Taro (Colocasia esculenta) breeding, as other root crop breeding, is based on the production and evaluation of large numbers of hybrids. The selection of parents is based on their phenotypic value in the absence of information concerning general combining ability (GCA), specific combining ability (SCA), or genetic distances between varieties. By combining data from heritability trials and from genetic diversity studies conducted with AFLP and SSR markers, we aimed at studying the relationship between hybrid vigour and genetic dissimilarity between parents. The traits studied included number of suckers, corm weight, corm dimensions, and dry matter content. Correlation coefficients between hybrid gain and dissimilarity values were calculated. The prediction of hybrid performance based on the mid-parent value was compared to the prediction based on a modified expression that takes into account the genetic relationships between parents. Correlations were all but one positive but not statistically significant for all traits, with the exception of the number of suckers, when using SSR markers for dissimilarity calculations. Accordingly, the genetic dissimilarities in the prediction of hybrid performances did not increase the correlation between predicted and observed hybrid vigour values. However, large differences were observed among the residual means from the regression between predicted and observed values when using AFLP or SSR markers, mainly due to the much higher polymorphism revealed by the latter. Models need to be further adapted to the type of molecular marker used, since their ability to reveal different rates of polymorphism will have a direct incidence on the calculation of genetic dissimilarities between genotypes. Nevertheless, since SSR markers are more polymorphic and more informative than AFLP markers, they should be preferentially used for these studies. Low genetic dissimilarity of parents yielded weak heterosis effects and future studies need to be