The primary objective of this study is to estimate the potential capacity of U-turns at unsignalized median openings on six-lane streets. To achieve the research objective, data were collected at seven unsignalized median openings in the Tampa Bay area of Florida. By using the maximum likelihood method and the Siegloch method, the research team estimated the critical headway and follow-up times for U-turns on six-lane streets. The critical headway was found to be 5.6 s. The follow-up time was found to be 2.3 s. With the estimated critical headway and follow-up time obtained, Harders's model was used for estimating the potential capacity of U-turns. The research team also estimated the conflicting traffic volume for U-turns. Model validation results show that the Harders model, which uses a weighted conflicting traffic volume, provides the best capacity estimates. The weighted conflicting traffic volume for U-turns on six-lane streets equals 2.2 times the average opposing major street traffic volume in each lane. Using the weighted conflicting traffic volume, Harders's model yields a mean absolute percent error (MAPE) of 17.8%. Even though the MAPE is relatively high, Harders's model does not systematically overestimate or underestimate the capacity of U-turns on six-lane streets. The results suggest that, with the parameters estimated in this study, the current capacity estimation methods provided by the Highway Capacity Manual can be applied to estimating U-turn capacity at unsignalized median openings on six-lane streets.