There has been great interest in using working memory (WM) training regimens as an alternative treatment for ADHD, but it has recently been concluded that existing training regimens may not be optimally designed because they target the primary memory component but not the secondary component of WM capacity. This conclusion requires the ability to accurately measure changes in primary and secondary memory abilities over time. The immediate free recall task has been used in previous studies to measure these changes; however, one concern with these tasks is that the recall order required on training exercises may influence the recall strategy used during free recall, which may in turn influence the relative number of items recalled from primary and secondary memory. To address this issue, previous training studies have explicitly controlled recall strategy before and after training. However, the necessity of controlling for recall strategies has not been explicitly tested. The present study investigated the effects of forward-serial-order training on free recall performance under conditions in which recall strategy was not controlled using a sample of adolescents with ADHD. Unlike when recall order was controlled, the main findings showed selective improvement of the secondary memory component (as opposed to the primary memory component) when recall order was uncontrolled. This finding advances our understanding of WM training by highlighting the importance of controlling for recall strategies when free recall tasks are used to measure changes in the primary and secondary components of WM across time.