A theory was developed to describe the equilibrium partitioning of flexible, linear polymers between a concentrated polymer solution and a medium consisting of a randomly oriented array of fibers (e.g., a fibrous membrane or gel). The fibers were represented as long cylinders, and the solute was modeled as a chain of cylindrical segments joined at specified bond angles. A Monte Carlo technique combined with excluded-volume concepts was used to describe steric interactions among chains and between chains and fibers. The partition coefficient ((D, the solute concentration in the fibrous material divided by that in the bulk solution) was predicted to increase markedly when the bulk solution concentration of the polymer was elevated. This sensitivity of (D to bulk concentration was found even at concentrations that were well below critical overlap values. The predicted partition coefficient decreased significantly as fiber volume fraction or molecular size was increased but tended to be less sensitive to the structural details of the chain (segment number, segment radius, and bond angle). Partition coefficients for dextran in agarose gels were measured as a function of agarose volume fraction, dextran size, and dextran concentration. The trends in the data were in good agreement with the theory. Over the range of dextran concentrations studied, as much as a 5-fold increase in Phi was observed.