(1) An artifice in which non-exponential service times are represented as a sum of exponential random variables, each of which is referred to as a stage. If the stages are independent and identically distributed, then we have an Erlang probability function as the distribution of the sum; if the stages are only independent, the resultant density is called a generalized Erlang. Further extensions of this sort of device lead to Coxian and phase-type distributions. Queueing theory. (2) The subdivisions of a dynamic programming problem where a decision is required. Each stage has a number of states associated with it, and the decision at any stage describes how the state at the current stage moves to a state at the next stage. Dynamic programming.