Data from three ryegrass spaced plant variety trials conducted at Crossnacr
eevy, Co. Down in each of 1989, 1990 and 1991 were investigated for evidenc
e of spatial dependence. Spatial dependence was most evident in variates me
asuring the overall dimensions of the plants, especially late season variat
es. However, the 'date of ear emergence' variate, which identifies most of
the variety distinctnesses that the trials are intended to detect, was not
spatially dependent. On average, 45 % of the residual variation in the plot
means of the spatially dependent variates was estimated to be of a spatial
nature. Consequently, it is likely that the efficiency of analysis of thes
e variates would be improved by using spatial analysis instead of non-spati
al analysis or by using alpha (incomplete block) designs instead of complet
e block designs. The spatial information was used to determine the optimal
incomplete block sizes and the likely increase in efficiency due to using a
lpha designs instead of complete block designs. The optimal incomplete bloc
k size for a trial with 10 plants per plot was compared with the optimal in
complete block size for the same trial with single plant plots at the same
spacing and was found to be smaller. The gains in efficiency were larger in
the trial with 10 plants per plot than in the trial with single plant plot
s.