AFLP markers were evaluated for their usefulness in the genetic analysis of sugarbeet and wild Beta species. Accessions of ten different sugarbeet breeding lines and five wild beets were screened using 256 primer combinations. Of the 11 309 bands investigated, 96.4% were polymorphic among the accessions. A strong positive correlation was found between the number of polymorphisms and AT content of the selective bases of the primer combinations. Random subsets of primer combinations were used to produce genetic distance trees. Permutation tests showed that, for the wild beets, 500 AFLP bands sufficed to obtain the best topology of the tree with a probability at any given node of more than 99%. Ten times as many bands were necessary to obtain support values of the same order of magnitude for the sugarbeet lines. The reproducibility of AFLP for seven primer combinations was investigated by repeated analysis of all steps from DNA isolation to data scoring. For 5088 comparisons, the overall reproducibility was 97.6%. Robustness to genotyping errors was investigated by including an artificial F1 (1 : 1 DNA mixture) of two sugarbeet lines in the screen for polymorphisms. For the 3160 cases of polymorphism between the two lines, 0.2% genotyping errors were found. The general reliability and usefulness of AFLP markers are discussed in relation to the results obtained.