Visual working memory (VWM) is a fundamental cognitive capacity that allows us to temporarily hold visual information, but storage is effortful and content-fragile. Rather than loading VWM to the maximum, individuals usually rely on the external world and access information just in time. However, participants do rely on VWM more as access costs to external information increase. This phenomenon is commonly investigated with so-called copy tasks, which differ across paradigms, manipulations, and dependent variables. We here present findings of a meta-analysis into the reliability and consistency of shifts in the assumed trade-off between storing and sampling across manipulations and dependent variables, using data from 28 experiments. We found that all cost manipulations led to substantial shifts from external sampling to storage in VWM. Cost manipulations did not differ in their effect across studies even though such differences are reported within studies. All dependent variables were associated with clear but different strong effects. We argue that the differences observed between indicators are not only due to sensitivity differences but also due to differential aspects of behavior that are measured. New variables and techniques might now pave the way to understanding the trade-off between storing and sampling more in-depth. Collectively, our findings suggest that the reliance on VWM or the external world shifts consistently as access cost is increased, is largely irrespective of cost manipulations, and expresses itself reliably across dependent variables. With this work, we seek to help establish standards and comparability across this growing body of work.