Species Selection: Theory and Data

Species selection in the broad sense—also termed species sorting—shapes evolutionary patterns through differences in speciation and extinction rates (and their net outcome, often termed the emergent fitness of clades) that arise by interaction of intrinsic biological traits with the environment. Effectmacroevolution occurs when those biotic traits, such as body size or fecundity, reside at the organismic level. Strict-sense species selection occurs when those traits are emergent at the species level, such as geographic range or population size. The fields of paleontology, comparative phylogenetic analysis, macroecology, and conservation biology are rich in examples of species sorting, but relatively few instances have been well documented, so the extent and efficacy of the specific processes remain poorly known. A general formalization of these processes remains challenging, but approaches drawing on hierarchical covariance models appear promising. Analyses integrating paleontological and neontological data for a single set of clades would be especially powerful. 501 A nn u. R ev . E co l. E vo l. Sy st . 2 00 8. 39 :5 01 -5 24 . D ow nl oa de d fr om a rj ou rn al s. an nu al re vi ew s. or g by U N IV E R SI T Y O F C H IC A G O L IB R A R IE S on 1 1/ 17 /0 8. F or p er so na l u se o nl y. ANRV360-ES39-24 ARI 10 October 2008 9:58 Aggregate traits: species properties that are characteristics of individual organisms (body size) or combined measurements of organisms (mean body size) INTRODUCTION: HIERARCHY AND EVOLUTION Complex systems naturally fall into hierarchies. Among the many hierarchies found in biological systems, the one incorporating genes, bodies, populations, species, and clades has incited the greatest controversy, centering on the dynamics of units at different levels and the consequences. Selfish or parasitic DNA sequences were quickly seen in this hierarchical light, subject to selection and other evolutionary forces within the genome and with effects cascading both to and from higher levels. In contrast, the potential for analogous processes operating on species within clades has had a longer but more troubled history. However, the view that differential speciation and extinction rates can shape evolutionary trajectories and trait distributions across clades is increasingly accepted theoretically and is supported by a growing but scattered body of data. Here I review the different definitions of species selection in the context of multilevel selection (MLS) theory, discuss potential approaches toward the still-incomplete formalization of species selection, and give an overview of the empirical evidence bearing on the operation of the processes that fall under the different concepts of species selection. I will not provide a full historical review, but discuss progress in developing operational approaches and an empirical basis for this aspect of MLS. Most workers now accept that broad-sense species selection can occur in principle [including such early critics of different forms of MLS as Dawkins (1989), Williams (1992), and Maynard Smith (1998)]. The question is thus an empirical one: when does species selection occur, and how can we recognize it and quantify its effects? The growing acceptance of species selection (in the broad sense) as a potential evolutionary force derives in part from an expansion in the range of evolutionary questions being addressed. The focus is no longer so exclusively on changes in allele frequencies and how those might drive the origin of features such as eyes and horns, but instead it more actively includes the frequencies of such features among clades (how many species have horns), the fates of those features (how long do horns last, in evolutionary terms), and origination and extinction rates of species and clades (why do clades wax and wane, and do horns determine clade dynamics in some way, or they are carried along in association with other features). Simple extrapolation from organismic fitnesses cannot always fully account for the dynamics of clades (see overviews by Jablonski 2000, 2005, 2007, 2008a), and so hierarchical approaches, comprising several elements including species selection, have increasingly been invoked. This does not mean that organismic-level processes are unimportant, of course, but suggests that effects of scale and hierarchy also enter into the equation for long-term, large-scale evolutionary patterns. SPECIES SORTING AND SELECTION Emergent Properties and Emergent Fitness The term species selection has been used in both broad and narrow senses, sometimes by a single researcher. In its original broad sense, species selection referred to the differential origination or persistence of species—together considered the emergent fitness of species within clades— owing to interaction with the environment. (By contrast, species drift can occur by differences in origination and extinction in the absence of environmental