PURPOSE: To measure directly the grid performance of mammography units for the range of breast thicknesses and x-ray tube potentials encountered in clinical practice. MATERIALS AND METHODS: Contrast improvement factors and Bucky factors were determined for four mammographic units as a function of x-ray tube potential (25, 30, and 35 kVp), phantom thickness (2, 4, and 8 cm) and, on one unit, three target-filter combinations. Three units used a linear grid; one, a cellular grid. Two methods were used for nongrid measurements. RESULTS: For all units tested, contrast improvement factor increased with increased phantom thickness and with increased kilovolt peak level for the 8-cm-thick phantom and changed little with kilovolt peak level for 2- and 4-cm-thick phantoms. At 25 and 30 kVp, contrast improvement factor performance with the linear grids was comparable; with the cellular grid, it was 5%-10% higher. In all cases, the Bucky factor increased with increased phantom thickness and decreased with increased tube potential. CONCLUSION: Differences in grid performance exist. At 25 and 30 kVp, the cellular grid exhibited superior contrast improvement factor performance, whereas one of the linear grids exhibited superior Bucky factor performance. Measured contrast improvement and Bucky factors are dependent on nongrid technique. Cassette tunnels introduce scatter and should not be used with nongrid or magnification techniques.