Slaying dragons: limited evidence for unusual body size evolution on islands

Transcription

1 Journal of Biogeography (J. Biogeogr.) (2011) 38, ORIGINAL ARTICLE Slaying dragons: limited evidence for unusual body size evolution on islands Shai Meiri 1 *, Pasquale Raia 2 and Albert B. Phillimore 3 1 Department of Zoology, Faculty of Life Sciences, Tel Aviv University, Tel Aviv, Israel, 2 Dipartimento di Scienze della Terra, Università Federico II, Naples, Italy, 3 Division of Biology, Imperial College at Silwood Park, Ascot SL5 7PY, UK ABSTRACT Aim Island taxa often attain forms outside the range achieved by mainland relatives. Body size evolution of vertebrates on islands has therefore received much attention, with two seemingly conflicting patterns thought to prevail: (1) islands harbour animals of extreme size, and (2) islands promote evolution towards medium body size ( the island rule ). We test both hypotheses using body size distributions of mammal, lizard and bird species. Location World-wide. Methods We assembled body size and insularity datasets for the world s lizards, birds and mammals. We compared the frequencies with which the largest or smallest member of a group is insular with the frequencies expected if insularity is randomly assigned within groups. We tested whether size extremes on islands considered across mammalian phylogeny depart from a null expectation under a Brownian motion model. We tested the island rule by comparing insular and mainland members of (1) a taxonomic level and (2) mammalian sister species, to determine if large insular animals tend to evolve smaller body sizes while small ones evolve larger sizes. Results The smallest species in a taxon (order, family or genus) are insular no more often than would be expected by chance in all groups. The largest species within lizard families and bird genera (but no other taxonomic levels) are insular more often than expected. The incidence of extreme sizes in insular mammals never departs from the null, except among extant genera, where gigantism is marginally less common than expected under a Brownian motion null. Mammals follow the island rule at the genus level and when comparing sister species and clades. This appears to be driven mainly by insular dwarfing in large-bodied lineages. A similar pattern in birds is apparent for species within orders. However, lizards follow the converse pattern. Main conclusions The popular misconception that islands have more than their fair share of size extremes may stem from a greater tendency to notice gigantism and dwarfism when they occur on islands. There is compelling evidence for insular dwarfing in large mammals, but not in other taxa, and little evidence for the second component of the island rule gigantism in small-bodied taxa. *Correspondence: Shai Meiri, Department of Zoology, Faculty of Life Sciences, Tel Aviv University, Tel Aviv, Israel. uncshai@post.tau.ac.il Keywords Birds, dwarfism, evolution, gigantism, island biogeography, island rule, lizards, mammals. INTRODUCTION Giant tortoises, enormous flightless birds and huge bears, alongside minute deer, tiny lizards, dwarf elephants and, lately, pygmy humans all spring to mind when the body sizes of island vertebrates is discussed. Body size evolution on islands is perceived to be fast (Lister, 1989; Millien, 2006) and has produced extreme phenotypes, with the smallest or the largest 89 doi: /j x

2 S. Meiri et al. species of many clades being insular (Hooijer, 1967; Berry, 1998; Greer, 2001; Glaw et al., 2006; Whittaker & Fernández- Palacios, 2007; Hedges, 2008; Losos & Ricklefs, 2009). For example, the St Helena earwig, Labidura herculeana, the Indonesian stick insect, Pharnacia serratipes, and the New Zealand giant wetas (Deinacrida spp.) are probably the largest representatives of their clades (Chown & Gaston, 2010). Similarly, the world s largest bat (Smith et al., 2003) is the Philippine-endemic golden crowned flying fox, Acerodon jubatus, the largest Quaternary bird was the Madagascan elephant bird, Aepyornis maximus, and the largest raptor was the New Zealand endemic Haast s eagle, Harpagornis moorei (Worthy et al., 2002; Murray & Vickers-Rich, 2004). Among reptiles, the largest living lizard (the Komodo dragon, Varanus komodoensis) and tortoises (the giant tortoises of Aldabra and the Galapagos, Geochelone gigantea and Geochelone elephantopus) are insular endemics (Arnold, 1979; Meiri, 2008). Islands also harbour the smallest members of several clades: the smallest bird is believed to be the Cuban bee hummingbird, Mellisuga helenaei (although Dunning, 2008b, suggests that Thaumastura cora from mainland Peru is smaller), and species of Caribbean Leptotyphlops and Sphaerodactylus are the world s smallest snakes and lizards, respectively (Hedges, 2008; Meiri, 2008). This perceived abundance of insular size extremes is usually thought to be a response to the low intensity of competition and predation both within and across taxa: the absence of carnivorous mammals is most often quoted as allowing the evolution of large size in birds (mainly through the evolution of flightlessness; e.g. Bunce et al., 2005; Murray & Vickers-Rich, 2004), reptiles (e.g. Case, 1978; Meiri, 2008) and small mammals (e.g. Angerbjörn, 1986; Adler & Levins, 1994). Alternatively, this perceived pattern may simply reflect an ascertainment bias, i.e. we may be more likely to notice animals of extreme body size when they happen to live on islands (Whittaker & Fernández-Palacios, 2007). By contrast, the island rule suggests that, rather than showing size extremes on islands, insular populations should be closer to the clade-wide median body size than their mainland counterparts (Lomolino, 1985, 2005). According to the island rule, populations of small species will evolve larger size, populations of large species will dwarf, and populations of average-sized species will show little size evolution on islands (Lomolino, 2005; Welch, 2009). Viewed at the clade level, the island rule predicts stabilizing selection: on islands an individual should not be either too large or too small. Once the optimum is reached (via directional selection on the founding population), stabilizing selection should maintain phenotypes around it. Interestingly, an opposite pattern of disruptive selection and increased variance is often thought to prevail within populations on islands (Van Valen, 1965; Scott et al., 2003; cf. Meiri et al., 2005a). The island rule is