Left, right or both? Estimating and improving accuracy of one-side-only geometric morphometric analyses of cranial variation

Abstract

Procrustes-based geometric morphometric analyses of bilaterally symmetric structures are often performed using only one side. This is particularly common in studies of cranial variation in mammals and other vertebrates. When one is not interested in quantifying asymmetry, landmarking one side, instead of both, reduces the number of variables as well as the time and costs of data collection. It is assumed that the loss of information in the other half, on which landmarks are not digitized, is negligible, but this has seldom been tested. Using 10 samples of mammalian crania and a total of more than 500 specimens, and five different landmark configurations, I demonstrate that this assumption is indeed easily met for size. For shape, in contrast, one-side landmarking has potentially more severe consequences on the estimates of similarity relationships in a sample. In this respect, microevolutionary analyses of small differences are particularly affected, whereas macroevolutionary studies are fairly robust. In almost all instances, however, a simple preliminary operation improves accuracy by making one-side-only shape data more similar to those obtained by landmarking both sides. The same operation also makes estimates of allometry more accurate and improves the visualization. This operation consists in estimating the missing side by a mirror reflection of bilateral landmarks. In the Supporting Information, I exemplify how this can be easily done using free user-friendly software. I also provide an example data set for readers to repeat and learn the steps of this simple procedure.