We conducted dual-observer trials to estimate detection probabilities (probability that a group that is present and available is detected) for fixed-wing aerial surveys of wading birds in the Everglades system, Florida. Detection probability ranged from <0.2 to ~0.75 and varied according to species, group size, observer, and the observer's position in the aircraft (front or rear seat). Aerial-survey simulations indicated that incomplete detection can have a substantial effect on assessment of population trends, particularly over relatively short intervals (≤3 years) and small annual changes in population size (≤3%). We conclude that detection bias is an important consideration for interpreting observations from aerial surveys of wading birds, potentially limiting the use of these data for comparative purposes and trend analyses. We recommend that workers conducting aerial surveys for wading birds endeavor to reduce observer and other controllable sources of detection bias and account for uncontrollable sources through incorporation of dual-observer or other calibration methods as part of survey design (e.g., using double sampling).