Using previously published atlas data for 122 mixed-species wading bird colonies on islands along the Atlantic coast (Maine to Florida, 1976-77), we examined relationships between population sizes of 11 species of egrets, herons, ibises, and wood storks (Mycteria americana) and nine habitat variables. On nautical charts, we measured four island characteristics (area, length, width, shape), three isolation factors (distances to nearest island, mainland, and a water barrier),, and two variables related to potential feeding habitat within 5 km of the center of the colony (wetland area and land-water interface, i.e., the linear distance between the marsh/upland and all water bodies within the same 5-km radius). One univariable and five multivariable .procedures were used to determine which habitat features were best related to population size .(all species combined). Multicollinearity problems among the variables limited interpretation for most procedures. Both univariable and the multivariable procedures indicated that land-water interface was the most important of the nine variables, but for all models, less than 10% of the total variance was explained (rz is less than 0.10). The size of the colony was not related to the amount of wetland area (within 5-km).per se. Colony data showed better 'structure' when examined on the basis of geographic and disturbance gradients. Population sizes of colonies near man-altered habitats were compared with those surrounded by relatively natural habitats in three geographic zones: north, middle, and south. Significant differences were found in colony size among the three zones (south largest) and between disturbance types. Surprisingly, in all three zones, colonies near man-altered areas were larger on average than those near more natural habitats in this region. A possible reason for this difference is suggested.