thought to manifest the combined effects of lower predation pressures on islands (Heaney, 1978), character release in the absence of competitors (Dayan & Simberloff, 1998) and the paucity of resources on islands driving dwarfism in large-bodied forms (Lomolino, 2005). Empirical evidence from terrestrial vertebrate studies provides mixed support for this rule (Clegg & Owens, 2002; Boback & Guyer, 2003; Lomolino, 2005; Meiri, 2007), with the best support coming from data for mammals (Lomolino, 1985; Price & Phillimore, 2007; Welch, 2009; but see Meiri et al., 2004, 2006, 2008). Size evolution is often hypothesized to be most drastic on small islands, with island area showing complex interactions with body size (Heaney, 1978), but empirical patterns are equivocal (e.g. Meiri et al., 2005b; Wu et al., 2006; Schillaci et al., 2009). These two hypotheses, that islands should harbour extreme sizes, and that they should harbour taxa that are closer to a cladewide mode, need not necessarily be contradictory (Fig. 1). If a clade, such as mammals, has a single size attractor (e.g. Brown et al., 1993) then members of a subclade may evolve to a size extreme: insular members of a subclade of large-bodied animals (e.g. elephants) will be the smallest within this subclade, but not across the larger clade as a whole. Similarly, insular members of a subclade of small-bodied animals (e.g. shrews) may be the largest members of this subclade, but not the largest mammals overall. If, however, each subclade has its own optimal size that its insular members evolve towards (Lomolino, 2005) then islands should harbour few size extremes (Fig. 1c). The seemingly conflicting discussion of insular size extremes on the one hand, and insular medium sizes on the other hand, stems in part from the different phylogenetic and temporal scope of studies dealing with them. Studies of size extremes are usually conducted at the inter-specific level (Glaw et al., 2006; Hedges, 2008), and often deal with extinct taxa (Kurten, 1953; Sondaar, 1977; Steadman et al., 2002; Raia et al., 2003). The study of evolution towards medium sizes usually involves intra-specific studies of insular and mainland populations of extant species (e.g. Lomolino, 1985; Boback & Guyer, 2003; Meiri, 2007). Evolutionary processes above and below the species level may differ through, for example, species sorting and adaptive radiation in the former versus inter-island and island mainland gene flow in the latter (Jablonski, 2008). As far as we are aware McClain et al. (2006) and Welch (2010) present the only purely inter-specific studies of the island rule, comparing mean sizes within genera of deep sea (= insular in McClain et al.) and shallow sea ( mainland ) species, rather than using comparisons within single species. An argument for extending tests of the island rule to the species level is that the selection pressures thought to promote convergence on a median body size for island populations should act similarly on island species. The major difference between the two scenarios is that there should be less gene flow between species than between populations, meaning that evolution should proceed more rapidly in the former. However, there is no obvious reason why the type of selection and adaptive optima should differ in these two contexts. A major advantage of intra-specific studies of insular size evolution is that they compare very closely related taxa (i.e. different populations within a species), usually from areas that are in geographic proximity. Intra-specific comparisons thus control for much of the variation in size that is unrelated to insularity. Restricting ourselves to intra-specific studies, however, we may be missing 90 Journal of Biogeography 38,

3 Island vertebrates and body size extremes No island rule, no insular size extremes No island rule, insular size extremes Island rule, no insular size extremes (a) (b) (c) Body size of island taxa Body size of mainland taxa Island rule, insular size extremes No island rule, insular size extremes Reversed island rule, insular size extremes (d) (e) (f) Body size of island taxa Body size of mainland taxa Figure 1 Schemes showing possible ways in which patterns of body size on islands and mainland within taxa (size extremes or average sizes) can combine to produce patterns of body size across taxa (slope <1, slope = 1, slope >1). x-axis, mainland body size; y-axis, island body size; dashed line, a line with an intercept of 0 and a slope of 1; solid line, slope returned by standardized major axis regression (can mask the dashed line). The body size frequency distributions depict insular (top) and mainland members of a clade (bottom). The mainland mean size is indicated by the vertical, dashed line. (a) The null distribution. No insular size extremes, no island rule (slope of 1, similar insular and mainland size ranges). (b) There are size extremes because island taxa have wider size distributions, but no island rule because mean sizes are similar (slope of 1). (c) The island rule holds (small animals evolve larger size and large ones dwarf, slope <1), but no size extremes because island taxa have narrower size distributions. (d) The island rule holds (slope <1), and insular clades are of extreme size members of large-bodied clades are smallest in their clade and members of small-bodied clades are the largest in their clade. (e) Island members are proportionally always larger than mainland ones, hence there is one type of size extreme, but no island rule (no dwarfing of large animals, slope = 1). (f) There are size extremes, but in the opposite direction than predicted under the island rule. Smallsized insular animals are extremely small, large ones are extremely large (slope >1). the more dramatic cases of insular size evolution, where changes are drastic enough to merit different specific status. Thus intra-specific studies will omit from consideration, for example, Elephas falconeri, which evolved to just 1% of the mass of its mainland ancestor (Roth, 1992), and even studies conducted on species within genera will miss members of insular endemic genera such as the c. 100 kg insular rodent Amblyrhiza inundata (Biknevicius et al., 1993) and the largest gecko, Hoplodactylus delcourti (Russell & Bauer, 1986). Here we examine, in a purely inter-specific fashion, whether: (1) species with extreme body sizes (i.e. the largest and smallest species) in a subclade are more often insular endemics than expected by chance given the number of insular endemics in the subclade; and (2) if there is evidence for a pattern consistent with the island rule above the species level using mean sizes of insular and mainland species within clades. We use nearly complete datasets of species body sizes of mammals, birds and lizards. For mammals (only) a comprehensive species-level phylogeny and an excellent late Pleistocene fossil record exist. This meant that we were also able to use a sister-clade comparison, and include data for species that went extinct since the end of the last glacial, which include some of the most pronounced cases of insular size evolution. MATERIALS AND METHODS Data Body size data are species specific. Body size (snout vent lengths in mm) and insularity data for lizards (Appendix S1a in the Supporting Information) are from an updated version of the Journal of Biogeography 38,

