Using an online sample experiment, the study described herein addresses the fervent debate about the relationships between working memory and long-term memory (LTM). We manipulated refreshing opportunities and LTM status of memoranda by varying, respectively, the type of span tasks (simple span with short or long lists versus complex span tasks) and the frequency of the memory words (low versus high). In accordance with the hypothesis that refreshing is involved particularly in complex rather than simple span tasks, the frequency effect in immediate recall tests was reduced in the former. Moreover, contrary to previous studies in which refreshing increases LTM effects in delayed recall tests, our data point in the opposite direction. However, the frequency effect was also reduced in the simple task with short lists, suggesting that refreshing might not be the only process underlying the reduction of frequency effect in delayed tests. Finally, no differences in delayed recall were found between the complex span task that affords refreshing opportunities and the other tasks, suggesting that another process than refreshing, probably consolidation, might be involved in delayed recall.