The hierarchical clustering technique was used to differentiate college dropouts from persisters, and to determine how student clusters differ from one another on relevant outcome variables. Subjects were 618 freshmen who entered a community college with the intention of completing a two-year associate degree. There were 432 persisters who returned for the last three semesters, and 186 non-persisters who did not. Four variables which were found to differentiate persisters and non-persisters were entered into a hierarchical clustering program (PROC CLUSTER in SAS): (1) age; (2) degree objective; (3) scores on writing placement tests; and (4) scores on mathematics placement tests. Five clusters were developed, according to academic readiness, performance, and persistence. Analysis of variance and chi square tests were performed. Dependent variables included age; sex; ethnic status; grade point average; placement test scores in reading, mathematics, and writing; and percentage of first semester courses completed. A comparison of the predictive value of the clustering technique and of placement tests indicated that clustering was not superior in diagnostic sensitivity or specificity. Although the distribution of persisters and non-persisters was significantly different across clusters, the hierarchical clustering patterns did not appear to have sufficient predictive validity as a screening method. (GDC)