1. Direct voltage-gated (voltage-dependent Ca2+ release, VDCR) and Ca2+ influx-gated (Ca2+-induced Ca2+ release, CICR) sarcoplasmic reticulum (SR) Ca2+ release were studied in feline ventricular myocytes. The voltage-contraction relationship predicted by the VDCR hypothesis is sigmoidal with large contractions at potentials near the Ca2+ equilibrium potential (ECa). The relationship predicted by the CICR hypothesis is bell-shaped with no contraction at ECa. 2. The voltage dependence of contraction was measured in ventricular myocytes at physiological temperature (37 C), resting membrane potential and physiological [K+]. Experiments were performed with cyclic adenosine 3',5'-monophosphate (cAMP) in the pipette or in the presence of the beta-adrenergic agonist isoproterenol (isoprenaline; ISO). 3. The voltage-contraction relationship was bell-shaped in Na+-free solutions (to eliminate the Na+ current and Na+-Ca2+ exchange, NCX) but the relationship was broader than the L-type Ca2+ current (ICa,L)-voltage relationship. 4. Contractions induced with voltage steps from normal resting potentials to -40 mV are thought to represent VDCR rather than CICR. We found that cAMP and ISO shifted the voltage dependence of ICa,L activation to more negative potentials so that ICa,L was always present with steps to -40 mV. ICa,L at -40 mV inactivated when the holding potential was decreased (VŁ = -57.8 +/- 0.49 mV). 5. ISO increased inward current, SR Ca2+ load and contraction in physiological [Na+] and a broad bell-shaped voltage-contraction relationship was observed. Inhibition of reverse-mode NCX, decreasing ICa,L and decreasing SR Ca2+ loading all decreased contractions at strongly positive potentials near ECa. 6. The voltage-contraction relationship in 200 microM cadmium (Cd2+) was bell-shaped, supporting a role of ICa,L rather than VDCR. 7. All results could be accounted for by the CICR hypothesis, and many results exclude the VDCR hypothesis.