Abstract
Read's (1974) discussion of contingency table analysis is amplified. Procedures for model formulation and testing advocated by Fienberg (1970) and Goodman (1968, 1969, 1970) are outlined, contrasted, and used to analyze a corpus of archaeological data. It is argued that Fienberg's approach is better suited to situations in which the behavior of the data can be accurately predicted within fairly narrow limits. The Goodman approach is more appropriate to situations in which the behavior of the data is completely unknown.