Hierarchic and non-hierarchic grouping methods are compared and the reallocation property of the latter is identified as an important advantage as it overcomes the problem of chaining, which is inherent in hierarchic methods. The two methods are used to produce regionalizations of crop composition data for the counties of Ohio in 1940. The grouping strategies are applied to the orthogonal crop dimensions from a Principal Components Analysis. Various data treatments are explored and their crucial methodological role is noted in determining the group structure of the final classifications. Different regionalizations were produced by each method and each data treatment. The non-hierarchic method gave simpler classifications. Stepwise discriminant analysis was used as a test of the efficiency of each method. Non-hierarchic classifications were more similar to the ‘optimal’ classifications produced by the discriminant analysis than those of the hierarchic method. The effect of subjective decisions in grouping strategies are often underestimated and the role of alternative methodologies in classification problems needs more careful evaluation.