparative method analyses in the book (showing a strong relationship between hindlimb length and perch diameter and between hindlimb length and run frequency). In the other one, Kirsch and Lapointe looked at the relationship between diet (percentage nectarivory) and tooth anatomy in Old World fruitbats and found a strong correlation that held up after taking phylogenetic nonindependence into account. Interestingly, the authors repeated these analyses on molecular trees, a morphological study of Springer's and a “tree” derived from a 1912 paper, and found significant correlations on only the molecular trees. It was disappointing to see no botanical case study of the statistical comparative method in this book, but there have been woefully few published anywhere. On the other hand, the methodologies used for reconstructing character evolution are nearly state-of-the-art. Barrett and Graham's analysis is a good example. They use parsimony reconstruction's on several alternative trees and employ MacClade's equivocal-cycling option to obtain all equally parsimonious reconstructions. Other authors are not quite so exhaustive but most consider both ACCTRAN and DELTRAN optimizations in an attempt to bracket extremes. Kambysellis and Craddock's chapter on Hawaiian Drosophila wins the award for sheer artistry in displaying detailed drawings/photographs next to their MacClade reconstructions of character evolution. Given the rarity of contributions that use the comparative method (sensu Harvey and Pagel) or the historical framework of adaptation (i.e., Gould and Vrba), what use are all these phylogenetic reconstructions of character evolution? I suspect that phylogenies are being used to discover homoplasy (multiple evolution of the same trait) in characters, and homoplasy is being taken as prima facie evidence of adaptation. The idea, of course, is that only traits that are under strong selection will evolve many times among closely related species. In the orchid genus Platanthera, for example, placement of pollinia on the eyes of pollinating moths has evolved at least seven times (Hapeman and Inoue's chapter). Despite their own argument that placement on the proboscis ought to result in more frequent successful pollination than placement on the eyes, because eye-placement evolves so many times, the authors conclude that it must have “some selective advantage.” Presumably, the converse need not be true: a trait evolving only once need not be neutral! This is extremely indirect evidence of adaptation. The use of homoplasy in the study of adaptation was reviewed recently in a paper by Larson and Losos (1996). They argue that homoplasy allows replicated tests of general hypotheses of adaptation. However, their prescription goes beyond merely counting the number of times a trait evolves in a clade. Each instance must pass some reasonably stringent set of criteria before it can be regarded as an adaptation, and only then does the existence of replicated specific adaptations suggest a general pattern of adaptation. According to Hodges (Chapter 13), nectar spurs have evolved at least 15 times in angiosperms. Although it is quite easy to imagine that they are adaptive in each case, because of the different morphologies and pollinators involved, one would like a separate test of the adaptedness of each. If all or many were demonstrably adaptive, only then would the idea that “nectar spurs are adaptive in angiosperms” be well supported. Moreover, although a trait evolving numerous times may be suggestive of adaptation, it seems neither necessary nor sufficient in principle to have high levels of morphological homoplasy in an adaptive radiation. Given enough truly novel innovations in a clade, high levels of divergence can be achieved with low levels of homoplasy. Conversely, large species-rich (nonadaptive?) radiations may show extremely high levels of homoplasy on account of repeated reassortment of the same character states in similar allopatric environments. The several thousand species in the legume genus Astragalus have evolved many of the same traits involving fruit morphology, life cycle, leaf morphology, and indumentum countless times, but most of them would probably not be regarded as spectacular innovations to the average botanist. It is a purely empirical question whether or not adaptive radiations show higher levels of homoplasy (or just plain disparity) than other clades. This book cries out for objective and quantitative comparisons of homoplasy and divergence in adaptive radiations. Many superlatives are used throughout the book to describe divergence patterns (“spectacular” for two radiations; “remarkable” for three; “unparalleled”, “tremendous”, and “dramatic” for one each!), and they are probably apt in most cases, but if phylogenetics has disinterred one issue in comparative biology, it is the potential role of artifacts in obscuring evolutionary patterns. Givnish et al. (Chapter 8) strive for this kind