The concept of probabilistic reactor dynamics is formalized in which deterministic reactor dynamics is supplemented by the fact that deterministic trajectories in phase-space switch to other trajectories because of stochastic changes in the structure of the reactor such as a change of state of components as a result of a malfunction, regulation feedback, or human error. A set of partial differential equations is obtained under a Markovian assumption from the Chapman-Kolmogorov equation giving the probability π(x, i, t) that the reactor is in a state x where vector x describes neutronic and ther-mohydraulic variables, and in a component state i at time t. The integral form is equivalent to an event tree where branching occurs continuously. A backward Kolmogorov equation allows evaluation of the probability and the average time for x(t) to escape from a given safety domain.