Investment in defense and cost of predator-induced defense along a resource gradient

Abstract

An organism's investment in different traits to reduce predation is determined by the fitness benefit of the defense relative to the fitness costs associated with the allocation of time and resources to the defense. Inherent tradeoffs in time and resource allocation should result in differential investment in defense along a resource gradient, but competing models predict different patterns of investment. There are currently insufficient empirical data on changes in investment in defensive traits or their costs along resource gradients to differentiate between the competing allocation models. In this study, I exposed tadpoles to caged predators along a resource gradient in order to estimate investment in defense and costs of defense by assessing predator-induced plasticity. Induced defenses included increased tail depth, reduced feeding, and reduced swimming activity; costs associated with these defenses were reduced developmental rate, reduced growth, and reduced survival. At low resource availability, these costs predominately resulted in reduced survival, while at high resource availability the costs yielded a reduced developmental rate. Defensive traits responded strongly to predation risk, but did not respond to resource availability (with the exception of feeding activity), whereas traits construed as costs of defenses showed the opposite pattern. Therefore, defensive traits were highly sensitive to predation risk, while traits construed as costs of defense were highly sensitive to resource allocation tradeoffs. This difference in sensitivity between the two groups of traits may explain why the correlation between the expression of defensive traits and the expression of the associated defense costs was weak. Furthermore, my results indicate that genetic linkages and mechanistic integration of multiple defensive traits and their associated costs may constrain time and resource allocation in ways that are not addressed in existing models.