Non invasive blood pressure monitoring using an automatic Nippon Colin recorder was performed in 20 hospitalized hypertensive patients every 15 min during a 4-day study period. A chronobiologic analysis of the data was performed in order to test the consistency of the whole-day blood pressure pattern. Halberg's method of single cosinor was used to determine the values and the confidence limits of the following parameters of 24-hour rhythmic changes: mesor (rhythm-adjusted 24-hour average), amplitude (24-hour variability estimate), acrophase (clock hour of maximum BP level). The variation of the above parameters during the four 24-hour periods was found to be within 9 mmHg for the systolic mesor, 6 mmHg for the diastolic mesor, 5 mmHg for the systolic amplitude, 4 mmHg for the diastolic amplitude, 55 min for the systolic acrophase, 60 min for the diastolic acrophase. The same parameters were also calculated in 48-hour, 72-hour and 96-hour long time series, and with different sampling periods (30, 60, 120 min). The length of the monitoring span did not influence significantly the values of the rhythm parameters only when 15-min or 30-min sampling periods were used. The widths of the 95% confidence limits of amplitude and acrophase were significantly narrower in the 2-day than in the 1-day series. Further lengthening of the monitoring span from 2 to 4 days resulted in a much smaller additional narrowing of the confidence limits. Sampling periods longer than 30 min caused a loss of sensitivity in rhythm detection.(ABSTRACT TRUNCATED AT 250 WORDS)