The biological significance of variation in such complex morphological characters as the bills of birds is open to question. Van Valen (1965) has argued that variation in bill width is adaptive in itself, and is not part of the genetic or phenotypic load. Soule and Stewart (1970) have argued against this interpretation, and have suggested instead that a large degree of variation indicates a transient release of variation associated with a deterioration of canalization. This happens, they argue, in small isolated populations living in simplified environments, as on islands. The difference might be resolved if it was known whether populations on islands generally exhibit a large degree of variation. It is difficult to anticipate whether they do or do not from a consideration of either the generation or maintenance of variation. If an island is colonized by one or a few individuals from a mainland region the island population starts with only a portion of the total genetic variety of the mainland population, because not all of the alleles are present in the island population. If this reduced genetic variety is translated into a reduced phenotypic variance, one might anticipate that the island population exhibits less morphological variation than its mainland counterpart, even after it has reached an equilibrium population size in the island environment. However, as Lewontin (1965, p. 481) has pointed out, the genetic variance upon which selection acts will not be substantially lower in the island population than in the mainland population. Indeed, while the population on the island is increasing, considerable genetic variation due to recombination and mutation might appear if selection is weaker here than on the mainland. But this is a temporary phenomenon. How might selection act when the population has reached equilibrium size? Again it is not easy to make an a priori judgment on the basis of ecological considerations. On the one hand the diversity of habitats and food is less on islands than on the mainland, conditions which might lead one to anticipate reduced morphological variation on islands as a result of selection. On the other hand the diversity of competitors and predators is less on islands than mainland, which might allow for greater ecological and morphological variation in a population on an island. In the face of such uncertainty it is helpful to proceed empirically. Amadon (1950) and Lack (1947) noted that morphological variation in birds from oceanic islands was approximately the same as that in distantly related species of birds from continents. The populations of six species Van Valen (1965) studied were from mainland and island regions. With one predictable exception, all the island populations exhibited greater morphological variation than did the mainland populations. However this result cannot be used to make a general statement about variation in birds on islands and mainland, because the examples were carefully selected on the basis of pre-existing information on the relative size of the ecological niche of each species in the island and mainland regions, and because the mainland areas were far from the closest point to the islands. Grant (1965a, 1967) attempted to discover if morphological variation was larger or smaller on islands than on mainland by using data from most of the 34 land bird species on the Tres Marias Islands and on the adjacent mainland of western Mexico. Variation in length of wing, tail, tarsus and