The evidence for deterministic chaos in normal heart rhythms is examined. Electrocardiograms were recorded of 29 subjects falling into four groups—a young healthy group, an older healthy group, and two groups of patients who had recently suffered an acute myocardial infarction. From the measured R–R intervals, a time series of 1000 first differences was constructed for each subject. The correlation integral of Grassberger and Procaccia was calculated for several subjects using these relatively short time series. No evidence was found for the existence of an attractor having a dimension less than about 4. However, a prediction method recently proposed by Sugihara and May and an autoregressive linear predictor both show that there is a measure of short-term predictability in the differenced R–R intervals. Further analysis revealed that the short-term predictability calculated by the Sugihara–May method is not consistent with the null hypothesis of a Gaussian random process. The evidence for a small amount of nonlinear dynamical behavior together with the short-term predictability suggest that there is an element of deterministic chaos in normal heart rhythms, although it is not strong or persistent. Finally, two useful parameters of the predictability curves are identified, namely, the ‘first step predictability’ and the ‘predictability decay rate,’ neither of which appears to be significantly correlated with the standard deviation of the R–R intervals.