Parallel episodes of phyletic dwarfism in callitrichid and cheirogaleid primates

Authors


Correspondence: Stephen H. Montgomery, Department of Genetics, Evolution and Environment, University College, London, Gower Street, London, WC1E 6BT, UK.

E-mail: stephen.montgomery@cantab.net

Abstract

The Callitrichidae are the smallest anthropoids, whereas the Cheirogaleidae include the smallest of all primates. Using species-level analyses, we show that these are derived conditions; both neonatal and adult body mass decreased in a gradual, phyletic manner in parallel across callitrichids, and across cheirogaleids. We identify lineages with particularly rapid decreases and highlight the pygmy marmoset, Callithrix pygmaea, as a phenotypic outlier. The life-history traits associated with body-mass reduction in each clade suggest that the convergent evolution of small body size was achieved by changes in different ontogenetic stages. Body-size reduction in callitrichids appears to be almost exclusively due to alterations in prenatal growth rate, whereas body-size reduction in cheirogaleids may have been largely due to reduced duration of growth phases. Finally, we use these results to discuss some of the debates surrounding the evolution of Homo floresiensis and suggest potential parallels between the evolution of H. floresiensis and callitrichids.

Introduction

It has historically been accepted that mammals tend to increase in size through time (‘Cope's Rule’; Alroy, 1998; Polly, 1998). Although the evidence for such an extensive trend is disputed (Boucot, 1976; Monroe & Bokma, 2010), evolutionary reductions in body mass, or episodes of ‘dwarfism’ (or ‘nanism’) are considered rare, but may be more frequent on islands (‘The Island Rule’; Foster, 1964) where resource limitation selects for smaller body size in large species (Marshall & Corruccini, 1978). The generality of the Island Rule is disputed (Lomolino, 2005; Meiri et al., 2008), but the available data suggest that it may hold true in primates (Bromham & Cardillo, 2007; Welch, 2009; but see also Meiri et al., 2008).

Although much of the focus around dwarfing has concerned islands, phyletic dwarfing on larger land masses may be a more common phenomenon than previously thought (Prothero & Sereno, 1982) with independent episodes occurring in several mammalian lineages (e.g. Kurtén, 1959; Marshall & Corruccini, 1978; MacFadden, 1986; Perry & Dominy, 2009). It has previously been argued that callitrichid primates are dwarfs (Ford, 1980). Callitrichids are small (120–650 g), neotropical primates allied to the Cebidae (including Cebus, Saimiri and Aotus; 700–2000 g; Wildman et al., 2009; Perelman et al., 2011). They share a number of distinguishing features in addition to small body size including higher rates of twinning, the general absence of third molars and hypocones and the presence of claws (Ford, 1980). These were historically viewed as the retention of primitive characteristics (Hershkovitz, 1977), but Ford (1980) argued that they constitute an ‘adaptive suite’ (Moynihan, 1976), which was the product of an evolutionary reduction in body size. The status of callitrichids as evolutionary dwarfs has been complicated by uncertainty in the phylogenetic relationships of extant genera, which has only recently been resolved (Wildman et al., 2009; Arnold et al., 2010; Perelman et al., 2011), and the paucity of the fossil record (Martin, 1992). Hence, the status of Callimico, which neither twins nor lacks the third molar, and the relationships of Saguinus, Callithrix and the pygmy marmoset, which some raise to the genus level (Cebuella pygmaea; Ford et al., 2009), have prompted questions over whether these traits evolved in synchrony during one evolutionary event or evolved during parallel episodes of body-size reduction (Martin, 1992; Ah-King & Tullberg, 2000). Regardless of the extent of dwarfism during callitrichid evolution, morphological diversification in callitrichids resulted from strong positive selection on size, indicating that body size is a key adaptive trait in this clade (Marroig & Cheverud, 2005, 2010).

