The interest in developmental sequences and learning hierarchies is growing. One approach to the study of such sequences is to gather data on several variables, each of which corresponds to a stage, step, or phase in the sequence and to examine the associations between the variables as displayed in a contingency table. If the variables are associated in ways predicted by the hypothesized sequence, then the data lend support to the sequence. Goodman's loglinear model for developmental or learning sequences is presented and illustrated on number concept data gathered by Brainerd and Fraser. Where its strong assumptions are satisfied, the model provides a probabilistic framework within which to: (1) test the plausibility of an hypothesized developmental sequence or learning hierarchy; (2) compare several hypothesized sequences on the same data; (3) estimate the proportion of subjects who do not conform to the sequence; and (4) estimate the proportion of subjects at each step in the sequence. (Author/RL)