1. A comparison has been made between progeny tests in special stations and those available from field records. The variation between sires in milk yield is much greater at the stations even when they are compared with herds at the same production level.

2. In respect of yield the repeatability of station tests in the field is not good. The regression of field results on station test is only about 0·2 in herds at all production levels. For fat percentage, the regression is 0·5.

3. The tests in special stations (averaging seventeen daughters per bull) were compared with those in high herds (average twenty-seven daughters) in their accuracy of prediction of the performance of bulls in the poorer herds. The field test was slightly (but not significantly) better for both milk yield and fat content.

4. In milk yield, there is undoubtedly some extra variation between progeny groups at the stations which is not repeated in the field. Perhaps the tying of the heifers in distinct daughter groups in the byre may be a cause of this. However, even without this, it is doubted whether the test stations can give as much useful information on the numerical aspects of performance as the field records usually available. It is suggested that the splitting of daughter groups within the byre might increase the accuracy of the special station in the assessment of yield.