Across ten experiments, we investigated the role of attentional control in the allocation of visual short-term memory resources to relevant visual information when multiple feature dimensions were present. In each experiment, a change detection task was embedded within a task switching paradigm wherein the relevant feature dimension on which to detect changes could either repeat (e.g., colour-colour) or switch (e.g., colour-orientation) on a given trial. Results show a consistent impact of dimension switching across a range of measures and experimental manipulations (e.g., set size, feature overlap). Our findings suggest that these dimension switch costs originate from an origin distinct to that observed with typical task switching costs. Manipulations of cue-stimulus (Experiment 3a) and response-cue (Experiment 3b) intervals—which are associated with moderation of the magnitude of task switch costs—had no influence on dimension switch costs, and cue switches—associated with priming accounts of task switch costs—did not contribute to the dimension switch cost to any great extent (Experiment 4). We discuss the results in relation to theoretical models of visual short-term memory in terms of the impact of dimension switching on efficient resource allocation.