Data from the National Center for Health Statistics were used to describe the decline in age-adjusted stroke mortality rates from 1962 to 1987 for North Carolina and for six regions of the state. For the state as a whole, stroke mortality decreased from 0.0050 to 0.0018 (64% decrease) for white males, 0.0039 to 0.0016 (59% decrease) for white females, 0.0072 to 0.0030 (58% decrease) for black males, and 0.0064 to 0.0025 (61% decrease) for black females. Under the hypothesis that the rate of decline in stroke mortality will slow and approach a "floor," a four-parameter logistic model was fit to the data. This model suggests that most of the decline in North Carolina stroke rates had occurred by 1987 and that the rates are now approaching the floor. For example, the estimated 1987 white male stroke mortality rate was 0.00190 and the floor 0.00176, suggesting only a 7% decline in the future. Similar percentage differences between 1987 levels and the floor were estimated for the three other race-gender groups. This analysis also suggested that while the absolute disparity by sex and race decreased between 1962 and 1987, the ratio of black to white mortality rates and the ratio of male to female mortality rates remained relatively constant. Estimates of 1-year case-fatality rates from two surveys of hospitalized stroke patients in a high-mortality five-county region of North Carolina decreased from 51% in 1970 to 38% in 1980. This decrease in stroke fatality rate is sufficient to account for 64% of the decrease in mortality estimated for this region over the same period.