Morphometric comparison of robust australopithecines One Or twO species ? A mOrphOmetric cOmpArisOn between rObust AustrAlOpithecines frOm KrOmdrAAi And swArtKrAns

The type specimen of Paranthropus robustus (TM 1517, including a partial cranium) was discovered at Kromdraai near Sterkfontein in 1938 and described by Robert Broom as a new species. Subsequently, more robust australopithecines were discovered at the nearby site of Swartkrans. These Swartkrans hominins were described by Broom as Paranthropus crassidens. Many palaeoanthropologists currently regard the robust australopithecines from Kromdraai and Swartkrans as one species, but consensus has not been reached on this issue. A morphometric analysis has been undertaken to assess the probability that specimens attributed to P. crassidens represent the same species as that which is represented by TM 1517, the holotype of P. robustus. Our results failed to reject the null hypothesis that both sites sample the same, single species of robust australopithecine.


INTRODUCTION
The type specimen of Paranthropus robustus (TM 1517, see Figure 1) was discovered at Kromdraai B in the Cradle of Humankind World Heritage Site in South Africa and described by Broom 1 as a new genus and species.Subsequently, additional specimens of 'robust' australopithecines were discovered at Swartkrans, within four kilometers of Kromdraai B, and described as a distinct species, P. crassidens. 2 Many palaeoanthropologists such as Tobias 3 , Brain 4 , Fuller 5 , and Kaszycka 6 , regard the Kromdraai B and Swartkrans 'robust' australopithecine specimens as representing a single species.Others such as Grine 7,8,9,10 and Howell 11 , and more recently, Schwartz and Tattersall 12 , regard the two as distinct at a species level, P. robustus and P. crassidens, respectively.To date, this issue has not been addressed by a morphometric analysis of both cranial and dental material.The objective of this study is to test the hypothesis that the robust australopithecine crania from Swartkrans and Kromdraai are morphometrically similar enough to belong to the same species, P. robustus.

MATERIALS
Craniodental measurements were taken on original specimens: TM 1517a from Kromdraai, and SK 12, SK 46, SK 48, SK 79, SK 83, and SKW 11 from Swartkrans.The set of cranial measurements was adapted from that of Lockwood 13 , and is listed in Table 1.Dental measurements included only the maximum labiolingual/buccolingual breadth, because heavy interproximal wear obscures maximum mesiodistal lengths in most specimens. 5,14Where a trait was present on both the right and left sides, the average of the two values was taken.

METHODS
Thackeray and colleagues 15,16 described a method whereby measurements of pairs of specimens were compared to one another, in order to assess probabilities of conspecificity.Measurements obtained from extant, conspecific male-female pairs were plotted against each other, with the female on the x-axis and male on the y-axis.Morphometrically similar pairs tended to exhibit a relatively low degree of scatter about a least-squares regression line associated with the equation y = mx + c.This degree of scatter, or degree of dissimilarity, was quantified by calculating the standard error of the slope, m (here, "se m ").Using modern conspecifics as a frame of reference, Thackeray and colleagues 15 found that this se m statistic displayed a normal distribution when log-transformed (base 10).Thus, pairs of specimens that were morphometrically very different from one another − specimens of different species − can be expected to have a relatively high degree of scatter about the regression line, and so a relatively high se m .Conversely, pairs of conspecifics tended to have a lower degree of scatter and hence a relatively low se m .Using an extant reference sample of over 1 400 specimens of vertebrates and invertebrates, Thackeray 16 presented a log se m value of -1.61 ± 0.23 as a statistical definition of a species expressed in terms of probabilities.
The log se m method described by Thackeray and colleagues 15 was developed further by Wolpoff and Lee 17,18 , who used a statistic which they referred to as the 'standard error test of the null hypothesis' (STET).Like log se m , STET is based on the standard error of the slope of the least-squares regression line in bispecimen comparisons.However, Wolpoff and Lee 18 argued against the dampening effects of logging the se m statistic.Additionally, they noted that regressing specimen x on specimen y produced different slopes and therefore different standard errors, as compared to regressions of y on x.Thus, STET utilises the standard error of the slopes of both possible pair-wise regressions: STET = [(s.e.mx ) 2 + (s.e.my ) 2 ] 1/2 , where mx and my are the least-squares slopes of each possible regression.In this way, the STET statistic is invariant of how specimens are regressed on one another, and makes no assumptions about sex, which is often difficult to determine in fossil specimens.Low STET values indicate very similar overall shape (low scatter about the regression lines).
For this analysis, we used the STET statistic to assess the likelihood that TM 1517 and the Swartkrans crania represent the same species.If TM 1517 represents a different species from that which is represented by the Swartkrans robust australopithecines, STET values computed for Swartkrans specimens (compared

