The circulation and internal structure of the oceans exert a strong influence on Earth's climate because they control latitudinal heat transport and the segregation of carbon between the atmosphere and the abyss (Sigman et al., 2010, doi:10.1038/nature09149). Circulation change, particularly in the Atlantic Ocean, is widely suggested (Bartoli et al., 2005, doi:10.1016/j.epsl.2005.06.020; Haug and Tiedemann, 1998, doi:10.1038/31447; Woodard et al., 2014, doi:10.1126/science.1255586; McKay et al., 2012, doi:10.1073/pnas.1112248109) to have been instrumental in the intensification of Northern Hemisphere glaciation when large ice sheets first developed on North America and Eurasia during the late Pliocene, approximately 2.7 million years ago (Bailey et al., 2013, doi:10.1016/j.quascirev.2013.06.004). Yet the mechanistic link and cause/effect relationship between ocean circulation and glaciation are debated. Here we present new records of North Atlantic Ocean structure using the carbon and neodymium isotopic composition of marine sediments recording deep water for both the Last Glacial to Holocene (35-5 thousand years ago) and the late Pliocene to earliest Pleistocene (3.3-2.4 million years ago). Our data show no secular change. Instead we document major southern-sourced water incursions into the deep North Atlantic during prominent glacials from 2.7 million years ago. Our results suggest that Atlantic circulation acts as a positive feedback rather than as an underlying cause of late Pliocene Northern Hemisphere glaciation. We propose that, once surface Southern Ocean stratification (Sigman, et al., 2004, doi:10.1038/nature02357) and/or extensive sea-ice cover (McKay et al., 2012, doi:10.1073/pnas.1112248109) was established, cold-stage expansions of southern-sourced water such as those documented here enhanced carbon dioxide storage in the deep ocean, helping to increase the amplitude of glacial cycles.
