The explanation for the substantial decline in stroke death rates can be investigated only by measuring trends in stroke incidence and case-fatality. Two community-based studies carried out in Auckland, New Zealand, in 1981 and 1991 used comparable methods and definitions, met criteria for well-designed studies, and had the power to detect small changes in incidence and case-fatality rates. 703 events (representing 50% of all strokes) were registered in 1981 and 1735 events in 1991. 521 (74.1%) and 1255 (72.3%) events in 1981 and 1991, respectively, were first-ever (in a lifetime) strokes. Although there was no change in overall stroke incidence between 1981 and 1991, there were changes in age and sex groups. The incidence rate among women younger than 75 years rose by a fifth (rate ratio 1.23 [95% CI 1.04-1.47]), whereas that in men of 75 years and older fell by a third (rate ratio 0.67 [0.54-0.82]). The 28-day case-fatality declined from 27.1 (21.7-32.6)% to 21.9 (18.1-25.7)% in men and from 37.6 (31.8-43.5)% to 25.8 (22.3-29.4)% in women from 1981 to 1991, but the decline was not statistically significant in any age or sex group. These findings suggest that we need to reappraise strategies for the prevention of stroke and assess the implications of improved survival in elderly stroke patients.