Conclusions about the adaptive nature of repeatable variation in behavior (i.e., “personality”) are often based on laboratory-based behavioral assays. However, the expression of genetic variation differs between laboratory and field. Laboratory-based behavior might not predict field-based behavior thus, cross-context validation is required. We therefore estimated the correlation between an established laboratory and field assay of activity in a novel environment applied to wild great tits (Parus major), both of which have been used as proxies for ‘exploration tendency’. Both behaviors had similar repeatability (R = 0.35 vs. 0.37) but differed in heritability (h² = 0.06 vs. 0.23), implying differences in evolvability. Unexpectedly, they were independent, thus reflecting expressions of two distinct underlying characters. Post hoc simulations show that sampling bias did not explain the lack of correlation.  Laboratory-based assays may reflect fear and exploration, but field-based assays escape behavior instead. Thus, in great tits, activity behaviors expressed in laboratory vs. field setups are modulated by multiple quasi-independent characters that differ in evolvability potential. The lack of cross-context correlation shown here may more generally apply to other setups, repeatable behaviors and taxa. Our study implies care should be taken in labeling behaviors prior to firm validation.

All column names should be descriptive enough to understand what variable it corresponds to. All data should also be understandable on their own (or with reading the article) apart from 'Sex' column where 1 is for females and 2 for males.