Are animal personality, body condition, physiology and structural size integrated? A comparison of species, populations and sexes, and the value of study replication.

A major question in behavioural ecology is why behaviour, physiology and morphology are often integrated into syndromes. In great tits, Parus major, for example, explorative males are larger (vs. smaller) and leaner (vs. heavier) compared to less explorative individuals. Unfortunately, considerable debate exists on whether patterns found in specific studies are replicable. This debate calls for study replication among species, populations and sexes. We measured behavioural (exploration), physiological (breathing rate) and morphological traits (body mass, tarsus length, wing length, bill length) in two species (great vs. blue tits Cyanistes caeruleus), two populations (Forstenrieder Park vs. Starnberg) and two sexes (males vs. females). We then tested whether the same pattern of integration characterized all unique combinations of these three biological categories (hereafter called datasets). We used a multi-year repeated measures set-up to estimate among-individual trait correlation matrices for each dataset. We then used structural equation modelling to test for size-dependent behaviour and physiology, size-corrected (i.e. size-independent) behaviour-physiology correlations and size-corrected body mass-dependent behaviour and physiology. Finally, we used meta-analyses to test which structural paths were generally (vs. conditionally) supported (vs. unsupported). We found general and consistent support for size-dependent physiology and size-corrected body mass-dependent physiology across datasets: faster breathers were smaller but heavier for their size. Unexpectedly, condition-dependent behaviour was not supported: explorative birds were neither leaner, nor was this relationship heterogeneous across datasets. All other hypothesized patterns were dataset-specific: the covariance between size and behaviour, and between behaviour and physiology differed in sign between datasets, and both were, on average, not supported. This heterogeneity was not explained by any of our moderators: species, population or sex. The specific pattern of size- and condition-dependent physiology reported for a unique combination of species, population, and sex, thus generally predicted those in others. Patterns of size- or condition-dependent behaviour (i.e. 'personality'), or behaviour-physiology syndromes reported in specific datasets, by contrast, did not. These findings call for studies revealing the ecological background of this variation and highlight the value of study replication to help understand whether patterns of phenotypic integration reported in one study can be generalized.