Two experiments were conducted to determine how performance is influenced by intellectual abilities of learners and by similarity between training and transfer stimuli. Participants trained for automatization by repeatedly performing a task on the same stimuli, and then worked on the same task as in training but with novel stimuli;In Experiment 1, gifted and average-ability youth and college students trained on alphabet arithmetic (A + 4 = E) and worked on transfer equations: commutativity (4 + A = E), reflexivity (E = A + 4), subtraction (E - A = 4), and add-one (A + 5 = F). Response-time data were linearly regressed on digits of equations. A decreasing pattern in slopes would indicate a shift from counting to remembering and was used as evidence for automatization. During training, all groups exhibited parallel shifts from counting to remembering, suggesting that these groups had comparable abilities for automatization. For all groups, slopes for commutativity, reflexivity, and subtraction were comparable to slope for final training block, suggesting that all groups were capable of using memory traces that accumulated during training in order to respond to novelty. For gifted youth and college students, slope for add-one was comparable to slope for final training block; for average-ability youth, slope for add-one was steeper than slope for final training block. Thus, gifted youth and college students utilized available memory traces during transfer to add-one, but average-ability youth did not;In Experiment 2, gifted youth and college students trained on number and object comparisons (21 21; MOUSE larger than COIN), and then worked on new comparisons (less-than and smaller-than comparisons not included in training). Response-time data (RT) of each participant were fitted to the power law: RT = a+bN-c, (N = number of trials). The parameter c measured rate of automatization. The two groups had comparable rates of automatization and parallel profiles. With numbers, c did not change across different stages. With objects, c did not change from training to reversed stage, but increased from reversed to new stage. Thus, increased efficiency in magnitude comparison was evident with objects but not with numbers;Implications of the current research to intellectual precocity and to the instance theory of automaticity are discussed.