A challenge for primary classroom teachers is to maintain students' engagement with learning tasks while catering for their diverse needs, capabilities and interests. Multiple pedagogical approaches are employed to promote on-task behaviours in the mathematics classroom. There is a general assumption by educators that games ignite children's on-task behaviours, but there is little systemically researched empirical data to support this claim. This paper compares students' on-task behaviours during non-digital game-playing lessons compared with non-game-playing lessons. Six randomly selected grade 5 and 6 students (9-12 year olds) were observed during ten mathematics lessons. A total of 2,100 observations were recorded via an observational schedule and analysed by comparing the percentage of exhibited behaviours. The study found the children spent 93% of the class-time exhibiting on-task engagement during the game-playing lessons compared with 72% during the non-game-playing lessons. The game-playing lessons also promoted greater incidents of student talk related to the mathematical task (34%) compared with the non-game-playing lessons (11%). These results support the argument that games serve to increase students' time-on-task in mathematics lessons. Therefore, it is contended that use of games explicitly addressing the mathematical content being taught in a classroom is one way to increase engagement and, in turn, potential for learning.