MEASUREMENTS of the attenuation of low-frequency sound in seawater (Fig. 1) show anomalously high values of attenuation below 10 kHz (ref. 1). The explanations of this anomaly include biological and particulate scattering, nonlinear propagational effects and chemical relaxation phenomena2 but because insufficient experimental evidence exists, a definite conclusion cannot be reached. Laboratory experiments are being carried out to determine if any of the solutes dissolved in seawater relax near 1 kHz (refs. 3, 4). We have attempted to clarify the situation by making comparative low-frequency measurements in large natural freshwater lakes. Earlier measure ments in Lake Superior5, where a very weak sound channel and relatively shallow bottom (175 m average depth) introduced complications into the data analysis, have suggested that more appropriate test sites were Lake Tanganyika in central Africa or Lake Baikal in southern Siberia.