Developing clinical tools that assess bone matrix quality could improve the assessment of a person's fracture risk. To determine whether Raman spectroscopy (RS) has such potential, we acquired Raman spectra from human cortical bone using microscope- and fiber optic probe-based Raman systems and tested whether correlations between RS and fracture toughness properties were statistically significant. Calculated directly from intensities at wavenumbers identified by second derivative analysis, Amide I sub-peak ratio I1670/I1640, not I1670/I1690, was negatively correlated with Kinit (N = 58; R2 = 32.4%) and J-integral (R2 = 47.4%) when assessed by Raman micro-spectroscopy. Area ratios (A1670/A1690) determined from sub-band fitting did not correlate with fracture toughness. There were fewer correlations between RS and fracture toughness when spectra were acquired by probe RS. Nonetheless, the I1670/I1640 sub-peak ratio again negatively correlated with Kinit (N = 56; R2 = 25.6%) and J-integral (R2 = 39.0%). In best-fit general linear models, I1670/I1640, age, and volumetric bone mineral density explained 50.2% (microscope) and 49.4% (probe) of the variance in Kinit. I1670/I1640 and v1PO4/Amide I (microscope) or just I1670/I1640 (probe) were negative predictors of J-integral (adjusted-R2 = 54.9% or 37.9%, respectively). While Raman-derived matrix properties appear useful to the assessment of fracture resistance of bone, the acquisition strategy to resolve the Amide I band needs to be identified.