The ventricular response in atrial fibrillation is often described as "chao
tic," hut this has not been demonstrated in the strict mathematical sense.
A defining feature of chaotic systems is sensitive dependence on initial co
nditions: similar sequences evolve similarly in the short term but then div
erge exponentially We developed a nonlinear predictive forecasting algorith
m to search for predictability and sensitive dependence on initial conditio
ns in the ventricular response during atrial fibrillation. The algorithm wa
s tested for simulated R-R intervals from a Linear oscillator with and with
out superimposed white noise; a chaotic signal (the logistic map) with and
without superimposed white noise, and a pseudorandom signal and was then ap
plied to RR intervals from 16 chronic atrial fibrillation patients. Short-t
erm predictability was demonstrated for the linear oscillators, without los
s of predictive ability farther into the future. The chaotic system demonst
rated high shortterm predictability that declined rapidly further into the
future. The pseudorandom signal was unpredictable. The ventricular response
in atrial fibrillation was weakly predictable (statistically significant p
redictability in 8 of 16 patients), without sensitive dependence on initial
conditions. Although the R-R interval sequence is not completely unpredict
able, a low-dimensional chaotic attractor does not govern the irregular ven
tricular response during atrial fibrillation.