To determine whether the proportion of severe dengue cases increased with the yearly seasonal increase in dengue incidence, we examined reports of disease symptoms in case surveillance data and laboratory testing results in Puerto Rico from January 1989 to July 1992. A computer algorithm was designed to identify severe cases, i.e., those that fulfilled three or all four of the World Health Organization criteria for dengue hemorrhagic fever (DHF). A monthly severity index (SI) was defined as the ratio of severe cases to laboratory-positive and indeterminate (all non-negative) cases for each month, while a more restrictive severity rate (SR) was defined as the ratio of severe laboratory-positive cases to the total number of laboratory-positive cases for each month. Monthly SI and SR were compared in two ways: within an epidemic cycle, and month-by-month. Linear regression analysis was performed over the monthly averages of the SI and SR. For a month-by-month examination of SI and SR, we examined the 43-month sequence by means of a linear model with autocorrelated disturbances. We found no statistically significant or cyclical change in the proportion of severe cases from month to month in this period. Our conclusions differ from the observations during the Cuban DHF epidemic of 1981, in which case severity was shown to increase markedly as the epidemic progressed; they agree with the conclusions of most previous studies in that dengue severity does not change significantly throughout a period of increased incidence.