At the next closest breeding colony, 150 km south of Año Nuevo at Point Piedras Blancas, pups were first born in 1992 and no brands have been seen.1 A few reports of branded animals elsewhere along the coast were sent by other observers and are included in analyses (Fig. 1). We consider all sightings of females age three or older between mid-December and early March at a breeding colony as breeding events, whether a pup was seen with the female or not. The vast majority of females (>97%) on colonies during that period have pups (Le Boeuf et al. 2011). We refer to males seen during the winter at age five and above as breeding, since they are sexually

mature at that age, though they are find more not physically mature until age 8–10. We used PFT�� cell line the Cormack-Jolly-Seber mark-recapture method to estimate annual survival rate as a function of age (Cormack 1964, Jolly 1965, Seber 1965, Cameron and Siniff

2004, Hastings et al. 2011). The method allows a different survival estimate for every age, but we sought parametric models describing smooth shifts in survival with age. There are two advantages to a parametric model: first, hypotheses about maturity and senescence are about consistent changes with age, and second, parametric models add power. The model we chose was piecewise linear regression (McGee and Carleton 1970), describing linear change in three separate age categories. Because the divisions between age categories are estimated along with the regression coefficients, there are no a priori assumptions about when survival increases or decreases (Sibly et al. 1997), and regression parameters provide explicit statistical tests for age-related shifts. We tested several other models allowing increases and decreases with age: piecewise regression with two or four categories, piecewise logistic Urocanase regression, and the Siler model, and all produced broadly similar results (Appendix S1, S2). We include two alternative models in the main presentation: the model with survival differing at every age, for graphical comparison with the piecewise regression results, and a model

with constant survival across age ranges identified by piecewise regression (5–16 yr in females, 1–15 yr in males) intended to produce the best-supported estimates for future modeling studies. All models were run separately for males and females. The piecewise regression model starts with two ages, β1 and β2, with β1 <  β2, to serve as break points defining three age categories. There is a regression slope αi relating survival to age within each category i and a single intercept, π = S(m), the survival rate at age x  = m, where m is an arbitrary age, usually the midpoint of the age range (we used m = 10 yr in females, m = 8 in males). A single intercept suffices due to the constraint that fitted lines join at break points.