Antarctica was divided into four areas which, in temperature terms, are as dissimilar as possible; from 29 stations 4 were chosen to represent these areas. The stations are Faraday (near 65°S 65°W), Scott Base (near 77°S 167°E), Amundsen-Scott (90°S) and Mirny (near 66°S 93°E). Despite being chosen for dissimilarity, the stations were quite similar. First the interannual variability of monthly mean temperature was large. The seasonal patterns, however, were found to be stable and both mean and scale correction were required in the deseasonalization of the data. Second temperature trends over 5–10 years were, in relative terms, remarkably in phase between all stations. This coherence tends to override the smallness of these trends and underlines the fact that any long term trend would be hard to detect. Indeed, fitted linear trends over the whole 30-year data period were statistically insignificant. Finally, autocorrelation coefficients were found to be significant only over the first few mouths except at Faraday, where significance was found to 19 months. Low order autoregressive models were fitted but found to be inadequate with respect to the observed spectral structures. Therefore, despite the prediction that “greenhouse” warming will increase polar temperatures much more than elsewhere, Antarctica, with its large variability in mean temperature and relatively large short term trends, will not easily exhibit long term trends.