HOW IMPORTANT IS IMAGE INTERPRETATION TO PALEONTOLOGY? CASE STUDIES IN DATA AQUISITION FIDELITY AND RETRODEFORMATION USING BILATERALLY SYMMETRIC GRAPTOLITES (ISOGRAPTUS AND PSEUDISOGRAPTUS) AND TRILOBITES (TRIARTHRUS BECKI AND T. EATONI)

Whether too big, too small, or too numerous, many fossil studies rely on representational images rather than the original fossils themselves for quantitative data collection. The fact that human error is present in the procedures employed to form and interpret images is little studied. We examine error due to three main categories. 1) Drawing fidelity: errors that occur during the procedures used to draw or reproduce the shape or outline of the fossils. 2) Landmark fidelity: error produced by landmark-based geometric morphometric data acquisition techniques. 3) Deformation: distortion of form by taphonomic and tectonic process. In each case, error may result in a loss of biological information and consequent reduction in accuracy of interpretation.

This study examines qualitatively the consequences of operator or technique specific data acquisition bias within a paleontological study based on taxonomy. Variance comparisons and contrasts within and between 10 sets of repeated image representations of Isograptus colony shape (drawing fidelity) and 5 repeated landmark sets of 13 rock slabs consisting of 54 individual Isograptus and Pseudisograptus colonies (landmark fidelity) were statistically scrutinized in order to assess magnitude and effect of the operator or technique specific data acquisition bias. The deformation error analysis used the same landmark fidelity data sets and 3 data sets of Triarthrus becki and 1 set of T. eatoni to asses the Sheets/ Dunlavey retrodeformation method (introduced at GSA 2004) when the effects of landmark error are known.

Preliminary conclusions suggest operator and technique specific bias does impact the reliability of a study. The choice of image acquisition techniques (drawing fidelity) does matter. The process of landmark acquisition (landmark fidelity) and subsequent conversion of landmark data to quantitative data demonstrated various amounts of error. The deformation study demonstrated if the amount of possible error associated with a taxonomic study is compounded and complicated, when all three sources of error are used in conjunction with one another.