Differences between linear program based timber harvest schedules that use decision variables based on stand types (homogeneous but generally noncontiguous areas) and management units (generally heterogeneous but contiguous areas) were investigated. It was proposed that (i) optimal harvest schedules identified using stand type decision variables should have larger present net value objective function values than those identified using models with management unit decision variables, (ii) optimal present net value objective function values in management unit models should decline as management unit size is increased, and (iii) as the number of management choices increases, differences between stand type and management unit optimal present net values should decrease. The propositions were tested using 48 linear programming timber harvest scheduling models constructed for the University of California's Blodgett Forest Experiment Station. These models, which form 12 model groups, differ in the numbers and type of prescriptions considered for existing or regenerated stands, and harvest flow or ending inventory policies. The results generally supported the propositions and indicate that the number of management choices considered in the timber harvest scheduling model is probably a more important factor influencing the optimal harvest schedules than is land classification.