interaction, as noted below.) Early proponents were concerned mainly with differential survival of species on the basis of organismal, aggregate traits present in all members of the species, and many still favor this usage (e.g., Arnold & Fristrup 1982; Coyne & Orr 2004; de Vries 1905, p. 800; Eldredge & Gould 1972; Gould 2002; Gould & Lloyd 1999; Morgan 1903, pp. 462–64; Okasha 2006; Stanley 1975a, 1979; Wright 1956; and see Lloyd 1988, pp. 101–7, 2000; Lloyd & Gould 1993; Stidd & Wade 1995; Williams 1992, pp. 26–27). Because traits affecting speciation and extinction rates can reside at any hierarchical 502 JablonskiAnnu.Rev.Ecol.Evol.Syst.2008.39:501-524.Downloadedfromarjournals.annualreviews.org byUNIVERSITYOFCHICAGOLIBRARIESon11/17/08.Forpersonaluseonly. ANRV360-ES39-24 ARI 10 October 2008 9:58 Emergent traits:species properties thatare not reducible toorganismic traits (e.g.,geographic range sizeor populationstructure)level, organismic attributes such as body size or metabolic rate can promote broad-sense speciesselection as readily as species-level attributes such as geographic range or genetic populationstructure (emergent traits; see below). The key requirements are that (a) a trait exhibits little or novariation within species relative to the variation among species (see Gould 2002, pp. 664–65 for hisclearest statement on this point), and (b) speciation and/or extinction covaries consistently acrossone or more clades with that trait. Mammalian body size is often viewed in this light: Species tendto exhibit modal sizes, and a cross-level discordance may exist in the evolutionary consequencesof size in that short-term organismic selection might often favor larger body size (cf. Kingsolver& Pfennig 2004), but larger bodied species or clades may be more extinction-prone over longertime scales (e.g., Van Valen 1975, Van Valkenburgh et al. 2004). Of course, broad-sense speciesselection need not oppose selection at the organismic level, although this is analytically moretractable; selection might as readily operate in the same direction at multiple levels (see Grantham1995, Lloyd & Gould 1993; such concordance assumes some intraspecific variation in a focal trait,which still could be modest relative to among-species variation).An alternative approach applies the more general term species sorting to nonrandom differencesin emergent fitness (i.e., broad-sense species selection), and attempts to specify the hierarchicallevel of the traits that confer those differences. Effect-macroevolution then signifies differentialrates governed by organism-level—that is, aggregate—traits, whereas strict-sense species selectionrefers only to differential rates that are governed by emergent, heritable properties at the specieslevel (see Damuth & Heisler 1988; Grantham 1995, 2001; Jablonski 1987, 2000; Vrba 1984, 1989;Vrba & Gould 1986). In Campbell’s (1974) famous terms, for effect-macroevolution the focallevel of selection is the organism, but upward causation to higher levels via effects on speciation orextinction probabilities can drive clade dynamics in unexpected ways, as in the body-size exampleabove (see Vrba & Eldredge 1984, Vrba & Gould 1986). For strict-sense species selection, thefocal level is the species, with downward causation influencing the frequencies of organismictraits among clades (a process termed species hitchhiking by Levinton et al. 1986) and upwardcausation again shaping overall clade composition. This narrower definition of species selectionmore stringently incorporates the classical recipe for evolution by selection codified by Lewontin(1970)-–differential production or survival of units owing to interaction of heritable traits with theenvironment—and so requires the identification of both emergent species-level traits and theirheritability among species.Identifying emergent properties of species is not straightforward, and by some accounts emer-gent and aggregate traits are extremes on a continuum, depending on the complexity of theinteractions among organism-level properties (and environmental effects, as in any phenotypictrait) in generating the species-level property (Okasha 2006; see also Wimsatt 2007, chapter 12).Just as we often think of selection as operating on phenotypes determined by the interaction ofmany genes (and the environment) rather than as operating solely at the genic level, the interactionof many factors to determine a geographic range size, a genetic population structure, or a level ofintraspecfic variation can yield an emergent property whose evolutionary effects are not reducibleto any single organismic property.The philosophy literature contains many criteria for emergence, often related to the mind–bodyproblem (e.g., see Brandon 1990, 1996; Sober 1984, 1992, 1999; Wimsatt 2007; and Grantham’s2007 review), but these have been difficult to apply to clade dynamics. One simple operational ap-proach is for a feature to be emergent at a given level if its evolutionary consequences