Jenkins, Myerson, Hale, and Fry (1999) showed that slopes relating complex spans to simple spans were considerably smaller than one, indicating that persons with higher simple spans suffered more interference when the span task was combined with a processing demand. They argued that this finding ruled out accounts of working memory based on interference and/or inhibition of interfering information. We demonstrate that the effect is mainly an artifact from regression to the mean, owing to the low reliability of span scores as used by Jenkins et al. Data from 133 young adults for two verbal and two spatial span tasks show that the slopes relating complex to simple performance are considerably higher for sum scores than for span scores. Furthermore, an adequate test for an interference or an inhibition account of working memory is to predict interference from complex span tasks, not from simple span tasks. Interference effects in the verbal span tasks were negatively correlated with an independent measure of working memory capacity, consistent with the interference/inhibition account.