Multicellular models were developed to predict contaminant dispersion in a three-dimensional (3-D) space. The method utilized in this study was patterned after models which have been developed for water pollution in lakes and streams. This approach involved (1) design of cell structures to accommodate the geometry and mass flow characteristics of the 3-D space and (2) approximation of dispersion coefficients to describe contaminant transport in addition to that resulting from mass flow between cells. The dispersion model utilized a mass balance equation to predict contaminant dispersion as a function of time. The computer model was evaluated against experimental data for oxygen deficiency inside a ventilated confined space (CS) model. Eight test cases of ventilation design for the CS model were tested for each of three contaminant release (nitrogen to cause oxygen deficiency) characteristics: (1) purging (oxygen recovery from an initial deficiency); (2) steady state; and (3) variable rate. The dispersion model did a reasonably good job of predicting oxygen concentrations for different locations (cells) in the CS model. The primary limitations of this multicellular method are associated with experimental approximations of flow patterns and dispersion coefficients.