Recidivism is defined as reincarceration and excludes convictions involving other penal sanctions. Prisoners serving sentences in police lockups, remanded in custody, or sentenced to imprisonment prior to July 1975 were also excluded. Estimates of the ultimate probability of recidivism for this population and various subgroups describe recidivism for race, sex, age, offense, and other variables, confirming a classic failure profile with strong distinctions by these factors, particularly race. Overall, the probability of recidivism for aboriginal male prisoners was 80 percent (SE 2 percent) and for nonaboriginal males was 48 percent (SE 2 percent) and the median times to fail for these groups were 11 months and 18 months, respectively. Female aboriginal recidivism was 75 percent (SE 6 percent) and female nonaboriginal recidivism was 29 percent (SE 7 percent) and the median times to fail were 16 months and 19 months, respectively. Higher recidivism was observed for male, and young prisoners. Other variables considered include major offense (at first receival), actual time served, sentence type, prison (at exit), and marital, educational, and employment status. Lower recidivism was observed for nonaboriginal prisoners incarcerated longer and for more serious offenses. Participation in work release and financial assistance on release was also associated with lower recidivism. The time to fail was consistently much shorter for aboriginals across all factors than for nonaboriginal prisoners. A general downward trend in recidivism was found between the years 1975 through 1976 and 1978 through 1979 although it was less marked for aboriginal prisoners and did not continue beyond the period 1980-1981. The contribution of aboriginal recidivism to high rates of imprisonment in Western Australia is discussed in relation to the utility of imprisonment. The limitations and merits of failure rate analysis in recidivism studies are considered. (Author abstract)