Recent work in linguistics and psycholinguistics has sought to understand the cognitive foundations of semantic meaning for function words like most (Pietroski, Lidz, Hunter, & Halberda, 2009; Hackl, 2009; Lidz, Pietroski, Halberda, & Hunter, 2011). This work has put forth a variety of theories about the meaning of most, including that it naturally relies on cardinality, its verification process is closely tied to its underlying semantics, and that logically equivalent meanings of most have been distinguished using simple psychophysical tasks. Here, we show that subjects’ performance in truth judgments tasks for most likely relies on a wide family of strategies, with a bias towards using cardinality-based strategies. The choice between these strategies is context-dependent, variable between participants, and sensitive to task-level factors like the number of trials participants are asked to complete. These results indicate that there is unlikely to be a single, simple formalization how most is computed, even in controlled laboratory tasks involving dot arrays. Instead, progress may be made by studying the way in which language users flexibly deploy a variety of different verification procedures.