Fossil species classified as being monophyletic with extant callitrichids are all from the mid-late Miocene and are estimated to have had body masses between 1000 and 1300 g, above the range of living callitrichids (Fleagle, 1999). However, extant callitrichids do not have large molars relative to their body size, a common feature seen in dwarfed species, questioning their status as dwarfs (Gould, 1975; Martin, 1992; Ravosa, 1992; Plavcan & Gomez, 1993), although this negative allometry is also absent in some other dwarf species (Marshall & Corruccini, 1978). A possible explanation for why callitrichids do not show negative allometry in tooth size has recently been proposed by Marroig & Cheverud (2009). It was previously noted that the relationship between body size and gestation length in callitrichids is weak (Martin, 1992). Marroig & Cheverud (2009) demonstrated that Callithrix and Callimico show paedomorphic morphological traits and suggested that the constant gestation length must be associated with a slowdown in prenatal growth rate, potentially explaining how body size was reduced. Such a change in ontogeny is rare, however, with the majority of episodes of dwarfism being explained by changes in post-natal growth rate or gestation length (Gould, 1975; Plavcan & Gomez, 1993; Webster et al., 2004).

The evolution of body size in the Cheirogaleidae (mouse lemurs, dwarfed lemurs and fork-crowned lemurs) is less frequently discussed. They too are small-bodied (30–450 g) species nested within larger bodied (600–8500 g) clade (Perelman et al., 2011). Mouse lemurs (Microcebus) are often been cited as a potential model of the ancestral primate (Gebo, 2004; Martin et al., 2007). However, recent analyses suggest the small body size of Microcebus may not be a retention of a primitive feature, but the product of evolutionary convergence and dwarfism (Masters et al., 2007; Montgomery et al., 2010).

Insular dwarfing has also been invoked to explain the small body size of a recently described hominin, Homo floresiensis (Brown et al., 2004; Niven, 2007). Controversy surrounds the description of H. floresiensis as a new species, with some arguing that it may be a pathological modern human (Jacob et al., 2006; Hershkovitz et al., 2007; Delaval & Doxsey, 2008; see also Aiello, 2010 for review). Understanding the evolutionary scenarios that bring about dwarfism in independent primate clades may offer new perspectives on these problems.

To test hypotheses surrounding the evolution of dwarfism, it is necessary to incorporate phylogenetic history to test for coevolution between traits and to reconstruct the timing and distribution of evolutionary events (Hansken & Wake, 1993; Gould & MacFadden, 2004). Although ancestral-state reconstructions are dependent on the data and model of evolution available, they may provide insights into the evolutionary history of a trait which is not well preserved in the fossil record. Such a full phylogenetic, comparative analysis has only been made possible recently with a confirmation of genus-level relationships and the estimation of species-level phylogenies (Wildman et al., 2009; Arnold et al., 2010; Perelman et al., 2011). In this study, we performed a species-level analysis of body-size evolution in callitrichids and cheirogaleids. We examined the hypothesis that body-size reduction in callitrichids is driven by changes in prenatal growth rate and ask whether or not the convergent evolution of small-body size in callitrichids and cheirogaleids share changes in the same stages of ontogeny.

Materials and methods

Data

We utilized the consensus species-level, extant primate tree from the recently released 10 k Trees Project (Arnold et al., 2010). To explore the effects of variation in the phylogenetic hypothesis adopted, the analysis was repeated using the species phylogeny from Bininda-Emonds et al. (2007); the results are nearly identical, so we only present the results obtained using the 10 k Trees phylogeny. Data on neonatal and adult body mass were obtained from the PanTHERIA database (Jones et al., 2009). Data were available for both traits for a total of 101 species represented in the 10 k Tree, of which 11 are callitrichids (1 Leontopithecus, 1 Callimico, 4 Callithrix and 5 Saguinus) and five are cheirogaleids (2 Cheirogaleus, 1 Mirza and 2 Microcebus). Data on gestation length and age at sexual maturity were taken from the same source. Sixty-eight species had data for all variables of which seven are callitrichids and three are cheirogaleids. Using these traits, we also calculated:

  • Prenatal growth rate = neonatal body mass [g]/gestation length[days]
  • Post-natal growth rate = (adult body mass [g] − neonatal body mass [g])/age at sexual maturity[days]

