The present experiment was designed to complete a series of ATI research on establishing a method for determining the optimal instructional treatment when two aptitude informations and multiple treatments are combined. Under four types of instructional treatments, Text-Numerical, Text-Graphical, Program-Numerical, and Program-Graphical, junior high school pupils were taught a learning task of understanding and applying a formula to summate arithmetic progression. Six aptitudes were measured including specific and general ones. By revising the pretest and the measure of payoff used in the preceding experiment, more decisive conclusions were drawn that specific aptitude informations were more advantageous in yielding significant ATIs than general ones, quite contrary to the result of the similar preceding experiment. Possible causes of this discrepancy were discussed in relation to the subtlety ATI has.