1. Isolated rat ventricular myocytes were whole-cell voltage clamped using electrodes containing fluorescent Ca2+ indicators. Cytosolic [Ca2+] ([Ca2+]i) was estimated with calcium green-2 in combination with carboxy SNARF-1 to remove movement artifacts, or with indo-1. 2. Sarcoplasmic reticulum (SR) Ca2+ was depleted using 20 mM caffeine in Na(+)-containing superfusion solution, and cells were Ca2+ loaded by voltage clamp depolarizations applied during superfusion with Na(+)-free 2 mM Ca2+ solution. Ca2+ currents (ICa) and fluorescence transients elicited by these depolarizations were measured, and the releasable Ca2+ content of the Sr was estimated from the amplitude of fluorescence transients elicited by the rapid application of 20 mM caffeine. 3. Depolarization-induced [Ca2+]i transients increased in amplitude and duration during superfusion with Na(+)-free 2 mM Ca2+ solution, independent of changes in peak ICa. Caffeine application confirmed that the SR Ca2+ content increased during this manoeuvre. 4. With increased Ca2+ loading, the fraction of releasable SR Ca2+ involved in depolarization-induced transients increased, and the gradation in [Ca2+]i transient amplitude produced by beat-to-beat variation of voltage clamp pulse duration (10-100 ms) was progressively lost. This duration dependence of [Ca2+]i transients was maintained during Ca2+ loading when the Ca2+ buffering capacity of the electrode solution was increased with 100 microM BAPTA, 150 microM EGTA, or 60 microM indo-1. 5. These data suggest that Ca2+ released from the SR during a stimulated [Ca2+]i transient promotes further SR Ca2+ release to a degree which is smoothly graded with SR Ca2+ content. The effects of exogenous Ca2+ buffers suggest that this positive feedback is mediated, at least in part, by [Ca2+]i.