Pre- and post-natal growth rates were log10 transformed to improve normality. It should be noted that post-natal growth rate includes adult body mass in the numerator, therefore, when regressed against adult body mass to test for associations with dwarfism, it is possible that any positive association could be driven by autocorrelation. However, as detailed below, the results do not indicate this is a problem. All phenotypic data are reproduced in Table S1. The phylogenetic relationship of the callitrichids and cheirogaleids included in the neonatal/adult body mass data set are shown in Fig. 1, together with their respective outgroups. The full phylogeny is shown in Fig. S1.

Figure 1.

Phylogeny of (a) Callitrichidae and (b) Cheirogaleidae with outgroups. Branches are drawn proportional to time. All species shown have data for adult and neonatal body mass. Species tips ending with black squares also have data for life-history variables. Nodes referred to in the text are labelled. Those nodes which are underlined are present in the tree when species for which no life-history trait data are available are pruned from the phylogeny. The Callitrichidae and Cheirogaleidae are shown in bold with their respective outgroups in grey.

Evolutionary analysis

The evolutionary history of body size in callitrichids and cheirogaleids was analysed in three ways. First, ‘phylogenetic t-tests’ (Organ et al., 2007) were used to statistically test whether callitrichids and cheirogaleids have significantly smaller neonatal and adult body masses than other anthropoids and platyrrhines or other strepsirrhines and lemurs, respectively, and whether or not there are significant differences in gestation length (a measure of the length of prenatal growth), prenatal growth rate, age at sexual maturity (a measure of the length of post-natal growth) or post-natal growth rate. The phylogenetic t-tests (Organ et al., 2007) test for a phylogenetically corrected association between a binary variable (0 or 1), assigned to the two groups under consideration, and the phenotype of interest. The significance of the association is equivalent to a standard t-test, but corrects for nonindependence of the data points (Organ et al., 2007). This was done using Phylogenetic General Least Squares (PGLS) in BayesTraits (Pagel et al., 2004; Pagel & Meade, 2006). With PGLS, the phylogeny is converted into a matrix that resembles a variance–covariance matrix, where the diagonal of the matrix gives information on the path length from root to tips and the off-diagonal values of the matrix provide information on the shared evolutionary history of any pair of species (Pagel, 1997, 1999; Freckleton et al., 2002). In a PGLS regression, the variance–covariance-like matrix is included into the error term of the regression model, and the estimated regression parameters are ‘phylogenetically controlled’ (Pagel, 1997, 1999; Freckleton et al., 2002). The rate parameter lambda (see below) was estimated for all PGLS regressions and unless otherwise stated was not significantly different to one.

In the second set of analyses, we used PGLS regressions to test whether variation in body mass is associated with variation in gestation length, prenatal growth rate, age at sexual maturity or post-natal growth rate. Finally, we conducted ancestral-state reconstructions, implemented in BayesTraits, to estimate the ancestral state of adult and neonatal body mass, gestation length and age at sexual maturity at each node labelled in Fig. 1. The full data set used in the analysis includes representatives of all the major primate clades (see supplementary information), and therefore allows more robust estimates of the basal nodes on callitrichids and cheirogaleids. These estimates were then used to calculate ancestral states of prenatal and post-natal growth rates (see above). To reconstruct ancestral states, Bayes Traits assumes a constant-variance Brownian motion model, but adopts a model-building approach to test and account for deviation from this null-model. The constant variance random walk model has only one parameter, alpha, which describes the instantaneous variance of evolution (Pagel, 1997, 1999). This model represents the default model with all branch length scaling parameters are set as one (Pagel, 1997, 1999). The scaling parameters account for deviation from the null model; lambda reveals to what extent the phylogeny correctly predicts the pattern of covariance among species for a trait (the phylogenetic signal), kappa stretches and compresses branch lengths and tests for stasis in longer branches and delta scales path lengths and tests for adaptive radiations or a greater importance of temporally early change. These parameters were estimated using maximum likelihood. Where a parameter is significantly different from the default value of 1, as determined using a likelihood ratio test (−2(Log[Lh(null model)] − Log[Lh(alternative model)])), they are estimated in the final model (Organ et al., 2007). Once the final ‘best model’ is obtained, this model is used to estimate ancestral states. Hence, although all ancestral-state reconstructions are dependent of the data, phylogenetic hypothesis and model of evolution employed, the approach adopted in Bayes Traits attempts to improve the fit of the data and phylogeny to the model and therefore produces more reliable ancestral-state estimates (Pagel et al., 2004; Pagel & Meade, 2006; Organ et al., 2007). Full information on the rate parameter estimates for each trait are given in the supplementary information.

