Standard working memory (WM) updating tasks confound updating requirements with generic WM functions. We introduce a method for isolating a process unique to WM updating, namely the removal of no-longer relevant information. In a modified version of an established updating paradigm, to-be-updated items were cued before the new memoranda were presented. Longer cue-stimulus intervals—that is, longer removal time—led to faster updating, showing that people can actively remove information from WM. Well-established effects of item repetition and similarity on updating RTs were diminished with longer removal time, arguably because representational overlap between out-dated and new information becomes less influential when out-dated information can be removed prior to new encoding. The benefit of removal time was found only for partial updating, not for complete updating of entire memory sets. We conclude that removal of out-dated information can be experimentally isolated, and that removal is a unique, active WM updating process.