Transplantation of whole bone marrow (BMT) as well as ex vivo-expanded mesenchymal stromal cells (MSCs) leads to striking clinical benefits in children with osteogenesis imperfecta (OI); however, the underlying mechanism of these cell therapies has not been elucidated. Here, we show that non-(plastic)-adherent bone marrow cells (NABMCs) are more potent osteoprogenitors than MSCs in mice. Translating these findings to the clinic, a T cell-depleted marrow mononuclear cell boost (> 99.99% NABMC) given to children with OI who had previously undergone BMT resulted in marked growth acceleration in a subset of patients, unambiguously indicating the therapeutic potential of bone marrow cells for these patients. Then, in a murine model of OI, we demonstrated that as the donor NABMCs differentiate to osteoblasts, they contribute normal collagen to the bone matrix. In contrast, MSCs do not substantially engraft in bone, but secrete a soluble mediator that indirectly stimulates growth, data which provide the underlying mechanism of our prior clinical trial of MSC therapy for children with OI. Collectively, our data indicate that both NABMCs and MSCs constitute effective cell therapy for OI, but exert their clinical impact by different, complementary mechanisms. The study is registered at www.clinicaltrials.gov as NCT00187018.