To explore the actual behavioral performance of subjects in multitasking training games, we designed a VR game including a Target-tracking task (TTT) of continuously moving "Player" to track "Targets" and a randomly appearing Color-discrimination task (CDT) requiring discriminating whether "Player" and "Monster" have the same color, and recorded subjects' pupillary changes to reflect mental effort. By analyzing the mean pupil diameter change (MPDC) of different groups, we found that the high group presented pupil dilation during the post-event stage, reflecting that they engaged in psychological processing of CDT during the event, whereas the low group had no pupil dilation during part of the post-event stage, reflecting the possibility of ignoring the appearance of CDT, and such behaviors hardly raise good expectations for training effect. Our study suggests that MPDC mirrors not only the actual behavior of the different groups treating the multitasking paradigm, but also the influence of game design.