Results are reported from estimating the parameters of a model in which offenders have a probability of desisting from further participation in crime following a conviction and, if they persist, a rate of crime commission. The probability of desisting and the rate of commission are seen as varying with offenders' personal characteristics and criminal records. The difficulty of estimating models is discussed in which failure to commit a new crime might be attributable either to termination of the criminal career or to a censored followup period. Both successul and unsuccessful estimation attempts are reported, and complications when distinguishing empirically between duration and frequency are discussed. 1 figure, 8 tables, 26 references. (Author abstract modified)