Neuronal correlates of brain functions are typically characterized by a high degree of variability in firing activity that manifests itself both within and between trials. These fluctuations are themselves of functional relevance and reflect computational dynamics that build on probabilistic transitions between attractor states. This suggests that probabilistic behavior is intrinsically linked with the stochasticity at cellular levels. In this chapter, we review the theoretical framework of stochastic neurodynamics that allows us to investigate the roles of noise and neurodynamics in the computation of probabilistic behavior. This is important for neuroscience, because it provides a theoretical framework that goes beyond the traditional noiseless neurodynamical analyses, and for neurophysiology, because transitions between states and not just averages across states are important to analyze.