The ability of the diaphragm to generate pressures at different lung volumes (VLs) in humans may be determined by the following factors: 1) its in vivo three-dimensional shape, radius of curvature, and tension according to Laplace law; 2) the relative degree to which it is apposed to the rib cage (i.e., zone of apposition) and lungs (i.e., diaphragm dome); and 3) its length-force properties. To gain more insight into these factors we have reconstructed from nuclear magnetic images the three-dimensional shape of the diaphragm of four normal subjects under supine relaxed conditions at four different VLs: residual volume, functional residual capacity, functional residual capacity plus one-half of the inspiratory capacity, and total lung capacity. Under our experimental conditions the shape of the diaphragm changes substantially in the anteroposterior plane but not in the coronal one. Multivariate regression analysis indicates that the zone of apposition is dependent on both diaphragm shortening and lower rib cage widening with lung inflation, although much more on the first of these two factors. Because of the changes in anteroposterior shape and expansion of the insertional origin at the costal margin with lung inflation, the data therefore suggest that the diaphragm may be more accurately modeled by a "widening piston" (Petroll's model) than a simple "piston in a cylinder" model. A significant portion of the muscular surface is lung apposed, suggesting that diaphragmatic force has radial vectors in the dome and vectors along the body axis in the zone of apposition. The muscular surface area of the diaphragm decreased linearly by approximately 41% with VL from residual volume to total lung capacity. Diaphragmatic fibers may shorten under physiological conditions more than any other skeletal muscle. The large changes in fiber length combined with limited shape changes with lung inflation suggest that the length-twitch force properties of the diaphragm may be the most important factor for the pressure-generating function of this respiratory muscle in response to bilateral phrenic shocks at different VLs.