Family selection has recently been adopted in some sugarcane breeding programs. It is used as a method of indirect selection of superior clones in early stages of selection among seedling populations. In sugarcane a potential advantage of family selection compared with individual clonal selection is that performance across several sites may be assessed in the earliest stages of selection. This may be advantageous if genotype × environment interactions are large in the targeted region. In this study, 48 sugarcane crosses (families) were evaluated at seven sites in the Herbert Region of North Queensland, Australia. A plant and one or two ratoon crops were grown at each site. The aims were to determine the magnitude and nature of family × environment interactions in order to develop optimal methods for selecting among sugarcane families in breeding programs. There was large family × site interaction for cane yield and sugar yield, with the component of variance for each being of similar magnitude to that for family main effects. Family × crop-year and family × crop-year × site interactions were also significant for cane yield and sugar yield but were small compared with family × site interactions and family main effects. For commercial cane sugar (CCS), all interaction components were small compared with the main effects of families. There was a small negative correlation (−0.24) between CCS and cane yield for family main effects, and a larger negative correlation between these two components for family × site effects. Estimates of predicted responses to selection indicated that the testing of families at two sites with three replicates per site would represent near optimal allocation of resources. An optimal selection index, balancing the relative economic importance of cane yield and CCS (related to commercially extractable sugar content), the relative heritabilities of these two components and their genetic correlation, was estimated as 12 × CCS + cane yield (t/ha). This index had a high genetic correlation with sugar yield and net merit grade (a selection criterion commonly used for selection in Australian sugarcane breeding programs, based on sugar yield relative to standard cultivars). This suggests that selection criteria based on sugar yield or net merit grade should be acceptable in most situations.