A meta-analysis of correlated behaviors with implications for behavioral syndromes: relationships between particular behavioral traits

Behavioral syndromes predict that individuals display behaviors consistently across different ecological situations, resulting in correlations among functionally different individual-specific behaviors (e.g., activity, exploration, aggression, and risk taking). Such consistencies can arise because of the common innate government of traits (i.e., temperament). However, different behaviors can be mediated by different selection regimes and/or measured with different errors. Furthermore, contextual overlap among traits may also vary. These possibilities can cause dissimilarities in the pair-wise relationship between particular traits. To determine the relationships among the most studied behaviors, we performed a modern meta-analysis, in which we assessed the strength of correlations in each possible combination of traits. Relying on data from 81 scientific papers, we found that the correlations among behaviors were generally weak and that they varied in magnitude across comparisons (e.g., novel environment exploration and activity: r = 0.345; novel object exploration and activity: r = 0.074). The partial correlations among traits revealed that certain relationships (e.g., novel environment exploration/activity and the novel object exploration/risk taking) were independent of the covariation with other traits, whereas certain relationships (e.g., aggression/novel environment exploration) consistently weakened after controlling for covariance. Some relationships were affected by contextual overlap: the effect sizes were systematically higher when the behaviors were assayed in the same experimental compartment (e.g., same test room or aquarium). Different correlations are unlikely to emerge due to differences in repeatabilities that are associated with the measurement of different traits, as we found that averaged repeatabilities vary around the same intermediate magnitude for each behavior. We suggest that the most commonly assessed behavioral traits do not necessarily form equally independent domains.