We present measurements of the stable carbon isotope ratio in air extracted from Antarctic ice core and firn samples. The same samples were previously used by Etheridge and co-workers to construct a high precision 1000-year record of atmospheric CO2 concentration, featuring a close link between the ice and modern records and high-time resolution. Here, we start by confirming the trend in the Cape Grim in situ δ13C record from 1982 to 1996, and extend it back to 1978 using the Cape Grim Air Archive. The firn air δ13C agrees with the Cape Grim record, but only after correction for gravitational separation at depth, for diffusion effects associated with disequilibrium between the atmosphere and firm, and allowance for a latidudinal gradient in δ13C between Cape Grim and the Antarctic coast. Complex calibration strategies are required to cope with several additional systematic influences on the ice core δ13C record. Errors are assigned to each ice core value to reflect statistical and systematic biases (between ± 0.025‰ and ± 0.07‰); uncertainties (of up to ± 0.05‰) between core-versus-core, ice-versus-firn and firn-versus-troposphere are described separately. An almost continuous atmospheric history of δ13C over 1000 years results, exhibiting significant decadal-to-century scale variability unlike that from earlier proxy records. The decrease in δ13C from 1860 to 1960 involves a series of steps confirming enhanced sensitivity of δ13C to decadal timescale-forcing, compared to the CO2 record. Synchronous with a ‘‘Little Ice Age’′ CO2 decrease, an enhancement of δ13C implies a terrestrial response to cooler temperatures. Between 1200 AD and 1600 AD, the atmospheric δ13C appear stable.