Four problems characterize much of the past research on the relation of social factors to juvenile delinquency: measurement of social background, one-dimensional indexes of delinquency, bias in measurement of delinquency, and nonparametric statis tical tests. These problems may be overcome in part by using specific, proximal measures of background; multiple indexes of delinquency; anonymous self-reports of delinquency incidence; and parametric statistics such as components analysis and regres sion or analysis of variance. The present paper demonstrates the use of these methods in a pilot study of about 400 public high- school students in Chicago. The components analysis shows the factor pattern of delinquency varies in samples of boys and girls; hence multiple rather than single indicators must be used. The regression analysis shows that different types of delinquency are functions of different combinations of family background, ethni city, and sex; for example, boys, especially white boys, in the sample more often reported car thefts; and Negro girls reported more frequently than white girls of having engaged in gang fights. Several methodological implications are drawn from the findings for future research on delinquency.