Instantaneous error (IE) and predictability time (PT) are two model predictability measures. The IE measure is adopted to evaluate deterministic (stochastic) dynamical systems with known initial errors (known initial error distributions). The PT measure is used to evaluate stochastic dynamical systems without prior initial error information. We define PT (t1) and its variability (t2) for stochastic dynamical systems using the first-passage time and the minimum (noise level) and maximum (tolerance level) model error limits. We developed a theoretical framework (backward Fokker-Planck equation) to determine t1 and t2. Since the initial error becomes the independent variable of the backward Fokker-Planck equation, no prior information about the initial error is required.