ABSTRACT The tanker oil spill risk assessment model (TRAM) allows a user to investigate the way in which oil spill probabilities, and amounts spilled, vary with such factors as fleet composition, navigational aids, or particular properties of the route. Such a comparison of alternatives is generally required for an environmental impact statement. TRAM operates through multiplication of a series of matrices: P, Q, R, and S. The rows and columns of all matrices are parameterized by accident type (such as collision or grounding) and location (such as at a pier, or in a harbor). The probability of an accident (P) and of a spill following an accident (Q) are based on world tanker fleet data. The matrix S expresses the conditional probability that if there is a spill, the vessel will be a total loss. This enables catastrophic spills in which spill size can be related to vessel size to be distinguished from minor spills in which vessel size is generally not a factor. The matrix R contains most of the unique features of the model. It accounts for vessel and route-specific features that reasonably may be expected to alter the world tanker fleet data. Examples are given from the literature for the elements of R corresponding to: fleet composition (domestic/foreign carrier); navigation aids such as vessel traffic system; vessel age, and construction (double hull, inerting system, segregated ballast, etc.); and various features of the tanker route, such as channel width and traffic density, based on Macduff's causation probability formalism. The practical usage of the model is demonstrated by applying it to a hypothetical project involving tanker oil transportation. The extension of this analysis technique to other forms of risk analysis studies (including operations other than oil transport) is also discussed.