4 S. Meiri et al. dataset of Meiri (2008). Lizard taxonomy follows Uetz et al. (2009). Body mass data (in g) for birds are from Dunning (2008). Where mass data were reported for multiple subspecies the mean of these measurements was taken (intra-specific data were first log 10 -transformed). For each species overall mean body mass was estimated across means for female, male and unsexed birds. We obtained data on avian insularity from McCall (1997), and supplemented and verified them using Avibase ( avibase.bsc-eoc.org) and regional guides (Appendix S1b). Taxonomy follows Clements (1998), and we included species that had been extirpated since the extinction of the dodo (17th century). As we were interested in the effect of insularity on birds, we excluded birds that forage at sea, which we defined as all members of the Alcidae, Diomedeidae, Fregatidae, Hydrobatidae, Laridae, Pelecanidae, Pelecanoididae, Phaethontidae, Phalacocoracidae, Procellariidae, Rynchopidae, Spheniscidae, Stercoraridae, Sternidae and Sulidae families. Body mass and insularity data for mammals are from the 2007 version of Smith et al. (2003) (data kindly provided by Felisa Smith), supplemented with literature data (see Appendix S1c for species for which we obtained data from sources other than the Smith et al. database). Mammal insularity was verified using Wilson & Reeder (2005), regional guides and palaeontological accounts of fossil species. We use only fossil mammals that went extinct since the end of the last glacial. Only fossil species known to be confined to islands during the last glacial were considered to be insular. All analyses were conducted on log 10 -transformed measures of body size, to bring the intra-class distribution of body size closer to a normal distribution and to remove a relationship between the mean and variance of a group (Lynch & Walsh, 1998). Phylogeny Complete species-level phylogenetic data exist only for mammals, and we thus only use phylogenetic analysis on this group. We used the species-level mammal supertree of Bininda- Emonds et al. (2007). The tree was modified as follows: we resolved taxonomic discrepancies between the tree and our data using the taxonomy of Wilson & Reeder (2005), and excluded species for which we had no size data and fossil species whose island endemism was uncertain. Finally, where the phylogenetic affinities of extinct species are well understood we added them to the tree. Our modified tree includes 3961 species. Although this tree excludes some new phylogenetic data, it is the only comprehensive species-level tree currently available for any vertebrate class. Statistical analysis Size extremes For each of the four datasets we counted the number of genera for which either the largest or the smallest species was insular. Only genera having at least one insular member and one continental member were included. We repeated this procedure at the family and order levels (family level only for lizards). We then assessed whether the observed values departed from the null expectation if insular status were randomly assigned by randomly selecting n individuals per taxon (i.e. genus, family or order), where n is the number of insular species in that taxon. We then calculated the number of taxa for which the largest or smallest representative had been randomly assigned insular status. This process was repeated 10,000 times to obtain a distribution of the null expectation for the frequency of insular gigantism or dwarfism. We calculated the proportion of the null distribution that was: (1) greater than or equal to the observed value, and (2) smaller than or equal to the observed value. The smaller of these proportions was multiplied by 2 to give a two-tailed P-value. Our randomization test for size extremes on islands should be conservative for two reasons. First, we classify all species that have at least one mainland population as mainland, even though in some of these species the largest or smallest populations may be insular (e.g. Kodiak brown bears, Ursus arctos middendorffi). Second, if insular taxa are clustered within a subclade (as is the case in mammals; Raia et al., 2010), the variance in body size among insular members of a subclade will be reduced, reducing the potential for the evolution of size extremes. However, it is possible that the ancestral members of subclades that colonize islands are themselves biased in size with respect to the mainland representatives of the subclade (e.g. they may be very large; Lomolino, 1985) and this could generate an increased incidence of insular size extremes without requiring further evolution on islands. In the absence of adequate fossil data we are unable to test this hypothesis. Using the mammalian phylogeny we were able to compare the observed incidence of size extremes with those generated under a phylogenetically explicit null model. We conducted 1000 simulations of body size evolution across the whole phylogeny following a Brownian motion (random walk with constant variance) model using the evolve.phylo function in the R library ape (Paradis et al., 2004). A recent study on body size evolution across the mammalian tree identified a weak but significant signature of early burst evolution (i.e. decelerating rates of phenotypic evolution; Cooper & Purvis, 2010). However, a constant-rate Brownian motion model did not perform much worse, and in the context of our study should represent an unbiased means of generating a phylogenetically explicit null expectation. The fossil record shows that mammals have increased in body size through time according to Cope s rule (e.g., Alroy, 1998), which conflicts with a Brownian model. Nonetheless, with respect to the hypotheses being tested here, adding directionality (universally across island and mainland lineages) to a random walk should not bias our tests. For each simulation we quantified the incidence of gigantism and dwarfism and in this manner generated a null distribution. We then used two-tailed tests to establish the 92 Journal of Biogeography 38,

