The fact that circum-Antarctic ocean circulation and the Antarctic cryosphere developed in parallel during the Cenozoic era was revealed by results from Antarctic and Southern Ocean Deep Sea Drilling Project (DSDP) legs in the 1970s. This association provided the basis for the paradigm that development of the Antarctic Circumpolar Current (ACC) reduced meridional heat transport, isolated the continent within an annulus of cold water, and was thus the main cause of intensifying glaciation (e.g. Kennett, 1977). From 1970s DSDP results it was already clear that the main glacial-climatic thresholds had been crossed near the Eocene-Oligocene boundary, during the Middle Miocene and during the Late Pliocene. Subsequent more detailed studies and increasingly refined timescales have narrowed down the times of these steps in environmental evolution to 34.0–33.6, 14.2–13.8 and 3.3–2.5 Ma. A challenge to the paradigm emerged ten years ago from a coupled climate-ice sheet modelling study (DeConto and Pollard, 2003). The modelling results suggested that ice sheet feedbacks associated with declining atmospheric pCO2 may have been sufficient to cause rapid growth of the Antarctic Ice Sheet at around the time of the Eocene-Oligocene boundary in the absence of any changes in ocean circulation. Remarkably, these results also suggested ice sheet growth in two steps within 0.5 Myr, which was subsequently confirmed by detailed palaeoceanographic data (Coxall et al., 2005). DeConto & Pollard (2003) did, however, also attempt to model the effect of ACC onset by using a parameterized 20% increase in southward ocean heat transport to simulate ice sheet growth with Drake Passage closed. The results suggested that ACC onset resulting from opening of Drake Passage could have been the critical factor controlling the timing of ice sheet inception, but only if atmospheric pCO2 was between 2.5 and 3 times pre-industrial level (i.e. in the range 700–840 ppmv). A recent estimate of past atmospheric pCO2 levels derived from alkenones in marine sediments suggests that they were decreasing rapidly at around the time of the Eocene-Oligocene boundary, but may have been within this critical range during the period of rapid ice sheet growth (Pagani et al., 2011). Other modelling studies incorporating more realistic ocean components have cast doubt on the notion that ACC onset caused a significant reduction in southward ocean heat transport (e.g. Huber and Nof, 2006). In contrast, another study incorporating an ocean general circulation model suggests that southward ocean heat transport and Antarctic air temperatures would have been significantly greater with Drake Passage closed at higher atmospheric pCO2 (Sijp et al., 2009). Indeed, Sijp et al. (2009) conclude that Drake Passage being closed is a necessary condition for ice-free Antarctic conditions at high pCO2. Even if future, more sophisticated modelling studies rule out the possibility that ACC onset could have caused a significant reduction in southward ocean heat transport, it remains possible that ocean circulation changes were the primary cause of earliest Oligocene global environmental changes through their impacts on the carbon cycle. The ACC is today the strongest current in the world’s ocean and the main pathway of nutrient exchange between the Pacific, Atlantic and Indian oceans. Thus, onset of ACC flow may have increased productivity and accelerated CO2 drawdown. The Early Oliogocene expansion of the zone of accumulation of biosiliceous facies in the Southern Ocean could have resulted from such increased productivity. Onset of ACC flow through Drake Passage could, however, only have been a contributory factor to earliest Oligocene Antarctic Ice Sheet growth if a deep-water pathway opened at or before 34 Ma. Estimates of when a deep gap first opened extend over a wide range of ages, from Late Eocene to earliest Miocene. Reconstructions of major plate motions indicate divergence between South America and Antarctica started in the Eocene, but the details of how this divergence was manifested in Drake Passage also depend on the evolution of local, short-lived small plates, which shunted continental fragments around and formed small basins. Moreover, even if a deep gap had formed in the vicinity of Drake Passage by 34 Ma, continental fragments to the east and subduction-related topography around the early Scotia Arc may have delayed ACC development further (Barker and Burrell, 1982). Thus, considerable uncertainty remains about the timing of onset of ACC flow through Drake Passage and its possible role in environmental changes in the earliest Oligocene. Further geophysical and drilling investigations are needed to resolve the early history of this important ocean gateway. Middle Miocene development of both the western Scotia Sea and the East Scotia Sea is well- constrained by undisputed interpretations of marine magnetic anomalies. Debate continues, however, about the age of linear magnetic anomalies in part of the central Scotia Sea lying to the southwest of South Georgia, with interpretations of their ages ranging from Cretaceous to Miocene. Regardless of the true age of this part of the Scotia Sea, it is clear that at the beginning of the Middle Miocene continental and remnant arc fragments formed almost continuous barriers along the North Scotia Ridge and South Scotia Ridge. It is likely that the Scotia Sea was also enclosed to the east by a palaeo-South Sandwich arc, which would have extended southwards from the southern tip of the South Georgia continental block to Discovery Bank. It is not clear whether or not the narrow, >3000 m-deep gap in the North Scotia Ridge at 48 ̊W, through which the modern Polar Front passes, had already formed by then. If this gap is a product of subsequent strike-slip displacement along the Scotia–South America plate boundary, then ACC flow at the start of the Middle Miocene might have been more limited than it is now. The early Middle Miocene step in global environmental evolution coincides with the time that a deep-water gap is likely to have opened to the east of South Georgia as a result of back-arc spreading in the East Scotia Sea (Larter et al., 2003), and we suggest the possibility that there might be a causal link merits further investigation. Opening of deep-water gaps along the South Scotia Ridge in the Middle Miocene allowed Weddell Sea Deep Water to flow into the Scotia Sea. Although this deep inflow must have uplifted less dense water masses in the Scotia Sea, it remains unclear what its effects were on the upper parts of the water column and on flow of the ACC. The Middle Miocene step also followed a slow-down of spreading in the western Scotia Sea, which resulted in formation of a median valley along the ridge axis and generation of rougher sea-floor topography on its flanks. The increase in sea-floor roughness would have increased vertical mixing, which could have enhanced productivity. The median valley along the remnant West Scotia Ridge today steers the Polar Front to the gap in the North Scotia Ridge at 48 ̊W. Thus, for mation of the median valley could have caused the course of the Polar Front to switch from a previous, more southerly path. This would in turn have increased flow over rugged topography to the east of the Falkland Plateau, and thus potentially have increased vertical mixing, enhanced productivity and accelerated CO2 drawdown. Since the Late Miocene the only significant changes in the overall configuration of the Scotia Sea have been further widening of the East Scotia Sea, slow sinistral strike-slip motion along both the North Scotia Ridge and South Scotia Ridge, and slow deformation resulting from transpression along the Shackleton Fracture Zone. It seems unlikely that these changes could have had any significant effect on ACC flow. Therefore, we concur with suggestions that the step in global environmental evolution during the Late Pliocene probably resulted from a decrease in atmospheric pCO2 and rapid variations in insolation that triggered intensification of Northern Hemisphere glaciation.