To investigate whether acquisition of the perspective-marking grammar of complement-clause constructions supports progression in children’s false-belief reasoning, we conducted a training study with 76 English-speaking three-year-olds from the North-West of England (age range: 3;0-3;10 years, 50% female, 80% White). Children were randomly assigned to one of three maximally comparable training conditions, and in a four-week eight-session program, all children participated in the same training activities with mental-state contrasts. Depending on condition, activities were mediated linguistically with either simple clauses, first-person complements or third-person complements. The study addressed critical confounds in previous training studies by avoiding the use of complement clauses in false-belief tests and controlling individual differences in memory, executive functioning, general language and pretest proficiency with complement clauses. The results yielded strong support for the hypothesis of a causal influence of complement-clause exposure on false-belief progression, as children trained with first-person complements advanced significantly more in false-belief reasoning from pretest to posttest than children trained with simple clauses. Examining the roles of first- and third-person complements, a direct comparison between progression in the two complement-clause conditions showed no significant difference, but only children trained with first-person complements progressed significantly more than children in the control condition trained with simple clauses. Follow-up analyses suggested that first- and third-person complements each support false-belief progression at different stages of development.