Bone is an exceptional material that is lightweight for efficient movement but also exhibits excellent strength and stiffness imparted by a composite material of organic proteins and mineral crystals that are intricately organised on many scales. Experimental and computational studies have sought to understand the role of bone composition and organisation in regulating the biomechanical behaviour of bone. However, due to the complex hierarchical arrangement of the constituent materials, the reported experimental values for the elastic modulus of trabecular and cortical tissue have conflicted greatly. Furthermore finite element studies of bone have largely made the simplifying assumption that material behaviour was homogeneous or that tissue variability only occurred at the microscale based on grey values from micro-CT scans. Thus, it remains that the precise role of nanoscale tissue constituents and microscale tissue organisation is not fully understood and more importantly that these have never been incorporated together to predict bone fracture or implant outcome in a multiscale finite element framework. In this paper, a three-scale finite element homogenisation scheme is presented which enables the prediction of homogenised effective properties of tissue level bone from its fundamental nanoscale constituents of hydroxyapatite mineral crystals and organic collagen proteins. Two independent homogenisation steps are performed on representative volume elements which describe the local morphological arrangement of both the nanostructural and microstructural levels. This three-scale homogenisation scheme predicts differences in the tissue level properties of bone as a function of mineral volume fraction, mineral aspect ratio and lamellar orientation. These parameters were chosen to lie within normal tissue ranges derived from experimental studies, and it was found that the predicted stiffness properties at the lamellar level correlate well with experimental nanoindentation results from cortical and trabecular bone. Furthermore these studies show variations in mineral volume fraction, mineral crystal size and lamellar orientation could be responsible for previous discrepancies in experimental reports of tissue moduli. We propose that this novel multiscale modelling approach can provide a more accurate description of bone tissue properties in continuum/organ level finite element models by incorporating information regarding tissue structure and composition from advanced imaging techniques. This approach could thereby provide a preclinical tool to predict bone mechanics following prosthetic implantation or bone fracture during disease.