Sixty-one paired radiocarbon-racemization (AAR) analyses of Holocene and Pleistocene Mercenaria from beach and shelf sites between New Jersey and Florida provide rigorous calibration of Holocene racemization rates and an aminostratigraphic definition of the regional Holocene-Pleistocene boundary. These data are useful for geologic framework studies in the coastal zone, and also yield insights into paleoclimate, relative sea level history, and possible transport histories of the dated samples.

Most of the racemization data are consistent with 14C results, deviations being explained by heating and/or contamination. In spite of careful sampling and AMS analysis, some shells that are definitely Pleistocene by AAR yield finite 14C dates (30-40kyr), and some shells thought to be Holocene based on AAR returned near infinite 14C dates.

Latitude trends of rate constants for Holocene samples suggest an activation energy of 25-28 kcal/mole, consistent with experimental observations. A subtle offset in apparent rates for samples from southeastern North Carolina suggests a 1.5° difference in effective water temperature, consistent with modern ocean temperature differences in this region. This offset could be incorporated into correlation of Pleistocene aminozones along the coastal plain, although the Quaternary temperature history of emergent units is expected to be dominated by terrestrial rather than marine paleoclimate.

Comparison of Holocene kinetics with results for Pleistocene shells permits estimates of late Quaternary temperature changes for the region. These estimates are both model and calibration dependent. Using an incremental parabolic model that computes D/L values in 1 kyr intervals for both low and high frequency temperature variations, we estimate extreme late Pleistocene temperature reductions of 12° for eastern North Carolina, with average full glacial temperature reductions of about 8°. This same modeling approach for AAR data from southern California mollusks with better chronological control yields extreme and average temperature reductions of 6 and 3.5°, respectively, providing a good test of the modeling approach.