The River and Stream Continuity Partnership has developed assessment protocols for evaluating the barrier effects of road-stream crossings. In addition to these protocols, the University of Massachusetts Amherst created an online database and an algorithm for scoring crossing structures according to the degree of obstruction they pose to aquatic organisms. In 2008, The Nature Conservancy and University of Massachusetts collected data to test the robustness and repeatability of the protocol and compare results with those of other fish passage models. A stratified random sample of 317 road stream crossings in Connecticut, Vermont, New Hampshire and Massachusetts that had previously been assessed using the protocols were re-evaluated by trained technicians. Of these sites, 111 were assessed in sufficient detail to evaluate them using FishXing, the California Salmonid Stream Habitat Restoration course screen and course screens developed by Seth Coffman. We excluded from analyses 51 crossings (16.1 %) because it appeared likely that the resurvey assessments were not conducted on the same crossings evaluated in the previous surveys. Some of these crossings appear to have been repaired or replaced since the previous surveys were conducted. However, most of these probably represent mismatched locations. Repeat surveys of crossing structures indicate relatively high rates of discrepancies for particular data fields (range from 3.8 to 53.8 percent) but small overall effects on crossing scores. Overall, 46.2 percent of crossings had the same score for the previous survey as for the resurvey and 73.7 percent of previous survey scores fell within one score of those for corresponding resurveys. Flow Conditions was one data field that showed a substantial number of differences (60.2%) between previous surveys and resurveys. We found no significant differences in changes in scores for crossings assessed at different flows as compared to changes in scores for crossings assessed under similar flow conditions. The two assessment methods that yielded the most similar results were Coffman‘s Coarse Screens and the Crossings Database. Scores for these two methods were in agreement for 39.6 percent of crossings, with 68.5 percent falling within one unit and 83.8 percent within two units of each other. Comparison of the Crossings Database scores with those derived from FishXing yielded the same scores at only 20.7 percent of crossings and at only about two-thirds of sites did scores fall with two units of each other. There also was little agreement between scores derived from FishXing and those from the Coffman Coarse Screens The scores from these two methods were the same for 23.4 percent of crossings with little over half (52.3 %) falling with one unit and 70.3 percent within two units of each other. It is not clear which of these three methods yields the most credible results. Results of these analyses are being used to revise and improve the assessment protocols as well as the scoring algorithm used by the database and will ultimately increase confidence and credibility in the data collected as part of the River and Stream Continuity Project.