Recent studies showed that proactive interference (PI) impairs visual working memory (WM), as performance is better when the memory items are unique rather than repeated throughout the experiment. To scrutinize the mechanisms driving this effect, we tested how it affects the stages of encoding, retention, and testing. Experiment 1 instructed participants to make speeded responses and found that reaction time was slower in the repeated than in the unique condition, suggesting memory source confusion during testing. Experiment 2 showed that a retention interval manipulation did not affect the magnitude of the PI effect. Finally, Experiment 3 found that the PI effect did not increase when the encoding interval was short. Instead, the difference between the repeated and unique conditions was reduced, probably because the increased familiarity with the memory items in the repeated condition facilitated consolidation. Taken together, these results suggest that the main locus of PI in visual WM is during testing and that encoding and retaining information in visual WM are hardly affected by no longer needed information. Consequently, these data support the involvement of long-term memory in the effect. (PsycInfo Database Record (c) 2021 APA, all rights reserved).