Ancestral-state reconstructions were performed for each trait to obtain sets of estimates for each node labelled in Fig. 1. The change in trait along each branch was calculated by taking the difference between consecutive nodes. These branch-specific changes, which are independent of each other, were then used in a nonparametric Spearman's rank correlation, implemented using GenStat (VSNi, Hemel Hempstead, UK), to test for time-dependent associations between major changes in body mass and other traits. Rates of evolution were calculated by dividing branch-specific changes by branch length estimates (in millions of years).

Preliminary analyses suggested the extremely small body size of Callithrix pygmaea drags ancestral nodes towards it, with closely related lineages ‘rebounding’ after diverging from the C. pygmaea lineage, suggesting that this species has an unduly large influence on the estimated ancestral states (see supplementary information for further details). To avoid this effect, the ancestral-state reconstructions for both adult and neonatal body mass were performed excluding C. pygmaea. To incorporate the change in body mass along the C. pygmaea lineage into the analyses, we estimated the trait value at Node F (Fig. 1) by setting the branch length of the C. pygmaea terminal branch to 1 × 10−6 and estimating the tip value. This data point is therefore effectively an estimate of the trait value at the point of divergence between C. pygmaea and C. argentata/humeralifera. No other traits suffer from this effect. Neonatal and adult body mass were initially analysed using the full data set, but were subsequently reanalysed using only the species where life-history traits were available. Comparisons of estimates of body mass and life-history trait estimates used the latter and are therefore not affected by differences in sampling.

Unless otherwise stated, the p-values reported below are from PGLS regressions. A summary of the results is provided in Table S2 including p-values corrected for multiple testing. Only the raw p-values are reported in the main text as all analyses reported as significant in the text are robust to correction for multiple testing.

Results

The evolution of body mass in callitrichids

Phylogenetic t-tests confirm that callitrichids have significantly smaller adult body masses (t69 = 2.658, P = 0.009, R2 = 0.093) and neonatal body masses (t69 = 3.012, P = 0.004, R2 = 0.116) than other anthropoids and other non-callitrichid Platyrrhines (adult body mass: t23 = 2.502, P = 0.019, R2 = 0.214; neonatal body mass: t23 = 2.696, P = 0.012, R2 = 0.240). The estimated ancestral state of the last common ancestor of the Cebidae (Node A, Fig. 1) is below the range observed among extant species and suggests that callitrichids descended from a larger bodied ancestor with an adult body mass of 1246.88 g and a neonatal body mass of 108.31 g (95% confidence intervals, adult: 1121.89–1385.80 g; neonate: 108.00–108.63 g). The reconstructed adult body mass is consistent with the body-mass estimates for the mid-late Miocene primates assigned to the Callitrichidae (Fleagle, 1999). Estimated branch-specific changes in adult and neonatal body mass are correlated (Spearman's correlation: t19 = 4.26, P < 0.001, rs = 0.699). Decreases in both traits are widespread across the callitrichid phylogeny (Fig. 2; Table S3). Adult body mass decreased along all but two branches and decreases account for 92.17% of evolutionary time. Neonatal body mass also decreased on all but two branches and decreases account for 96.95% of total evolutionary time. The percentage reduction in neonatal/adult body mass is particularly high (>20%) along six branches of 21, including the stem branch of the callitrichids, the stem of Saguinus and the stem of Callithrix, suggesting that small body size has evolved in parallel across callitrichids. The largest decrease is seen on the terminal C. pygmaea branch with a 60.00% reduction in neonatal body mass and a 69.50% reduction in adult body mass. The second largest decrease occurs along the stem callitrichid branch with a 41.7% decrease in neonatal body mass and a 42.00% decrease in adult body size. Branches with particularly high rates of evolution include the branches leading to the origin of the Leontopithecus-Callimico-Callithrix, the Callimico-Callithrix and the Callithrix clades, and the terminal C. pygmaea lineage. The highest rate of change for both adult and neonatal body mass is the terminal C. pygmaea lineage, where the rate is 5.45 and 6.27 times faster than the mean for neonatal and adult body mass respectively.

