We present strontium (Sr) isotope ratios that, unlike traditional Sr-87/Sr-86 data, are not normalized to a fixed Sr-88/Sr-86 ratio of 8.375209 (defined as delta Sr-88/86 = 0 relative to NIST SRM 987). Instead, we correct for isotope fractionation during mass spectrometry with a Sr-87-Sr-84 double spike. This technique yields two independent ratios for Sr-87/Sr-86 and Sr-88/Sr-86 that are reported as (Sr-87/Sr-86*) and (delta Sr-88/86), respectively. The difference between the traditional radiogenic (Sr-87/Sr-86 normalized to Sr-88/Sr-86 = 8.375209) and the new Sr-87/Sr-86* values reflect natural mass-dependent isotope fractionation. In order to constrain glacial/interglacial changes in the marine Sr budget we compare the isotope composition of modern seawater ((Sr-87/Sr-86*, delta Sr-88/86)(Seawater)) and modern marine biogenic carbonates ((Sr-87/Sr-86*, delta(88)/Sr-86)(Carbonates)) with the corresponding values of river waters ((Sr-87/Sr-86*, delta Sr-88/86)(River)) and hydrothermal solutions ((Sr-87/Sr-86*, delta(88)/Sr-86)(HydEnd)) in a triple isotope plot. The measured (Sr-87/Sr-86*, delta(88)/Sr-86)(River)). values of selected rivers that together account for similar to 18% of the global Sr discharge yield a Sr flux-weighted mean of (0.7114(8), 0.315(8)parts per thousand). The average ((Sr-87/Sr-86*, delta Sr-88/86)(HydEnd) values for hydrothermal solutions from the Atlantic Ocean are (0.7045(5), 0.27(3)parts per thousand). In contrast, the (Sr-87/Sr-86*, delta Sr-88/86)(Carbonates) values representing the marine Sr output are (0.70926(2), 0.21(2)parts per thousand). We estimate the modern Sr isotope composition of the sources at (0.7106(8), 0.310(8)parts per thousand). The difference between the estimated (Sr-87/Sr-86*, delta Sr-88/86) and (Sr-87/Sr-86*, delta(88)/Sr-86)(output) values reflects isotope disequilibrium with respect to Sr inputs and outputs. In contrast to the modern ocean, isotope equilibrium between inputs and outputs during the last glacial maximum (10-30 ka before present) can be explained by invoking three times higher Sr inputs from a uniquely "glacial" source: weathering of shelf carbonates exposed at low sea levels. Our data are also consistent with the "weathering peak" hypothesis that invokes enhanced Sr inputs resulting from weathering of postglacial exposure of abundant fine-grained material. (C) 2010 Elsevier Ltd. All rights reserved.