An examination of existing studies indicates that reports of generalized delinquency often coincide with the use of self-report measures containing fewer answer categories, while specialized delinquency is more apparent when more answer categories are used. The present study systematically manipulated the answer categories of a 41item self-report measure of delinquency. Subjects were 205 Oregon boys in grades 4, 7, and 10. Data were subject to various factor analytic techniques comparing three-answer and continuous-answer categories. When the number of factor loadings is the criterion, a three-answer version produced more evidence of generalized delinquency than the continuous-answer version, particularly in the case of principal component analysis or varimax rotation. When factor labels served as the criterion, the opposite was found. The content of the factors was very much a function of the truncation of the answer categories. Thus, manipulation of answer categories produces considerably divergent results, which easily could lead to substantially different conclusions regarding the presence or absence of delinquency specialization or generalization. 6 tables and 39 references. (Author abstract modified)