Mathematical "chaos" appears to occur in the heartbeat time-series. Low-dimensional chaos can be visualized as a geometric or spatial pattern in a temporal process. Several methods are available for representing these geometric patterns, including the mutual information content, the correlation dimension, and the type of pattern recurrences. Application of these methods to the heartbeat following coronary artery occlusion shows that prior to the initiation of lethal arrhythmogenesis the mutual information increases, the correlation dimension decreases, and the recurrence patterns shift from periodic clustering to local clustering. These observations not only have implications for new theoretical approaches to understanding how the heartbeat is generated and how arrhythmias occur, but also for the development of new biomedical tests and the manufacture of new devices for use in clinical cardiology.