Previous error detection research focused on the effectiveness of different checking methods. In this paper, we focus on the psychological mechanisms on error detection. We conceptualize working memory (WM) as a critical cognitive component in error detection and two studies were carried out to investigate the effects of WM load and capacity on error detection performance and the detection of different error types. Study I found a significant interaction effect of WM load × capacity: low WM capacity participants performed significantly worse in higher WM load condition, however, high WM capacity participants' performances were unaffected by higher WM load. Study II employed think-aloud technique to gain insights into detectable error types and generated novel predictions about the effect of WM demands on detecting different errors. These predictions allow for a new research direction in error detection. Copyright is held by the author/owner(s).