A Comparative Synthesis of Game-Based Learning Approaches for K–12 Computational Thinking

<p>Game-based learning improves computational thinking, but no synthesis has established which kind of game-based instruction to choose, because existing reviews compare gamebased learning against ordinary teaching rather than against itself. This article argues that the conventional approach categories do not predict outcomes and that a different distinction does. Drawing on twenty seven empirical studies and six syntheses, with every reported statistic verified against the source publication for both its value and what it measured, we sort design changes by whether they serve an instructional function, teaching a specific construct, procedure, or the fact that a solution is wrong, or a merely contextual one. The distinction tracks outcomes. Substituting one delivery medium for another produced null results in three head-to-head comparisons. Adding a constraint, an optional hint system, or a narrative wrapper to a functioning game produced no cognitive gain in three studies and measurable harm in two. Supports that taught a specific construct did produce gains, and in one case log data show scaffolded learners writing significantly more conditionals and loops. Removing the means of checking whether a solution runs cost learners measurably, in a quasi-experiment and in a randomised trial of 198 preschoolers. The contribution is a falsifiable account of why approach-level comparisons have failed to converge, together with a documented verification procedure that identified reporting errors in four published sources and three studies whose secondary summaries invert their findings.</p>