Figure 2.

Changes in (a) adult body mass; (b) neonatal body mass: the topology of the tree is as shown in Fig. 1, but branch lengths are drawn according to the difference in log(mass) between consecutive nodes such that at any point in the tree, the phenotypic state can be read off the x-axis. Branches in solid black show decreases in size and are ‘reversed’ in direction, branches in dashed grey show increases. The percentage change in mass is shown for branches with a percentage change ≥20%. The figure begins with the ancestral state for the last common ancestor (LCA) of the Cebidae (black square).

The evolution of life history in callitrichids

Although gestation length and adult body size are significantly associated across primates (t66 = 5.695, P < 0.001, R2 = 0.330) and callitrichids are significantly smaller than other anthropoids, gestation length is not significantly shorter in callitrichids than other anthropoids (t47 = 1.794, P = 0.079, R2 = 0.063) or other non-callitrichid platyrrhines (t18 = 1.561, P = 0.136 R2 = 0.119) as previously suggested (Martin, 1992; Marroig & Cheverud, 2009). Within callitrichids, gestation length is not significantly associated with adult body mass (t5 = 0.573, P = 0.592, R2 = 0.062), nor is age at sexual maturity (t5 = 0.301, P = 0.775, R2 = 0.022) suggesting that changes in the duration of pre- or post-natal growth do not account for the reduction in body size in this clade.

Prenatal growth rate is significantly slower in callitrichids when compared with other anthropoids (t47 = 3.845, P < 0.001, R2 = 0.239) or just non-callitrichid platyrrhines (t18 = 3.818, P = 0.001 R2 = 0.447). In contrast, post-natal growth rate showed no significant difference for either comparison (anthropoids: t47 = 1.277, P = 0.208, R2 = 0.034; just non-callitrichid platyrrhines: t18 = 1.257, P = 0.225, R2 = 0.081). Post-natal growth rate is significantly associated with variation in body mass (t5 = 3.809, P = 0.013, R2 = 0.743), but the significance is lost if C. pygmaea is excluded (t4 = 0.543, P = 0.616, R2 = 0.069). In contrast, prenatal growth rate is significantly associated with adult body size with (t5 = 10.121, P < 0.001, R2 = 0.953) and without C. pygmaea (t4 = 4.671, P = 0.009, R2 = 0.845). When both pre- and post-natal growth rate are included in a multiple regression using PGLS including C. pygmaea, prenatal growth rate is significantly associated with body size, but post-natal growth rate is not (prenatal: t3 = 4.473, P = 0.021; post-natal: t3 = 0.600, P = 0.591).

Ancestral-state reconstructions support the hypothesis that shifts in prenatal growth rate occurred concurrently with episodes of body-mass reduction. Estimated branch-specific changes in prenatal growth rate are significantly correlated with changes in adult body mass (Spearman's correlation: t11 = 2.743, P = 0.027, rs = 0.637). In contrast, changes in post-natal growth rate are not correlated with changes in adult body mass (Spearman's correlation: t11 = 0.968 P = 0.320).

