Quantifying personality in the terrestrial hermit crab: Different measures, different inferences

There is much interest in studying animal personalities but considerable debate as to how to define and evaluate them. We assessed the utility of one proposed framework while studying personality in terrestrial hermit crabs (Coenobita clypeatus). We recorded the latency of individuals to emerge from their shells over multiple trials in four unique manipulations. We used the specific testing situations within these manipulations to define two temperament categories (shyness-boldness and exploration-avoidance). Our results identified individual behavioral consistency (i.e., personality) across repeated trials of the same situations, within both categories. Additionally, we found correlations between behaviors across contexts (traits) that suggested that the crabs had behavioral syndromes. While we found some correlations between behaviors that are supposed to measure the same temperament trait, these correlations were not inevitable. Furthermore, a principal component analysis (PCA) of our data revealed new relationships between behaviors and provided the foundation for an alternate interpretation: measured behaviors may be situation-specific, and may not reflect general personality traits at all. These results suggest that more attention must be placed on how we infer personalities from standardized methods, and that we must be careful to not force our data to fit our frameworks. ⺠Different measurements of personality should be correlated within a specific dimension. ⺠We studied personality in terrestrial hermit crabs (Coenobita clypeatus) and focused on different ways to measure boldness and exploration. ⺠We found consistent individual behavior (personality) in both categories, but we did not find strong support that different methods to measure them were consistent. ⺠Principal components analysis suggests that these behaviors may relate to specific testing conditions, not traits. ⺠This variation suggests that more attention must be placed on how we infer personalities from standardized methods.