Many studies have shown that reaching movements to visual targets can rapidly adapt to altered visual feedback of hand motion (i.e., visuomotor rotation) and generalize to new target directions. This generalization is thought to reflect the acquisition of a neural representation of the novel visuomotor environment that is localized to the particular trained direction. In these studies, participants perform movements to a small number of target locations repeatedly. However, it is unclear whether adaptation and generalization are comparable when target locations are constantly varied and participants reach to visual targets one time only. Here, we compared performance for reaches to a 30° counter-clockwise visuomotor rotation to four targets, spaced 90° apart across four areas of workspace 18 times each (repeated practice (RP)) with one time only reaching movements to 72 targets, spaced 5° apart (varied practice (VP)). For both training groups, participants performed 18 reaches to radial targets (either at the repeated or varied location) in a specific area of the workspace (i.e., one of four quadrants) before reaching in the adjacent workspace. We found that the RP group adapted more completely compared to the VP group. Conversely, the VP group generalized to new target directions more completely when reaching without cursor feedback compared to the RP group. This suggests that RP and VP follow a mainly common pattern of adaptation and generalization represented in the brain, with benefits of faster adaptation with RP and more complete generalization with VP.