Drug offenders were selected for the study because drug offenses have been those on which the deterrence doctrine seems to have had the most significant legislative impact during the last 30 years. Using various statistical approaches, the researchers examined whether empirical tests of special deterrence based on criminal records are consistent. The study also examined the importance of varying ways of operationalizing recidivism for this issue. The Norwegian Ministry of Justice granted researchers access to data in the National Register of Convictions relating to persons convicted under the Medicines Act and/or Section 162 of the Criminal Code and sentenced to an unconditional term of imprisonment during the period 1980 to 1986. The study population was limited to persons of Norwegian nationality who had completed their index sentence by the end of 1986. The findings show that the probability of recidivism declined with offenders' increasing age, which conforms to the findings of other similar studies. Also, the younger the age at first offense, the greater the probability of recidivism. The various analytical approaches and operational concepts of recidivism produced conflicting results. This lack of consistency may point to a general problem for this type of study. The researchers note the possibility that the statistical association found does not refer to a causal link between variables but to differences between sub-populations in the way their members act and react. If a statistical association between severity of punishment and recidivism possibly refers as much to differences between sub-populations as to a causal link between the variables, it would be important to make use of some type of measure of recidivism that can capture the degree of seriousness of the criminal act. 7 tables and 65 references