of quantitative statement when they assert that Brocchinia has evolved more sorts of nutrient capture adaptations than any other angiosperm genus (p. 261), but consider the possible artifacts in this apparently unambiguous claim: the age of the taxa, their species-richness, and the caprice of taxonomic circumscription of genera. Hopefully. this volume will inspire further quantitative analyses. A recurring idea in this book is that molecular systematic data permit reconstruction of phylogenetic history more accurately than morphological data (p. xiii). This has less to do with adaptive radiation as a process and more to do with how to make reliable inferences about it. Assuming phylogenies are useful, it makes sense to obtain good ones. At the root of this theme is the notion that morphological data are more homoplastic than molecular data. Several of the contributors support this idea. For example, both the papers on diversification in orchids comment on how classical taxonomic reliance on floral traits has been misleading (e.g., Chase and Palmer, p. 334). Hapeman and Inoue (Chapter 15) say, “Given the frequency of convergence and reversal in floral morphology, it is clear that floral characters are unsuitable for inferring phylogeny in Platanthera.” Reinthal and Meyer argue for the superiority of DNA sequence data relative to “homoplasy-ridden morphology” in East African cichlid fish species flocks (p. 376). These are bald and unsubstantiated assertions. Givnish and Sytsma, however, make an elaborate and lengthy case for the higher accuracy of molecular data (Chapter 2). Their meta-analysis of 134 published phylogenetic analyses concludes that molecular data have significantly less homoplasy than morphological data (at least RFLP data—sequence data are annoyingly homoplastic too, cf. p. 79). Then, by appealing to the results of a simulation study published elsewhere (Givnish and Sytsma, 1997), which concludes that higher homoplasy means lower accuracy, they infer that morphological data are less accurate than molecular data. This conclusion is couched as a rebuttal to previous meta-analyses of homoplasy (published by Michael Donoghue and myself over the last 10 years), but the argument seems a bit irrelevant at the end of the 1990s. Thumbing through the collection of phylogenies in this book, I find myself much more interested in the support values for clades on trees than I am with whether or not the tree is based on molecular or morphological data. I want to know, for example, whether to believe Chase and Palmer's inference that the radiation of orchid twig epiphytes started with vegetative and life history traits and later moved on to floral diversification in a “leapfrog” pattern. These conclusions depend on the tree being right. Had the authors merely reported that it was based on molecular data and that on average molecular data are more reliable than morphological data—but no bootstrap values were reported—my inclination would have been to turn the page. It seems reckless (and potentially wasteful) to use the class of data as a surrogate for the reliability of any particular data set or any particular clade, especially in view of the tremendous range of variation exhibited by molecular (and morphological) data sets (see fig. 2.2 in Chapter 2) and given that we have tools for assessing the robustness of phylogenetic hypotheses whatever their source. The theme of the intrinsic superiority of morphological data is closely paired with the notion that to understand adaptive radiation, which involves morphological diversification, it is necessary to reconstruct phylogenies without using morphology, to avoid circularity (e.g., pp. 6, 450). In principle it could be dealt with merely by excluding from any analysis those specific morphological characters that are implicated in the adaptive divergence of the taxon. However, Givnish and Sytsma (Chapter 2) argue for the relegation of all morphology to second-class citizenship because of the possible existence of “concerted homoplasy,” in which suites of interacting and integrated characters evolve multiple times by convergent adaptation. Because we do not know ahead of time whether a given morphological character is involved in such concerted homoplasy, we should wonder about all morphology. As a result they propose (p. 90) that morphological data are best suited for recognizing species and suggesting broad patterns of relationships, and that molecular data—by dint of the greater number and higher consistency of characters involved—are a more precise guide to detailed phylogenetic relationships, barring any problems arising from hybridization, introgression and lineage sorting. The hypothesis that concerted homoplasy is enough of a problem to abandon or at least downweight an entire class of phylogenetic data is testable. It predicts significant, widespread incongruence between molecular and morphological data sets. My reading of the plant molecular systematics