Nineteen normal subjects and five patients with atrial fibrillation underwent transesophageal and transthoracic echocardiographic studies to evaluate the normal pulmonary venous flow pattern, compare right and left pulmonary venous flow and assess the effect of sample volume location on pulmonary venous flow velocities. Best quality tracings were obtained by transesophageal echocardiography. Anterograde flow during systole and diastole was observed in all patients by both techniques. Reversed flow during atrial contraction was observed with transesophageal echocardiography in 18 of the 19 subjects in normal sinus rhythm, but in only 7 subjects with transthoracic echocardiography. Two forward peaks during ventricular systole were clearly identified in 14 subjects (73%) with transesophageal echocardiography, but in none with the transthoracic technique. The early systolic wave immediately followed the reversed flow during atrial contraction and was strongly related to the timing of atrial contraction (r = 0.78; p less than 0.001), but not to the timing of ventricular contraction, and appeared to be secondary to atrial relaxation. Conversely, the late systolic wave was temporally related to ventricular ejection (r = 0.66; p less than 0.001), peaking 100 ms before the end of the aortic valve closure and was unrelated to atrial contraction time. Quantitatively, significantly higher peak systolic flow velocities were obtained in the left upper pulmonary vein compared with the right upper pulmonary vein (60 +/- 17 vs. 52 +/- 15 cm/s; p less than 0.05) and by transesophageal echocardiography compared with transthoracic studies (60 +/- 17 vs. 50 +/- 14 cm/s; p less than 0.05). Increasing depth of interrogation beyond 1 cm from the vein orifice resulted in a significant decrease in the number of interpretable tracings.(ABSTRACT TRUNCATED AT 250 WORDS)