The need for improvement at a rail/highway crossing typically is based on the Expected Accident Rate (EAR) in conjunction with other criteria carrying lesser weight. In recent years new models for assessing the need for improvements have been developed, and in the research reported here five such models selected from a list established from a literature review and a user survey were evaluated. The selected models--the DOT, Peabody-Dimmick, NCHRP No. 50, Coleman-Stewart, and New Hampshire--were evaluated using a data base maintained by the Virginia Department of Highways and Transportation. Additionally, the performance of the methods in predicting the EAR were compared using the chi-square test and the power factor. The results indicated that the DOT formula outperformed the other four methods in both the evaluative and comparative analyses, and thus was recommended for use. The priority list produced by this formula is only one criterion used in determining the need to improve conditions at any crossing. It must be supplemented with information obtained by regular site inspections and with qualitative data that cannot feasibly be incorporated into a mathematical formula.