Size Variation in Small-Bodied Humans from Palau, Micronesia

Background Recent discoveries on Palau are claimed to represent the remains of small-bodied humans that may display evidence insular size reduction. This claim has yet to be statistically validated Methodology/Principal Findings Published postcranial specimens (n = 16) from Palau were assessed relative to recent small-bodied comparative samples. Resampling statistical approaches were employed to test specific hypotheses relating to body size in the Palau sample. Results confirm that the Palau postcranial sample is indisputably small-bodied. Conclusions/Significance A single, homogenous body size morph is represented in early prehistoric postcrania from Palau. Small body size in early Palauans is an ancestral characteristic and was likely not a consequence of in-situ size reduction. Specimens from Palau have little bearing upon hypothesised insular size reduction in the ancestral lineage of Homo floresiensis.


Introduction
Body size variation in Southeast Asian terminal Pleistocene and Holocene humans is poorly understood. Present inhabitants of the SE Asian tropics display considerable body size variability [1][2][3]. 'Negritos' of the Philippines and surrounding Islands are distantly related to Polynesian peoples and approach African Pygmies in their diminutive size [1,4]. Archaeological material from Palau is pertinent to establishing a timeframe for the dispersal of smallbodied humans in Southeast Asia [5][6][7]. In a recent contribution, Berger et al. [8] concluded that specimens from archaeological sites in the Rock Islands display evidence of chronological size reduction. This was argued to be a function of significant ecological selection in a reproductively-isolated population [8].
Berger et al. [8] propose that postcranial elements recovered from Ucheliungs and Omodokel caves (n = 61) sample the lowest extremes of extant size variation in Homo sapiens (Table 1). This proposal has been the focus of contention within scientific circles [9,10]. Statistical appraisal of size variability within the skeletal sample from Palau [7,8] is imperative to understanding patterns and homogeneity of body size distribution in early prehistoric inhabitants of Micronesia and in validating any hypothesis of insular dwarfism [8].
This correspondence focuses upon two explicit hypotheses concerning body size in prehistoric humans from Palau. Does the skeletal sample from Palau [7,8] fall within observed morphological size ranges of extant small-bodied humans and is the skeletal sample homogenous or heterogenous? An explicit assessment of an associated femur and tibia from Chelechol ra Orrak [7] is pertinent. Simply stated, is there overwhelming evidence for a single, smallbodied morphotype among the earliest inhabitants of Palau?

Results
Results are consistent irrespective of whether bootstrapping or randomization is preferred [Supplementary Data 1]. African Pygmies and Southeast Asian Negritos are remarkably similar in their distal humerii but differ significantly in lower limb size profiles (Figures 1 and 2 [Tables S1 and S2]). The African Pygmy and Southeast Asian Negrito samples are significantly smaller than African and European samples [Tables S3 and S4]. Direct comparisons confirm that a majority of the Palau postcrania derive from individuals whom can be accommodated within the observed size ranges of small-bodied humans (Tables 2 and 3 [11] there is little support for the hypothesis that these specimens exceed the lower size range in recent humans. In contrast, postcranial dimensions of the LB1 hominin cannot be sampled from recent small-bodied humans and are truly diminutive (Tables 2 and 3). Results confirm previous conclusions that the Rock Island specimens are those of small-bodied humans [8].
The CV* for the distal humerii approaches the significance criterion of African Pygmy and SE Asian Negrito CV*s but remains insignificant (Table 4 [ Figures S1 & S2]). The Grand Mean of the Palau series (n = 16) is significantly greater than approximates derived from African Pygmies but not Southeast Asian Negritos (Table 4; Figures 3 and S3). Bootstrapped standard deviations (SD's) confirm that the variation within the Palau postcranial series is not significant and is generally consistent with that observed in small-bodied human references (Table 4). Statistical comparisons of the associated femur and tibia from Chelechol ra Orrak [7] confirms that these derive from a smallbodied individual [ Figures S4 and S5]. Pairwise randomization and bootstrapped t-tests reject the hypothesis that the Chelechol ra Orrak specimens exceed the size range of small-bodied humans. The observed homogeneity of prehistoric humans from Palau confirms the hypothesis that a single size morph is represented at c3000 BP. Early prehistoric Palauans were indisputably 'smallbodied'.

Discussion
Ongoing investigations confirm an initial human presence on the southern Islands of the Palau archipelago by c3000 BP [5][6][7][8]. Berger et al. [8] proposed that the earliest inhabitants of the Rock Islands were small-bodied and this is confirmed by statistical analyses of available postcranial specimens. Two lower limb elements are unarguably 'very small' but even these small individuals do not approximate the diminutive size of the LB1 H. floresiensis holotype [11,12].
Analyses of within-group variability confirm that Palauan postcrania are neither excessively variable nor excessively small. All specimens included in this study approximate the size ranges of African Pygmies and Southeast Asian Negritos [13]. Statistical tests further confirm the remarkable homogeneity of size in prehistoric Palauans. Upper and lower limb elements from Chelechol ra Orrak are not consistently larger than corresponding specimens from the Rock Island sites [7,8]. An associated femur and tibia from Chelechol ra Orrak are within the size range of small-bodied humans. Results of this study provide no support for the hypothesis that two distinct size 'morphs' are represented at c3000 BP. Evidence overwhelmingly supports the conclusion that the pioneer colonists of Palau were small-bodied.
Berger et al. [8] hypothesized body size reduction among the Rock Island samples. Current contextual and chronological ambiguity at Omedokel cave compromises a hypothesis of insular dwarfism [8]. Published AMS dates are wildly disparate and contrast with the chronological controls evident at Ucheliungs [8]. Results overwhelmingly support the proposal that the earliest colonists of Palau were small-bodied and that within-situ size reduction cannot be substantiated by current evidence. Body size reduction in the ancestral population of earliest colonists of Palau is likely unconnected to the prolonged genetic isolation envisaged for the Middle Pleistocene of Flores [11,12,[14][15][16].  Table 1 for inventory) are from Nelson

Materials and Methods
Distal humeral, innominate, proximal femoral and proximal tibial specimens from archaeological sites dating to c3000 BP and younger (Table 1; 7, 8) and corresponding dimensions from Liang Bua 1 [11] were compared with two small-bodied reference samples. Resampling approaches were utilised [17,18]. Articular dimensions were the focus of this analysis [19][20][21][22]. African Pygmy (n = 34) and SE Asian Negrito (n = 44) specimens are derived from several Institutions [Supplementary Data 1]. Resampling comparisons were performed with and without replacement [17,18]. Initially, absolute deviations of individual specimens (table 1) from the arithmetic mean of the reference samples were randomization over n = 5000 iterations. These comparisons were performed using Rundom Projects 2 [23]; http://pjadw.tripod.com]. Computational analyses of bootstrapped Student's t' in the case of a single  Table 1 [25]. The CV*, modified for small sample sizes [24], for available distal humerii (n = 6) can facilitate direct assessment within a single dimension. A bootstrapping approach repeatedly selected six specimens from the comparative series to calculate the CV* over 5000 iterations [26,27]. The test statistic is a ratio specifying the position of the observed CV* within a generated distribution of CV*s of the comparative distributions (n = 5000). A modified version of the approach outlined in Green et al. [28] was used to assess variation within the postcranial series (n = 16). Geometric Means were calculated [24] for and the Grand Mean and standard deviation (SD) were used as proxies of within-sample variance. Grand Means and SD's were bootstrapped 5000 times. Femoral and tibial lengths of the associated skeleton from Chelechol ra Orrak [7] were subjected to the test procedures as outlined previously.