This paper describes a probabilistic approach to estimate the conditional probability of release of hazardous materials from railroad tank cars during train accidents. Monte Carlo methods are used in developing a probabilistic model to simulate head impacts. The model is based on the physics of impact in conjunction with assumptions regarding the probability distribution functions of the various factors that affect the loss of lading. These factors include impact velocity, indenter size, tank material, tank diameter, effective collision mass, and tank thickness. Moreover, each factor is treated as a random variable characterized by its assumed distribution function, mean value, and standard deviation (or variance). Reverse engineering is performed to back-calculate the mean values and standard deviations of these random variables that reproduce trends observed in available accident data. The calibrated model is then used to conduct a probabilistic sensitivity analysis to examine the relative effect of these factors on the conditional probability of release. Results from the probabilistic sensitivity analysis indicate that the most significant factors that affect conditional probability of release are impact velocity, effective collision mass, and indenter size.