Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) doi: 10.1111/j.1463-6395.2008.00364.x BFlackwell Puoblishing Lstd sils provide better estimates of ancestral body size than do extant taxa in fishes James S. Albert,1 Derek M. Johnson1 and Jason H. Knouft2 Abstract 1Department of Biology, University of Albert, J.S., Johnson, D.M. and Knouft, J.H. 2009. Fossils provide better Louisiana at Lafayette, Lafayette, LA estimates of ancestral body size than do extant taxa in fishes. — Acta Zoologica 70504-2451, USA; 2Department of (Stockholm) 90 (Suppl. 1): 357–384 Biology, Saint Louis University, St. Louis, MO, USA The use of fossils in studies of character evolution is an active area of research. Characters from fossils have been viewed as less informative or more subjective than comparable information from extant taxa. However, fossils are often the Keywords: continuous trait evolution, character state only known representatives of many higher taxa, including some of the earliest optimization, morphological diversification, forms, and have been important in determining character polarity and filling vertebrate taphonomy morphological gaps. Here we evaluate the influence of fossils on the interpretation of character evolution by comparing estimates of ancestral body Accepted for publication: 22 July 2008 size in fishes (non-tetrapod craniates) from two large and previously unpublished datasets; a palaeontological dataset representing all principal clades from throughout the Phanerozoic, and a macroecological dataset for all 515 families of living (Recent) fishes. Ancestral size was estimated from phylogenetically based (i.e. parsimony) optimization methods. Ancestral size estimates obtained from analysis of extant fish families are five to eight times larger than estimates using fossil members of the same higher taxa. These disparities arise from differential survival of large-bodied members of early branching lineages, and are not statistical or taphonomic artefacts. Estimates of ancestral size obtained from a limited but judicious selection of fossil fish taxa are more accurate than estimates from a complete dataset of extant fishes. Journal: Acta Zoologica, 11th International Symposium on Early Vertebrates: Symposium volume James S. Albert, Department of Biology, University of Louisiana at Lafayette, Lafayette, LA, USA. E-mail: [email protected] fossils are often fragmentary and exhibit large amounts of Introduction non-randomly distributed missing data (Wiens 2006). All of The vast majority (> 99%) of species that have ever existed these sources of error contribute to uncertainty about the are now extinct (Simpson 1952) and whole branches of the phenotypes and phylogenetic positions of fossil taxa (Gauthier tree of life are known only from fossil forms (e.g. trilobites, et al. 1988; Wilkinson 2003). placoderms, plesiosaurs). Consequently, fossils represent a Considering even limited sample sizes and problems unique resource for evolutionary studies. However, inter- associated with preservation of individuals, fossils can still preting the morphology of fossil taxa is regarded as more provide irreplaceable information regarding the tempo and subjective and less informative than data derived from living mode of character evolution. Fossils are often the only members of the same higher taxon (Patterson 1981; Ax exemplars from the earliest radiations in many higher taxa, 1987). In particular, the morphology of extant taxa can be providing critical information in determining character polarity studied in greater detail than in fossils, including aspects of (Conway Morris 1993; Budd and Jensen 2000; Briggs and soft anatomy, and usually using larger sample sizes. By Fortey 2005). Because the taxa they represent are often contrast, and despite occasionally exceptional preservation, extinct, fossils may represent taxa on shorter genealogical © 2009 The Authors 357 Journal compilation © 2009 The Royal Swedish Academy of Sciences Fossils estimates of ancestral body size in fishes • Albert et al. Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) branches than their living relatives, and as such are more Ordovician radiations (c. 488 Ma; Long 1995; Janvier 1996). likely to preserve less derived character states. Fossils often The second dataset is a compilation of mean body size for sample lineages closer in time to relatively deep splitting all 515 families of living (Recent) fishes using data from events, and frequently display character state combinations FishBase (Froese and Pauly 2005). A main conclusion is that not observed among extant forms (Gauthier et al. 1988; estimates of ancestral body size obtained from a limited Donoghue et al. 1989; Wilson 1992; Santini and Tyler 2004). but judicious selection of fossil taxa are more accurate than Because of the added information from extinct lineages, estimates from an (almost) complete dataset of all extant fishes. inclusion of fossils in phylogenetic analyses has substantially The results invite caution when interpreting the conclusions improved understanding of phylogeny (O’Leary 1999; of character state optimization studies based on examination Gatesy and O’Leary 2001), character state evolution (O’Leary of extant taxa alone. These limitations persist even when the 2001) and phylogenetic trends (Finarelli and Flynn 2006; terminal taxa represent a complete (or almost complete) Cobbett et al. 2007). sampling of the living biota, and when they have been Body size is among the most easily acquired and directly analysed in a robust phylogenetic context. These results are a comparable attributes of organisms for which reliable esti- reminder that patterns of organismic diversification arise mates may be obtained from the fossil record (Stanley 1998). from the processes of both speciation and extinction, and Adult body size is also a central feature of organismal design, that only study of fossils allows the sampling of character imposing constraints on many aspects of life history, espe- states in extinct clades. cially critical scaling functions related to growth, metabolism and fecundity (Peters 1983; Schmidt-Nielsen 1984; Haldane Materials and methods 1985; McNab 2002). Consequently, understanding the evolution of body size