The evolution of body mass in cheirogaleids

Cheirogaleids have significantly smaller adult body masses than other strepsirrhines (t28 = 2.809, P = 0.009, R2 = 0.233) and just other lemurs (t19 = 3.106, P = 0.006, R2 = 0.362). They also have significantly smaller neonatal body masses than other strepsirrhines (t28 = 2.844, P = 0.009, R2 = 0.237) and just other lemurs (t19 = 3.843, P = 0.001, R2 = 0.465). The last common ancestor of Lepilemur and the cheirogaleids (Node L) is estimated to have had an adult body mass of 700.84 g (95% CI: 697.00–704.70 g) and a neonatal body mass of 32.81 g (95% CI: 32.67–32.95 g), both of which are larger than the maximum size seen within cheirogaleids (Table S5). As is the case for callitrichids, decreases in both adult and neonatal body mass were widespread in cheirogaleids (Table S5) and strongly correlated (Spearman's correlation; t9 = 3.65, P = 0.014, rs = 0.773). The largest decreases in adult body mass are estimated to have occurred along the stem cheirogaleid branch (−46.81%), the terminal C. medius branch (−37.55%), the branch leading to the Mirza-Microcebus clade (−47.99%), the branch leading to Microcebus (−56.33%) and the terminal M. rufus branch (−42.85%). The same branches are highlighted when change in mass is controlled for branch length. Large decreases in neonatal body mass (>14%) are also observed on all these branches with the exception of the M. rufus branch, which shows a more modest decrease of 8.22% (Table S5). Interestingly, when body size has increased, it is primarily restricted to changes in adult body size; increases in neonatal body size of 5.67% and 7.84% are estimated to have occurred on the C. major and Mirza terminal branches, respectively, while adult body mass increased by 41.57% and 68.40% respectively. This suggests that decreases in neonatal body mass played a major role in body-size evolution, but post-natal body growth also contributed to both decreases (M. rufus) and increases (C. major and Mirza) in cheirogaleid body size.

Life-history data are somewhat limited in cheirogaleids, with data only available for C. medius, Mirza coquereli and Microcebus murinus. Based on these data, cheirogaleids have significantly shorter gestation periods than other strepsirrhines (t15 = 3.976, P = 0.001, R2 = 0.530) and other lemurs (t8 = 3.507, P = 0.008, R2 = 0.606), and reach sexual maturity at a younger age (strepsirrhines: t15 = 3.842, P = 0.002, R2 = 0.486; other lemurs: t8 = 3.498, P = 0.008, R2 = 0.605). After correcting for multiple testing, they do not have slower pre- or post-natal growth rates than other strepsirrhines or other lemurs (Table S2). There is insufficient power to test for associations between life-history traits and body size within cheirogaleids using PGLS due to the small data set, but the comparison with other strepsirrhines and branch-specific changes within cheirogaleids suggests that dwarfism in cheirogaleids is unlikely to have been brought about by a slowdown in prenatal growth rates (Table S5). Hence, the convergent evolution of small body size in these two clades may have been driven by changes in different ontogenetic stages.

Discussion

Our analysis supports the proposal that callitrichids are phyletic dwarfs, provides the first estimates of the degree and pace of change and suggests that dwarfism occurred in parallel across callitrichids. We also find strong evidence that cheirogaleids can be considered phyletic dwarfs, with extensive and progressive reductions in body mass occurring in this clade. The putative similarities between mouse lemurs and the ancestral primate (Gebo, 2004; Martin et al., 2007) are therefore probably the result of convergent evolution rather than the maintenance of ancestral traits.

The convergent evolution of dwarfism in these clades appears to be associated with evolutionary changes in different life-history traits. We provide robust support for the hypothesis that dwarfism across the callitrichids is associated with changes in prenatal growth rates and not in changes in the duration of prenatal or post-natal growth, or in post-natal growth rate (Martin, 1992; Marroig & Cheverud, 2009). Dwarfism via changes in prenatal ontogeny is considered rare (Gould, 1975; Plavcan & Gomez, 1993; Webster et al., 2004; Marroig & Cheverud, 2009). Indeed, our analysis suggests that dwarfism in cheirogaleids follows the more commonly expected pattern with reductions in gestation length and post-natal growth likely to have had a major influence on body growth. The route to body-size reduction along which callitrichids have progressed may therefore be unique among primates and rare across other mammals.