literature on incongruence suggests that there is at least as much incongruence between lines of evidence from different genes or genomes as there is between any gene and morphology. Much apparent incongruence is anecdotal and goes away when it is noted that the clades in dispute are not well supported in one or both analyses. This book itself provides some interesting examples. Springer et al. (Chapter 4) analyzed a morphological matrix of 102 characters in which only four clades were supported at better than the 50% bootstrap level. Comparing this tree to their composite phylogeny based on DNA–DNA hybridization data and sequence data from three genes reveals that the three best-supported morphological clades are all found on the DNA tree. Only the morphological clade supported at the 59% level was absent on the DNA tree. Horovitz and Meyer found only 3% incongruence between nuclear, mitochondrial, and morphological data sets for New World monkeys (no support measures for individual data sets were given). The authors used a combined analysis. Sakai et al. also compared a morphological data set to two RFLP data sets in Hawaiian Alsinoideae (Caryophyllaceae). No clade supported at the 80% bootstrap level in any data set was contradicted by another equally well-supported alternative in a different data set, and the remainder of the analysis used either a combined data set or the morphological data set. Smith and Littlewood also combined their four data sets (including one adult morphological and one larval morphological data set), after an analysis of congruence showed that the data set most incongruent with the others, larval morphology, was only incongruent at nodes supported at less than 56% bootstrap levels. The Brocchinia study of Givnish et al., compared nuclear and chloroplast RFLP data sets to a morphological data set and the authors suggested that the morphological data set disagreed with the molecular trees at several points (p. 289). However, the morphological data set had 19 characters for 19 taxa and bootstrap values were so low that they were not shown (caption to fig. 8.12; decay indices were all 1 within Brocchinia, with one outgroup node supported by a 2). The one fly in the ointment is Barrett and Graham's reference to significant incongruence uncovered between molecular and morphological data sets for Pontederiaceae (Graham et al., in press). Thus, in the six studies in this book in which incongruence between morphological and molecular data sets might have been discovered if concerted homoplasy were an important issue, significant incongruence was found in only one. Admittedly, the sample size is hardly compelling; what is needed is a really good meta-analysis of incongruence. Most of my comments have been aimed at themes that cropped up in several chapters. Other, equally interesting, ideas emerge singly in chapters without being echoed elsewhere in this book. Vogler and Goldstein present an elegant study on tiger beetles in which they attempt to distinguish two general modes of adaptive radiation from each other, the taxon-cycle and taxon pulse hypothesis. The taxon-cycle hypothesis assumes that speciation is tightly coupled to habitat shifts, whereas the taxon pulse hypothesis assumes that habitat shifts precede a pulse of speciation events. These hypotheses make dramatically different predictions about the phylogenetic pattern of diversification, and the authors convincingly argue that, at least for phenological syndromes, the historical pattern is more consistent with a taxon pulse than a taxon cycle. A completely different general mode of adaptive radiation is discussed by Chase and Palmer. They use the term “leapfrog radiation” to describe a sequence of events starting with the invasion of an adaptive zone (twigs by twig-epiphytic orchids) and succeeded by a secondary radiation involving floral diversification (see Stanley, 1990). These are just two examples that illustrate some of the intellectual variety and depth in this volume. Together with the heterodox ideas of its editors—supported by some contributors, refuted by others—these perspectives combine to make Molecular Evolution and Adaptive Radiation a truly thought-provoking and significant book. Anyone interested in visiting or revisiting this centerpiece issue in evolutionary biology would do well to start here. Tucked away in various places, this book considers a myriad of evolutionary processes and patterns that impinge on evolutionary diversification in general and adaptive radiation in particular. Although no consensus on what adaptive radiation is or how it should be studied emerges, the book conveys the sense that the subject is undergoing intense and vigorous scrutiny, holding out the promise that revealing syntheses will emerge in the future. Schematic diagram showing the possible combinations of processes of diversification in relation to adaptive radiation, following the definitions of Givnish (Chapter 2).