Two techniques for estimating past variations in sea surface temperature (SST) have been used to investigate climatic change in Biogeochemical Oceanic Flux Study (BOFS) core 31K (19 ¿N, 20 ¿W) from the eastern subtropical Atlantic. High-resolution SST records for the last 28 kyr have been produced using planktonic foraminiferal assemblages, based on the Imbrie-Kipp transfer function technique, and the U37K' index derived from abundances of C37 alkenones biosynthesized by prymnesiophyte algae. Modern observations suggest that these indices reflect particular hydrographic conditions in the upper ocean: the U37K' index corresponds to the temperature at the time of maximum coccolith productivity, typically late spring-early summer in the study area today, whereas the faunal transfer function is calibrated for seasonal maximum and minimum temperatures. In general, the faunal and biomarker paleotemperature records display comparable SST variations during the last glacial and deglacial, but although the overall trends are similar, differences exist in the magnitude and timing of these temperature changes. Most notably, the faunal Twarm and U37K' SST estimates diverge by 3 ¿C between 8 ka and 6 ka, and this offset persists through the late Holocene. This difference cannot be adequately explained by uncertainties associated with either the calibration data sets or fluctuating preservation levels. We therefore propose that the deviation in SST estimates is linked to a switch in the seasonal timing of maximum coccolith production from the summer in the glacial ocean to the late spring-early summer in the modern ocean. Our results suggest that a dual approach to SST estimation based on faunal and biomarker proxies can provide a valuable means of evaluating mixed layer and productivity changes associated with the movement of oceanographic frontal zones during the late Quaternary. ¿ American Geophysical Union 1996