Mean residence times (MRTs) have been classified into two main groups, namely system moment MRT [MRT(S,MO] and system matrix MRT [MRT(S,MA]. There are also MRTs of individual compartments [MRT(i)] such as central or plasma compartment, [MRT(P)], and tissue compartments and the MRT of an absorption site, [MRT(A)]. Much of the literature on MRTs does not clearly indicate which MRT is being discussed. MRT(S,MO) has been termed non-compartmental, but is really based on a structured model. There are really no model-independent MRTs or steady-state volumes of distribution. For the classical two-compartment open model with central compartment input, sampling and elimination MRT(S,MO) = MRT(S,MA) for a given set of microscopic rate constants. When elimination occurs from any but the central compartment then MRT(S,MO) is not equal to MRT(S,MA). For 'first-pass' drugs it is necessary to have a model where elimination occurs from a compartment different from the central and sampling compartment. Many of the methods of estimating MRTs which have been reported in the literature to date are reviewed and some generalizations are drawn. Some uses of MRTs are indicated. These uses involve both amounts of drug in the body as well as concentrations. The relationship between MRT(S,MO) and MRT(S,MA) for the Rowland two-compartment open model with peripheral compartment elimination is: MRT(S,MA)--MRT(S,MO) = 1/(k20 + k21). Thus the system matrix MRT is always larger than the system moment MRT for this linear model, which is most useful for 'first-pass' drugs. A general equation for MRT(S,MA) of all three two-compartment open models with input into either of the compartments is (lambda 1 + lambda 2 - ki0)/lambda 1 lambda 2 where i is the compartment (i = 1 or 2) into which input occurs.