A double-observer approach was used during colony surveys of large gulls to estimate nest detection probabilities and unbiased estimates of abundance. Several models were developed that incorporate different assumptions about the sources of variation in detection probabilities (observer, colony, vegetation height). Nine colonies were surveyed in which a total of 137 transects were conducted. Although the most likely model was a one with a constant nest detection probability, a constrained model with observer dependent nest detection probability had substantial support. Overall detection probability (probability that a nest was detected by at least one of the two observers) was 0.943 (SE ± 0.008). Detection probability for a single observer was lower (0.761, SE ± 0.016). These estimates indicate a relatively high underestimation of the population sizes of large gull species when not taking into account nest detectability.