Objective: Given previous research examining the relationship between physical activity (PA) and working memory in children has not taken into account of potentially confounders, this study specifically controlled for the effects of intelligence and fitness-related factors (i.e., aerobic and motor fitness) on this relationship. Method: Thirty-two children were assigned to either the higher PA group (HP) (n = 16) or lower PA group (LP) (n = 16) based on their amount of PA as measured by an accelerometer. Working memory was measured by the delayed-matching test, with concurrent recordings of event-related potentials including the P3 and the positive slow wave (PSW) components. Results: Data analysis controlling for the effects of potential confounders found that the HP group had higher response accuracy in both task conditions relative to the LP group. For reaction times (RT), the HP group showed shorter RT than the LP group in the delayed condition. Furthermore, the HP group had similar P3 amplitudes across task conditions, whereas amplitudes in the delayed condition were smaller than that in the non-delayed condition in the LP group. In addition, the HP group also showed larger PSW amplitudes than the LP group. Supplementary partial correlations analysis indicated negative correlations between levels of PA with RT in both task conditions, and positive correlations with P3 amplitude and PSW amplitude in the delayed condition. Conclusions: Children with higher levels of PA, while controlling for individual differences in intelligence and fitness-related factors, have better working memory as assessed by both behavioral and neuroelectric measures.