Visual working memory is a mental workspace used to temporarily maintain and manipulate visual information. Although we can perceive the visual world around us in seemingly infinite richness and detail, only a fraction of these details can be held in working memory at any given time. Thus, working memory is described as a capacity-limited system, and individual differences in working memory capacity strongly predict individual differences in fluid intelligence and academic achievement. Despite the importance of working memory capacity for cognition, the nature and definition of capacity limits are still under extensive debate. In this dissertation, I use novel behavioral and neural measures to interrogate the limits of visual working memory and to inform these ongoing debates. In Chapter 1, I summarize key questions related to the limitations of visual working memory, focusing on capacity limits in working memory and individual differences in these limits. In Chapter 2, I characterize the dynamic nature of visual working memory performance by measuring and modeling how fluctuations of attention influence working memory. In Chapter 3, I address critiques of capacity limits by theoretical models which propose that all items receive mnemonic resources (i.e. no truly uninformed guesses are necessary). I demonstrate how looking at within-trial mnemonic variability, rather than average data, can provide clear evidence against these capacity-unlimited models. Finally, in Chapter 4, I ask whether we can further bolster evidence for capacity limits by decoding the content of active working memory signals in EEG (the topography of alpha-band power). Together, these studies inform contemporary models of visual working memory, supporting a capacity-limited system whose efficacy varies dynamically under the influence of attention.