<p>The process of making perceptual information resistant to distraction in working memory (WM) is known as WM consolidation. The nature of WM consolidation is poorly understood, including basic aspects such as its speed and what can or cannot interrupt it. To evaluate the speed of consolidation, studies often use a dual-task approach with a WM sample (T1) preceding a sensorimotor decision (T2) by a variable interval. It is unclear what aspects of T1 and T2 lead to interference, particularly the retroactive interference that provides a cleaner measure of consolidation than proactive interference. These observations motivated the current goal of establishing the boundary conditions for the retroactive interference effect—that is, determining in what circumstances WM consolidation can be interrupted by a subsequent task. We investigated these boundary conditions via a series of modified versions of the dual-task consolidation interruption paradigm. The results demonstrate that retroactive interference is robust to changed WM probe demands. However, retroactive interference was less robust to modifications of the T2 task: it appears that a task switch (from WM encoding to another task) with an immediate speeded motor response is the most effective approach to evoke retroactive interference. These results are best understood in the context of central resource sharing models of dual-task interference.</p>