The dynamic coupling between heart rate intervals (RR) and ventricular repolarization (QT) is analyzed. The analysis is based on measurements of 11 patients with pacemaker. In each measurement, there are at least 4 abrupt changes of RR preset by the pacemaker. With such a protocol, RR changes are important and well defined while disturbing factors and noise sources (such as those related with motion of patient) are minimized. The QT/RR coupling was described by 3 parameters (a1, b2, b3) transfer function (TRF) selected on the basis of a statistical analysis of performances of different TRF models. We found that our model is by far the best in its class: with more parameters (higher order models) the residuals remain almost the same while the extra parameters display variability much larger than that of our parameters. For all measurements, our TRF model describes more than 70% of QT variability. Within the patient set, we found interesting differences concerning dynamic non-linearity (response times longer with decreasing RR intervals than with increasing RR).