One notable example where a change in prenatal growth rate is the major contributing factor to dwarfing comes from carnivores. Domestic breeds of dog vary greatly in body mass but, like the callitrichids (Martin, 1992), have almost invariant gestation lengths (Wayne, 1986a). Miniature breeds of dog show a number of paedomorphic morphological traits (Wayne, 1986b), as do callitrichids (Marroig & Cheverud, 2009). Post-natal growth rates do not vary greatly between breeds and together these observations suggest that dwarfism in the domestic dog was caused primarily by slowing foetal growth rate (Wayne, 1986a). Interestingly, two wild carnivores which are thought to have undergone dwarfism in association with an unpredictable food supply, the raccoon dog (Nyctereutes procyonoides) and the bush dog (Speothos venaticus), also have long gestation lengths relative to their body size and lower prenatal growth rates (Wayne, 1986a). This suggests that the slow prenatal growth rates of miniature dog breeds is not just an oddity caused by domestication, but is a route accessible to natural selection (Wayne, 1986a). The small-dog phenotype has a simple genetic basis, being associated with a single major locus (Sutter et al., 2007). Although it is unclear to what extent the genetic basis of phenotypic evolution under artificial selection is analogous to evolution under natural selection, it is possible that the rapid evolutionary reduction of body size in the callitrichids may have been facilitated by selection acting on a small number of loci.

The average rates of body-mass reduction in callitrichids and cheirogaleids are somewhat lower than rates of change in other well-studied episodes of dwarfism in Pleistocene mammals (Kurtén, 1959; Marshall & Corruccini, 1978). However, the timescale considered in these cases is typically thousands of years, rather than millions, and the percentage change is equal or greater than many other examples of insular dwarfing (Marshall & Corruccini, 1978). The range in rates of change in adult body mass overlaps with those seen in dwarfed lineages of fossil horses (MacFadden, 1986), which evolved over similar time periods (branch lengths 2–7 million years). Notably, the evolution of the pygmy marmoset (C. pygmaea) was clearly a major deviation from typical modes of body-mass evolution in primates. The rate of body-mass reduction during its descent far exceeds that in any other lineage in callitrichids and across primates in general (see also Eastman et al., 2011; Venditti et al., 2011). Whereas dwarfism in other callitrichids is primarily explained by reductions in prenatal growth rate, the evolution of the extremely small body mass of C. pygmaea likely involved major decreases in both pre- and post-natal growth rates. This suggests that the hypothesis that progenesis played a major role in the evolution of the pygmy marmoset (Groves, 1989; Marroig & Cheverud, 2010) is likely to be true. The extreme dwarfism in this species in particular may therefore serve as a useful model for exploring the selection pressures favouring dwarfism and its developmental basis, as well as the allometry of changes in other morphological and behavioural traits during dwarfism.

Our results may also have some bearing on the interpretation of the controversial hominin remains ascribed to a new species Homo floresiensis (Brown et al., 2004). It is notable that the proportional change in body mass during the dwarfing of C. pygmaea is greater than the predicted change for H. floresiensis under a number of evolutionary scenarios (Montgomery et al., 2010). H. floresiensis has been interpreted as an insular dwarf descended from an early hominin (Niven, 2007; van Heteren, 2008; Aiello, 2010). H. floresiensis shows several features consistent with insular dwarfism by paedomorphism (van Heteren, 2008), but interestingly, like callitrichids, lacks clear evidence of megadonty (Brown & Maeda, 2009; Aiello, 2010). Megadonty is thought to be a common feature associated with dwarfism caused by a reduction in post-natal body growth and negative allometry between body size and tooth size (Gould, 1975; Martin, 1992; Ravosa, 1992; Plavcan & Gomez, 1993). The apparent lack of megadonty in callitrichids and H. floresiensis could indicate some similarities in the developmental basis of any body-size reduction that occurred in these species (Gould, 1975; Martin, 1992; Brown & Maeda, 2009).

