This study applies a microscopic approach for modeling the risk of freeway rear-end accidents by considering the occurrence mechanism of rear-end accidents. The probability of occurrence of a rear-end accident is expressed by the product of the probability of a leading vehicle becoming an obstacle (Po) and the probability of the following drivers failure to avoid a collision (Pf). Both traffic flow and freeway design characteristics are included in the model. Compared with most existing models, this model has two major advantages: 1) it directly considers a drivers response time distribution; and 2) it applies a unique model structure that is capable of modeling the dual impacts of one variable to both Po and Pf. Using observed rear-end accident data in Washington State, this accident risk model is estimated by maximum likelihood using a modified negative binomial regression form. Eight different explanatory variables are found significant in the model. This model is potentially useful for accident risk prediction and safety plan evaluations for freeways.