Visual working memory (VWM) is a limited-capacity cognitive system that allows us to maintain and manipulate visual information over a brief period of time. It plays a critical role in many cognitive functions, including visual search, problem-solving, and decisionmaking. While the importance of VWM is not disputed, a critical issue addressed in research is whether visual objects bind all their component features during VWM entry or if task-relevant features are prioritised for representation. In a recent study, Zhu et al. (2022) proposed that attended sensory information does not automatically enter VWM, but instead is subject to an additional reselection process that determines whether items are selected for VWM entry. Furthermore, they suggested that reselection operates using a dimensional memory filter, such that when an individual feature value (e.g., red) is selected for entry into VWM, all attended feature values within the same feature dimension (e.g., all colours) automatically enter VWM. Across three experiments, we systematically investigated these hypotheses by conceptually replicating two of Zhu et al.’s experiments, while incorporating modifications to address our concerns about their methodology. In Experiment 1, we investigated Zhu et al.’s proposal that a reselection process operates to restrict the entry of features into VWM. We assessed whether the irrelevant colour of a target object captured more attention than a new colour when it appeared as a distractor singleton in an orthogonal visual search task. We found that this was the case, indicating that the irrelevant colour of the target object was held in VWM, contrary to the key result that Zhu et al. cited to support reselection. In Experiments 2 and 3, we investigated Zhu et al.’s claim that VWM consolidation operates via a dimensional memory filter. Specifically, we assessed whether memorising the colour of a fixation cross would cause the task-irrelevant colour of a target object to enter VWM. To determine whether this colour entered VWM, we measured its interference in an orthogonal shape change detection task. Experiment 2 used a 700 ms ISI, while in Experiment 3, we reduced the ISI to 100 ms. Unlike Zhu et al., who only reported trials where shape was consistent between displays, we included trials where shape changed between displays in our analysis and included trials in which the shape probe matched the memorised fixation colour. In Experiments 2 and 3, we found evidence that the target colour had entered VWM, consistent with a dimensional memory filter. However, we also found that the fixation colour had a stronger impact on shape change detection than the target colour, a finding that was not explicitly predicted by Zhu et al. Overall, our results align more closely with predictions from object-file theory (Kahneman et al., 1992) and event-file theory (Hommel, 1998, 2004), which propose object-specific binding occurs during encoding, and objectspecific benefits and costs emerge during object retrieval and review. The facilitation we observed when features were consistent between displays could reflect objectspecific repetition benefits, while the interference we observed when features partially changed between displays may indicate partial repetition costs.