can provide insights into the diversi- Body size was measured as total length (TL) in centimetres fication of biological form and function. Changes in body size from the tip of the snout to the posterior margin of the caudal can have allometric effects on morphology, physiology, fin directly from specimens, published photographs or behaviour (e.g. activity patterns, thermoregulation), and reconstructions of articulated specimens in primary sources. are widely used as an adaptation to novel physical (e.g. Appendix 1 presents maximum recorded total length and temperature extremes, hypoxia, desiccation) and biotic stratigraphic data for 465 fish species, including 425 species (e.g. predation, competition) environmental parameters known only as fossils and 40 species in clades for which (Hutchinson and Macarthur 1959; Strathdee and Bale 1998; fossils with reliable size data are lacking (e.g. Myxine; Huso). Burness et al. 2001; Leaper et al. 2001). Importantly, many Appendix 2 presents statistical measures of size (average total features associated with body size transcend the particularities length in cm), size variation (standard deviation, skew, of taxonomic design, and as such often exhibit repeated patterns kurtosis), and species richness (N) for all 515 recognized of evolution (Wake 1991; Mabee 2000; Bird and Mabee 2002; extant fish families, based on data for 24 259 species from Mabee 2002). FishBase. The taxonomic compositions of these two datasets Fishes (non-tetrapod craniates) provide numerous ex- are summarized in Tables 1 and 2, respectively. Total length is amples of taxa and circumstances in which to test theories highly correlated with other measures of overall size, including on the evolution of continuous traits such as size (Albert et al. maximum body weight, and size and age to first reproduction 1999; Knouft and Page 2003). Marine and freshwater fishes (Froese and Binohlan 2000). Among fishes, body mass in represent the largest component of contemporary vertebrate grams (g) may be estimated from total length in cm from the diversity, including more than 50% of all living vertebrate empirical equation: g = 0.0217 TL2.861 (Fig. 1). species, and inhabiting most of the Earth’s aquatic habitats The fossil taxa included in this analysis were selected to and geographical regions. Fishes also have a rich palaeonto- maximize representation of phylogenetically basal craniate logical record from throughout the Phanerozoic, with fossil taxa lineages (sensu Janvier 1996), and include a thorough sam- ranging in body size over more than three orders of magnitude. pling of higher fish taxa for which reliable estimates of size are Moreover, recent discoveries of early Palaeozoic fossils have currently available. Taxon sampling followed the basal exem- greatly expanded our knowledge of early vertebrate diversity plar approach which maximizes representation of phyloge- and phylogeny (see Mallatt and Chen 2003; Shu et al. 2003; netically basal clades (Prendini 2001; Prendini and Wheeler Janvier et al. 2006; references therein). 2004). The basal-exemplar approach is less sensitive to preser- To better understand the importance of fossil data in docu- vational biases than stratigraphically based taxon-counting menting patterns of diversification in relation to body size, methods (Lane et al. 2005). This sampling strategy produced we compare estimates of ancestral body size in fishes using a fossil dataset that is broadly representative of the fossil phylogenetically based methods of character state optimiza- record of fishes as a whole, including members of more than tion applied to two new (previously unpublished) and large half (51%) of all the 324 fish families known only as fossils datasets. The first is a palaeontological dataset representing (Benton 1993), 26% (164 of 622) of all fish families, living all principal clades of non-tetrapod craniates from through- and extinct, known as fossils, and 68% (71 of 105) of all non- out the Phanerozoic, including exemplars of all the early teleost actinopterygian genera known only as fossils. Fossil © 2009 The Authors 358 Journal compilation © 2009 The Royal Swedish Academy of Sciences Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Albert et al. • Fossils estimates of ancestral body size in fishes Table 1 Taxonomic summary of the fossil fish database Clade Fossil* Extant† Total Total (%) Cephalochordata 3 1 4 0.86 Yunnanozoa 2 0 2 0.43 Hyperotreti 3 7 10 2.15 Myllokunmingiida 2 0 2 0.43 Hyperoartia 5 5 10 2.15 Pteraspidomorphi 25 0 25 5.38 Thelodonti 8 0 8 1.72 Anaspida 6 0 6 1.29 Galeaspida 7 0 7 1.51 Pituriaspida 1 0 1 0.22 Osteostraci 15 0 15 3.23 Furcacaudiformes 2 0 2 0.43 Fig. 1—Length and weight are significantly correlated in fishes. Placodermi 39 0 39 8.39 Data are maximum recorded total length (cm) and mass (g) for 517 Chondricthyes 91 5 96 20.65 Acanthodii 17 0 17 3.66 extant fish species. The slope of the regression (m = 0.3267) on this Sarcopterygii 38 4 42 9.03 log–log plot is close to the theoretically expected value 0.33 (i.e. Actinopterygii 161 18 179 38.49 mass = length3). Size data from FishBase (Froese and Pauly 2005). Total 425 40 465 100.00 *Taxa known only as fossils. †Extant taxa for clades lacking fossils with to be closely related to crown Actinopteri giving a long reliable size data (e.g. Myxine; Huso). Data are maximum recorded total branch (c. 371–301 Ma). length (cm), geological age (Epoch or Series), and phylogenetic position, The taxa examined provide sufficiently broad temporal from multiple sources (see text for explanation). (107–108 MY) and taxonomic (102–104 species) scopes to avoid non-random sampling errors arising from community assembly processes, convergent evolution or investigator bias Table 2 Taxonomic summary of the extant fish species database. (Ackerly 2000; Pollock et al. 2002). Due to the limited Data are maximum recorded total length (cm) and taxonomic number of known fossils closely related to certain extant affiliation for more than 24 000 species from Froese and Pauly (2005) basal fish clades, size data for 13 terminal taxa are presented as an average of extant species from FishBase. These clades Clade Orders Families Species Total (%) include the seven extant myxiniform