We present a new speleothem record of atmospheric Δ 14 C between 28 and 44ka that offers considerable promise for resolving some of the uncertainty associated with existing radiocarbon calibration curves for this time period. The record is based on a comprehensive suite of AMS 14 C ages, using new low-blank protocols, and U–Th ages using high precision MC-ICPMS procedures. Atmospheric Δ 14 C was calculated by correcting 14 C ages with a constant dead carbon fraction (DCF) of 22.7±5.9%, based on a comparison of stalagmite 14 C ages with the IntCal04 (Reimer et al., 2004) calibration curve between 15 and 11ka. The new Δ 14 C speleothem record shows similar structure and amplitude to that derived from Cariaco Basin foraminifera (Hughen et al., 2004, 2006), and the match is further improved if the latter is tied to the most recent Greenland ice core chronology (Svensson et al., 2008). These data are however in conflict with a previously published 14 C data set for a stalagmite record from the Bahamas — GB-89-24-1 (Beck et al., 2001), which likely suffered from 14 C analytical blank subtraction issues in the older part of the record. The new Bahamas speleothem ∆ 14 C data do not show the extreme shifts between 44 and 40ka reported in the previous study (Beck et al., 2001). Causes for the observed structure in derived atmospheric Δ 14 C variation based on the new speleothem data are investigated with a suite of simulations using an earth system model of intermediate complexity. Data-model comparison indicates that major fluctuations in atmospheric ∆ 14 C during marine isotope stage 3 is primarily a function of changes in geomagnetic field intensity, although ocean–atmosphere system reorganisation also played a supporting role.