Definitions of hypertension have historically been based on at least one of three concepts. The first approach identifies thresholds of hypertension based on the frequency of occurrence in the population. The statistical approach designates a point in the distribution (e.g., the 95th percentile), as the threshold for hypertension. This distribution method identifies different limits for hypertension depending on the age, sex, and race, of the population, all of which affect the average pressure. Although distribution curves do not by themselves identify thresholds for intervention, they are useful for examining changes in population groups over time. The second approach to defining hypertension relates pressures to the risk of morbidity and mortality and is characterized by a continuously graded curve with no clear categorical thresholds. Studies correlating both diastolic and systolic pressures with cardiovascular complications demonstrate continuous risks from lowest to highest values for both sexes, all ages, and both blacks and whites in the United States. The blood pressure-risk relationship provides a compelling rationale for treatment but does not by itself define thresholds for the initiation of therapy. The third approach uses data from clinical intervention trials to identify thresholds where the benefits of therapy outweigh the costs and side effects of long-term treatment. Although results of large randomized trials have clearly demonstrated reductions in morbidity and mortality by lowering blood pressures, consensus on the lowest threshold within the mild range for which antihypertensive drug treatment is recommended has not been reached. Because an optimal definition of hypertension must encompass all three approaches and the resultant classification scheme must be sufficient for all purposes, attempts to refine and improve upon the presently recommended thresholds will undoubtedly continue.