The pulse pressure (PP) is proportional to the preceding interval (T) because of the restitution of contractility and the Starling mechanism, and is inversely proportional to the pre-preceding interval (T-l) because of potentiation of contractility. The aim of the present paper was to find if this relationship can be used for diagnostic purposes. Blood pressure was noninvasively and continuously recorded for 3 minutes (Pen^z method), in 26 healthy subjects, in 13 patients with congestive heart failure (NYHA I,II) and sinusoidal rhythms and in 21 patients with atrial fibrillations. By means of multidimensional regression analysis the coefficient D[T] and D[T-1] were calculated in each subject. D[T] expresses the relative role of the preceding, D[T-1] of the pre-preceding interval. The correlation between PP and T was small in subjects with sinusoidal rhythms. Subjects with particular correlation coefficients between PP and T-l higher than 0.5 were used for further analysis (18 controls, 7 patients). The difference between D[T-1] in controls (0.30 ± 0.20) and in patients (0.48 ± 0.19) was significant (Wilcoxon P<0.05). In subjects with atrial fibrillations both D[T] and D[T-1] were higher in decompensated patients (Wilcoxon P<0.05). The ratio D[T]/D[T-1] was higher in patients with mitral stenosis than in patients with ischaemic heart disease (t-test, P<0.05). The test can be usefully employed as a screening test in medical practice.