The increasing precision of laboratory measurements in radiocarbon dating renders it more than ever important that only reliable samples with very firm associations should be dated, and that laboratory pretreatments should be completely effective. Four key aspects are discussed in detail: Past variations in atmospheric C-14 and their investigation by other absolute dating methods; laboratory measurements; sample selection; and limitations. All calibrations should be regarded as approximations until there are reliable means of correlating radiocarbon and calendar years.