Seismic tomography is now a common approach to estimating velocity structure of the Earth, regardless of whether the data sources are earthquake recordings or controlled sources such as explosions, airguns or Vibroseis. Seismic tomography is convenient to implement because it requires little to no a priori knowledge of Earth structure and is much less time consuming than forward modeling schemes. Despite its convenience, the method still lacks satisfactory quantitative assessments of model reliability. Here we explore the viability of applying travel-time sensitivity testing that uses a modified Cauchy distribution as its statistical foundation to assessing the uncertainty in velocity models produced with seismic tomography. Using a crustal refraction survey as a test data set, we find that this approach produces a more realistic estimate of the velocity uncertainty than does either a resampling approach or travel-time sensitivity testing that uses a Gaussian distribution as its statistical foundation. The velocity uncertainty estimates provide an important complement to other estimates of model reliability including checkerboard tests.