Prompted by the failure to detect a clinically evident knee-joint effusion on the cross-table lateral view of an injured patient, a prospective study was carried out to compare the routine overhead and cross-table lateral knee views for detection of joint effusion in 18 patients with acute knee trauma. In every case, the size of the effusion as determined by the "fat-pad separation sign" was greater on the overhead view (p less than 0.001). In three patients the effusion would have been missed radiologically had the vertical-beam projection been omitted. The authors conclude that the cross-table lateral view is less sensitive than the routine overhead lateral view in the detection of knee-joint effusions because of fluid shift into the lateral recesses of the suprapatellar bursa with the patient in the supine position. This phenomenon is demonstrated by arthrography and computed tomography in one patient.