genera, the two extant Hyperotreti 1 1 69 0.28 petromyzontiform genera, one extant dipnoan (Protopterus Hyperoartia 1 2 40 0.16 with six spp.), and five extant actinopterygians (Polypterus, Chondrichthyes 13 46 826 3.40 Acipenser, Scaphyrinchus, Psuedoscaphyrinchus, Clupea). Taxa Sarcopterygii 3 4 10 0.04 for which morphologically mature specimens are not known Actinopterygii 45 462 23 314 96.10 were excluded, as were taxa for which adult body size cannot Total 63 515 24 259 100.00 be reliably estimated from known fossilized fragments (e.g. †Polymerolepis margaritifera, †Lophosteus sp.). Mature specimens are recognized by osteological criteria when available, that is, species were dated to epoch or series (e.g. Upper Devonian, the shape of bones in the sphenoid and palatoquadrate Palaeocene) with geological dates from Gradstein et al. regions of the skull, and the scapulocoracoid region of the (2004). Conodonts were excluded from analysis due to pectoral girdle (Arratia 1997). Size of some Palaeozoic forms uncertainties in body size and detailed phylogenetic informa- was estimated from large body fragments (e.g. †Andinaspis tion (Donoghue and Sansom 2002; Janvier 2003; Dong et al. suarezorum, †Pituriaspis doylei; Janvier, pers. comm.). 2005; Northcutt 2005; Wickstrom & Donoghue 2005). Composite tree topologies were constructed from litera- Hyperoartia data are from Janvier and Lund (1983), Gess ture sources. The phylogeny of principal craniate clades (i.e. et al. (2006) and Janvier et al. (2006). Triassic neoselachians with initial radiations during the Lower Ordovician, c. 488– are known only from teeth and were excluded from analysis 472 Ma; Fig. 2), of 465 fossil species, and of the 515 extant (Underwood 2006). Sarcopterygian data are largely from fish families (Fig. 3), largely follow Janvier (1996, 2003) and Clouthier and Forey (1991), Cloutier 1996), Cloutier and Long (1995), and references therein, with certain emendations Ahlberg (1996) and Clouthier (1997). Actinopterygian data noted by taxon in Appendix 1. These sources were used are largely from Coates (1998), Dietze (2000), Arratia to a construct a tree topology for fossil fishes with 843 (1997, 1999), Arratia and Cloutier (2002), Arratia and branches and 86 polytomies, or a tree that is c. 91% resolved. Clouthier (2004), Lund (2000) and Friedman and Blom (2006). Carboniferous actinopterygians are not considered †extinct taxa known only as fossils. © 2009 The Authors 359 Journal compilation © 2009 The Royal Swedish Academy of Sciences Fossils estimates of ancestral body size in fishes • Albert et al. Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Fig. 2—Interrelationships and stratigraphic ranges of the principal craniate clades (i.e. Ordovician radiations c. 488 Ma (Long 1995; Janvier 1996; Table 1). Thick lines represent known stratigraphic ranges. Extant craniates represent just five of the 14 principal craniate clades. In terms of numbers of clades and species, Actinopterygii (ray-finned fishes) dominates the marine ichthyofauna from the Carboniferous (c. 363 Ma) to the Recent, the global freshwater ichthyofauna from the Upper Cretaceous (c. 100 Ma) to the Recent, and includes 96.1% of living fish species. Fig. 3—Phylogenetic hypotheses of fish taxa with size-change optimized at all internal tree branches using Linear Parsimony (LP). —A. 465 fossil species, representing all 14 principal craniate clades (Ordovician radiations), with size optimized at 926 branches. —B. 515 extant fish families representing five principal craniate clades, with size optimized at 1031 branches. Tree topologies from Long (1995), Janvier (1996), Appendix 1 and references therein. Names of extinct clades (†) in grey font; extant clades in coloured fonts. Size data in cm log transformed before analyses. Branch lengths in MY estimated from stratigraphic data. © 2009 The Authors 360 Journal compilation © 2009 The Royal Swedish Academy of Sciences Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Albert et al. • Fossils estimates of ancestral body size in fishes Two chordate outgroup taxa were used to root the size optimizations: Cephalochordata and Yunnanozoa (Mallatt and Chen 2003). The phylogenetic positions of †Haikouichthys as a non-craniate deuterostome follows Shu (2003). Linear and least squared parsimony (LSP) methods were employed to optimize ancestral size using the Mesquite v.1.06 software package (Maddison & Maddison 2006). Linear Parsimony (LP) minimizes the total amount of trait change along tree branches such that the cost of a change from state x to y is |x – y| (Swofford and Maddison 1987). LSP, also referred to as Squared-Change Parsimony, follows a Brownian motion model of evolutionary change in which the cost of a change from state x to state y is (x – y)2 (Maddison 1991). LP differs from LSP and model-based (i.e. Bayesian and Likelihood) approaches to character state optimization in that it permits the reconstruction of discontinuous events, or of large changes in trait values (Butler and Losos 1997; Pagel 1999). Although evolutionary change is often considered as gradual, large differences in trait values between internal tree nodes may result from a variety of real biological processes, including punctuated evolution (Pagel et al. 2006) or extinction of taxa with intermediate trait values (Butler and Losos 1997; Albert et al. 1998). LP also permits the reconstruction of ambiguous ancestral state values when data are insufficient to provide an unambiguous resolution. Nevertheless, estimates of mean size among fossil fishes per epoch using LP and LSP are significantly correlated (P < 0.0001; Fig. 4). All ancestral reconstruction methods assume that trait evolution is conservative enough for node reconstruction techniques to be useful, even in the face of large standard errors (Polly 2001). Ancestral trait optimization was performed using 10 repli- cates on arbitrarily fully resolved trees using MacClade 4.07 (Maddison and Maddison 2005). The qualitative results of this analysis were similar in all replicates of arbitrary node Fig. 4—Estimates of mean body size (ln cm) per epoch from resolution. Available methods of character state reconstruc- phylogenetic optimization (LP and LSP) and stratigraphic tion