The direct measurement of the ratio of cosmogenic 10Be (T½ = 1.5 Ma) to stable terrigenously sourced 9Be in deep seawater or marine deposits can be used to trace water mass movements and to quantify the incorporation of trace metals into the deep sea. In this study a SIMS-based technique has been used to determine the 10Be/9Be ratios of the outermost millimetre of hydrogenetic ferromanganese crusts from the worlds oceans. 10Be/9Be ratios, time-corrected for radioactive decay of cosmogenic 10Be using 234U/238U, are in good agreement with AMS measurements of modem deep seawater. Ratios are relatively low in the North and equatorial Atlantic samples (0.4-0.5 X 10-7. In the Southwest Atlantic ratios increase up to 1 X 10-7, they vary between 0.7 and 1.0 X -7 in Indian Ocean samples, and have a near constant value of 1.1 ± 0.2 X 10-7 for all Pacific samples. If the residence time of 10Be 10Beγ10 in deep water is constant globally, then the observed variations in 10Be/9Be ratios could be caused by accumulation of 10Be in deep water as it flows and ages along the onveyor, following a transient depletion upon its formation in the Northern Atlantic. In this view both 10Be and 9Be reach local steady-state concentration in Pacific deep water and the global γ10Be ≅ 600 a. An alternative possibility is that the Be isotope abundances are controlled by local scavenging. For this scenario γ10Be would vary according to local particle concentration and would ≅ 600 a in the central Pacific, but γ10Be ≅ 230 a in the Atlantic. Mass balance considerations indicate that hydrothermal additions of 9Be to the oceans are negligible and that the dissolved riverine source is also small. Furthermore, aeolian dust input of 9Be appears insuffkient to provide the dissolved 9Be inventory. The dissolution of only a small proportion (2%) of river-derived particulates could in principle supply the observed seawater 9Be content. If true, ocean margins would be the sites for 9Be addition. Due to the particle-reactive nature of Be, these would also be the primary sites of Be removal. A possible net result of horizontal water masses passing through these marginal areas might be a decrease in seawater 10Be/9Be, and establishment of a relatively constant 9Be concentration. As γ10Be (~ 600 a) is less than the apparent age of deep water in the Pacific (~ 1500 a), the Pacific record of 10Be/9Be is not expected to show secular variations due to changes in deep-water flow, despite the large variations in 10Be/9Be between different water masses. Because of this insensitivity to deep-water flow, however, it is suggested that the 10Be/9Be ratio, determined in the authigenic phase of marine sediments or hydrogenetic precipitates, should be a suitable tool for monitoring changes in continental input or cosmic ray intensity on longer time scales.