Anomalies from the respective decadal means of the monthly mean temperatures for central England covering a period of 250 years are ranked to form a quintile distribution. Contingency tables are then prepared which show the percentage frequencies of the changes in the temperature anomalies from one calendar month to the following calendar month for each quintile. Correlation coefficients for a lag of one month are computed for each calendar month-to-month change throughout the years, and for lags 1–12 months inclusive for the whole data series.
It is shown that persistence of extreme anomalies at certain times of the year are double the frequency which would be expected by chance, while the changes from one extreme quintile to the other occur with very low frequency compared with chance.
It is suggested that application of the tables would produce results as good, and perhaps a little better, at forecasting temperature anomalies for a month ahead than the official long-range forecasts supplied by the Meteorological Office for central and southeast England.