The alkenone-pCO2 method is one of the most widely used approaches to reconstruct atmospheric CO2 in the Cenozoic. The method depends upon fractionation of stable carbon isotopes during algal photosynthesis, expressed as ep37:2, and a physiological scaling parameter, b, that accounts for biological factors such as growth rate, cell size, and membrane permeability. Alkenone-derived CO2 records for the late Pleistocene, however, are poorly correlated with ice core CO2, challenging the classic model that considers most of the CO2 used for coccolithophore photosynthesis to be acquired through simple diffusion. In this study, we investigate the nature of the b term and the underlying patterns of the sensitivity of ep37:2 to pCO2 changes. We generated two new ep37:2 records from the South China Sea (MD01-2392) and tropical Atlantic Ocean (ODP 668B) and compiled other published ep37:2 records over glacial-interglacial cycles. Using the ep37:2 data, ocean temperature estimates, and ice core CO2, we were able to back-calculate the corresponding values of b. At all locations, b varies over glacial cycles. The highest values of b correspond to peak interglacial stages, indicating that the phytoplankton growth rate is faster or cell size is smaller during interglacials than during glacial periods. We further show that the range of ep37:2 between glacial and interglacial conditions, Dep37:2, scales with growth conditions, consistent with the predictions of the carbon isotope fractionation model based on CO2 diffusion. In other words, the sensitivity of ep37:2 to pCO2 changes increases where the modern b values are large, contradicting the recommendations that oligotrophic sites are the best for alkenone-CO2 applications because of the presumed stability of b. Using the average back-calculated b value for each site, the composite pCO2 estimates from MD01-2392 and ODP 668B – the two sites with adequate Dep37:2 sensitivity – show broad agreement with the ice core CO2 record.