This paper describes a uniformity trial with irrigated broadcast rice and shows the large inter- and intra-field variation which exists.

The standard deviations decrease with the increased size of plots and vary from 20–1 per cent, with 1/436 acre plots to 7–8 per cent, with 1/22 acre plots.

The decrease of the standard deviation is small with plots larger than 1/87 acre. This size of plot is considered to be the most practicable for field trials.

The influence of proximity to bunds, irrigation channels and drains on increasing yields is discussed and methods of reducing errors due to this cause are suggested.

It is shown that the use of a Latin square extending over more than one field has not produced a more accurate experiment than the use of randomised blocks. The latter method would therefore appear to be the more suitable for manurial and other trials, under the particular conditions of rice growing.