This study used accumulated data from the 5-year University of California, Los Angeles Study of Adolescent Growth to test models for the development of traditional law abidance or, conversely, rebelliousness. Longitudinal data were analyzed with latent variable causal modeling techniques to develop and test models that explain patterns of general rebelliousness and their psychosocial causes and consequences. The most important models tested are those that simultaneously examine the influences of many factors on the changes in levels of rebelliousness. The models demonstrated the consistent effects of the personality tendency of law abidance of the social factors both within and across time. There were strong tendencies for frequent law violators to be heavily involved with their peers and to be friendly with peers and adults who use and support the use of a variety of drugs. The testing failed to find that the social factors had causal influences on a person's level of law abidance. Presumably, this means that by the time the study first measured the trait in early adolescence, the levels had become crystallized and not generally susceptible to change by social agents. Beyond specific findings, this study illustrates ways that latent-variable causal models can be used to test hypotheses in personality research that had previously been considered virtually impossible outside the laboratory. Tabular data and 28 references are provided. (Author summary modified)