This paper describes the development of an ultrasonic technique for the characterization of membrane morphology. Ultrasonic frequency-domain reflectometry (UFDR) using a 90 MHz focused immersion transducer has been employed to obtain characteristic acoustic responses from microporous polymeric membranes with nominal pore sizes in the range of 0.1-0.6 mum. Systematic and statistically significant differences in signal amplitude are observed in the frequency domain for reflections from the back surface of poly(vinylidene fluoride) (PVDF) and mixed cellulose ester (MCE) membranes. As the pore-size of the membranes increases, the amplitude of the reflected signal at high frequencies decreases due to increased scattering of the ultrasonic wave. These UFDR differences correspond well with those obtained from independent measurements including scanning electron microscopy (SEM) and gas-liquid porometry. A simple phenomenological artificial neural network (ANN) model has been developed to predict membrane pore-size based upon the signal amplitude at five frequencies. Such predictive capability suggests that the combination of UFDR and ANN may have significant utility for membrane quality control applications. (C) 2004 Elsevier B.V. All rights reserved.