We examine the utility of the uranium (U) content of planktonic foraminifera tests as an indicator of past changes in seawater U content. The U/Ca ratio in foraminifera from Atlantic and Caribbean cores is constant in the Holocene and decreases by ~25% during the last glacial period. Magnesium/calcium (Mg/Ca) ratios of the same samples show similar trends. While the timing of the U/Ca changes appears to be associated with glacial-interglacial changes, the magnitude of the change is too large to be caused by variations in the extent of anoxic or suboxic sediments or by changes in riverine input. We assume that the same process produced changes in both U/Ca and Mg/Ca ratios because of a strong correlation between the two ratios. Partial dissolution of the calcite is ruled out, because we observe the same changes in well-preserved cores from basins with opposite dissolution histories. We also reject exchange between foraminiferal and pore water U because of the oxic depositional environment of both cores and because of the consistency in the U/Ca trends from cores in different parts of the ocean. We suggest that the observed foraminiferal U/Ca and Mg/Ca trends may be the result of a temperature effect on the incorporation of these metals. If this is true, it introduces the possibility of a new paleotemperature indicator but complicates the use of U/Ca ratios in planktonic foraminifera tests as an indicator of past seawater U changes. ¿ American Geophysical Union 1996