N. Shackleton (1977) first proposed that changes in the marine δ13C record (Δδ13C) could be used to infer ice age changes in carbon storage on land. The previously published best estimate from the marine record is equivalent to about 490 Gt (0.32 Δδ13C). However, Adams et al. (1990) utilized a pollen database to estimate a 1350 Gt change in carbon storage, which would cause a Δδ13C of about 0.90‰. The nearly trillion ton difference in estimates amounts to almost half of the total carbon stored on land. To address the nature of this discrepancy, I have reexamined the terrestrial carbon record based on a new pollen database compiled by R. Webb and the Cooperative Holocene Mapping Project (COHMAP) group. I estimate about 750–1050 Gt glacial-interglacial change in terrestrial carbon storage, with the range reflecting uncertainties in carbon storage values for different biomes. Additional uncertainties apply to rainforest and wetland extent and presence of C4 plants, which have a significantly different isotopic signature than C3 plants. Although some scenarios overlap a new estimate of carbon storage based on the oceanic Δδ13C record (revised to 0.40‰ and 610 Gt), most estimates seem to fall outside the envelope of uncertainty in the marine record. Several possible explanations for this gap involve: (1) a missing sink may be involved in land-sea carbon exchange (e.g., continental slopes); (2) the geochemistry of the exchange process is not understood; (3) carbon storage by biome may have changed under ice age boundary conditions; or (4) the standard interpretation of whole ocean changes in the marine δ13C record requires reevaluation. This latter conclusion receives some support from comparison of the δ13C records for δ18O Stages 2 and 6. For the Stage 6 glacial, the δ13C changes are 50–60% larger than for the Stage 2 glacial. Yet implications of increased aridity are not supported by longterm trends in atmospheric dust loading. To summarize, the above analysis implies that, despite the uncertainties remaining in estimates of terrestrial carbon storage changes, a case can be made that our understanding of the transfer process is incomplete and that the eventual explanation may require clarification of factors affecting the marine δ13C record.