ABSTRACT
The type specimen of Paranthropus robustus (TM 1517, including a partial cranium) was discovered at Kromdraai near Sterkfontein in 1938 and described by Robert Broom as a new species.Subsequently, more robust australopithecines were discovered at the nearby site of Swartkrans.These Swartkrans hominins were described by Broom as Paranthropus crassidens.Many palaeoanthropologists currently regard the robust australopithecines from Kromdraai and Swartkrans as one species, but consensus has not been reached on this issue.A morphometric analysis has been undertaken to assess the probability that specimens attributed to P. crassidens represent the same species as that which is represented by TM 1517, the holotype of P. robustus.Our results failed to reject the null hypothesis that both sites sample the same, single species of robust australopithecine.against each other) should be significantly lower than the STETs derived from TM 1517-Swartkrans comparisons.A ranked-sums test 19 was used to determine whether the Swartkrans-only STET values were significantly different from the values computed for the TM 1517-Swartkrans comparisons.This non-parametric test was appropriate in this case because the two samples were not independent: The same Swartkrans crania were used to compute STET values for within-Swartkrans and TM 1517-Swartkrans comparisons.
Finally, because the STET statistic is more reliable with increasing numbers of variables, 18 the above procedure was repeated in a secondary analysis, omitting pair-wise comparisons that share less than ten variables, as in the case of the following pairs: TM 1517-SK 12, TM 1517-SK 79, SK 79-SK 12, and SK 12-SK 83.An advantage of the methods developed by Thackeray and colleagues 15,16 and Lee and Wolpoff 17,18 is that they facilitate assessment of fossil assemblages when certain specimens do not preserve all of the traits under study.Additionally, these methods provide a means of testing hypotheses about taxonomy, based on metric similarity.It should be noted that these methods only test null hypotheses − hypotheses of no significant difference.

RESULTS
The mean STET value for Swartkrans-only was 0.156 ± 0.118, and that of TM 1517-Swartkrans was 0.119 ± 0.104 (see Table 2).The Wilcoxon test showed the distributions are not significantly different (p = 0.3) The TM 1517-Swartkrans distribution was within the range of the Swartkrans distribution, with the exception of the TM 1517-SKW 11 comparison, which had a lower STET value than any comparison within Swartkrans alone (see Figure 2).That the TM 1517-Swartkrans distribution had a lower mean STET value than the Swartkrans-only distribution, highlights the morphological similarities of TM 1517 with many specimens from the Swartkrans sample.
Figure 2 shows that within the Swartkrans-only distribution there is a major outlier, the comparison of SK 12-SK79; this is one of the comparisons preserving fewer than 10 variables in common.The secondary analysis, omitting the comparisons with less than 10 variables in common, produced similar results to the whole-sample analysis (see Figure 3).The mean STET for Swartkrans-only was 0.120 ± 0.070, and that of TM 1517-Swartkrans was 0.053 ± 0.012.In the reduced dataset, the TM 1517-Swartkrans STET values span an even smaller part of the lower range of the within-Swartkrans variation, further underscoring the similarity between TM 1517 and the Swartkrans fossils.A Wilcoxon ranked-sums test on this reduced dataset showed the TM 1517-Swartkrans and Swartkrans-only STET distributions to be significantly different (p = 0.02) at the p < 0.05 level.Because the TM 1517-Swartkrans STET values were generally lower than those within Swartkrans, this significant result is surprising, and further emphasises the morphological similarity between the crania from these two sites, as well as the great variation within Swartkrans itself.

