The stability of child conduct and oppositional defiant behaviors during the period from 7 to 15 years was studied in a birth cohort of New Zealand children. These data were analyzed using two methods. In the first method the observed state to state changes in childhood behavioral tendencies were analyzed using empirical transition matrices. These results suggested that children classified as cases showed high rates of symptom remission, with approximately 50% of cases being classified as noncases 2 years later. In the second approach the data were analyzed using a latent Markov model which took account of errors of measurement in the classification of children. This analysis suggested the presence of strong continuities in childhood problem behaviors, with only 14% of children showing remission of behavioral problems within a 2-year period. The differences in the estimates yielded by the empirical transition matrices and the latent analyses were explained by the fact that there were relatively high probabilities that children who were cases were misclassified as a result of measurement errors.