Research has shown that mental rotation (MR) can be improved through training. However, studies with preschool children are very scarce, due in part to the lack of consensus as to the age at which this ability arises and can be trained, and due to the difficulties of working on the understanding of this ability when it begins to develop. The present study was designed to observe the effect of an MR training on 1st (3–4-year-old children) and 3rd year (5–6-year-old children) of Early Childhood Education (preschool), as well as the development of this ability between both courses. Finally, this study aimed to analyze the differential increase of the training effect in relation to the initial MR ability of the participants. The results showed a significant improvement in the participants who underwent training in 3rd year of preschool, with the trained group showing a marginal improvement in 1st year of preschool. The older group showed lower error rates in training performance than the younger group, the latter having a linear decrease in performance as the angular disparity increased. In addition, in relation to training, a greater increase of MR was observed in the 3rd year preschoolers with lower scores in the pretest. These results suggest that MR is in full development and that it is a spatial ability that can be trained at preschool ages. In addition, the possibility of enhancing this ability to a greater extent in preschoolers who exhibit lower initial MR level is especially relevant.