National Assessment of Educational Progress (NAEP) Mathematics Assessment (1982) data were analyzed using latent class models to determine patterns of distinct skills required by different exercises and to estimate the pattern distributions. The populations were 9-, 13-, and 17-year-old examinees. Skills were assumed to be intermediate between objectives and report topics, such as "solving quadratic equations," and were treated as dichotomous - an examinee either did or did not possess the skill. At age 9 and 13, one assessment booklet was selected and nine clusters of three exercise parts were chosen. Twenty pairings of clusters yielded a 6-item set for analysis. At age 17, six exercise parts of apparently common skills were drawn from each of six booklets. Item clusters which could be collapsed and organized hierarchically were indicated by the latent classes. For each cluster, all analyses including it were examined together, yielding separate estimates of the proportion of examinees able to solve items in that cluster. The distributions of these estimates were an indication of the cluster's conformity to the assumption of skill dichotomies. Student mathematics skills at each age level are reported. Results of the use of NAEP data tapes are reported and improvements in the methodology are suggested. Primary type of information provided by the report: Results (Secondary Analysis). (CM)