A collection of pottery sherds is used to show the use of log linear and logit analyses as appropriate statistical techniques to model categorical data (i.e., nominal and binary data). Contingency table analysis, logit modeling, and logit discriminant analysis are treated. Indications of the types of hypotheses that can be examined are given, and approaches to some of the difficulties that may be encountered are described. Finally, a small simulation is done to show the effect on the models of possible selection bias.