Magnification and resolution in dental microwear analysis using light microscopy

Abstract

One potentially problematic aspect of dental microwear analysis is sensitivity to the resolution (fineness of detail) at which dental wear surfaces are viewed. Magnification is one of many variables that determine resolution. Microwear studies based on light microscopy generally report magnifications ranging from 30X-100X, although ambiguities in reported methods prohibit duplication of the resolutions of many studies. Moreover, magnification settings have been arbitrary and, thus far, the biasing effects of different resolutions have not been tested. We tested sensitivity of dental microwear analysis to resolution by manipulating pixel density and magnification in photomicrographs taken under a light microscope using the molars of a browser (moose), grazer (zebra), and frugivore/hard object feeder (peccary). Resolution affected the number of observable microwear features and distorted the proportional frequencies of large and small features. Nonetheless, two observers independently found similar differences between the species throughout a range of resolutions. At no resolution did observers recover statistically undifferentiated data, although interobserver correlation was best when resolution was decreased to 20-40% of the initial resolution. Observer correlation for microwear features with a maximal dimension >20 μm was substantially better than for smaller features. We conclude that (1) dimensionless aspects of microwear data (e.g., proportional numbers of scratches between species) are more robust to resolution than raw frequency data, (2) higher resolution does not produce higher quality data, and (3) the optimal resolution may be dependent upon the size of the microwear features. Further testing of the interaction of resolution, microwear feature size, and the observer will increase repeatability of results and lead to more robust paleodietary interpretations.