The question attacked is whether soil variations are sufficiently constant from year to year to give useful corrections to the yields of experimental plots from their yields under previous uniformity trials, and the data investigated were the published results of uniformity trials carried out on two fields at Aarslev (Denmark) between 1906 and 1911. In one case the plots did tend to keep constant in their relative yields, and the precision of an experiment would be increased by nearly 150 per cent. if the regression on the mean yield in the three previous years were used: with the other field, however, the plots showed no constancy in yield (when the variation due to strips was taken out as in modern experimental methods), and consequently previous uniformity trials could give no assistance.

The work described here was done whilst I was enjoying the hospitality of the Rothamsted Laboratory: it gives me much pleasure to acknowledge my great indebtedness to Dr R. A. Fisher, F.R.S., who suggested the problem, and guided my unsteady steps throughout.