Methods of palaeoclimate reconstruction from pollen are built upon the assumption that plant–climate interactions remain the same through time or that these interactions are independent of changes in atmospheric CO2. The latter may be problematic because air trapped in polar ice caps indicates that atmospheric CO2 has fluctuated significantly over at least the past 400,000 yr, and likely the last 1.6 million yr. Three other points indicate potential biases for vegetation-based climate proxies. First, C3-plant physiological research shows that the processes that determine growth optima in plants (photosynthesis, mitochondrial respiration, photorespiration) are all highly CO2-dependent, and thus were likely affected by the lower CO2 levels of the last glacial maximum. Second, the ratio of carbon assimilation per unit transpiration (called water-use efficiency) is sensitive to changes in atmospheric CO2 through effects on stomatal conductance and may have altered C3-plant responses to drought. Third, leaf gas-exchange experiments indicate that the response of plants to carbon-depleting environmental stresses are strengthened under low CO2 relative to today. This paper reviews the scope of research addressing the consequences of low atmospheric CO2 for plant and ecosystem processes and highlights why consideration of the physiological effects of low atmospheric CO2 on plant function is recommended for any future refinements to pollen-based palaeoclimatic reconstructions.