Records for milk, fat, solids-not-fat and fat percentage over forty-eight complete lactations were used as uniformity data to determine the relative efficiencies of a variety of experimental designs used in dairy husbandry experiments.

In general, it was found that:

1. Experiments conducted during the earlier part of the lactation are more efficient than those conducted during the later part.

2. Three-weekly periods are generally more efficient than 5-weekly periods.

3. If 3-weekly periods are used, then extending the time of the experiment over four such periods instead of three reduces the error. The converse is the case with 5-weekly periods since the experiment is carried over into the highly variable later part of the lactation.

4. Grouping the animals according to the yields recorded during the first 40 days of lactation does not appear to reduce the error.

An analysis of covariance using the pro-experimental records of the performance of the animals has been carried out on ordinary group trials and it is hoped to publish the results at a later date, but investigations on the data have shown that both reversal and Latin square designs are more efficient than ordinary group trials even when such an analysis is performed.