This study taught the passive to two children with specific language impairment (aged 8;1 and 8;2). It employed usage-based principles including "constructional grounding"; using short structures as the basis for acquiring long structures, and "construction conspiracy"; encouraging analogies between partially overlapping constructions. Significant improvements in production and comprehension were observed, though generalizability is limited by factors including experimental design and control over treatment fidelity. It is suggested that interventions modelled on usage-based approaches could prove effective for treating complex structures. (Contains 2 tables and 2 figures.)