This paper gives an overview of the ecomorphology of birds; particular points are illustrated with analyses primarily of our own data. The preferred method is to analyse morphological data from systematically close knit groups (genera, families). In all the examples presented, data, insofar as they were measurements rather than counts, were corrected for size by dividing them by the cube root of the lean body weight. With the exception of variables which could take on values of zero or less, they were then transformed to their logarithms. In Principal Component Analyses (PCA), correlation matrices were favoured over covariance matrices to weigh characters equally. Most examples refer to our own measurements of external and skeletal characters in Old World warblers, Sylviidae, chat-like thrushes, and tits. Fig. 1 shows the standard set of 32 measures used. The various problems of morphometrical analysis and functional interpretation are treated at different levels of organization, i.e. within species, between species, among genera, and in bird communities. Stepwise Discriminant Analysis was used to elucidate the differences between the willow tit and the marsh tit, Parus montanus and P. palustris, and to search for the most relevant characters (Tab. 1). Although the measurements were not adjusted for size, shape characteristics dominate the canonical axes. Willow tits have longer, more graduated tails and smaller feet than marsh tits. A plot (Fig. 2) of the canonical scores of separate analyses of hindlimb and forelimb characters shows the relationship between inter-species and intra-species variation. The Dartford warbler and Marmora's warbler, Sylvia undata and S. sarda are differentiated by their flight apparatus and their hindlimbs, which reflects the preference of the Dartford warbler for low vegetation and for skulking in a cluttered habitat. The same analysis applied to four species of Phylloscopus warblers reveals size and shape differences in the first canonical axis, which are mainly associated with long-distance movements, and morphological differences in the second, which correlate with increase use of arboreal habitats (Fig. 3). To analyse variation within genera, Sylvia with 11 species and Parus with 6 species served as examples; PCAs of within species means of 32 characters were carried out. The first component of morphological variation of the warblers can be explained by differences in migratory behaviour and the vegetation of their breeding habitats, and the second by the diversity of locomotory habits which include the contrast between hopping and skulking and the ability to take off quickly or to break efficiently on the wing (Tab. 2, Fig. 4). In the first component of a PCA of Parus the contrast between square tails combined with large feet and long graduated tails with small feet dominates (Fig. 5). This component reflects differences in habitat structure while the second is mainly associated with the morphology of the bill and its diverse use. A PCA of sylviids demonstrates the importance of hindlimb characters for separating the various genera along the first component (Fig. 6). The second for the most part represents mainly differences in long distance migration. Cluster Analysis of the same data describes intergeneric and intrageneric similarities quite well, but is not amenable to straightforward ecological interpretation (Fig. 7). An analysis of the community of migrants in a southwest-German stopover-site illustrates several important points. Most of the morphological variation represented by the first and the second PC, which together explain 60 % of the total variance, concerns characters related to locomotion. The first component expresses the contrast between aerial hunters on the low end and species which climb vertical stalks on the high end (Fig. 8). The systematic heterogeneity of the species set hampers more definitive interpretations. Morphological variation within genera does not affect all characters equally. The demands of migration determine variation in wing morphology of those genera containing migratory species. Genera comprised of residents and short distance migrants vary more only in bill measurements, whereas canopy dwellers (Parus, Phylloscopus) develop a variety of hindlimb features (Tab. 3). There are many interactions between single functional complexes of morphological characters. Important constraints arise for birds due to the problem of flight and weight. For example, there is a general tendency to appoint less mass to the hindlimb when flight musculature is well developed (Fig. 9). As other studies of fruit-eating birds demonstrate, interactions between grasping for fruits with the bill and the use of perches lead to correlations between bill and hindlimb characteristics. The incompatibility of some character combinations restrains the set of possible adaptations even within functional complexes. In warblers of the genus Sylvia, for example, features suited for long distance migration are not congruent with certain modes of habitat use (Fig. 10). Yet characters within functional complexes may show diverse correlation structures in different systematic groups, so that the functional interpretation of the variation of single characters depends on the context of covariation with other characters (Fig. 11). This aggravates generalizations and poses a principal difficulty for the application of ecomorphological methods to problems of community ecology. Correlations between morphology and ecology can only be understood if the relevant behaviour is known. Bill shape can be modified according to specific feeding techniques. Seasonal variation in behaviour can match changes in morphology. Sexual differences and social status are other variables which may influence ecomorphological relationships. Valuable insights can be gained when morphological, ecological and ethological data about several species are available. Regression Analysis and Canonical Correlations can be used to describe the relationships and to select the appropriate characters for interpretation and further studies (Fig. 12). Examples for relations between feeding techniques and morphology are presented in Figs. 13, 14 and 15. Besides such studies of correlations, experimental corroboration of the causal interplay of behaviour, morphology, and ecology is needed. Some examples for such studies are reviewed. In conclusion, it must be stressed that many ecomorphological studies, particularly those concerned with community ecology, fail to recognize the importance of selecting a representative set of characters which also have some functional meaning. The results of studies of the type advocated here may stimulate ecologists as well as morphologists. More ecomorphological work needs to be devoted to the role of morphology in constraining behaviour, to the problem of morphological changes caused by differential use, to the generalist-specialist continuum, and to convergent evolution.
