A long-standing hypothesis is that feelings of anger and anxiety increase the risk for essential hypertension. Most studies examining this hypothesis have been cross-sectional in design or undertaken with men only. We tested this hypothesis along with determination of the other behavioral and biological predictors of increases in systolic (SBP) and diastolic (DBP) blood pressure from baseline to a follow-up examination 3 years later in a prospective study of 468 middle-aged women whose blood pressure at the baseline examination was less than 140/90 mmHg. Analyses showed that increases in the Spielberger Trait Anger Scale between the baseline and 3-year follow-up examination, as well as Framingham Tension scores (a measure of anxiety) at baseline, independently predicted an increase in SBP (P less than 0.01). Other factors that independently predicted an increase in SBP were baseline fasting insulin, parental history of hypertension and increases in body mass index and in alcohol intake across the 3 years of follow-up. Increases in the Spielberger Trait Anger Scores independently predicted increases in DBP (P less than 0.02), as did black race, increases in body mass index and hematocrit and decreases in potassium intake. Although menopausal status and hormone replacement therapy were unrelated to changes in blood pressure, postmenopausal women on hormone replacement therapy did show significant increases in DBP in the univariate analysis. Anxiety at baseline, along with parental history of hypertension, baseline fasting insulin and baseline body mass index, predicted a later onset of hypertension, i.e. on pharmacologic treatment for hypertension, in the univariate analysis.(ABSTRACT TRUNCATED AT 250 WORDS)