Evaluating the use of pairwise dissimilarity metrics in paleoanthropology

Performance of pairwise shape dissimilarity morphometrics on nonmammalian taxa (Insecta: Neuroptera: Mantispidae)

Morphometric dissimilarity metrics aim to quantify the variation between compared specimens such that inferences about their relatedness and alpha taxonomy can be made. Recently, the technique has developed metrics that purport to quantify shape dissimilarity between specimens-employing the use of least squares regression analysis. These metrics have been well applied by studies in the hominin fossil record with an arguably unsubstantiated backing for the technique. Originally postulated was the log10 sem metric which subsequently led to the standard error test of the null hypothesis metric. Following this, the standard deviation of logged ratios (SLR ) metric arose as a pairwise dissimilarity metric that constrains the regression to a zero-intercept, that is, a significant development in the robustness of the technique. This metric was tested on extant primates in order to evaluate its effectiveness alongside the two other metrics. It was shown to be the most reliable for comparisons between specimens of primates, but was unable to discriminate between heterospecific and conspecific comparisons. Arguably, an alternative model organism with which to compare the technique is lacking. This study considers shape dissimilarity metrics with respect to a group of nonmammalian organisms (mantidflies) and tests the metrics against three lines of evidence (morphology, CO1-DNA, and geographic distribution) that can delimit the species-level taxonomy for the group. It is shown that the metrics are unable to discriminate between pairwise comparisons of closely related species, resulting in biologically erroneous groupings, and contradicting the groupings derived from morphological, CO1-DNA, and distributional comparisons. It is thus asserted that the technique is unsuitable for use in alpha taxonomy as an additional line of evidence in mantidflies. It is further supposed that morphometrics in general should be employed with caution in studies of evolutionary history as phylogeny is not the only information contained within morphometric data.

Morphometric analyses of hominoid crania, probabilities of conspecificity and an approximation of a biological species constant

Thackeray has previously explored the possibility of using a morphometric approach to quantify the "amount" of variation within species and to assess probabilities of conspecificity when two fossil specimens are compared, instead of "pigeon-holing" them into discrete species. In an attempt to obtain a statistical (probabilistic) definition of a species, Thackeray has recognized an approximation of a biological species constant (T=-1.61) based on the log-transformed standard error of the coefficient m (log sem) in regression analysis of cranial and other data from pairs of specimens of conspecific extant species, associated with regression equations of the form y=mx+c where m is the slope and c is the intercept, using measurements of any specimen A (x axis), and any specimen B of the same species (y axis). The log-transformed standard error of the co-efficient m (log sem) is a measure of the degree of similarity between pairs of specimens, and in this study shows central tendency around a mean value of -1.61 and standard deviation 0.10 for modern conspecific specimens. In this paper we focus attention on the need to take into account the range of difference in log sem values (Δlog sem or "delta log sem") obtained from comparisons when specimen A (x axis) is compared to B (y axis), and secondly when specimen A (y axis) is compared to B (x axis). Thackeray's approach can be refined to focus on high probabilities of conspecificity for pairs of specimens for which log sem is less than -1.61 and for which Δlog sem is less than 0.03. We appeal for the adoption of a concept here called "sigma taxonomy" (as opposed to "alpha taxonomy"), recognizing that boundaries between species are not always well defined.