The retro-cue paradigm is frequently used to study how attention interacts with visual working memory (WM). In this paradigm, a single cue presented during the retention interval directs attention to a memorized item, thereby improving subsequent memory performance for that item. Recent studies have extended this approach to investigate the multiple retro-cues effect, i.e., simultaneously presenting two or more retro-cues to enhance memory performance for multiple cued items. However, the capacity limit and neural basis of this effect remain unclear. Through four experiments, we revealed that the physical distance between cued items can modulate the capacity of multiple retro-cues. Experiment 1 compared one, two neighboring (cued items presented at two contiguous locations), three neighboring, and neutral cues, revealing performance benefits only for one and two neighboring retro-cues. Experiment 2 directly compared the effect of two neighboring versus two separated cues (with one or two intervening items) and found a significant benefit only for two neighboring cues. Experiment 3 directly manipulated the spatial scope of the memory array (wide vs. narrow display) and showed that separated cues became effective when the array was presented in a narrow scope, where the absolute physical distance between cued items was reduced. Experiment 4 provided neural evidence by showing larger contralateral delay activity (CDA) with multiple cues and a complex pattern of N2pc which suggested strategic adjustment. Together, these findings delineate the quantitative boundary and spatial constraints of the multiple retro-cues effect and support the “attentional zoom-lens” mechanism of attention modulation.