The Late Quaternary, and more particularly the Last Glacial (Marine Isotope Stages 4, 3 and 2, defined here as 73,500–14,700 calendar years, 73.5–14.7 ka; ka is used solely for calendar years and kyr BP refers to 14 C ages), was characterised by millennial-scale climate oscillations of irregular periodicity. The onset of these oscilla-tions was abrupt, with most of the change in climate accomplished within 10–200 years (Steffensen et al., 2008) and the magnitude of the change was large (of the order of 8–15 C in Greenland) (Huber et al., 2006). Two types of rapid climate changes have been described: Dansgaard–Oeschger (D-O) cycles (Dansgaard et al.,1984), associated with abrupt warming and subsequent cooling in Greenland, and cold phases associated with the formation of ice-rafted debris (IRD) deposits (Heinrich layers) (Heinrich, 1988) in the North Atlantic. Dansgaard–Oeschger cycles are clearly registered in the Greenland ice-core record (e.g. Johnsen et al., 1992; North GRIP Members, 2004) and the traces of both of these climate oscillations are recorded in a variety of marine and terrestrial records worldwide (e.g. Bond et al., 1993; Allen et al., 1999; Wang et al., 2001; Gonzalez et al., 2008). The geographic pattern of registration of individual oscillations (Sanchez Goñ i et al., 2008), and the magnitude, nature and length of the component phases of each recorded oscillation, appear to vary (Johnsen et al., 1992). However, documentation of regional changes has been hampered by problems of the synchronisation of individual chronologies, and our understanding of the mechanisms underlying these climate oscillations is still far from complete. Analysis of the mechanisms and impacts of large and rapid climate changes in the past is given additional impetus by the possibility that such events might occur in the future. Although the precise causes are different, investigation of the impact of the warming events at the beginning of D-O cycles or of iceberg melting during Heinrich intervals on the Meridional Overturning Circula-tion (MOC) in the North Atlantic speaks directly to the impact of future changes in the MOC on regional climate. Coupled ocean– atmosphere model simulations show a reduction of the MOC during the 21st century, in some cases by up to 50%, as a conse-quence of greenhouse-gas-induced polar warming (Gregory et al., 2005). Simulations with simpler climate models have shown that complete shutdown of the MOC can occur if the slowdown reaches a crucial threshold (which differs in different models) (Stocker and Schmittner, 1997; Stouffer and Manabe, 2003; Stouffer et al., 2006). Clearly, the coupled models do not reach the critical threshold as a result of the gradual change in greenhouse-gas forcing during the 21st century but might do so if the additional forcing due to even partial melting of the Greenland ice sheet were taken into account. There have been several recent attempts to synthesise millen-nial-scale climate change during the glacial (e.g. Broecker and Hemming, 2001; Alley et al., 2002; van Andel, 2002; Voelker, 2002), but most focus on marine and ice-core records. In this issue, we focus on documenting regional changes in vegetation indicated by pollen records from both marine and terrestrial cores (Fletcher et al., in this issue; Takahara et al., in this issue; Jimé nez-Moreno et al., in this issue; Heßler et al., in this issue). There are multiple reasons why this is timely. First, there has been a very rapid increase during the last decade in the number of pollen records with high temporal resolution. However, there is no global compi-lation of the pollen data, nor a synthesis of vegetation reconstruc-tions based on these data. Second, the development of a new and coherent ice-core reference chronology (GICC05: Svensson et al., 2006, 2008; see also Wolff et al., in this issue) makes it possible to achieve a better synchronisation between documented changes in Greenland and the pollen records. Third, a regional synthesis of charcoal records from North America (Marlon et al., 2009) shows that fire regimes respond to abrupt climate changes during the last deglaciation – but little is known about the response of fire globally to millennial-scale climate variability and associated vege-tation changes during earlier intervals (Daniau et al., in this issue). Fourth, investigation of the impact of changes in vegetation cover, including wetland extent, and in fire regimes is important for understanding the rapid and extremely large (up to ca 200 ppb) changes in methane during D-O cycles (Blunier and Brook, 2001; Flü ckiger et al., 2004) and the potential climate feedback. Finally, modelling groups have recently begun to explore the impact of, e.g., changes in freshwater forcing under glacial conditions on regional climates (e.g. Crucifix et al., 2001; Ganopolski and Rahm-storf, 2001; Claussen et al., 2003; Knutti et al., 2004; Flü ckiger et al., 2006, 2008; see also Kageyama et al., in this issue) but more detailed documentation of observed changes is required for the evaluation of these experiments. A multiplicity of terms is used in discussing rapid climate changes and millennial-scale climate variability during the glacial, and this has led to some confusion particularly in relating records from different regions. Alley et al. (2002), in a definition adopted by the IPCC (Meehl et al., 2007), define abrupt climate change as one that takes place more rapidly than the underlying forcing, pointing out that this kind of behaviour can only occur when the climate system crosses a critical threshold defining the limit between two different climate states. This definition provides a theoretical basis for understanding abrupt climate changes (e.g. Kageyama et al., in this issue) but the rapidity of the climate change