do not dependon how the feature is generated at lower levels ( Jablonski 2000, 2007; Jablonski & Hunt 2006).This approach is related to Brandon’s (1990) application of the statistical concept of screening-off(where variation at the lower level is “screened off” from selection at the higher level) and to themultiple-realizability criterion for emergence (where a higher-level trait can be realized by many www.annualreviews.org • Species Selection 503Annu.Rev.Ecol.Evol.Syst.2008.39:501-524.Downloadedfromarjournals.annualreviews.org byUNIVERSITYOFCHICAGOLIBRARIESon11/17/08.Forpersonaluseonly. ANRV360-ES39-24 ARI 10 October 2008 9:58 distinct combinations of lower-level properties); both are controversial in some situations but thegeneral approach can provide a working hypothesis in the present context (see Brandon et al. 1994,Grantham 2007, Sober 1999, Sterelny & Griffiths 1999, Wimsatt 2007 for discussions). Thus,geographic range is a species-level property (see references in Table 1), not simply because mostgeographic ranges are determined by the overall distribution of conspecifics rather than by themovements of individual organisms, but because the macroevolutionary consequences—the pos-itive relation between geographic range and extinction-resistance—tend to be similar regardlessof how those ranges are mediated/attained at the organismal level (see below). Only geographicrange size has been subjected empirically to this criterion, and so the effective hierarchical levelof most others remains theoretical. However, only a few of the properties in Table 1 need to beconfirmed to open a large domain for strict-sense species selection: all species have a geographicrange, a genetic population structure, and so on. What we really need to know is which traits aremost likely to screen off, override, or significantly reinforce selection and other forces operatingat other levels, and in what circumstances.Cutting across the aggregate/emergent distinction is the separation of sorting processes intomultilevel selection (MLS) 1 and 2 (a viewpoint developed by Arnold & Fristrup 1982 and Sober1984 and formalized by Heisler & Damuth 1987; see also Damuth & Heisler 1988, Okasha2006). In MLS1, the organisms are the focal level; changes in the frequency of organismic traitsdepend on selection both within and between larger units; that is, an individual’s fitness dependsin part on the phenotypes of other individuals in the population so that change in organismicproperties both is affected by and will alter the larger units to which they belong. Most modelsfor group selection (e.g., for the evolution of altruism) take this form (e.g., Damuth & Heisler1988, Okasha 2006, Wilson & Wilson 2007), and I will discuss this no further, except to notethat some researchers have applied the term group selection to selection on any unit above theorganism level [e.g., Wright 1932, 1945 (as intergroup selection); Okasha 2003; Rice 2004; VanValen 1975]. In contrast, for MLS2 the higher-level units such as species are the focal level, andthe key variables are the origin and demise of those higher-level units, for example, speciation andextinction. The traits that influence speciation and extinction might reside at the organismic level( = effect-macroevolution) or at the species level ( = strict-sense species selection).Disputes over different senses of species selection are largely terminological and secondaryto the broader theoretical implications of MLS: any higher-level sorting process irreducible towithin-population evolutionary forces conforms to a hierarchical evolutionary theory. However,if we are to move from pattern to process, then the hierarchical level of the target of selection—that is, the nature of the interaction between the focal unit and its environment and not just thedifferential proliferation of species—is a key consideration. Upward Causation: Effect-Macroevolution Aggregate characters are expressed at the organismic level but may influence speciation and ex-tinction rates and thus mediate effect-macroevolution, that is, differential origination and/or ex-tinction owing to organismic traits. An impressive array of aggregate traits has been hypothesizedor demonstrated to play such a role, from niche breadth to the intensity of sexual selection toclonality to propagule size (Table 2). Three striking features of this list, aside from its length, areas follows: 1. Many of these traits involve macroevolutionary trade-offs such that high origination ratesare often expected to be accompanied by high species-extinction rates (e.g., Coyne & Orr2004, p. 435; Gould & Eldredge 1977; Stanley 1979, 1990). For example, intense sexual 504 JablonskiAnnu.Rev.Ecol.Evol.Syst.2008.39:501-524.Downloadedfromarjournals.annualreviews.org byUNIVERSITYOFCHICAGOLIBRARIESon11/17/08.Forpersonaluseonly. ANRV360-ES39-24 ARI 10 October 2008 9:58 Table 1 Proposed species-level traits and their hypothesized impact on rates of origination (O) and/or extinction (E)