We have used cross-sectional contrast echocardiography with peripheral venous injection in 3 groups of patients: A) 23 patients with atrial septal defect (ASD); B) 19 patients with various forms of heart disease, but no evidence of ASD; C) 11 patients with proven interatrial communication, but no significant left-to-right shunt. The contrast studies were performed with a mechanical sector scanner and were evaluated retrospectively but "blindly", i.e. without knowing the patient's name and diagnosis and without knowing that part of the echocardiogram, which had been recorded before injecting contrast. The sound track of the video recorder was turned off. Left-to-right atrial shunting could be demonstrated in 20/23 (87%) patients with ASD, and in all 14 patients in whom both the parasternal and the apical examination technique had been used. The left-to-right shunt was seen as a stream of contrast-free blood flowing from the left atrium into the right atrium and washing away the contrast echoes at the site of the defect. This "negative contrast" or "wash-out" effect was accompanied by turbulences in the right atrium. All 22 patients with this sign had interatrial communications, but interestingly there were 2 patients in whom no definite left-to-right shunt could be demonstrated by oxymetry. Thus the "wash-out" phenomenon seems to be specific for a defect, but not necessarily a significant left-to-right shunt. Right-to-left atrial shunts were diagnosed in 65% of the ASD group. A left-to-right or right-to-left shunt was never diagnosed in any of the control patients. Seemingly turbulent blood flow, the intermittent appearance of contrast-free blood (presumably from the coronary sinus or the inferior vena cava) in the right atrium and a "wash-out" effect at the site of the entry of the inferior vena cava are potential sources of error. However, our blind analysis has shown that these pitfalls can be avoided.