Hematopoietic bone marrow in the distal femur of the adult may be mistaken for a pathologic marrow process in magnetic resonance imaging of the knee. We investigated the incidence of hematopoietic marrow in the distal femur in a series of 51 adult patients and compared spin-echo (TR/TE in ms: 500/35, 2000/80) and opposed-phase gradient-echo (0.35 T, TR/TE in ms: 1000/30, theta = 75-degrees) magnetic resonance images. Zones with intermediate to low signal intensity on T1-weighted spin-echo and opposed-phase gradient-echo sequences representing hematopoietic marrow within high signal intensity fatty marrow were observed in 18 of the 51 patients. Five patterns of marrow signal reduction were identified; type 0: uniform high signal, i.e., no signal change; type I, focal signal loss; type II, multifocal signal loss without confluence; type III, confluent signal loss; and type IV, complete homogeneous reduction in marrow signal. Opposed-phase gradient-echo sequences demonstrated markedly greater red-yellow marrow contrast than conventional spin-echo sequences. Follow-up studies in three patients using a gradient-echo sequence with TE varying from 10 to 21 ms at 1-ms increments showed a cyclic increase and decrease in red and yellow marrow signal intensity depending on the TE. The contribution of intravoxel chemical shift effects on red-yellow marrow contrast in opposed-phase gradient-echo images was verified by almost complete cancellation of the TE-dependent marrow signal oscillation with use of a chemically selective pulse presaturating the water protons. Hematopoietic marrow in the adult distal femur in the absence of hematologic abnormalities is found primarily in women of menstruating age. It may be residual and may represent a biologic variation in the normal adult pattern of red-yellow marrow distribution. Reconverted red marrow appears to be related to increased erythrocyte demand. Residual and reconverted red marrow should not be mistaken for bone marrow malignancy. Opposed-phase gradient-echo imaging is easily implemented and appears ideally suited to monitor the distribution of hematopoietic marrow as a function of age and erythrocyte demand in vivo.