### abstract

- Advance in crop genomics have led to a demand for more accurate and precise phenotyping of a large number of genotypes for discovering genes and mechanisms underpinning important agronomic traits. The use of unbordered plots for field phenotyping is one approach for reducing the resources needed, but competition effects between neighboring lines may confound varietal performance. Four field experiments were conducted in Benin to examine whether grain yields of 14 diverse upland rice (Oryza sativa spp.) varieties determined in unbordered one-row or two-row plots differ from those measured in self-bordered four-row plots and to examine if statistical models including covariates based on plant characteristics (height, panicle number, and days to heading) for correcting competition effect can improve the estimation of the yield in unbordered plots. Mean grain yield across all varieties ranged from 118 to 378 g m(-2) in four experiments. There was no significant variety ' row number interaction effect on grain yield, except for the highest yielding experiment. In that experiment, the variety ' row number interaction was significant for one-row versus four-row plots, but not for two-row versus four-row plots. In one-row plots in this high-yielding experiment, the neighborhood covariate model based on panicle number improved residual mean square by 20%, but relative selection intensity by 3% only. Similarly, the covariate models based on height or panicle number in both one-and tworow plots in the other experiments improved just 4%. We conclude that unbordered one-or tworow plots can provide reasonable estimates of grain yield of upland rice without any bias due to competition effects, except for high-yielding one-row plots (>350 g m(-2)).
- Advance in crop genomics have led to a demand for more accurate and precise phenotyping of a large number of genotypes for discovering genes and mechanisms underpinning important agronomic traits. The use of unbordered plots for field phenotyping is one approach for reducing the resources needed, but competition effects between neighboring lines may confound varietal performance. Four field experiments were conducted in Benin to examine whether grain yields of 14 diverse upland rice (Oryza sativa spp.) varieties determined in unbordered one-row or two-row plots differ from those measured in self-bordered four-row plots and to examine if statistical models including covariates based on plant characteristics (height, panicle number, and days to heading) for correcting competition effect can improve the estimation of the yield in unbordered plots. Mean grain yield across all varieties ranged from 118 to 378 g m−2 in four experiments. There was no significant variety × row number interaction effect on grain yield, except for the highest yielding experiment. In that experiment, the variety × row number interaction was significant for one-row versus four-row plots, but not for two-row versus four-row plots. In one-row plots in this high-yielding experiment, the neighborhood covariate model based on panicle number improved residual mean square by 20%, but relative selection intensity by 3% only. Similarly, the covariate models based on height or panicle number in both one- and two-row plots in the other experiments improved just 4%. We conclude that unbordered one- or two-row plots can provide reasonable estimates of grain yield of upland rice without any bias due to competition effects, except for high-yielding one-row plots (>350 g m−2).