The ability of the [15-O] water PET activation technique to localize statistically significant changes in regional cerebral blood flow is dependent on activity injected and task performed. Statistically insignificant changes may occur if insufficient activity is injected, of the task results in an undetectable change in blood flow. For a given PET scanner and level of injected activity, increasing the cognitive load of a task may result in larger blood flow changes. In a series of studies with the EXACT HR+, a state-of-the-art high resolution PET tomograph, the authors explored the relationship between injected activity and the statistical significance attained with a cognitive task that imposed increasing working memory demand. Activity levels of 8-15 mCi are typically injected to optimize HR+ performance in 3D. For this study, 5 mCi, 10 mCi and 15 mCi of [15-O] water were injected into seven normal volunteers. At each activity level the volunteers were required to perform four different N-back tasks in which a letter displayed on a TV monitor is matched with the letter displayed N letters previously. With increasing N, this task places increased load on working memory. The complete sequence of four tasks (N=0, 1, 2, 3) at three activity levels was repeated twice, for a total of 24 injections. The authors show that robust activation maps can be obtained with as little as 5 mCi injected dose, even though the maximum noise equivalent count rate for the HR+ is attained at injected activity levels of [15-O] water in excess of 15 mCi.