Four hundred and thirteen continuous, high-frequency-sampled gastric pH-metries were subdivided into six groups (normal subjects and patients with duodenal ulcer, gastric ulcer, chronic gastritis, prior cholecystectomy, or antisecretory treatment). The frequency distribution of those pH fluctuations that were greater than the quantitation error of instruments did not differ significantly (p greater than 0.30) between groups. Five per cent of these real pH fluctuations lasted less than 3 min. As the sampling rate increased from 1 to 60 min, the percentage of subjects whose 24-h median of pHs differed by no more than +/- 10% from that calculated on the raw high-frequency-acquired data progressively decreased (99.8%-42%). This error propagation was more marked in the case of [H+] integrals (99%-10%). A sampling rate equal to or faster than 1 min is necessary to provide a proper representation of the circadian pH pattern and to calculate accurate acidity indexes, regardless of the physiologic or pharmacologic nature of the profile.