Covariance between predation risk and nutritional preferences confounds interpretations of giving‐up density experiments

Giving‐up density ( GUD ) experiments have been a foundational method to evaluate perceived predation risk, but rely on the assumption that food preferences are absolute, so that areas with higher GUD s can be interpreted as having higher risk. However, nutritional preferences are context dependent and can change with risk. We used spiders and grasshoppers to test the hypothesis that covariance in nutritional preferences and risk may confound the interpretation of GUD experiments. We presented grasshoppers with carbohydrate‐rich and protein‐rich diets, in the presence and absence of spider predators. Predators reduced grasshopper preference for the protein‐rich food, but increased their preference for the carbohydrate‐rich food. We then measured GUD s with both food types under different levels of risk (spider density, 0–5). As expected, GUD s increased with spider density indicating increasing risk, but only when using protein‐rich food. With carbohydrate‐rich food, GUD was independent of predation risk. Our results demonstrate that predation risk and nutritional preferences covary and can confound interpretation of GUD experiments.