Uncertainties in the pattern of fire development are caused by multiple interactions among physical and chemical processes evolved by a variety of burning materials arranged in different ways. Hence, in a general sense, the spread of fire in a building is non-deterministic and the likely pattern of spread can only be predicted within limits of confidence expressed in probabilistic terms. Non-deterministic models are of two typesprobabilistic and stochastic. The object of this paper is to explain, with examples, the distinguishing features of these two types.In probabilistic models, critical events occurring during fire spread are treated as entirely random and independent. These models deal with the final outcomes, e.g., area damaged, which is sufficient in fire protection and insurance problems not requiring a detailed knowledge of the underlying physical processes. Probabilistic models discussed include probability distributions, logic-probability trees and probabilistic version of a deterministic model.In stochastic models, critical events occurring seauentially in space and time form a chain and are interconnected by transition probabilities. A state-transition stochastic model is discussed in detail with suggestions for improvement. Other models reviewed briefly include epidemic theory, branching processes, random walk and percolation process.