Senescence is defined in terms of age, and theories for the evolution of senescence were developed with age-dependent demographic models, in which the selection gradient against mortality declines (often dramatically) with age. Many types of organisms, however, pass through strikingly different stages in their life histories, and making age an inadequate or irrelevant individual state variable. Stage-classified models are often used to capture this life cycle complexity, but doing so makes it difficult to evaluate senescence per se, because age does not appear in the model. This talk answers two questions (1) how to compute selection gradients on age-specific mortality from a stage-specific demographic model and (2) what kinds of patterns of age-specific mortality appear in perennial plant life histories.
(1) Age- and stage-dependence can be included in a model constructed using vec-permutation matrix methods. The model is based on three sets of matrices, describing stage transitions, aging, and reproduction. Selection gradients can be calculated from the resulting model using matrix calculus. (2) In contrast to age-classified models, models for perennial plants including both age and stage can generate selection gradients that increase, rather than decrease, with age. Selection on senescence in stage-classified populations may thus differ fundamentally from that in age-classified populations.