In human placenta, 3D spatial relationships between villi and the maternal vascular bed determine intervillous porosity and this, in turn, influences haemodynamics and transport. Recently-developed stereological methods were applied in order to examine and quantify these relationships. Placentas were collected after 37 wk from control pregnancies and those associated with maternal diabetes mellitus classified according to duration and severity (White classification scheme). Two principal questions were addressed: (1) are normal spatial arrangements maintained in well-controlled diabetes mellitus? and (2) do arrangements vary between diabetic groups? To answer these questions, tissue sections cut at random positions and orientations were generated by systematic sampling procedures. Volume densities of villi (terminal + intermediate), intervillous spaces and perivillous fibrin-type fibrinoid deposits were estimated by test point counting and converted to global volumes after multiplying by placental volumes. Design-based estimates of the sizes (volume- and surface-weighted volumes) of intervillous 'pores' were obtained by measuring the lengths of point- and intersection-sampled intercepts. From these, theoretical numbers of pores were calculated. Model-based estimates (cylinder model) of the hydraulic diameters and lengths of pores were also made. Second-order stereology was used to examine spatial relationships within and between villi and pores and to test whether pair correlation functions deviated from the value expected for 'random' arrangements. Estimated quantities did not differ significantly between diabetic groups but did display some departures from control values in non-insulin-dependent (type 2) diabetic placentas. These findings support earlier studies which indicate that essentially normal microscopical morphology is preserved in placentas from diabetic subjects with good glycaemic control. Therefore, it is likely that fetal hypoxia associated with maternal diabetes mellitus is due to metabolic disturbances rather than abnormalities in the quantities or arrangements of maternal vascular spaces.