The authors investigated quarterly trends in survival after the diagnosis of Pneumocystis carinii pneumonia for 19,607 patients in California in the decade from January 1, 1983, through December 31, 1992. Subjects included all cases for whom P. carinii pneumonia was the initial (and only) acquired immunodeficiency syndrome (AIDS)-defining diagnosis as reported to the California human immunodeficiency virus/AIDS surveillance registry. There was a period of rapidly improving survival from approximately June 1986 until April 1988, coincident with the widespread Introduction of antiretroviral therapy (zidovudine) and the Institution of P. carinii pneumonia prophylaxis (with cotrimoxazole and pentamidine). There was no evidence, however, of meaningful improvements in survival for these patients after that period. The association of several covariates (risk transmission group, gender, race/ethnicity, certainty of P. carinii pneumonia diagnosis, age, region of residence, availability of CD4 count, and level of CD4 count) were also studied both by proportional hazards regression and by recursive partitioning (i.e., tree-based) survival analysis. The availability of a CD4 count (regardless of its level) was the single factor most strongly associated with survival (median survival 36 months among those with and 14 months among those without reported CD4 counts, p <0.05). Data from this large, population-based surveillance registry of AIDS in California suggest that, despite earlier improvements in survival after the diagnosis of P. carinii pneumonia, the long-term survival of these patients remains poor (39% alive 2 years after diagnosis) and that no improvement in survival has occurred since 1988. Am J Epidemiol 1997;146:115–27.