One of the most elegant mechanisms forwarded for late Quaternary atmospheric CO2 variability concerns the sensitivity of calcium carbonate preservation in deep ocean sediments to the relative delivery rates of calcium carbonate and particulate organic carbon (the CaCO3:POC “rain ratio”). It was implicitly assumed that any change in the CaCO3:POC rain ratio of biogenic material produced in the surface ocean will be communicated directly to the sediments. This would allow relatively subtle shifts in ecosystem composition to affect sedimentary CaCO3 preservation (and thus atmospheric CO2). However, recent research into the controls on the transport of POC to depth suggests that the rain ratio “seen” by the sediments may instead be buffered against any perturbation occurring at the surface. This casts doubt on the viability of hypotheses envisaging ecological changes as a means of accounting for the observed glacial-interglacial CO2 signal.
 Ice cores recovered from the Antarctic ice cap and analyzed for air bubble gas composition reveal that the atmospheric mixing ratio of CO2 (xCO2) at the height of the last glacial was only about 190 ppm, compared with 260–270 ppm at the start of the Holocene [Delmas et al., 1980; Neftel et al., 1982]. Although numerous different possible biogeochemical mechanisms for controlling CO2 have been identified, an accepted explanation for an increase of this magnitude has proved surprisingly elusive [Archer et al., 2000]. Arguably the most elegant candidate mechanism involves the influence that the ratio of biogenic calcium carbonate to particulate organic carbon (the CaCO3 to POC flux (or “rain”) ratio) has on the preservation of CaCO3 in deep sea sediments [Archer, 1991]. In order to maintain steady state (weathering inputs balanced by loss through CaCO3 burial in sediments) in the ocean carbon cycle, the effect of any perturbation able to drive an initial decrease in CaCO3 burial is ultimately compensated for by an increase in the carbonate ion concentration ([CO32−]) of the deep sea (which, in turn, equates to a reduction in atmospheric xCO2). A reduction in the production and export rate (relative to POC) of CaCO3 is one possible means of initiating such a chain of events, and forms of the basis of the “rain ratio” hypothesis [Archer and Maier-Reimer, 1994]. The operation of this is shown schematically in Figure 1a.
 Coupled ocean-sediment models have confirmed the plausibility of this mechanism, and indicate that a mean global reduction of 40% in the CaCO3:POC rain ratio at the sediment surface is sufficient to drive atmospheric xCO2 down by about 70–90 ppm [Archer and Maier-Reimer, 1994; Ridgwell et al., 2002]. This is equivalent to the entire amplitude of observed glacial-interglacial CO2 change, making this mechanism potentially key to resolving the “glacial CO2” question. How might a glacial decrease in rain ratio have arisen? While the remains of all phytoplankton species contribute toward the settling flux of POC leaving the base of the euphotic zone, it is primarily only coccolithophorids (together with their zooplankton equivalent – foraminifera) that contribute to CaCO3 export. Thus a change in the CaCO3:POC export ratio can, in theory, be brought about simply through a shift in ecosystem composition [Dymond and Lyle, 1985] - one involving a change in the relative contribution made by coccolithophorids to total ecosystem productivity. Probably the most powerful means of achieving this is via the “silicate switch” [Ridgwell et al., 2002], whereby increased silicic acid (H4SiO4) availability enables diatoms to further “out compete” smaller and more tightly grazing-controlled phytoplankton species (such as the coccolithophorids) for limiting nutrients common to all phytoplankton species (such as nitrate and iron). Just such a control has been expounded in a number of recent studies, in which lower atmospheric xCO2 is (at least partly) explained as a consequence of greater H4SiO4 availability. The larger oceanic H4SiO4 inventory necessary to achieve this could arise due to either increased Si supply rates to the ocean derived from aeolian dust [Harrison, 2000] or continental rock weathering [Tréguer and Pondaven, 2000], or due to decreased rates of Si removal by opal burial on continental shelves [Ridgwell et al., 2002]. Other modeling work has highlighted the possibility that rather than a whole-ocean change in H4SiO4 inventory, a meridonal redistribution of nutrients might have occurred. In this hypothesis, enhanced dust deposition to the Southern Ocean during glacial times enables greater (iron-driven) utilization efficiency of H4SiO4 by diatoms [Watson et al., 2000]. The result of this is an increase in the northward transport of “left over” (unutilized) H4SiO4 to the subtropics where it was previously in insufficient supply and limiting to diatom growth [Brzezinski et al., 2002; Ganeshram, 2002; Matsumoto et al., 2002].
 In each of these scenarios atmospheric xCO2 will be affected directly as a result of changes in CaCO3 export – lower CaCO3 export fluxes producing a reduction in the alkalinity gradient between surface and deep, and thus a lower partial pressure of CO2 in the surface ocean [Dymond and Lyle, 1985]. An analogous effect could also arise should POC originating from diatoms be transferred to the deep ocean more efficiently than POC originating from pico- and nano-plankton [Ganeshram, 2002], although in this case it is an increase in the dissolved inorganic carbon gradient that causes lower atmospheric xCO2. The dominant CO2 control in each glacial scenario, through, was assumed to be mediated indirectly via the “rain ratio” mechanism [Archer and Maier-Reimer, 1994]. However, research just recently published by Klaas and Archer  now casts doubt on the effectiveness of the “rain ratio” mechanism, and with it, the importance of this particular control on CO2.
