A model for the stress-dependent elastic wave velocity response of fractured rock mass is proposed based on experimental evidence of stress-dependent fracture normal and shear stiffness. Previously proposed models and previous experimental studies on stress-dependent fracture stiffness have been reviewed to provide a basis for the new model. Most of the existing stress-dependent elastic wave velocity models are empirical, with model parameters that do not have clear physical meanings. To propose the new model, the rock mass is assumed to have randomly oriented microscopic fractures. In addition, the characteristic length of microfractures is assumed to be sufficiently short compared to the rock mass dimensions. The macroscopic stress-dependent elastic wave velocity response is assumed to be attributed to the stress dependency of fracture stiffness. The stress-dependent fracture normal stiffness is defined as a generalized power law function of effective normal stress, which is a modification of the Goodman's model. On the other hand, the stress dependency of fracture shear stiffness is modeled as a linear function of normal stress based on experimental data. Ultrasonic wave velocity responses of a dry core sample of Berea sandstone were tested at effective stresses ranging from 2 to 55 MPa. Visual observation of thin sections obtained from the Berea sandstone confirms that the assumptions made for microstructure of rock mass model are appropriate. It is shown that the model can describe the stress-dependent ultrasonic wave velocity responses of dry Berea sandstone with a set of reasonable material parameter values. Published 2013. This article is a U.S. Government work and is in the public domain in the USA.