are limited to estimating ancestral trait values from (non-phylogenetic) methods. Stratigraphic estimates assessed within the limits of those observed in terminal taxa. LP directly as average log-transformed body size data of fossils per analysis may therefore perform poorly at detecting a consistent epoch. Phylogenetic estimates assessed as averages of interior node underlying trend like Cope’s rule. The reader is referred to values per epoch using LP and LSP optimization on the phylogeny Albert (2006) for a discussion of the limits and assumptions of fossil fishes (Fig. 3A). All R2 values are significant at P < 0.0001. of different ancestral trait reconstruction methods. Strati- Note stratigraphic estimates are more highly correlated with LP than graphic data of fossils were used to constrain minimum age LSP, due to the averaging nature of squared-change optimization. estimates for internal tree branches (Benton and Donoghue 2007). Branch lengths were estimated from stratigraphic mates using fossil members of the same higher taxa (Fig. 5). data from fossils following Benton (1993, 2005). Branch This result is consistent for all of the 14 craniate clades with lengths were estimated as the absolute difference in MY origins during the Lower Ordovician, including taxa with between nodes. In several taxa known only from Recent broad disparities in date of clade origin (c. 550–450 Ma), organisms, branch lengths were estimated from biogeo- clade duration (c. 50–500 MY), body size at origin (c. 5– graphical information among sister taxa (see Appendix 1). 25 cm), habitat (i.e. marine, freshwater, euryhaline) and geography (i.e. tropical, extratropical, cosmopolitan). Plesio- morphic size estimates from the fossil dataset for Craniata, Results and discussion Hyperotreti, Vertebrata and Hyperartia are 5.0–8.0 cm, as Ancestral size estimates obtained from analysis of the 515 compared with 45–50 cm from the extant dataset for these extant fish families are five to eight times larger than esti- same taxa. Similarly, plesiomorphic size estimates from the © 2009 The Authors 361 Journal compilation © 2009 The Royal Swedish Academy of Sciences Fossils estimates of ancestral body size in fishes • Albert et al. Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) radiations (Webby et al. 2004), being limited to just five of the six clades that survived the Late Devonian crisis (c. 375 Ma; Fig. 2). This extinction event was a strong filter on the size as well as taxonomic composition of surviving fish faunas (Janvier 1996; McGhee 1996). Third, having persisted for longer periods of geological time, living members of a clade may be expected to have accrued on average more changes than lineages of the same clade cut short by extinction. As a result, fossil species often preserve plesiomorphic states with more fidelity than related extant species (Donoghue et al. 1989). Why are size estimates derived from analysis of living fishes so much larger than those derived from fossil representatives of the same higher taxa? Such disparities could arise from systematic biases in methods used to assemble the fossil data- set or conduct the optimization analysis, reflecting tapho- nomic or statistical artefacts from size-selective preservation, recovery or identification of fossils. Alternatively, the dis- parities might arise from real differences in the evolutionary histories of taxa which have become extinct vs. those which have persisted to the Recent. Consideration of the available data suggests the different estimates of ancestral body size result from the persistence of phylogenetically basal taxa with large size among living fish clades. This result also reinforces the claim that early branching lineages do not necessarily retain primitive or ancestral traits (Crisp and Cook 2005). The disparity in body size estimates from the fossil and extant datasets does not appear to be a taphonomic artefact arising from size-selective preservational biases. Size-related biases on the preservation, recovery and identification of fos- Fig. 5—Ancestral size estimates from analysis of extant fishes (515 sils may provide potentially confounding signals in inferring extant families) are five to eight times larger than estimates using fossil members of the same higher taxa. —A. Inferred ancestral sizes size evolution from palaeontological data (Barton and Wilson using LP optimization. —B. Size estimates from extant (= Recent) 2005; Northwood 2005). Large specimens are more subject vs. fossil members of the same higher taxa using LP and LSP to disarticulation and dispersal through hydrodynamic trans- optimization. Note the averaging effect of LSP results in somewhat port and physical wear through abrasion (Long and Langer less disparity in size estimate from Recent and fossil taxa. 1995; Butler and Schroeder 1998; Butler 2004). As a result, large fishes are less likely to be preserved as intact skeletons, and preserved isolated elements are less likely to be recovered fossil dataset for Gnathostomata, Chondrichthyes, Osteich- and identified, thus hindering accurate estimates of body thyes, Sarcopterygii and Actinopterygii are 20–25 cm, as size. In this regard, Phanerozoic escalation of predation and compared with 120–140 cm from the extant dataset. Clearly bioturbation rates (Vermeij 1994) could influence long-term there has been a strong filter on the size distribution of living trends in the size-frequency distributions of fossil fishes taxa. A similar bias in the persistence of taxa with larger sizes through time. On the other hand, larger skeletal elements are has been observed in mammalian carnivores (Finarelli and usually more robust, more resistant to abrasion, and often Flynn 2006). have more readily observed diagnostic morphologies. As a Which of the two datasets, palaeontological or contemporary result they are more likely to be preserved, recovered and provides more accurate estimates of size evolution among correctly identified (Behrensmeyer 1978; Kidwell and Flessa the principal craniate clades? Three features of craniate 1995; Alroy 2000). Indeed large-bodied taxa are better phylogeny and diversity suggest that information from fossils represented in many vertebrate palaeofaunas (Cooper et al. is