Single plot trials owing to the magnitude of the experimental errors, are practically useless as a test for comparing the yields of grain of one variety of oat with that of another. At best the results are only applicable to the particular experiment in question. For the old Scotch varieties of oat the probable error on trials of this kind amounts to about 18 per cent, of the yield of grain and to about 20 per cent, for the new varieties. Adopting single plot trials an 18 per cent, error means that when determining the superiority in the yield of grain of one variety over that of another and when the difference is not likely to be more than about 5 per cent, it would be necessary to have 214 centres with no duplication of plots at any centre in order to endeavour to obtain a conclusive result from a single year's trials. When the difference is likely to be about 10 per cent, it would be necessary to have 53 centres.

Data are supplied showing that the experimental error for different crops also for the grain and straw of the same crop is not the same. It is greatest for the grain and lowest for the hay crop.