We used the Luo and Rudy (LR) mathematical model of the guinea pig ventricular cell coupled to experimentally recorded guinea pig ventricular cells to investigate the effects of geometrical asymmetry on action potential propagation. The overall correspondence of the LR cell model with the recorded real cell action potentials was quite good, and the strength-duration curves for the real cells and for the LR model cell were in general correspondence. The experimental protocol allowed us to modify the effective size of either the simulation model or the real cell. 1) When we normalized real cell size to LR model cell size, required conductance for propagation between model cell and real cell was greater than that found for conduction between two LR model cells (5.4 nS), with a greater disparity when we stimulated the LR model cell (8.3 +/- 0.6 nS) than when we stimulated the real cell (7.0 +/- 0.2 nS). 2) Electrical loading of the action potential waveform was greater for real cell than for LR model cell even when real cell size was normalized to be equal to that of LR model cell. 3) When the size of the follower cell was doubled, required conductance for propagation was dramatically increased; but this increase was greatest for conduction from real cell to LR model cell, less for conduction from LR model cell to real cell, and least for conduction from LR model cell to LR model cell. The introduction of this "model clamp" technique allows testing of proposed membrane models of cardiac cells in terms of their source-sink behavior under conditions of extreme coupling by examining the symmetry of conduction of a cell pair composed of a model cell and a real cardiac cell. We have focused our experimental work with this technique on situations of extreme uncoupling that can lead to conduction block. In addition, the analysis of the geometrical factors that determine success or failure of conduction is important in the understanding of the process of discontinuous conduction, which occurs in myocardial infarction.