5 Island vertebrates and body size extremes probability of obtaining the observed incidences of gigantism and dwarfism under the null Brownian model. Mean size We assessed whether the island rule applies above the species level. We calculated the mean log 10 body size of insular and mainland species within each genus. We tested the null hypothesis that there are no differences between patterns of mainland and insular size evolution across genera (Welch, 2009) by examining the slope of the island versus mainland taxon means using standardized major axis (= reduced major axis) regression. A slope <1 is expected under the island rule, which predicts gigantism in small-bodied taxa and dwarfism in large-bodied ones and a reduction of variance amongst insular taxa as compared with their mainland counterparts (Price & Phillimore, 2007). Different taxa within a taxonomic level will vary in age, meaning that the expected variance of phenotypes among them is likely to vary. This heterogeneity of expected variance violates an assumption of standardized major axis (SMA) regression. We therefore used a distribution-free variant of the SMA test, as proposed by Welch (2009). The SMA correlation coefficient (r) between x + y (where x is body size on the mainland and y is body size on an island) and x ) y was the test statistic. The observed correlation coefficient was then compared with that observed when the identity of insular means was randomized with respect to x and y (Welch, 2009). Ten thousand randomizations were conducted and the twotailed P-value was the smaller of twice the proportion of randomized correlation coefficients that were either greater than or equal to or less than or equal to the observed correlation. If the observed SMA correlation between x + y and x ) y was significantly greater than expected at random this would support the island rule. At the family level we adjusted the protocol to control for clade species richness, so that rather than each species contributing equally to the family mean, each species contributed equally to the genus mean and each genus then contributed equally to the family mean. The same procedure was repeated at the order level with the addition of taking the mean across families in an order. Tests with species treated independently (i.e. simple means of all species in a family or order) gave qualitatively the same results (not shown). Using the modified mammal tree we located nodes that represented a partition between a solely insular and a solely mainland clade and estimated the mean body size for each clade. Clades including polytomies were excluded. The sister clades we identified are shown as Appendix S2. As the clades involved in each island versus mainland comparison vary in age, then under a Brownian motion model of evolution the expected variance of the mean values of these clades is expected to vary, violating an assumption of SMA regression (Welch, 2009). Consequently, we estimated the statistical significance of a departure from a 1:1 relationship between the size of island and mainland taxa by applying 10,000 randomizations of the identity of insular and mainland clades using the distribution-free randomization approach of Welch (2009) described above. All statistical analyses were conducted in R (R Development Core Team, 2008) and all tests were two-tailed. RESULTS Lizards Of 5380 lizards in our dataset, 1636 are insular endemics (Appendix S1a). The largest lizard in 9 of 19 families is an insular endemic (see size frequency histograms in Appendix S3a), while the median expected number is five (P = 0.07). At the genus level the number of largest insular species is no different from the null expectation. The number of insular species that are the smallest species in their genus or family is likewise not statistically different from that expected by chance (Table 1). The SMA slope between the taxon mean masses on islands versus mainland does not depart significantly from 1 at the genus level (Fig. 2a). At the family level the slope estimate of 1.24 [95% confidence interval (CI) = ] (Fig. 2b) is significantly steeper than 1 using both SMA (r = 0.49, P < 0.05) and the distribution-free randomization test (P < 0.05, similar to the pattern depicted in Fig. 1f). Table 1 Actual and expected numbers of taxa in which insular members are the largest or the smallest, in different taxa at different taxonomic levels. Clade Level Number of clades Largest Smallest Lizards Genus (36) 36 (35) Lizards Family 19 9* (5) 7 (5) Birds Genus ** (71) 73 (71) Birds Family (21) 21 (21) Birds Order 19 3 (3) 6 (4) Mammals Genus (36) 38 (35) Mammals Family 39 9 (11) 10 (12) Mammals Order 12 4 (3) 2 (3) Mammalsà Genus (40) 41 (39) Mammalsà Family (14) 12 (14) Mammalsà Order 14 5 (3) 3 (3) Mammals [BM] Genus 91 28* (35) 38 (35) Mammals [BM] Family 39 9 (11) 10 (11) Mammals [BM] Order 12 4 (3) 2 (3) Mammalsà [BM] Genus (37) 41 (37) Mammalsà [BM] Family (14) 12 (14) Mammalsà [BM] Order 14 6 (3) 3 (3) *, **Denote statistical significance at the a = 0.1 and 0.01 levels, respectively. The median expected values derived from our randomizations are in parentheses. àincluding extinct species. [BM] denotes analyses conducted using Brownian motion simulations rather than randomizations. Note that because we only have phylogenetic data for c. 92% of mammalian species observed numbers of size extremes can differ between non-phylogenetic and phylogenetic analyses. Journal of Biogeography 38,

