We use a global ocean general circulation model (OGCM) with low vertical diffusion and isopycnal mixing to simulate the circulation in the Atlantic Ocean at present-day and the Last Glacial Maximum (LGM). The OGCM includes d18O as a passive tracer. Regarding the LGM sea-surface boundary conditions, the temperature is based on the GLAMAP reconstruction, the salinity is estimated from the available d18O data, and the wind-stress is derived from the output of an atmospheric general circulation model. Our focus is on changes in the upper-ocean hydrology, the large-scale horizontal circulation and the d18O distribution. In a series of LGM experiments with a step-wise increase of the sea-surface salinity anomaly in the Weddell Sea, the ventilated thermocline was colder than today by 2 3°C in the North Atlantic Ocean and, in the experiment with the largest anomaly (1.0 beyond the global anomaly), by 4-5°C in the South Atlantic Ocean; furthermore it was generally shallower. As the meridional density gradient grew, the Antarctic Circumpolar Current strengthened and its northern boundary approached Cape of Good Hope. At the same time the southward penetration of the Agulhas Current was reduced, and less thermocline-to-intermediate water slipped from the Indian Ocean along the southern rim of the African continent into the South Atlantic Ocean; the 'Agulhas leakage' was diminished by up to 60% with respect to its modern value, such that the cold water route became the dominant path for North Atlantic Deep Water (NADW) renewal. It can be speculated that the simulated intensification of the Benguela Current and the enhancement of NADW upwelling in the Southern Ocean might reduce the import of silicate into the Benguela System, which could possibly resolve the 'Walvis Opal Paradox'. Although d18Ow was restored to the same surface values and could only reflect changes in advection and diffusion, the resulting d18Oc distribution came close to reconstructions based on fossil shells of benthic foraminifera.