2003 Seattle Annual Meeting (November 2–5, 2003)

Paper No. 13
Presentation Time: 8:00 AM-12:00 PM


DUNLAVEY, Tammy L., Dept of Geology, SUNY at Buffalo, 876 Natural Science Complex, Buffalo, NY 14260, MITCHELL, Charles, Dept. of Geology, SUNY at Buffalo, 876 Natural Sciences Complex, Buffalo, NY 14260 and SHEETS, David, Dept of Physics, Canisius College, Buffalo, NY 14208, dunlavey@buffalo.edu

For decades camera lucida and photographic techniques have proven useful for data acquisition. Paleontologists faithfully utilize these techniques under the assumption that the resulting images reliably represent the intended sample. However, the reliability of such images in the context of landmark based geometric morphometric methods has not be rigorously tested. The goal of this presentation is to statistically evaluate the relative reliability of camera lucida and photographic image data acquisition techniques. Image interpretations employed on three specimens of the graptolite Isograptus lunatas, from Early Ordovician rocks of Newfoundland, Canada. Each of the three fossil samples are drawn ten times by three different artists using a camera lucida. These drawings were prepared one or more days apart, yielding a total of thirty representative images per artist. The three fossil samples additionally were photographed ten times each, resulting in a total of thirty representative images. This compiled database of one hundred twenty images was graphically processed utilizing TPSDig32 landmark digitizing computer software. Comparisons and contrasts of colony shape within and between acquired camera Lucida drawings and photographic images were investigated by analysis of geometric morphometrics (Procreates alignment) and multivariate statistical analysis using principle components analysis (PCA). Statistically, we test whether the comparisons and contrasts of colony shape within and between acquired camera lucida drawings and photographic images exhibit operator or technique specific bias. The study of the data acquisition techniques, along with the subsequent role of operator error, yields indexes of similarity and these were used to determine the statistical degree of confidence of the resulting images.