Current clinical bone diagnostic measures rely predominantly on X-ray-based contrast and are primarily sensitive to bone mineral content. Since bone also contains collagen and water components, which are heavily implicated in fracture resistance, these X-ray measures are micro-anatomically incomplete and do not identify individuals who will fracture. This dissertation aims to improve clinical bone fracture risk assessment through the use of novel magnetic resonance imaging (MRI) methods, which provide quantitative measures of the non-mineral bone components. The overall goal is to advance our understanding of 1H nuclear magnetic resonance (NMR) relaxation in human cortical bone to the point that diagnostically-relevant parameters may be extracted from in vivo bone MRI measurements. To accomplish this, custom NMR hardware was first developed for a rigorous, NMR relaxation-based characterization of ex vivo cortical bone. Such characterization was used to identify the micro-anatomical origins of cortical bone NMR signal components, which included collagen, bound water, and pore water protons. These signal components correlated well with various bone mechanical properties, indicating diagnostic relevance. Using the well-characterized cortical bone relaxation characteristics, novel and clinically practical methods for quantitative, diagnostic bone MRI were developed and validated. Collectively, this work represents a biophysical basis for cortical bone MRI, which stands ready for translation to clinical and research studies.