more reliable for this purpose. First, the plesiomorphic 2006; Valentine et al. 2006), and may actually serve to inflate size estimates of Craniata obtained from LP optimization perceptions of trends to larger size. The aggregate effect of of the fossil dataset are similar to the sizes (c. 3–5 cm) of these confounding taphonomic influences on size evolution closely related (fossil and extant) craniate outgroups (Mallatt remains poorly understood (Madin et al. 2006). and Chen 2003). Second, the extant diversity of fish clades Global patterns of diversity may also reflect variation in the represents only a fraction of the original (Ordovician) craniate nature of the fossil record and fossil bearing sediments (Alroy © 2009 The Authors 362 Journal compilation © 2009 The Royal Swedish Academy of Sciences Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Albert et al. • Fossils estimates of ancestral body size in fishes of this study are similar regardless of the parsimony-based optimization approach employed. The phylogenetic distribution of body size among living fishes strongly suggests a history in which the surviving members of basal taxa attain much larger sizes than did their fossil relatives. Among the principal craniate clades that emerged during the Ordovician and which have survived to the Recent, in all cases the living representatives are sub- stantially larger than are the earliest fossils. To cite some examples, extant hagfishes (Myxiniformes, avg. 51 cm, n = 69 species) are larger than the Pennsylvanian †Myxinkela siroka (7 cm) or †Myxineides gononorum (15 cm), extant lampreys (Petromyzontiformes, avg. 31 cm, n = 40 species) are larger than the Mississippian †Hardistiella montanensis (10 cm) or Fig. 6—Numbers of taxa per epoch in the fossil fish dataset Pennsylvanian †Mayomyzon pieckoensis (6 cm), extant hetero- (Appendix 1). The limited number of Late Palaeozoic dontiform sharks (avg. 118 cm, n = 8 species) are larger than (c. 318–251 Ma) fossils reflects a major trough in documented the Jurassic †Heterodontus falcifer (28 cm) or †Paracestracion ichthyofaunas from the Pennsylvanian to Permian (Hurley et al. 2007). Family level diversity for 645 fish families from Benton zitteli (15 cm), extant coelacanths (avg. 154 cm, n = 2 species) (1993, 2005). are larger than the Middle Devonian †Miguashaia bureau (50 cm) or Upper Devonian †Diplurus newarki (25 cm), extant lungfishes (Dipnomorpha, avg. 111 cm, n = 8 species) et al. 2001). For example, the dearth of fossil fish taxa from are larger than the Upper Devonian †Rhinodipterus ulrichi the Pennsylvanian to the Permian (318–251 Ma; Fig. 6) (28 cm), and extant non-teleost actinopterygians (avg. 165 cm, corresponds to a major trough in fish diversity and docu- n = 53 species) are larger than Palaeozoic actinopterygians mented ichthyofaunas known from this interval (Sepkoski (e.g. Lower Devonian †Dialipina salgueiroensis at 25 cm, 2002; Hurley et al. 2007). Among stratigraphic intervals Middle Devonian †Cheirolepis tralii at 25 cm, Middle examined (Table 1), there are no significant correlations Devonian †Stegotrachelys finlayi at 10 cm, or Middle between mean body size and species richness or the propor- Devonian †Moythomasia nitida at about 10 cm). Some tion of articulated skeletons (P > 0.1). Madin et al. (2006) Palaeozoic actinopterygians did attain somewhat larger sizes found that escalatory trends did not drive Phanerozoic macro- (although not approaching modern values), especially during evolutionary patterns in a large dataset of fossil benthic the Middle Devonian (e.g. †Cheirolepis canadensis, 55 cm) marine invertebrates. Similarly, body size has not been found and Upper Devonian (e.g. †Howqualepis rostridens, 95 cm). to be correlated with other long-term geological trends, for Large size in these taxa is apparently derived (Lund and example, sedimentary rock volume (Peters and Foote 2001; Poplin 2002; Arratia and Cloutier 2002; Arratia and Clouthier Crampton et al. 2003), bioturbation rates (Crimes and 2004; Friedman and Blom 2006). Droser 1992; Vermeij 1993) or mean size of top predators To summarize, the available information pertaining to (Janvier 1996; Twitchett et al. 2005). In combination we take body size and phylogeny among the principal clades of fishes these results as evidence that taphonomic effects have not indicates differential survival of large-bodied members of been the primary factor influencing the assessment of size early branching lineages. It is important to note these results distributions of fossil fishes over the Phanerozoic. pertain to phylogenetic patterns only, and do not directly If, as predicted by Cope’s rule, there was a persistent address potential underlying evolutionary processes. In other and general tendency to increase body size within lineages, words, we were not able to reject hypotheses of long-term ancestral size estimates obtained from analysis of terminal anagenetic change (e.g. Cope’s rule; Hone and Benton (fossil or extant) taxa would be systematically overestimated 2005), or of the effects of body size on relative rates of diver- (Stanley 1973; Polly 1998; Hone and Benton 2005). For sification (Brown 1999; Gillooly et al. 2001). The patterns of example, estimates from terminal taxa are limited to the size evolution observed in fishes closely resemble those of range of values observed, and are not capable of estimating other vertebrate clades examined to date with comparable ancestral values smaller than that of the smallest terminal taxonomic and temporal resolution (Gardezi and da Silva taxon. This overestimate in the value of internal tree nodes 1999; Laurin 2004; Smith et al. 2004; Webster et al. 2004). would arise regardless of optimization method used (i.e. LP vs. LSP). However, among 23 Phanerozoic epochs, estimates Acknowledgements of internal node values from LP and LSP approaches are significantly correlated (P < 0.001) with those of a direct We thank John Eisenberg, William Eschmeyer, William Fink, statigraphic approach that does not use phylogenetic methods Gavin Hanke, David Julian, Brian McNabb, Joseph Neigel, (Fig. 4). In other words, the principal qualitative results Joseph Nelson, Larry Page, David Pollock, Gerald Smith, Ed © 2009 The Authors 363 Journal compilation © 2009 The Royal Swedish Academy of Sciences Fossils estimates of ancestral body size in fishes • Albert et al. Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Wiley and Miriam Zelditch for data and ideas. JSA acknowl- Budd, G. E. and Jensen, S. 2000. A critical reappraisal of the fossil edges support from U.S. National Science Foundation DEB record of the bilaterian phyla. – Biological Reviews 75: 253–295. 0138633, 0215388, and 0614334. Burness, G. P., Diamond, J. and Flannery, T. 2001. Dinosaurs, dragons, and dwarfs: The evolution of maximal body size. – Proceedings of the Academy of Nat Sciences of the United States of America 98: 14518–14523. References Butler, M. A. and Losos, J. B. 1997. Testing for unequal amounts of Ackerly, D. D. 2000. Taxon sampling, correlated evolution, and evolution in a continuous character on different branches of a phy- independent contrasts. – Evolution 54: 1480–1492. logenetic tree using linear and squared-change parsimony: An Albert, J. S. 2006. Phylogenetic character reconstruction. In Kaas, example using lesser antillean Anolis lizards. – Evolution 51: 1623– J. H. (Ed.): Evolution of Nervous Systems Volume 1: History of Ideas, 1635. Basic Concepts, and Developmental Mechanisms, pp. 41–54. Aca- Butler, V. L. 2004. Fish remains. In Hicks, B. A. (Ed.): Marmes demic Press, Oxford. Rockshelter A Final Report on 11 000 Years of Cultural Use, pp. 319– Albert, J. S., Froese, R., Bauchot, R. and Ito, H. 1999. Diversity of 337. Washington State University Press, Pullman. brain size in fishes: Preliminary analysis of a database including Butler, V. L. and Schroeder, A. R. 1998. Do digestive processes leave 1174 species in 45 orders. In Seret, B. &. Sire, J. Y. (Eds) 5th Indo- diagnostic traces on fish bones? Journal of Archaeological Science Pacific Fish Conference Proceedings, pp. 647–656. Société Francaise 25: 957–971. d’Ichthyologie, Paris. Clouthier, R. 1996. The primitive actinistian Miguashaia bureaui Albert, J. S., Lannoo, M. J. and Yuri, T. 1998. Testing hypotheses of Schultze (Sarcopterygii) In Schultze, H. P. and Clouthier, P. neural evolution in gymnotiform electric fishes using phylogenetic (Eds): Devonian Fishes and Plants of Miguasha, Quebec, Canada, character data. – Evolution 52: 1760–1780. pp. 227–247. Verlag Dr. Friedrich Pfeil, München. Alroy, J. 2000. New methods for quantifying macroevolutionary Clouthier, R. 1997. Morphologie et variations du toit crânien du patterns and processes. – Paleobiology 26: 707–733. Dipneuste Scaumenacia curta (Whiteaves; Sarcopterygii), du Alroy, J., Marshall, C. R., Bambach, R. K., Bezusko, K., Foote, M., Dévonien supérieur du Québec. – Geodiversitas 19: 59–105. Fursich, F. T. et al. 2001. Effects of sampling standardization on Clouthier, R. and Forey, P. L. 1991. Diversity of extinct and living estimates of Phanerozoic marine diversification. – Proceedings of actinistian fishes (Sarcopterygii). – Environmental Biology of Fishes the National Academy of Sciences of the United States of America 98: 32: 59–74. 6261–6266. Cloutier, R. and Ahlberg, P. E. 1996. Morphology, characters, and Arratia, G. 1997. Basal teleosts and teleostean phylogeny. – Palaeo the interrelationships of basal sarcopterygians. In Stiassny, M. L., Ichthyologica 7: 1–168. Parenti, L. R. and Johnson, G. D. (Eds): Interrelationships of Fishes, Arratia, G. 1999. The monophyly of the Teleostei and stem-group pp. 445–479. Academic Press, San Diego. teleosts. In Arratia, G. and Schultz, H.-P. (Eds): Mesozoic Fishes 2 Coates, M. I. 1998. Actinopterygians from the Namurian of – Systematics and Fossil Record, pp. 265–334. Verlag Dr. Friedrich Bearsden, Scotland, with comments on early actinopterygian Pfeil, Munchen. neurocrania. – Zoological Journal of the Linnean Society 122: 27– Arratia, G. and Clouthier, R. 2004. A new cheirolepidid fish from 59. the Middle-Upper Devonian of Red Hill, Nevada, USA. In Arra- Cobbett, A., Wilkinson, M. and Wills, M. A. 2007. Fossils impact tia, G., Wilson, M. V. H. and Cloutier, R. (Eds): Recent Advances as hard as living taxa in parsimony analyses of morphology. – in the Origin and Early Radiation of Vertebrates, pp. 583–598. Verlag Systematic Biology 56: 753–766. Dr. Friedrich Pfeil, München. Conway Morris, S. 1993. The fossil record and the early evolution Arratia, G. and Cloutier, R. 2002. Cheirolepiform fish from the of the Metazoa. – Nature 361: 219–225. Devonian of Red Hill, Nevada. Journal of Vertebrate Paleontology Cooper, R. A., Maxwell, P. A., Crampton, J. S., Beu, A. G., Jones, 22: 33A. C. M. and Marshall, B. A. 2006. Completeness of the fossil Ax, P. 1987. The Phylogenetic System. John Wiley and Sons, Chichester. record: Estimating losses due to small body size. – Geology 34: Barton, D. G. and Wilson, M. V. H. 2005. Taphonomic variations in 241–244. Eocene fish-bearing varves at Horsefly, British Columbia, reveal Crampton, J. S., Beu, A. G., Cooper, R. A., Jones, C. M., Marshall, 10 000 years of environmental change. – Canadian Journal of B. A. and Maxwell, P. A. 2003. Estimating the rock volume bias Earth Sciences 42: 137–149. in paleobiodiversity studies. – Science 30: 358–360. Behrensmeyer, A. K. 1978. Taphonomic and ecologic information Crimes, T. P. and Droser, M. L. 1992. Trace fossils and bioturbation: from bone weathering. – Paleobiology 4: 150–162. The other fossil record. – Annual Review of Ecology and Systematics Benton, M. J. 1993. The Fossil Record 2. Chapman & Hall, London. 