Silk fibroin is a potent alternative to other biodegradable biopolymers for bone tissue engineering (TE), because of its tunable architecture and mechanical properties, and its demonstrated ability to support bone formation both in vitro and in vivo. In this study, we investigated a range of silk scaffolds for bone TE using human adipose-derived stem cells (hASCs), an attractive cell source for engineering autologous bone grafts. Our goal was to understand the effects of scaffold architecture and biomechanics and use this information to optimize silk scaffolds for bone TE applications. Silk scaffolds were fabricated using differ- ent solvents (aqueous vs. hexafluoro-2-propanol (HFIP)), pore sizes (250–500 lm vs. 500–1000 lm) and structures (lamellar vs. spherical pores). Four types of silk scaffolds combining the properties of interest were systematically compared with respect to bone tissue outcomes, with decellularized trabecular bone (DCB) included as a ‘‘gold standard’’. The scaffolds were seeded with hASCs and cultured for 7 weeks in osteogenic medium. Bone formation was evaluated by cell proliferation and differentiation, matrix pro- duction, calcification and mechanical properties. We observed that 400–600 lm porous HFIP-derived silk fibroin scaffold demonstrated the best bone tissue formation outcomes, as evidenced by increased bone protein production (osteopontin, collagen type I, bone sialoprotein), enhanced calcium deposition and total bone volume. On a direct comparison basis, alkaline phosphatase activity (AP) at week 2 and new calcium deposition at week 7 were comparable to the cells cultured in DCB. Yet, among the aqueous- based structures, the lamellar architecture induced increased AP activity and demonstrated higher equi- librium modulus than the spherical-pore scaffolds. Based on the collected data, we propose a conceptual model describing the effects of silk scaffold design on bone tissue formation.