There is a theoretical ideal to be achieved in the construction of a simple, practical membrane oxygenator. Direct contact between blood and oxygen common to many current artificial lungs is unphysiologic and entails the obvious possibility of air embolism in addition to suspected but not well-defined damage to blood elements at gas-fluid interfaces.1 Contemporary membrane oxygenators2-4 necessitate relatively high pressure pumping of venous blood into sandwich-like devices which must use membranes which are undesirably thick. In all of the currently employed machines the primary problem would appear to be that of securing enough contact between blood and oxygen without excessive recurrent damage to the blood which must be returned to the patient's systemic circulation. Our primary intent has been to employ membrane substances which are as thin as possible. As suggested by Crescenzi5,6 and Kylstra,7 these require that only gravity flow be used to avoid rupture