6 S. Meiri et al. Insular mean SVL (log 10 scale) (a) (b) Continental mean SVL (log 10 scale) Figure 2 Standardized major axis regression of mean body size [log 10 -transformed snout vent length (SVL) in mm] on islands versus continents for species in lizard genera (a) and families (b). The dashed line has a slope of 1 and an intercept of 0. The solid line represents the standardized major axis regression slope estimate. Insular mean mass (log 10 scale) (a) (b) (c) Continental mean mass (log 10 scale) Figure 3 Standardized major axis regression of mean body size (log 10 -transformed mass, in g) on islands versus continents for species in bird genera (a), families (b) and orders (c). The dashed line has a slope of 1 and an intercept of 0. The solid line represents the standardized major axis regression slope estimate. Birds Of 8069 bird species in our dataset, 1347 are insular endemics (we have maximum mass data for all, and for 7552, including 1268 insular endemic species, we have data on mean size; Appendix S1b). In 90 out of 223 avian genera the largest member of the clade is an insular endemic. This is significantly more than expected under our randomizations (median expected = 71; P < 0.01, Table 1). However, at the family and order levels (see size frequency histograms in Appendix S3b) the largest member of a clade is no more often an insular endemic than expected at random. The frequency with which the smallest member of a clade is an insular endemic does not depart from the null expectation for any taxonomic level. Moreover, when we repeated the analysis at the genus level using the maximum rather than mean body size (data are the maximum reported for a species in Dunning, 2008), neither the frequency of gigantism nor dwarfism departed from the null expectation (67 observed insular maxima across 219 genera, median expected = 71, P = 0.53). There was no evidence for the island rule at the genus and family level in birds, with slopes equal to 1.00 and 0.98, respectively (Fig. 3a,b). The slope at the order level, however, was significantly shallower than 1 (SMA slope = 0.76, 95% CI = , r = )0.70, P < 0.01; P randomization < 0.01; Fig. 3c), even when ratites are excluded (SMA slope = 0.78, 95% CI = , r = )0.56, P < 0.05; P randomization < 0.05). Mammals Of 3961 extant mammal species, 670 are insular endemics (778 of 4213 when extinct species are included). The frequency of insular endemics that are either the largest or smallest members of their clades does not depart from the null expectation derived by randomization at any taxonomic level, either including or excluding extinct taxa (Table 1, Appendix S3c). In agreement with the results from randomizations, the frequency of gigantism and dwarfism tended not to exceed the null expectation generated under a single-rate Brownian model on the mammalian phylogeny. This was true for all taxonomic levels, both including and excluding extinct species, except for 94 Journal of Biogeography 38,

7 Island vertebrates and body size extremes extant mammals at the genus level, where the frequency of gigantism was marginally lower than expected (P = 0.1). Mammals show a general tendency to follow the island rule at the genus level. The observed slope of mean insular clade mass versus mean continental clade mass is significantly less than 1 for genera both including and excluding extinct species (extant only: SMA slope = 0.96, 95% CI = , r = )0.24, P < 0.05; P randomization < 0.05, Fig. 4a; extant + extinct: slope = 0.92, 95% CI = , r = )0.41, P < 0.01; P randomization < 0.05, Fig. 4d; see also Fig. 1c). The slope for families is not significantly different from 1 (slope = 1.01, 95% CI = , and 0.95, 95% CI = , with and without extinct species, respectively, Fig. 4b,e). The slope for orders likewise is not significantly different from 1 (slope = 1.01, 95% CI = , and 0.87, 95% CI = , with and without extinct species, respectively; Fig. 4c,f). When analyses were conducted using the mammalian phylogeny, regressing mean body sizes within insular clades on the mean size within their mainland sister clades returned a slope significantly shallower than 1 using SMA (n = 91, slope = 0.931, 95% CI = , r = )0.25, P < 0.01), but not using randomizations (P = 0.155; Fig. 5a). Restricting the analysis to insular versus mainland sister species, we get a slope consistent with the island rule using both tests (n = 57, slope = 0.870, 95% CI = , r = )0.49, P < 0.01; P randomization < 0.05, Fig. 5b). DISCUSSION The largest species in bird and lizard taxa tend to be insular more frequently than expected were insular status assigned to species at random. Interestingly, however, this holds within bird genera (but not when species maximum rather than mean masses are used), whereas lizards only show gigantism among families. Furthermore, the lizard giants are often the result of insular radiations (e.g. Hoplodactylus, Gallotia, Cyclura) on islands lacking mammalian carnivores (here New Zealand, the Canaries and the Antilles, respectively). Such absence of mammalian predation has been hypothesized to promote lizard gigantism by allowing for more foraging, as less time is spent hiding from predators, and by enabling lizards to evolve the role of top predators (Case, 1982; Meiri, 2008). In birds, the largest members of genera that are insular seem to be quite (a) (b) (c) Insular mean mass (log 10 scale) (d) (e) (f) Continental mean mass (log 10 scale) Figure 4 Standardized major axis regression of mean body size (log 10 -transformed mass, in g) on islands versus continents for species in mammalian genera (a, d), families (b, e) and orders (c, f). Plots (a) (c) are for extant taxa only and (d) (f) include extinct taxa (see Materials and Methods). The dashed line has a slope of 1 and an intercept of 0. The solid line represents the standardized major axis regression slope estimate. Journal of Biogeography 38,

8 S. Meiri et al. Insular mean mass (log 10 scale) (a) (b) Continental mean mass (log 10 scale) Figure 5 Standardized major axis regression of mean body size (log 10 -transformed mass, in g) on islands versus continents within (a) mammalian sister clades and (b) a subset of the data in (a): only sister species [only clades in (a) where both mainland and insular sample sizes are 1]. The dashed line has a slope of 1 and an intercept of 0. The solid line represents the standardized major axis regression slope estimate. evenly distributed between continental shelf, volcanic and continental plate islands (e.g. the Philippines, New Guinea, Tasmania, Hawaii and New Zealand). This suggests that release from predation on islands has often promoted gigantism in lizards and perhaps gigantism associated with flightlessness in birds (e.g. the New Zealand moas, Sylviornis; Russell, 1877; McNab, 1994), but perhaps a different mechanism drove gigantism among birds that retained their power of flight (e.g. high population density; Blondel, 2000). Size extremes in insular mammals show no departure from null