 Critical to a correct understanding of the “rain ratio” mechanism is the distinction between the CaCO3:POC export ratio (i.e., that measured at the base of the euphotic zone or stratified ocean surface) and the ratio measured at the sediment surface – it is not the export rain ratio that is the critical variable in determining preservation of CaCO3 in deep sea sediments, per se, but the sediment rain ratio. Any process that affects the efficiency (relative to CaCO3) with which POC is transferred from the surface to the deep sea will therefore impact on the operation of the “rain ratio” mechanism. It seems that “ballast” minerals play such a role – by increasing mean particulate density, the presence of CaCO3, opal, and dust (terrigenous silicate minerals) in aggregates with organic matter enhance the settling rate of particles through the water column [Armstrong et al., 2002; Klaas and Archer, 2002]. Mineral surfaces and clay inter-layer sites can also help protect (primarily dissolved) organic material from bacterial attack [Kennedy et al., 2002]. Either way, the presence of mineral particles will enhance the proportion of organic matter exported from the surface that reaches deep ocean depths [Armstrong et al., 2002]. Of the three candidate ballast minerals, analysis of sediment trap data suggests that CaCO3 is by far the most important, and on a global basis is responsible for 83% of the POC flux to the ocean floor [Klaas and Archer, 2002]. If the analysis of Klaas and Archer  is correct (with opal and dust, the other possible ballast minerals, playing only a minor role) any perturbation of surface ocean ecosystem composition that results in a decrease in CaCO3 export will also drive a reduction in the efficiency with which POC is transported to depth. Thus the CaCO3:POC rain ratio seen at the sediment surface is “buffered” against changes in the export ratio, and the effectiveness of the “rain ratio” mechanism diminished (shown schematically in Figure 1b). For instance, assuming that CaCO3 is responsible for 83% of the POC flux to the ocean floor [Klaas and Archer, 2002], a 40% reduction in export CaCO3:POC would result in only a 10% decrease in rain ratio at depth. This is sufficient to account for no more than 20–25% of the total glacial drawdown in CO2 – just a quarter of the magnitude previously assumed.
 There are also implications of this “ballast buffer” in the interpretation of certain paleoceanographic proxies. The search for a reliable proxy for past changes in deep ocean [CO32−] has assumed something of “Holy Grail” status in paleoceanography research in recent years. Such a proxy would be invaluable in deciphering the mechanisms responsible for the observed glacial-interglacial changes in atmospheric xCO2 (A. J. Ridgwell et al., Implications of coral reef buildup for the late Quaternary carbon cycle, manuscript submitted to Paleoceanography, 2003). Applied over timescales beyond the current ∼400 ka limit of ice core records [Petit et al., 1999], it would also greatly help in reducing uncertainty in reconstructions of past surface ocean pCO2 (and thus atmospheric xCO2) [Pearson and Palmer, 2000], which can rely heavily on assumptions regarding past ocean chemistry. One of the leading contenders for such a [CO32−] proxy involves the analysis of foraminiferal shell weight [Broecker and Clark, 2001]. Although interpretation is complicated by a number of currently poorly understood factors [Bijma et al., 2002], correction for arguably the most important variable, that of the dependence of initial shell weight on surface environmental conditions, is now possible [Barker and Elderfield, 2002]. However, the findings of Klaas and Archer  suggests that an additional correction will be required, and the effect of changes in ballasting (thinner, lighter shells, for instance, being able to “ballast” less organic matter to the sediments, with the result that dissolution loss of CaCO3 by organic carbon metabolism will be less) be taken into account. While this does not necessarily present an insurmountable problem for this proxy, it does add a further dimension to the interpretation.
 Considerable advances have been made over the past decade in how complex biological processes in the ocean surface and diagenetic transformation in deep-sea sediments are represented and coupled together in carbon cycle models [e.g., Archer and Maier-Reimer, 1994; Heinze et al., 1999]. However, the representation of ocean interior biogeochemical processes has tended to lag behind, and the simple exponential and power law functions, previously standard in describing the transfer of biogenic material from the ocean surface to the sediments, are now no longer sufficient in many current model applications. For instance, even when ambient water column conditions of temperature and silicic acid are used to calculate a depth-specific opal dissolution rate, model predictions still perform poorly against sediment trap observations [Ridgwell et al., 2002]. In this example, it is likely that still further factors such as microbial degradation of protective organic coatings [Bidle and Azam, 1999] are important and need to be taken into account. For CaCO3, the existence of microenvironments within individual “marine snow” aggregates may be critical in determining the overall degree of dissolution that takes place in the water column [Jansen et al., 2002]. Clearly, improved understanding of how the global carbon cycle responds to perturbation and the nature of its relationship to both past and future climatic change requires further focused research effort in this area. There may be further “regime change” to come in our view of how the global carbon cycle operates.
 AJR would like to thank the School of Environmental Sciences at the University of East Anglia and the Trusthouse Charitable Foundation for financial support.