ABSTRACTThe Great Plains experienced extreme fluctuations in precipitation and temperature throughout the Holocene, but these fluctuations have been difficult to quantify systematically across the region. Pollen has long been used as a proxy for reconstructing climate changes, but its power is limited if a region is devoid of modern pollen samples to facilitate comparison with known climate conditions. Here, we present a set of pollen‐climate transfer functions developed using weighted‐averaging partial least squares to reconstruct mean annual precipitation (MAP), mean temperature of the coldest month (MTcold) and mean temperature of the warmest month (MTwarm). At the foundation of these transfer functions is a new set of 141 modern pollen samples that specifically cover the climate space of the Great Plains. These functions quantify the relationship between pollen assemblages and modern climate in this region (r = 0.928, 0.838 and 0.897 for MAP, MTcold and MTwarm, respectively). We applied these functions to three previously established pollen records taken from lacustrine sites in the region – Fox Lake, MN; Moon Lake, ND; and Kettle Lake, ND – to reconstruct precipitation and temperature at these sites throughout the Holocene. Annual precipitation reconstructed with these transfer functions at Moon Lake and Kettle Lake does not capture a severe, prolonged period of drought during the mid‐Holocene as seen in many other proxy records from these sites and from the region. Reconciliation of these results involves both the autecology of Ambrosia and the inability of any pollen surface samples to capture seasonal and short‐term variability in precipitation. Thus, the transfer functions correctly reconstruct periodic wet summers and a disturbance factor favouring Ambrosia, but miss the effects of reduced winter precipitation and increased frequency of severe drought. The problem revealed here may be more general to biomes with higher frequency disturbance regimes and more weedy species.