Many verbal theories describe working memory (WM) in terms of physical metaphors such as information flow or information containers. These metaphors are often useful but can also be misleading. This article contrasts the verbal version of the author’s three-embedded-component theory with a computational implementation of the theory. The analysis focuses on phenomena that have been attributed to the focus of attention in WM. The verbal theory characterizes the focus of attention by a container metaphor, which gives rise to questions such as: how many items fit into the focus? The computational model explains the same phenomena mechanistically through a combination of strengthened bindings between items and their retrieval cues, and priming of these cues. The author applies the computational model to three findings that have been used to argue about how many items can be held in the focus of attention (Oberauer and Bialkova, 2009; Gilchrist and Cowan, 2011; Oberauer and Bialkova, 2011). The modeling results imply a new interpretation of those findings: The different patterns of results across those studies don’t imply different capacity estimates for the focus of attention; they rather reflect to what extent retrieval from WM is parallel or serial.