Endoscopic measurement of gastric blood perfusion by laser Doppler flowmetry (LDF) has been evaluated in 28 patients and 15 healthy volunteers. During the recordings it was necessary to keep the probe in light contact with the mucosa to obtain stable curves and to avoid artificial Doppler signals caused by relative movements between the gastric wall and the probe. Gastric distention by air insufflation did not influence the recorded flow level significantly when air insufflation was moderate. The intravenous injection of 0.6 mg atropine did not cause any significant alteration in recorded blood flow, and this drug may be used as premedication before endoscopic blood flow measurements. Recordings with both 4- and 12-kHz bandwidth of the LDF instrument showed a relative constant relationship for different flow levels, the flow values measured with 12 kHz being about twice the corresponding values measured with 4 kHz. With 12-kHz bandwidth more of the disturbance signal is recorded, which makes analysis of endoscopic recorded flow curves difficult and inaccurate. It is therefore recommended to use 4-kHz bandwidth during endoscopic measurements in conscious humans. Blood flow measurements from both sides of the gastric wall were consistently of the same order of magnitude (r = 0.91), and the endoscopically recorded output signal increased in three of five patients when a reflecting mirror was placed at the serosal side. The results indicate that endoscopic LDF usually represents blood perfusion in all layers of the gastric wall.