The efficiency of single plant selection in the initial selection stage of a potato breeding programme was examined. A random sample of potato clones was grown in single plant plots and also in tworeplicates of five plant plots. After harvest, each plot was visually assessed by three potato breeders; total tuber weight and number of tubers per plant were also recorded, from which mean tuber weight was calculated. The error variances of the single plant plots were found to be significantly greater than from the five plant plots for total tuber weight, mean tuber weight and number of tubers per plant, but not significantly different for breeders' preference. Coefficients of correlation between single and five plants plots were significantly greater than zero for all traits examined, but they were lower than the corresponding correlations between the two replicates of five plant plots. Although a random sample of clones was examined, each clone had been assessed for breeders' preference the previous year. When the clones were grouped according to the previous years preference scores, it was found that the correlation coefficients between single and five plant plots for breeders' preference were inversely related, in magnitude, to the mean preference score of each group. It was concluded that single plant selection was generally ineffective, particularly when only the ‘better’ clones, from the previous years assessment, are examined.