THE EFFECT OF WEAR ON THE PLACEMENT OF LANDMARKS ON MAMMALIAN DENTITIONS USING TWO- AND THREE-DIMENSIONAL GEOMETRIC MORPHOMETRICS

Geometric morphometrics is an important tool in paleontological studies of shape variation between and within species, and has been used extensively for identifying morphospace and ecospace in mammalian communities. Mammals are often identified based upon tooth shape differences, due to the complex nature and high interspecific and low intraspecific variation in tooth morphology, resulting in teeth as the primary element used in landmark-based studies. An assumption when using landmark-based analyses is that in order to collect a landmark, the morphology must be present and visible, meaning that when mammalian dentitions are measured, only unworn or little worn teeth are used (as wear erases tooth morphology). This assumption limits samples, especially for species and time bins that are poorly known, which may exclude them from any geometric morphometric analysis, thereby missing potentially important information when assessing a community. We set out to test the assumption that differing amounts of wear affect the placement of landmarks on mammalian dentitions in an attempt to determine whether or not a worn tooth can be used in an analysis. If landmarks can be accurately placed on worn teeth, then these analyses are likely to include all of the important morphological disparity in a community. To test the effect of wear, we made polyurethane casts of upper and lower molars of dentally conservative mammals with no wear. Five copies of each cast were made and five discrete wear stages were simulated on each copy, from no wear to extreme wear, following observed natural wear patterns. For each wear stage, two observers collected landmarks three times each to evaluate for intra- and inter-observer consistency. Results suggest that even on heavily worn specimens, landmark data can be sufficiently recognized as belonging to the correct genus. However, wear stages with increased wear exhibit higher variance within each generic group, compared to those with less wear. It is interesting to see, but not altogether surprising, that two-dimensional analyses appear to be less susceptible to the influence of wear than three-dimensional studies, with the x- and y-coordinates (corresponding to the occlusal position of the landmark) easier to accurately choose than the z-coordinate (corresponding to the height of the landmark).