This paper describes how a short, repeated and structured opportunity to reflect was integrated in the storyline of a serious game in order to stimulate the development of a meta-cognitive skill: the ability to self-assess the degree of confidence in own answers. An empirical validation of the approach, conducted with 28 secondary school pupils, delivers an uncommon pattern: while the cognitive benefits--the acquisition of academic knowledge in optics--are negligible and mixed up, the meta-cognitive gains present a raising tendency. The experiment also demonstrates that reflection does not necessarily hamper the game flow, if certain conditions, discussed in the paper, are met.