The Variation of soybean seed protein content has been investigated in
performance trials of maturity group 000 to I genotypes at different
locations in the eastern part of Austria between 1992 and 1994. In all
experiments evaluated, significant portions of spatial Variation were
identified which affected seed protein content in the range of -25 to
+ 30 g/kg. Lattice and neighbour analysis were more efficient than th
e randomized complete block design in reducing error variance in most
experiments. Neighbour analysis was superior to the lattice design whe
n soil trends were changing over short distances or in both longitudin
al and latitudinal directions. In most instances, neighbour analysis b
ased on two or three neighbouring plots at each side of a test plot wa
s more efficient than an analysis using only the nearest neighbours. R
esiduals derived from neighbour analysis were also used to visualize s
patial heterogeneity, and clear trends of protein content were identif
ied for particular experimental sites. The ranking of genetic entries
was considerably biased by spatial variations, which would have reduce
d the efficiency of selection for high protein content if genotype per
formance was not adjusted for field variation.