Single-row plots (SRP) are often used in early generation selection. For wh
eat, in such trials, it has already been demonstrated that inter-plot compe
tition leads to highly biased estimations of grain yield. An experiment inc
luding one SRP trial and one reference 6-row-plot trial were conducted over
two consecutive years in order to compare two statistical models for corre
cting competition effects. The neighbourhood covariate model relying on pla
nt height proved to be more efficient than the more general producer-compet
itor model. It provided significant improvement in yield estimation, slight
ly greater using height measured at heading than using final height as cova
riate. Moreover, relative efficiency of selection in SRP when corrected for
competition effects always reached about 85%, whatever the intensity of co
mpetition. The neighbouring covariate model based on plant height should be
used for routine analysis of single-row-plot trials.