CHARACTER SELECTION AND THE QUANTIFICATION OF MORPHOLOGY

Morphologic patterns and trends in disparity are strongly linked to a priori choices in the methodology used to quantify organismal form. Variation in the perceived patterns can be caused by differences between methodologies (e.g. landmark, outline, or discrete characters) as well as the choices made within a method. These choices are often justified based on the particular question being asked, but the role of potential biases needs to be explored. This is particularly true when using discrete characters to quantify morphology because the choice of characters, the amount of character dependence, the resolution of character states, and the number of characters can all potentially influence the results.

The effect of varying the number of characters used in the quantification of morphology was examined with a morphologic dataset of early Paleozoic crinoids. A set of characters was recently compiled as part of the Assembling the Echinoderm Tree of Life project and includes 178 multistate and binary characters that encompass the entire skeletal morphology of the organisms (holdfast, stem, calyx, tegmen, and arms). Two hundred early Paleozoic crinoids were coded and the resulting matrix was analyzed to produce a morphospace using principal coordinate analysis. A rarefaction analysis was conducted on the dataset in which a subset of characters was randomly chosen, analyzed, and the structure of the resulting morphospace was then compared to the morphospace built using the original character matrix in terms of the position of taxa along the primary axis, the amount of variance, and the relative distance between taxa. The results indicate that with this character suite, similar results can be obtained using only 10-15% of the character matrix, i.e. 20-30 characters out of 178. Therefore, large character sets may not be required to detect the major morphologic patterns that are of interest in paleontological studies.