A uniformity trial on groundnuts has been analysed and the results discussed. Plant number is less variable than yield and less sensitive to shape of plot and of block. For yield, long narrow plots are more efficient than shorter and wider plots, in all shapes and sizes of blocks and in Latin squares. The plots should not be arranged end to end along the contours, but side by side, either singly or in pairs, forming compact blocks.

The regression of the plot variance of the mean yield per unit area on size of plot approximately follows a linear logarithmic relationship. A similar relationship holds for plant number. The value of the regression coefficient b′ is low and it has been shown that, as expected, there is considerable gain from the use of small blocks. The efficiencies of various confounded and incomplete block designs relative to designs in larger blocks have been determined for some particular layouts, and values for other layouts, ignoring shape of plots and of blocks, have been obtained by interpolation.

The field used for this uniformity trial appears equally variable in all directions.