The link between atmospheric CO2 levels and global warming is an axiom of current public policy, and is well supported by physicochemical experiments, by comparative planetary climatology and by geochemical modelling. Geological tests of this idea seek to compare proxies of past atmospheric CO2 with other proxies of palaeotemperature. For at least the past 300 Myr, there is a remarkably high temporal correlation between peaks of atmospheric CO2, revealed by study of stomatal indices of fossil leaves of Ginkgo, Lepidopteris, Tatarina and Rhachiphyllum, and palaeotemperature maxima, revealed by oxygen isotopic (δ18O) composition of marine biogenic carbonate. Large and growing databases on these proxy indicators support the idea that atmospheric CO2 and temperature are coupled. In contrast, CO2–temperature uncoupling has been proposed from geological time–series of carbon isotopic composition of palaeosols and of marine phytoplankton compared with foraminifera, which fail to indicate high CO2 at known times of high palaeotemperature. Failure of carbon isotopic palaeobarometers may be due to episodic release of CH4, which has an unusually light isotopic value (down to –110%, and typically –60%δ13C) and which oxidizes rapidly (within 7–24 yr) to isotopically light CO2. Past CO2 highs (above 2000 ppmv) were not only times of catastrophic release of CH4 from clathrates, but of asteroid and comet impacts, flood basalt eruptions and mass extinctions. The primary reason for iterative return to low CO2 was carbon consumption by hydrolytic weathering and photosynthesis, perhaps stimulated by mountain uplift and changing patterns of oceanic thermohaline circulation. Sequestration of carbon was promoted in the long term by such evolutionary innovations as the lignin of forests and the sod of grasslands, which accelerated physicochemical weathering and delivery of nutrients to fuel oceanic productivity and carbon burial.