Nonlinear dynamics are certainly involved in tlie generation and control of cardiovascular signais. The paper presents a methodological and clinical review of nonlinear dynainic analysis in heart rate variability signals (HRV) obtained from normal subjects and pathological cases. The employed methods starts both from state space reconstruction procedures (Correlation Dimension, Lyapunov Exponents, False Nearest Neighborhood, Nonlinear Noise Filtering in the space-state) and by direct estimation of time series characteristics (Self-Similarity Hurst Exponent, Approximate Entropy). Analysis parameters confirm their usefulness in ihe classification of different patient groups. Examples reported for normal subjects, patients after orthotopic heart transplant, patients who recently had an anterior myocardial infarction episode, confirm the ability of these indexes in the HRV patterns classification both in short and long temporal Windows. Moreover, preliminary results in ICU patients indicate that long period indexes seem to be strongly related to the prediction of both positive and negative patient outcome. Computation of parameters from the nonlinear dynamic approach to the cardiovascular system can be a valid help for diagnosis of pathological heart States.