A significant amount of the debate surrounding H. floresiensis has focused on the scaling relationship between brain and body size during dwarfism (see Niven, 2007 for review). Based on several mammalian brain:body scaling relationships, including examples of mammalian insular dwarfism, it has been argued that the H. floresiensis brain is too small for its body mass, potentially indicating a pathological condition (Martin et al., 2006). However, if selection acted to bring about a reduction in body mass by targeting early periods of body growth when brain mass grows rapidly, this could bring about a greater reduction in brain size than would be the case if selection targeted post-natal body growth after the cessation of brain-size development (Weston & Lister, 2009). Hence, if the shared lack of megadonty is an indicator of a common developmental basis of dwarfism, callitrichid primates may be a useful model for contextualizing the debates surrounding H. floresiensis. An intergeneric analysis has previously shown that under certain phylogenetic scenarios, the pattern of brain-size reduction observed during the evolution of H. floresiensis is in line with that observed in callitrichids (Montgomery et al., 2010) and a more detailed species-level analysis also strongly supports this conclusion (Montgomery, 2011). The brain of LB1, the only individual of H. floresiensis with an intact cranium from which a virtual endocast could be created, displays a range of features not found in other hominins (Falk et al., 2005), which have been interpreted as derived traits indicative of a reorganization of the H. floresiensis brain potentially to permit maintained behavioural complexity while brain size decreased (Falk et al., 2005, 2009). A comparative study of these traits in other dwarfed primates may identify structural reorganizations generally associated with dwarfism and provide a useful framework for considering the brain of H. floresiensis.

Of course, a question of major interest is what selection pressures favour the evolution of dwarfism in these separate clades? Ford (1980) discussed at length the potential selection pressures favouring dwarfism in callitrichids. These remain largely untested, but it seems possible that resource limitation, perhaps driven by climatic fluctuations in the mid-late Miocene, drove the initial reduction in body size, facilitating an ecological radiation within a new small-bodied niche (Ford, 1980). Energetic constraints and resource limitations are often invoked to explain dwarfism (e.g. Niven, 2007; Weston & Lister, 2009) and have been linked to dwarfism in the raccoon and bush dogs (Wayne, 1986a) and H. floresiensis (Brown et al., 2004; Niven, 2007). However, why prenatal growth rates are targeted in some taxa undergoing dwarfism but not others is unclear. It has been suggested that dwarfism in callitrichids was a consequence of the evolution of twinning (Leutenegger, 1973), and Martin (1992) suggested the lag phase of slow growth at the beginning of in utero development is extended in Callithrix to delay the costs of pregnancy until weaning of the previous litter is completed. However, it is clear that major episodes of body-size reduction have occurred in callitrichids without changes in litter size. Further clarification of the selection pressures favouring dwarfism by alternative developmental mechanisms would clearly aid our understanding of life history and body-size evolution, including controversial examples such as H. floresiensis.

In conclusion, we have demonstrated that parallel episodes of phyletic dwarfism characterize the evolution of the smallest anthropoids, the callitrichids and the smallest strepsirrhines, the cheirogaleids. We demonstrate that dwarfism in these two clades is likely to have been produced by selection acting on different stages of development, with dwarfism in callitrichids being correlated with a slowdown in prenatal growth rate. This rare developmental change may explain why these species lack a common feature of dwarfs, megadonty, and suggests potential parallels with the evolution of H. floresiensis. If this is the case, the apparently peculiar nature of brain:body scaling in this species may be a product of selection on early stages of development (Weston & Lister, 2009). We suggest that comparisons of the selection pressures favouring dwarfism in callitrichids, and the morphological and neurological consequences of these selection pressures, offer a potential framework for interpreting the controversial remains of H. floresiensis.

Acknowledgments

We are grateful to Leslie Knapp, Chris Ponting and two anonymous reviewers for comments on aspects of this work. We thank the BBSRC, the Leverhulme Trust and Murray Edwards College for financial support.

Ancillary