The majority of contractile calcium in cardiac muscle is released from stores in the sarcoplasmic reticulum (SR), by a process of calcium-induced calcium release (CICR) through ryanodine receptors. Because CICR is intrinsically self-reinforcing, the stability of and graded regulation of cardiac EC coupling appear paradoxical. It is now well established that this gradation results from the stochastic recruitment of varying numbers of elementary local release events, which may themselves be regenerative, and which can be directly observed as calcium sparks. Ryanodine receptors (RyRs) are clustered in dense lattices, and most calcium sparks are now believed to involve activation of multiple RyRs. This implies that local CICR is regenerative, requiring a mechanism to terminate it. It was initially assumed that this mechanism was inactivation of the RyR, but during the decade since the discovery of sparks, no sufficiently strong inactivation mechanism has been demonstrated in vitro and all empirically determined gating schemes for the RyR give unstable EC coupling in Monte Carlo simulations. We consider here possible release termination mechanisms. Stochastic attrition is the spontaneous decay of active clusters due to random channel closure; calculations show that it is much too slow unless assisted by another process. Calcium-dependent RyR inactivation involving third-party proteins remains a viable but speculative mechanism; current candidates include calmodulin and sorcin. Local depletion of SR release terminal calcium could terminate release, however calculations and measurements leave it uncertain whether a sufficient diffusion resistance exists within the SR to sustain such depletion. Depletion could be assisted by dependence of RyR activity on SR lumenal [Ca(2+)]. There is substantial evidence for such lumenal activation, but it is not clear if it is a strong enough effect to account for the robust termination of sparks. The existence of direct interactions among clustered RyRs might account for the discrepancy between the inactivation properties of isolated RyRs and intact clusters. Such coupled gating remains controversial. Determining the mechanism of release termination is the outstanding unsolved problem of cardiac EC coupling, and will probably require extensive genetic manipulation of the EC coupling apparatus in its native environment to unravel the solution.