23: 339–360. Benton, M. J. 2005. Vertebrate Palaeontology, 3rd edn. Blackwell Crisp, M. D. and Cook, L. G. 2005. Do early branching lineages Science Ltd, Oxford. signify ancestral traits. – Trends in Ecology and Evolution 20: 122–128. Benton, M. J. and Donoghue, P. C. J. 2007. Paleontological evi- Dietze, K. 2000. A revision of paramblypterid and amplypterid dence to date the tree of life. – Molecular Biology and Evolution 24: actinopterygians from Upper Carboniferous-Lower Permian 26–53. lacustrine deposits of central Europe. – Palaeontology 43: 927– Bird, N. C. and Mabee, P. M. 2002. Homology, individuality, and 966. developmental morphology of the axial skeleton of the zebrafish. Dong, X. P., Donoghue, P. C. J. and Repetski, J. E. 2005. Basal Integrative and Comparative Biology 42: 1197. tissue structure in the earliest euconodonts: Testing hypotheses Briggs, D. E. G. and Fortey, R. A. 2005. Wonderful strife: System- of developmental plasticity in euconodont phylogeny. – atics, stemgroups, and the phylogenetic signal of the Cambrian Palaeontology 48: 411–421. radiation. – Paleobiology 31: 94–112. Donoghue, M. J., Doyle, J. A., Gauthier, J., Kluge, A. G. and Rowe, Brown, J. H. 1999. Macroecology: Progress and prospect. – Oikos T. 1989. The importance of fossils in phylogeny reconstruction. – 87: 3–14. Annual Review of Ecology and Systematics 20: 431–460. © 2009 The Authors 364 Journal compilation © 2009 The Royal Swedish Academy of Sciences Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Albert et al. • Fossils estimates of ancestral body size in fishes Donoghue, P. C. J. and Sansom, I. J. 2002. Origin and early evolution Lane, A., Janis, C. M. and Sepkoski Jr., J. J. 2005. Estimating paleo- of vertebrate skeletonization. – Microscopy Research and Technique diversities: A test of the taxic and phylogenetic methods. – 59: 352–372. Paleobiology 31: 21–34. Finarelli, J. A. and Flynn, J. J. 2006. Ancestral state reconstruction Laurin, M. 2004. The evolution of body size, Cope’s rule and the of body size in the Caniformia (Carnivora, Mammalia): The origin of amniotes. – Systematic Biology 53: 594–622. effects of incorporating data from the fossil record. – Systematic Leaper, R., Raffaelli, D., Emes, C. and Manly, B. 2001. Constraints Biology 55: 301–313. on body-size distributions: An experimental test of the habitat Friedman, M. and Blom, H. 2006. A new actinopterygian from architecture hypothesis. – Journal of Animal Ecology 70: 248–259. the Famennian of east Greenland and the interrelationships of Long, D. J. and Langer, M. R. 1995. Nassariid gastropods as Devonian ray-finned fishes. – Journal of Paleontology 80: 1186– destructive agents in preservation and fossilization of marine 1204. fishes. – Cellular and Molecular Life Sciences 51: 85–89. Froese, R. and Binohlan, C. 2000. Empirical relationships to Long, J. A. 1995. The Rise of Fishes: 500 Million Years of Evolution. estimate asymptotic length, length at first maturity and length at Johns Hopkins University Press, Baltimore & London. maximum yield per recruit in fishes, with a simple method to evaluate Lund, R. 2000. The new actinopterygian order Guildayichthyi- length frequency data. – Journal of Fish Biology 56: 758–773. formes from the Lower Carboniferous of Montana (USA). – Froese, R. and Pauly, D. 2005. Fishbase 2005: Concepts, Design and Geodiversitas 22: 171–206. Data Sources. ICLARM, Los Banos. Lund, R. and Poplin, C. 2002. Cladistic analysis of the relationships Gardezi, T. and da Silva, J. 1999. Diversity in relation to body size of the Tarrasiids (Lower Carboniferous Actinopterygians). – in mammals: A comparative study. – American Naturalist 153: Journal of Vertebrate Paleontology 22: 480–486. 110–123. Mabee, P. 2002. Evolution of median fin modules in the axial skeleton Gatesy, J. and O’Leary, M. A. 2001. Deciphering whale origins with of fishes. Integrative and Comparative Biology 42: 1270. molecules and fossils. Trends in Ecology and Evolution 16: 658– Mabee, P. M. 2000. Developmental data and phylogenetic system- 658. atics: Evolution of the vertebrate limb. – American Zoologist 40: Gauthier, J., Kluge, A. G. and Rowe, T. 1988. Amniote phylogeny 789–800. and the importance of fossils. – Cladistics 4: 105–209. Maddison, W. P. 1991. Squared-change parsimony reconstructions Gess, R. W., Coates, M. I. and Rubidge, B. S. 2006. A lamprey from of ancestral states for continuous-valued characters on a phyloge- the Devonian period of South Africa. – Nature 443: 981–984. netic tree. – Systematic Zoology 40: 304–314. Gillooly, J. F., Brown, J. H., West, G. B., Savage, V. M. and Charnov, Maddison, W. P. and Maddison, D. R. 2005. MACCLADE, Analysis of E. L. 2001. Effects of size and temperature on metabolic rate. – Phylogeny and Character Evolution, Version 4.07. Sunderland Asso- Science 293: 2248–2251. ciates, Inc, Massachusetts. Gradstein, F. M., Ogg, J. G., Smith, A. G., Agterberg, F. P., Bleeker, Maddison, W. P. and Maddison, D. R. 2006. MESQUITE: A Modular W., Cooper, R. A. et al. 2004. A Geologic Time Scale. Cambridge System for Evolutionary Analysis, Version 1.11. http://mesquite- University Press, Cambridge. project.org. Haldane, J. B. S. 1985. On Being the Right Size and Other Essays. Madin, J. S., Alroy, J., Aberhan, M., Fursich, F. T., Kiessling, W., Oxford University Press, Oxford. Kosnik, M. A. and Wagner, P. J. 2006. Statistical independence Hone, D. W. E. and Benton, M. J. 2005. The evolution of large size: of escalatory ecological trends in Phanerozoic marine How does Cope’s Rule work? Trends in Ecology and Evolution 20: invertebrates. – Science 312: 897–900. 4–6. Mallatt, J. and Chen, J. Y. 2003. Fossil sister group of craniates: Hurley, I. A., Mueller, R. L., Dunn, K. A., Schmidt, E. J., Fried- Predicted and found. – Journal of Morphology 258: 1–31. man, M., Ho, R. K., Prince, V. E., Yang, Z. H., Thomas, M. G. McGhee, G. R. 1996. The Late Devonian Mass Extinction: The and Coates, M. I. 2007. A new time-scale for ray-finned fish Frasnian/Famennian Crisis. – Columbia University Press, New York. evolution. – Proceedings of the Royal Society of B-Biological Sciences McNab, B. K. 2002. Physiological Ecology of Vertebrates: A View from 274: 489–498. Energetics. – Cornell University Press, Ithaca. Hutchinson, G. E. and Macarthur, R. H. 1959. A theoretical eco- Northcutt, R. G. 2005. The new head hypothesis revisited. – Journal logical model of size distributions among species of animals. – of Experimental Zoology Part B-Molecular and Developmental Evolu- American Naturalist 93: 117–125. tion 304B: 274–297. Janvier, P. 1996. Early Vertebrates. Oxford University Press, Oxford. Northwood, C. 2005. Early Triassic coprolites from Australia and Janvier, P. 2003. Vertebrate characters and the Cambrian their palaeobiological significance. – Palaeontology 48: 49–68. vertebrates. – Comptes Rendus Palevology 2: 523–531. O’Leary, M. A. 1999. Parsimony analysis of total evidence from Janvier, P., Desbiens, S., Willett, J. A. and Arsenault, M. 2006. extinct and extant taxa and