C4 photosynthesis had a mid-Tertiary origin that was tied to declining atmospheric CO2, but C4-dominated grasslands did not appear until late Tertiary. According to the ‘CO2-threshold’ model, these C4 grasslands owe their origin to a further late Miocene decline in CO2 that gave C4 grasses a photosynthetic advantage. This model is most appropriate for explaining replacement of C3 grasslands by C4 grasslands, however, fossil evidence shows C4 grasslands replaced woodlands. An additional weakness in the threshold model is that recent estimates do not support a late Miocene drop in pCO2. We hypothesize that late Miocene climate changes created a fire climate capable of replacing woodlands with C4 grasslands. Critical elements were seasonality that sustained high biomass production part of year, followed by a dry season that greatly reduced fuel moisture, coupled with a monsoon climate that generated abundant lightning-igniting fires. As woodlands became more open from burning, the high light conditions favoured C4 grasses over C3 grasses, and in a feedback process, the elevated productivity of C4 grasses increased highly combustible fuel loads that further increased fire activity. This hypothesis is supported by paleosol data that indicate the late Miocene expansion of C4 grasslands was the result of grassland expansion into more mesic environments and by charcoal sediment profiles that parallel the late Miocene expansion of C4 grasslands. Many contemporary C4 grasslands are fire dependent and are invaded by woodlands upon cessation of burning. Thus, we maintain that the factors driving the late Miocene expansion of C4 were the same as those responsible for maintenance of C4 grasslands today.