Four experiments examined the nature of forgetting and the processing--storage relationship during performance on a prevalent working memory task, the reading span test. Using two different presentation paradigms, Experiments 1 and 2 replicated Towse, Hitch, and Hutton's (1998, 2000) finding that the Short-Final lists, which presented a long sentence first and a short sentence last, led to better recall performance than the reverse-order Long-Final lists. This effect was still obtained when the retention duration for the target words was held constant and the amount of sentence processing required during that interval was varied (Experiment 3). However, the effect disappeared when the retention duration was varied while holding constant the amount of sentence processing required (Experiment 4). These results suggest that the amount of processing activities, not the sheer passage of time, may be the critical factor underlying the sentence order effect, thereby challenging purely time-based explanations of forgetting during reading span performance. In addition, the analysis of reading times (Experiment 1) revealed that the number of memory items had a subtle yet reliable negative effect on reading times, suggesting that the processing and storage requirements of the reading span test are not completely independent.