the Cetacean-Artiodactyl question Lamprey-like gills in a gnathostome-related Devonian jawless (Mammalia, Ungulata). – Cladistics 15: 315–330. vertebrate. – Nature 440: 1183–1185. O’Leary, M. A. 2001. The phylogenetic position of cetaceans: Fur- Janvier, P. and Lund, R. 1983. Hardistiella montanensis N. General et ther combined data analyses, comparisons with the stratigraphic sp. (Petromyzontida) from the Lower Carboniferous of Montana, record and a discussion of character optimization. – American with remarks on the affinities of the lampreys. – Journal of Verte- Zoologist 41: 487–506. brate Paleontology 2: 407–413. Pagel, M. 1999. Inferring the historical patterns of biological Kidwell, S. M. and Flessa, K. W. 1995. The quality of the fossil evolution. – Nature 401: 877–884. record: Populations, species, and communities. – Annual Review Pagel, M., Venditti, C. and Meade, A. 2006. Large punctuational of Ecology and Systematics 26: 269–299. contribution of speciation to evolutionary divergence at the Knouft, J. H. and Page, L. M. 2003. The evolution of body size in molecular level. – Science 314: 119–121. extant groups of North American freshwater fishes: Speciation, Patterson, C. 1981. Significance of fossils in determining evolution- size distributions, and Cope’s rule. – American Naturalist 161: ary relationships. – Annual Review of Ecology and Systematics 12: 413–421. 195–223. © 2009 The Authors 365 Journal compilation © 2009 The Royal Swedish Academy of Sciences Fossils estimates of ancestral body size in fishes • Albert et al. Acta Zoologica (Stockholm) 90 (Suppl. 1): 357–384 (January 2009) Peters, R. H. 1983. The Ecological Implications of Body Size. Cam- Strathdee, A. T. and Bale, J. S. 1998. Life on the edge: Insect ecology bridge University Press, Cambridge. in Arctic environments. – Annual Review of Entomology 43: 85– Peters, S. E. and Foote, M. 2001. Biodiversity in the Phanerozoic: 106. A reinterpretation. – Paleobiology 27: 583–601. Swofford, D. L. and Maddison, W. P. 1987. Reconstructing ances- Pollock, D. D., Zwickl, D. J., McGuire, J. A. and Hillis, D. M. 2002. tral character states under Wagner parsimony. – Mathematical Bio- Increased taxon sampling is advantageous for phylogenetic sciences 87: 199–229. inference. – Systematic Biology 51: 664–671. Twitchett, R. J., Feinberg, J. M., O’Connor, D. D., Alvarez, W. and Polly, P. D. 1998. Cope’s rule. – Science 282: 50–51. McCollum, L. B. 2005. Early Triassic ophiuroids: Their paleo- Polly, P. D. 2001. Paleontology and the comparative method: ecology, taphonomy, and distribution. – Palaios 20: 213–223. Ancestral node reconstructions versus observed node values. – Underwood, C. J. 2006. Diversification of the Neoselachii (Chon- American Naturalist 157: 596–609. drichthyes) during the Jurassic and Cretaceous. – Paleobiology 32: Prendini, L. 2001. Species or supraspecific taxa as terminals in 215–235. cladistic analysis? Groundplans versus exemplars revisited. – Valentine, J. W., Jablonski, D., Kidwell, S. and Roy, K. 2006. Assess- Systematic Biology 50: 290–300. ing the fidelity of the fossil record by using marine bivalves. – Prendini, L. and Wheeler, W. C. 2004. Assembling the scorpion tree Proceedings of the National Academy of Sciences of the United States of life: Phylogeny of extant scorpions based on morphology, of America 103: 6599–6604. molecules and exemplars. – Cladistics 20: 602–603. Vermeij, G. J. 1993. Evolution and Escalation: An Ecological History of Santini, F. and Tyler, J. C. 2004. The importance of even highly Life. Princeton University Press, Princeton. incomplete fossil taxa in reconstructing the phylogenetic relation- Vermeij, G. J. 1994. The evolutionary interaction among species: ships of the Tetraodontiformes (Acanthomorpha: Pisces). – Selection, escalation, and coevolution. – Annual Review of Ecology Integrative and Comparative Biology 44: 349–357. and Systematics 25: 219–236. Schmidt-Nielsen, K. 1984. Scaling Why is Animal Size so Important? Wake, D. B. 1991. Homoplasy – the result of natural-selection, or Cambridge University Press, Cambridge. evidence of design limitations. – American Naturalist 138: 543–567. Sepkoski, J. J. 2002. A compendium of fossil marine animal genera. Webby, B., Paris, D., Droser, M. L. and Percival, I. G. 2004. The In Jablonski, D. and Foote, M. (Eds): Bulletins of American Pale- Great Ordovician Biodiversification Event. Columbia University ontology, pp. 1–560. Paleontological Research Institution, Ithaca. Press, New York. Shu, D. G. 2003. A paleontological perspective of vertebrate Webster, A. J., Gittleman, J. L. and Purvis, A. 2004. The life history origin. – Chinese Science Bulletin 48: 725–735. legacy of evolutionary body size change in carnivores. – Journal of Shu, D. G., Morris, S. C., Han, J., Zhang, Z. F., Yasui, K., Janvier, Evolutionary Biology 17: 396–407. P., Chen, L., Zhang, X. L., Liu, J. N., Li, Y. and L., H. Q. 2003. Wickstrom, L. M. and Donoghue, P. C. J. 2005. Cladograms, phy- Head and backbone of the Early Cambrian vertebrate logenies and the veracity of the conodont fossil record. Conodont Haikouichthys. – Nature 421: 526–529. Biology and Phylogeny: Interpreting the Fossil Record, pp. 185–218. Simpson, G. G. 1952. How many species? Evolution 6: 342. The Palaeontological Association, Cambridge, UK. Smith, F. A., Brown, J. H., Haskell, J. P., Lyons, S. K., Alroy, J., Wiens, J. J. 2006. Missing data and the design of phylogenetic Charnov, E. L. et al. 2004. Similarity of mammalian body size analyses. – Journal of Biomedical Informatics 39: 34–42. across the taxonomic hierarchy and across space and time. – Wilkinson, M. 2003. Missing entries and multiple trees: Instability, American Naturalist 163: 672–691. relationships, and support in parsimony analysis. – Journal of Stanley, S. M. 1973. An explanation for Cope’s rule. – Evolution 27: Vertebrate Paleontology 23: 311–323. 1–26. Wilson, M. V. H. 1992. Importance for phylogeny of single and mul- Stanley, S. M. 1998. Macroevolution. Pattern and Process. The Johns tiple stem-group fossil species, with examples from fresh-water Hopkins University Press, Baltimore. fishes. – Systematic Biology 41: 462–470. © 2009 The Authors 366 Journal compilation © 2009 The Royal Swedish Academy of Sciences