DISCUSSION
The results show that one cannot reject the null hypothesis: That TM 1517 from Kromdraai and crania from Swartkrans sample the same species, Paranthropus (or Australopithecus) robustus.The TM 1517-Swartkrans STET values were generally lower than the within-Swartkrans values, indicating that not only is TM 1517 within the Swartkrans-only range of variation, but also that TM 1517 is generally more similar to Swartkrans crania than the Swartkrans crania are to each other.In fact, the lowest STET value from the analysis, indicating the two most similar specimens, is that of TM 1517 from Kromdraai and SKW 11 from Swartkrans Member 1's Hanging Remnant.In light of the fairly large variation displayed by the Swartkrans fossils, our results point to a relatively stable lineage of robust australopithecine in the Sterkfontein valley in the lower Pleistocene.
Our results indicate that the Kromdraai and Swartkrans robust australopithecines represent the same species.This is contrary to findings of Howell 11 , Grine 7,9,10 , and Schwartz and Tattersall 12 , whose conclusions were based largely on differences in dental sizes, proportions, and gross morphology between the two sites.Two issues arose with these authors' conclusions.Firstly, the decisions to make specific distinctions appear to be have been made from gross inspection of each site's summary statistics for given dental traits (e.g.M 3 mesiodistal length), rather than from statistical tests.Along these lines, Fuller 5 used a resampling procedure to compare the pooled Swartkrans-Kromdraai dental coefficients of variation (CV) to the CVs of modern humans, African apes, and fossil hominins.Her results showed that variation within the pooled Swartkrans-Kromdraai sample was statistically no greater than in most of the modern and fossil referents, leading her not to reject the hypothesis of a single species represented at the two sites.

42
Secondly, some Kromdraai specimens contributing to the seemingly specific differences between the two sites may actually relate to the presence of early Homo.For example, Grine 9,10 suggested that the teeth of a sub-adult from Kromdraai, KB 5223, represented a robust australopithecine species different from that which was represented at Swartkrans.However, Braga and Thackeray 20 have shown, using qualitative and quantitative criteria, that KB 5223 is probably attributable to early Homo.Placing all other Kromdraai B dental specimens in the P. robustus sample along with Swartkrans specimens, these authors found it improbable that the KB 5223 postcanine dentition came from the same P. robustus sample, but rather more likely indicated the presence of Homo at Kromdraai.In some non-metric traits, this specimen is similar to the TM 1536 M 1 and dm 1 , also from Kromdraai.However, the authors noted that TM 1536 lacked the key diagnostic features of Homo molars, such as the absence of the hypoconulid on the dm 1 and C6 on the dm 2 , present in KB 5223.Thus, the results of Braga and Thackeray 20 corroborated those of Fuller 5 , indicating a high probability that the 'robust' dental specimens from Kromdraai and Swartkrans, with the exception of the KB 5223 dentition, sample a single species.Our results further demonstrate the craniodental affinities of the Kromdraai and Swartkrans robust australopithecine fossils.

CONCLUSION
From our morphometric analysis, we failed to reject the null hypothesis, therefore there is a very high probability that the type specimen of P. robustus, TM 1517, is the same species of robust australopithecine that is present at Swartkrans.Moreover, TM

FIGURE 1
FIGURE 1Left lateral view of the TM 1517a cranium

FIGURE 2
FIGURE 2 Boxplot of the distributions of STET values Top plot is the set of TM 1517-Swartkrans comparisons, bottom are the Swartkrans-only comparisons.The thick black band in the box is the median.

FIGURE 3
FIGURE 3 Boxplot of the distribution of STET values for the secondary analysis with reduced datasetAs above, the top box in each plot is the set of TM 1517-Swartkrans comparisons, bottom is the Swartkrans-only set.

TABLE 1
List and description of cranial measurements used in this analysis

TABLE 2
STET values for all pair-wise Comparisons seen in Swartkrans, and is indeed morphometrically very similar to many of the Swartkrans fossils.Hence, the Swartkrans 'robust' australopithecines and TM 1517 can be considered conspecific, representing a single species, Paranthropus (Australopithecus) robustus. http://www.sajs.co.za