One of the basic characteristics of visual working memory (VWM) is that recall precision declines with the number of items to be memorised. This finding is interpreted as a consequence of the allocation of a limited pool of resources, whereby the amount of resources allocated to each item drops with each additional stimulus. Recent studies show that recall in VWM tasks also depends on the characteristics of the stimulus set which is to be memorised. Specifically, when memorising simple visual features of two stimuli (e.g. their orientations), the recall of one stimulus's features is marked by a systematic bias towards the other stimulus's features. In this study, we wanted to examine how robust this effect is to experimental manipulations of memory precision. In four experiments, participants (N = 33) memorised and recalled the orientations of two simultaneously presented stimuli. In the first experiment, the recall order was chosen at random. In the second, participants had to choose the recall order themselves. In the third and fourth experiments, recall order was again chosen at random, but participants were presented with a cue before (exp. 3) or after (exp. 4) the stimuli were displayed, which indicated which of the stimuli is most likely to be recalled first. The attraction effect was observed in all four experiments and affected the recall of both stimuli. This study demonstrates the robustness of the attraction effect in VWM, implying that the mechanisms underlying the effect are not susceptible to correction, or in other words, that it is a part of the fundamental processes of coding visual stimuli.