expectations. However, mammals conform to the island rule at the genus level, especially when fossil species are included. This pattern also emerges when we compare sister species, a better test of size evolution than comparing clade members ignoring their phylogenetic affinities. In all three taxa, deviations from a slope of 1 seem to stem from small (mammals and birds) or large (lizards) members of generally large-bodied forms (e.g. small elephants, artiodactyls, ratites and ducks, large iguanas). Extinct mammals show the most drastic cases of dwarfism. However, many of the recently extinct insular lizards and birds are extremely large (Pregill, 1986; Blondel, 2000) and extinction in these taxa was probably much more prevalent on islands than on mainlands, whereas extinctions of large mammals were common in mainland settings (Barnosky et al., 2004). Thus in early Holocene times there were giant insular birds in orders now exhibiting an overall tendency for small sizes on islands. Indeed very large, recently extinct, insular birds include members of the Falconiformes (e.g. Amplibuteo, Harpagornis moorei, Circus eylesi; Worthy et al., 2002; Suárez & Olson, 2007) and Strigiformes (e.g. Tyto riveroi, Ornimegalonyx oteroi; Alcover & McMinn, 1994), ratites (Dinornis, Aepyornis; Worthy et al., 2002), Anseriformes (e.g. Cygnus falconeri and the very large Hawaii goose ; Milberg & Tyrberg, 1993; Paxinos et al., 1999), Ciconiiformes (Threskiornis solitarius, perhaps a Pelecaniform; Mourer-Chauvire et al., 1995), Gruiformes (Diaphorapteryx hawkinsii, perhaps Aptornis; Holdaway, 1989; Raia, 2009), Galliformes (e.g. Megavitiornis, perhaps Sylviornis; Steadman, 2006) and Columbiformes (Raphus cucullatus, Pezophaps solitaria, Natunaornis gigoura; Steadman, 2006; Worthy, 2000). Thus, the apparent tendency for smaller mean size in insular members of large-bodied avian orders may be a result of human-mediated extinction (Steadman, 2006; Pavia, 2008) rather than a feature of natural insular evolution. The finding that bird genera harbour more insular giants than expected by chance is surprising, given that the mean size of species within bird genera does not seem to differ between islands and mainlands (Fig. 2a). A thorough study of the avian subfossil record may even reveal that when extinct taxa are included a similar pattern will be revealed within families and orders. A common explanation for insular gigantism in birds and reptiles (e.g. elephant birds, moas, Komodo dragons and the giant skinks of Cape Verde and New Caledonia) is that they have evolved large size on islands with no mammalian competitors or predators (Russell, 1877; Case, 1978; McNab, 2002; Meiri, 2008). While we view this as a highly likely explanation, we are not sure it can explain our finding that island birds tend to be the largest members of their genera more often than is expected by chance. Our impression from the data is that these largest members of avian genera are mostly found on large islands, rich in bird, reptile and often mammal species. Being classified as congenerics of mainland forms, insular giants seldom occupy niches vacated by mainland mammals, and usually differ relatively little from the size of their mainland relatives. Currently we are unable to sufficiently explain why this pattern prevails, or why it holds only for birds, and only within genera, and note that species maximum sizes [probably an inferior size measure because it is more sensitive to sample size (Meiri, 2007) but representing about 7% more species (Dunning, 2008)] do not show the same trend. The evidence we find for the island rule in mammals emerges primarily via insular dwarfism in large taxa. Curiously, the tendency of large mammals to dwarf on islands (see also Raia et al., 2010), which is corroborated by our phylogenetic tests, and when fossils are included, is also linked to the absence of predators and competitors, and seems more prevalent in herbivores than in carnivores (Raia & Meiri, 2006). McNab (2002) has claimed that gigantism in insular 96 Journal of Biogeography 38,

9 Island vertebrates and body size extremes birds is more likely in herbivorous taxa. Additionally, in lizards insularity is often associated with large size and herbivory (Troyer, 1983; Meiri, 2008). Gigantism may be favoured where resources are abundant (McClain et al., 2006), and the size of large carnivorous vertebrates may depend on the size of available prey; thus islands lacking large herbivorous mammals are likely also to lack large carnivores. Because mammals can grow much larger than either birds or lizards, one might say that even the largest avian and reptilian predators, Haast s eagle and the Komodo dragon, are not large predators compared with large mammalian carnivores. Thus low predation and competition pressures on islands may tend to produce both relatively small mammals and relatively large lizards. The nature of the islands that we study, in terms of their area and isolation, climate, geology (e.g. whether they are part of the continental shelf, part of a tectonic plate or volcanic) and biogeographic settings (e.g. realm, ocean) may all affect the mode of size evolution (Meiri et al., 2005b; Schillaci et al., 2009). Moreover, these attributes may interact with the ecological attributes of the different taxa themselves, such as their functional group or guild, their diet and microhabitat preferences, as well as their behaviour (Case, 1978; McNab, 1994; Raia & Meiri, 2006) in shaping the way that size evolves. Such attributes of islands and taxa offer promising avenues for research into size evolution on islands. CONCLUSIONS The evidence that insular conditions favour the evolution of extreme sizes within clades is restricted to gigantism in lizard families and, perhaps, bird genera, but is not found in these groups at other taxonomic levels, and neither does it apply to mammals. Furthermore, large insular lizards seem often to result from radiations on oceanic islands with no mammalian carnivores whereas giant insular bird species are scattered over highly variable set of islands (S.M., unpublished). We thus think it is unlikely that these two patterns have a common explanation. The island rule applies in a statistical sense to mammalian species within genera, and between sister species. Biologically, however, while dwarfism in large insular mammals seems prevalent, we find no evidence for the second component of the island rule a general tendency for gigantism