We recently demonstrated that salience can have a massive impact on visual working memory in a delayed continuous recall task (Constant & Liesefeld, 2021; https://doi.org/gjk9jh). We also showed that this effect is hard to erase or overrule with pre-/retro-cues or with manipulations of relevance, even at long encoding times (Constant & Liesefeld, 2023; https://doi.org/gr6xzr; Constant & Kerzel, 2024; https://doi.org/mvnk). Here, participants memorized three tilted bars (12°, 28° and 45°) presented for 500 ms among 33 vertical bars. After a delay of 1000 ms, a response display appeared and memory for one of the tilted bars was probed. In previous studies, participants adjusted the color of the probed bar in a grayscale version of the original display. Here, the probed bar was shown in its original tilt in the center, which had unexpected effects on memory performance. In the first experiment, participants first recalled the original position of the probed bar and then recalled its color on a colorwheel. As expected, the most salient bar with 45° tilt was remembered more accurately than the less salient ones, but in contrast to previous studies, there was no difference between the two less salient bars with tilts of 28° and 12°. The same pattern of results was observed when participants judged only color, ruling out dual-task load as an explanation. Finally, restoring a context of vertical bars around the tilted bar in the center of the response display did not change the results. These results may suggest that the most salient bar is maintained in an active state in VWM, whereas the less salient bars are stored in a silent state which is less resilient to sudden death or swapping.