In order to eliminate cells that are redundant, damaged, or infected, metazoan organisms have evolved the cell suicide mechanism termed apoptosis (Kerr et al. 1972). This genetic program is vital for normal development, for maintenance of tissue homeostasis, and for an effective immune system. Not surprisingly therefore, its disturbance is implicated in numerous pathological conditions, ranging from degenerative disorders to autoimmunity and cancer (Cory and Adams 2002; Cory et al. 2003). The cell’s death throes are choreographed by a set of previously dormant proteases, the caspases, which cleave several hundred cellular substrates (Thornberry and Lazebnik 1998). Two principal pathways to caspase activation have been recognized. One is triggered by engagement of so called “death receptors” on the cell surface (Strasser et al. 2000; Ashkenazi 2002). The other, of more ancient origin, is provoked by various forms of stress, including inadequate cytokine support and diverse types of intracellular damage. During stress, the cell’s decision on whether to invoke the suicide program—“to be or not to be”—rests primarily with the Bcl-2 family. Its interacting opposing members integrate developmental cues with the signals received from other cells and assess intracellular damage to determine whether to throw the caspase execution switch (Gross et al. 1999; Cory and Adams 2002; Cory et al. 2003). Early in stress-induced apoptosis of vertebrate cells, the outer membrane of the mitochondria is permeabilized, releasing cytochrome c and other death-promoting proteins (Newmeyer and Ferguson-Miller 2003). Because Bcl-2 can prevent this change, it has been widely assumed that commitment to all forms of stress-induced apoptosis, that is, all those regulated by the Bcl-2 family, requires this step to initiate all relevant caspase activation, and hence that Bcl-2 and its anti-apoptotic relatives function as guardians of mitochondrial integrity (Green and Reed 1998; Gross et al. 1999; Wang 2001). That viewpoint is challenged by recent findings. It will be argued here that in stress-induced apoptosis, as in death receptor signaling (Strasser et al. 2000; Ashkenazi 2002), caspases can be activated upstream or independently of mitochondria, and hence that the mitochondrial breach may often simply deliver the coup de grace to a cell already on the path to death. A related theme is that the Bcl-2 cohort must man checkpoints at several sites in the cell. Finally, the vexed unresolved issue of how the opposing factions of the Bcl-2 family govern caspase activation is addressed.