in smallbodied mammals. Within small-sized lizard families insular species are smaller than mainland ones, and within largebodied families insular species are larger than mainland ones, reversing the island rule. These findings are consistent with intra-specific studies (Lomolino, 1985; Meiri, 2007), suggesting that similar selection pressures may operate to produce patterns seen both within and between species. More comprehensive fossil data are needed to resolve the pattern of size evolution in island birds. The different courses of size evolution on islands taken by different taxa imply an important role for contingency, as animals differing in their ecology respond differently to the selective forces imposed by agents such as resource abundance, predation and competition, which in turn differ across different islands. ACKNOWLEDGEMENTS We thank Liz Butcher and Barbara Sanger from the Michael Way Library for their enormous help in obtaining literature sources for data used in this work. Felisa Smith kindly provided us with the latest version of the Integrating Macroecological Pattern and Processes across Scales (IMMPS) working group mammalian mass database. We thank Ian Owens for valuable discussion and Mark Lomolino, Craig McClain, John Welch and two anonymous referees for very helpful comments on earlier versions of this manuscript. REFERENCES Adler, G.H. & Levins, R. (1994) The island syndrome in rodent populations. Quarterly Review of Biology, 69, Alcover, J.A. & McMinn, M. (1994) Predators of vertebrates on islands. BioScience, 44, Alroy, J. (1998) Cope s rule and the dynamics of body mass evolution in North American fossil mammals. Science, 280, Angerbjörn, A. (1986) Gigantism in island populations of wood mice (Apodemus) in Europe. Oikos, 47, Arnold, E.N. (1979) Indian Ocean giant tortoises: their systematics and island adaptations. Philosophical Transactions of the Royal Society B: Biological Sciences, 286, Barnosky, A.D., Koch, P.L., Feranec, R.S., Wing, S.L. & Shabel, A.B. (2004) Assessing the causes of late Pleistocene extinctions on the continents. Science, 306, Berry, R.J. (1998) Evolution of small mammals. Evolution on islands (ed. by P.R. Grant), pp Oxford University Press, Oxford. Biknevicius, A.R., McFarlane, D.A. & MacPhee, R.D.E. (1993) Body size in Amblyrhiza inundata (Rodentia: Caviomorpha), an extinct megafaunal rodent from the Anguilla Bank, West Indies: estimates and implications. American Museum Novitates, 3079, Bininda-Emonds, O.R.P., Cardillo, M., Jones, K.E., MacPhee, R.D.E., Beck, R.M.D., Grenyer, R., Price, S.A., Vos, R.A., Gittleman, J.L. & Purvis, A. (2007) The delayed rise of present-day mammals. Nature, 446, Blondel, J. (2000) Evolution and ecology of birds on islands: trends and prospects. Vie et Milieu, 50, Boback, S.M. & Guyer, C. (2003) Empirical evidence for an optimal body size in snakes. Evolution, 57, Brown, J.H., Marquet, P.A. & Taper, M.L. (1993) Evolution of body size, consequences of an energetic definition of fitness. The American Naturalist, 142, Bunce, M., Szulkin, M., Lerner, H.R.L., Barnes, I., Shapiro, B., Cooper, A. & Holdaway, R.N. (2005) Ancient DNA provides new insights into the evolutionary history of New Zealand s extinct giant eagle. PLoS Biology, 3, Journal of Biogeography 38,

10 S. Meiri et al. Case, T.J. (1978) A general explanation for insular body size trends in terrestrial vertebrates. Ecology, 59, Case, T.J. (1982) Ecology and evolution of the insular giant chuckawallas, Sauromalus hispidus and Sauromalus varius. Iguanas of the world: their behavior, ecology and conservation (ed. by G.M. Burghardt and A.S. Rand), pp Noyes Publications, Park Ridge, NJ. Chown, S.L. & Gaston, K.J. (2010) Body size variation in insects: a macroecological perspective. Biological Reviews, 85, Clegg, S.M. & Owens, I.P.F. (2002) The island rule in birds: medium body size and its ecological explanation. Proceedings of the Royal Society B: Biological Sciences, 269, Clements, J.F. (1998) Birds of the world: a checklist, 5th edn. Ibis, Vista, CA. Cooper, N. & Purvis, A. (2010) Body size evolution in mammals: complexity in tempo and mode. The American Naturalist, 175, Dayan, T. & Simberloff, D. (1998) Size patterns among competitors: ecological character displacement and character release in mammals, with special reference to island populations. Mammal Review, 28, Dunning, J.B. (2008) CRC handbook of avian body masses, 2nd edn. CRC Press, London. Glaw, F., Kosuch, J., Henkel, F.W., Sound, P. & Bohme, W. (2006) Genetic and morphological variation of the leaftailed gecko Uroplatus fimbriatus from Madagascar, with description of a new giant species. Salamandra, 42, Greer, A.E. (2001) Distribution of maximum snout vent length among species of scincid lizards. Journal of Herpetology, 35, Heaney, L.R. (1978) Island area and body size of insular mammals: evidence from the tri-colored squirrel (Callosciurus prevosti) of Southeast Asia. Evolution, 32, Hedges, S.B. (2008) At the lower size limit in snakes: two new species of threadsnakes (Squamata, Leptotyphlopidae: Leptotyphlops) from the Lesser Antilles. Zootaxa, 1841, Holdaway, R.N. (1989) New Zealand s pre-human avifauna and its vulnerability. New Zealand Journal of Ecology, 12(Suppl.), Hooijer, D.A. (1967) Indo-Australian insular elephants. Genetica, 38, Jablonski, D. (2008) Species selection: theory and data. Annual Review of Ecology, Evolution, and Systematics, 39, Kurten, B. (1953) On the variation and population dynamics of fossil and recent mammal populations. Acta Zoologica Fennica, 76, Lister, A.M. (1989) Rapid dwarfing of red deer on Jersey in the last interglacial. Nature, 342, Lomolino, M.V. (1985) Body size of mammals on islands: the island rule reexamined. The American Naturalist, 125, Lomolino, M.V. (2005) Body size evolution in insular vertebrates: generality of the island rule. Journal of Biogeography, 32, Losos, J.B. & Ricklefs, R.E. (2009) Adaptation and diversification on islands. Nature, 457, Lynch, M. & Walsh, B. (1998) Genetics and analysis of quantitative traits. Sinauer Associates, Sunderland, MA. McCall, R.A. (1997) Biological, geographical and geological factors influencing biodiversity on islands. PhD Thesis, University of Oxford, Oxford. McClain, C.R., Boyer, A.G. & Rosenberg, G. (2006) The island rule and the evolution of body size in the deep sea. Journal of Biogeography, 33, McNab, B.K. (1994) Energy conservation and the evolution of flightlessness in birds. The American