The majority of Ca that initiate cardiac contraction is released from the sarcoplasmic reticulum (SR), via the process of Ca-induced Ca release (CICR)1 mediated by ryanodine receptors (RyRs). Since CICR by definition is a self-regenerating, SR Ca release would be expected to continue until the SR Ca pool is fully exhausted. However, compelling experimental evidence shows that CICR ends when intra-SR Ca ([Ca]SR) is only partially depleted2-4. Thus, a robust termination mechanism must exist to counteract the positive feedback of CICR. Despite a significant effort, the CICR termination mechanism remains unclear. This is significant because abnormal SR Ca release termination has been linked to Ca-dependent arrhythmias during cardiac pathologies such as heart failure (HF)5. In cardiac myocytes, SR Ca release occurs at specialized release sites that contain clusters of RyRs. Spontaneous elementary SR Ca release events, called sparks6, arise from the concerted opening of several RyRs at one release site. Sparks are measured as non-propagating local increases in cytosolic [Ca] ([Ca]i). The corresponding local [Ca]SR depletion can be measured and is called a Ca blink2, 4. To explore mechanisms that control CICR termination, we simultaneously monitored local changes of [Ca]i (spark) and [Ca]SR (blink) in permeabilized rabbit ventricular myocytes. In addition, single cardiac RyR activity was studied after reconstruction of the channel into lipid bilayers We analyzed Ca sparks and corresponding blinks over a wide range of SR Ca loads. To decrease SR Ca load, cells were treated with the SR Ca pump inhibitor thapsigargin. Upon thapsigargin application, SR Ca load and spark frequency progressively decreased until sparks ceased when load had declined to 60% of its starting value. We also found that irrespective of SR Ca load blinks terminate at a relatively fixed level of [Ca]SR (60% of initial [Ca]SR). Our data indicate that release during a spark reduces [Ca]SR to a critical level where RyRs in a cluster become inactive. A fall in local [Ca]SR may terminate a spark by two different mechanisms. First, [Ca]SR depletion can drive unbinding of Ca from an intra-SR RyR regulatory site and thus promote RyR closing7,8. Second, local [Ca]SR depletion can terminate a spark by reducing the unitary RyR current, decreasing [Ca]i around a cluster and thus breaking the local positive feedback of CICR9. The contribution of the latter mechanism to CICR termination has not been systematically explored. We studied effects of Mg and caffeine (compounds which change RyR sensitivity to [Ca]i) on Ca spark termination. Cytosolic Mg (0.7 mM) decreased RyR activity by 80%, whereas caffeine (0.2 mM) increased it by 110%. We found that RyR inhibition by Mg decreased spark amplitude, width and Ca release flux. Moreover, Ca sparks terminated at 85% of the initial [Ca]SR, much higher than for blinks observed in the absence of RyR inhibition (60%). On the other hand, RyR [Ca]i sensitization by caffeine had the opposite action on Ca spark termination. Caffeine (0.2 mM) increased Ca spark amplitude and decreased the [Ca]SR level at which sparks terminated (by 20%). We also explored how changes in cytosolic Ca buffering affect Ca spark termination. No substantial changes in single RyR function were observed in the presence of cytosolic BAPTA (a fast Ca buffer) if free [Ca] was kept constant. However, the same cytosolic [BAPTA] added to permeabilized myocytes significantly suppressed Ca spark activity. This BAPTA effect on sparks was not due to decrease in SR Ca load. Moreover, the fast cytosolic Ca buffer caused faster CICR termination at higher [Ca]SR. In contrast, lowering cytosolic Ca buffering (by decreasing [EGTA]) drastically increased Ca spark frequency and width, leading to generation of multifocal (or propagating) sparks. By enhancing local CICR, a low [EGTA] solution decreased the termination [Ca]SR level for Ca sparks. Finally, we studied if Ca spark termination was altered in myocytes from failing hearts, where remodeling processes lead to RyR malfunction. Recordings of RyR activity revealed that the channel sensitivity to [Ca]i is significantly increased in rabbit HF myocytes. In cells, this alteration of RyR function in HF causes termination of global Ca transients and local Ca sparks at significantly lower [Ca]SR compared to that in non-failing myocytes. We suggest that increased RyR sensitivity to [Ca]i allows HF myocytes to maintain systolic SR Ca release of nearly normal size even at a depleted SR Ca load. However, this RyR modification contributes to enhanced SR Ca leak and increased propensity of arrhythmogenic Ca waves. Collectively, our results indicate that cardiac Ca sparks terminate when [Ca]SR fall to a certain critical level, supporting the existence of a functional link between partial [Ca]SR depletion and CICR termination. This functional link likely involves falling single RyR Ca flux and Ca unbinding from an intra-SR RyR Ca regulatory site as local [Ca]SR falls. We conclude that multiple [Ca]SR-dependent termination mechanisms likely co-exist to ensure a stability of cardiac Ca cycling.