Naturalist, 144, McNab, B.K. (2002) Minimizing energy expenditure facilitates vertebrate persistence on oceanic islands. Ecology Letters, 5, Meiri, S. (2007) Size evolution in island lizards. Global Ecology and Biogeography, 16, Meiri, S. (2008) Evolution and ecology of lizard body sizes. Global Ecology and Biogeography, 17, Meiri, S., Dayan, T. & Simberloff, D. (2004) Body size of insular carnivores: little support for the island rule. The American Naturalist, 163, Meiri, S., Dayan, T. & Simberloff, D. (2005a) Variability and sexual size dimorphism in carnivores: testing the niche variation hypothesis. Ecology, 86, Meiri, S., Simberloff, D. & Dayan, T. (2005b) Insular carnivore biogeography: island area and mammalian optimal body size. The American Naturalist, 165, Meiri, S., Dayan, T. & Simberloff, D. (2006) The generality of the island rule reexamined. Journal of Biogeography, 33, Meiri, S., Cooper, N. & Purvis, A. (2008) The island rule: made to be broken? Proceedings of the Royal Society B: Biological Sciences, 275, Milberg, P. & Tyrberg, T. (1993) Naïve birds and noble savages a review of man-caused prehistoric extinctions of island birds. Ecography, 16, Millien, V. (2006) Morphological evolution is accelerated among island mammals. PLoS Biology, 4, Mourer-Chauvire, C., Bour, R. & Ribes, S. (1995) Was the solitaire of Réunion an ibis? Nature, 373, 568. Murray, P.F. & Vickers-Rich, P. (2004) Magnificent mihirungs. The colossal flightless birds of the Australian dreamtime. Indiana University Press, Bloomington, IN. Paradis, E., Claude, J. & Strimmer, K. (2004) APE: analyses of phylogenetics and evolution in R language. Bioinformatics, 20, Pavia, M. (2008) The evolution dynamics of the Strigiformes in the Mediterranean islands with the description of Aegolius martae n. sp. (Aves, Strigidae). Quaternary International, 182, Paxinos, E.E., James, H.F., Olson, S.L., Sorenson, M.D., Jackson, J. & Fleischer, R.C. (1999) mtdna from fossils reveals a radiation of Hawaiian geese recently derived from the 98 Journal of Biogeography 38,

11 Island vertebrates and body size extremes Canada goose (Branta canadensis). Proceedings of the National Academy of Sciences USA, 99, Pregill, G. (1986) Body size of insular lizards: a pattern of Holocene dwarfism. Evolution, 40, Price, T.D. & Phillimore, A.B. (2007) Reduced major axis regression and the island rule. Journal of Biogeography, 34, R Development Core Team (2008) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Available at: r-project.org/. Raia, P. (2009) Gigantism. Encyclopedia of islands (ed. by R.G. Gillespie and D.A. Clague), pp University of California Press, Berkeley, CA. Raia, P. & Meiri, S. (2006) The island rule in large mammals: paleontology meets ecology. Evolution, 60, Raia, P., Barbera, C. & Conte, M. (2003) The fast life of a dwarfed giant. Evolutionary Ecology, 17, Raia, P., Carotenuto, F. & Meiri, S. (2010) One size does not fit all: no evidence for an optimal body size on islands. Global Ecology and Biogeography, 19, Roth, V.L. (1992) Inferences from allometry and fossils: dwarfing of elephants on islands. Oxford surveys in evolutionary biology, Vol. 8 (ed. by D. Futuyma and J. Antonovics), pp Oxford University Press, Oxford. Russell, A.P. & Bauer, A.M. (1986) Le gecko geant Hoplodactylus delcourti et ses relations avec le gigantisme et l endemisme insulaire chez les Gekkonidae. Mesogee, 46, Russell, I.C. (1877) The giant birds of New Zealand. The American Naturalist, 11, Schillaci, M.A., Meijaard, E. & Clark, T. (2009) The effect of island area on body size in a primate species from the Sunda Shelf Islands. Journal of Biogeography, 36, Scott, S.N., Clegg, S.C., Blomberg, S.P., Kikkawa, J. & Owens, I.P.F. (2003) Morphological shifts in island-dwelling passerines: the roles of generalist foraging and niche expansion. Evolution, 57, Smith, F.A., Lyons, S.K., Morgan Ernest, S.K., Jones, K.E., Kaufman, D.M., Dayan, T., Marquet, P.A., Brown, J.H. & Haskell, J.P. (2003) Body mass of late Quaternary mammals. Ecology, 84, Sondaar, P.Y. (1977) Insularity and its effects on mammal evolution. Major patterns of vertebrate evolution (ed. by M.K. Hecht, P.C. Goody and B.M. Hecht), pp Plenum Press, New York. Steadman, D.W. (2006) Extinction and biogeography of tropical Pacific birds. University of Chicago Press, Chicago. Steadman, D.W., Pregill, G.K. & Burley, D.V. (2002) Rapid prehistoric extinction of iguanas and birds in Polynesia. Proceedings of the National Academy of Sciences USA, 99, Suárez, W. & Olson, S.L. (2007) The Cuban fossil eagle Aquila borrasi Arredondo: a scaled-up version of the great blackhawk Buteogallus urubitinga (Gmelin). Journal of Raptor Research, 41, Troyer, K. (1983) The biology of iguanine lizards: present status and future directions. Herpetologica, 39, Uetz, P., Goll, J. & Hallermann, J. (2009) The JCVI reptile database. Available at: (accessed 23 July 2009). Van Valen, L.M. (1965) Morphological variation and the width of the ecological niche. The American Naturalist, 99, Welch, J.J. (2009) Testing the island rule: primates as a case study. Proceedings of the Royal Society B: Biological Sciences, 276, Welch, J.J. (2010) The island rule and deep-sea gastropods: re-examining the evidence. PLoS ONE, 5, e8776. doi: /journal.pone Whittaker, R.J. & Fernández-Palacios, J.M. (2007) Island biogeography: ecology, evolution, and conservation, 2nd edn. Oxford University Press, Oxford. Wilson, D.E. & Reeder, D.M. (2005) (eds) Mammal species of the world, 3rd edn. Johns Hopkins University Press, Baltimore, MD. Worthy, T.H. (2000) The fossil megapodes (Aves: Megapodiidae) of Fiji with descriptions of a new genus and two new species. Journal of the Royal Society of New Zealand, 30, Worthy, T.H., Holdaway, R. & Morris, R. (2002) The lost world of the moa: prehistoric life of New Zealand. Indiana University Press, Bloomington, IN. Wu, Z., Li, Y. & Murray, B.R. (2006) Insular shifts in body size of rice frogs in the Zhoushan Archipelago, China. Journal of Animal Ecology, 75, SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article: Appendix S1 Body size and insularity status for (a) lizards, (b) birds and (c) mammals. Appendix S2 Insular and mainland mammal sister clades. Appendix S3 Size frequency histograms of species within (a) lizard families, (b) bird orders, (c) mammal orders. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be reorganized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Journal of Biogeography 38,