scientific correspondence Figure 1 Mortality of core and from the effects of overcrowding when 26 9 marginal offspring prove redundant . We marginal offspring. E denotes a Experimental Unmanipulated conducted field studies on red-winged black- broods broods clutch enlargement; R a clutch 50 birds (Agelaius phoeniceus) over seven field 19 reduction; Cs, sham control seasons at Norman, Oklahoma (1991– 93) 147 broods; Cu, unmanipulated con- and Winnipeg, Manitoba (1993–96). The 40 trol broods; NHF, unmanipulated clutch size of blackbirds at the two sites 199 nests with no hatching failure; ranged from two to five eggs (mode, four). HF, unmanipulated broods with 6 30 Core and marginal offspring were defined hatching failure. Vertical lines 52 on the basis of the day of hatching. Brood 32 represent 95% confidence inter- size at eight days post-hatch was used as an 20 28 vals with sample sizes shown index of fledging success. Site differences in 42 above. 24 the average mortality of core and marginal 156 10 226 chicks were small and non-significant; data of nestlings (%) Mortality rate on nestling survival were therefore pooled. 70 We experimentally enlarged or reduced 0 clutches by removing or adding single eggs E CsCu RECsCu R NHFHF NHF HF Core Marginal Core Marginal during egg-laying. In a small number of nestlings nestlings nestlings nestlings cases, we enlarged or reduced broods by moving a chick at hatching. Sham control broods without hatching failure was more asynchrony creates a caste of low-cost, dis- broods were created by reciprocally swap- than five times higher than for core offspring posable offspring. ping eggs across nests to control for any in unmanipulated nests with no hatching Scott Forbes, Suzanne Thornton potential effects of handling. The mortality failure (Fig. 1). But the phenotypic handicap Department of Biology, 515 Portage Avenue, of core and marginal chicks in sham control was reversible, as predicted under insurance University of Winnipeg, Winnipeg, broods was virtually identical to that in theory. Marginal offspring mortality was Manitoba R3B 2E9, Canada unmanipulated broods (Fig. 1), thus the lower in non-experimental broods with e-mail: [email protected] experimental treatment had no discernible hatching failure than in broods without Barb Glassey effect. To avoid pseudoreplication, the mor- hatching failure (D40.199, P40.002; Fig. Department of Zoology, University of Manitoba, tality of core or marginal chicks was averaged 1), and marginal offspring mortality was Winnipeg, Manitoba R3T 2N2, Canada for each brood, and two-tailed comparisons lower in experimentally reduced broods Margaret Forbes, Neil J. Buckley based on the difference in sample means (D) than that in unmanipulated control Department of Zoology, University of Oklahoma, were performed using approximate random- broods (D40.187, P40.012), sham control Norman, Oklahoma 73019, USA 10 4 4 ization : 95% confidence intervals were esti- broods (D 0.184, P 0.068), or enlarged 1. Mock, D. W. & Parker, G. A. The Evolution of Sibling Rivalry 11 mated by bootstrapping . broods (D40.302, P40.001). (Oxford Univ. Press, 1997). Mortality for marginal offspring in As predicted by insurance theory, the 2. Magrath, R. D. Biol. Rev. Camb. Phil. Soc. 65, 587–622 (1990). phenotypic handicap buffered core offspring 3. Stoleson, S. H. & Beissinger, S. R. Curr. Ornithol. 12, 191–270 Core addition Marginal addition (1995). +0.063 (n =17) +0.105 (n =16) +0.007 (n =11) –0.003 (n =10) from the effects of overcrowding when core 4. Dorward, E. F. Ibis B 103, 74–220 (1962). failure did not occur. Our experimental 5. Lundberg, S. Oikos 45, 110–117 (1985). enlargements and reductions of the core and 6. Cash, K. J. & Evans, R. M. Behav. Ecol. Sociobiol. 18, 413–418 marginal brood show that the number of (1986). 7. Mock, D. W. & Forbes, L. S. Trends Ecol. Evol. 10, 130–134 core offspring influenced the mortality of (1995). marginal offspring strongly but the reverse 8. Forbes, L. S. & Lamey, T. C. J. Theor. Biol. 180, 247–256 (1996). Core mortality Marginal mortality was not true. Core brood reductions, in par- 9. Forbes, L. S. J. Theor. Biol. 147, 345–359 (1990). [0.069 n = 156] [0.353 n =147] 10. Noreen, E. W. Computer-intensive Methods for Testing ticular, resulted in a sharp reduction of mar- Hypotheses: An Introduction. (Wiley, New York, 1989). ginal offspring mortality, whereas the 11. Efron, B. The Jackknife, the Bootstrap and Other Resampling number of marginal offspring had virtually Methods (Soc. Ind. Applied Math., Philadelphia, 1982). no effect on core offspring mortality, even 12. Ricklefs, R. E. Smithsonian Contr. Zool. 9, 1–48 (1969). when the marginal brood was experimental- ly enlarged (Fig. 2). Thus, whether marginal Correction +0.017 (n = 27)–0.310* (n = 23) +0.031 (n =15) –0.057 (n =9) Core reduction Marginal reduction chicks lived or died was rendered inconse- In “Tree-ring dating the 1700 Cascadia earthquake” quential to the fate of core offspring by the by David K. Yamaguchi et al. (389, 922–923; 1997) Figure 2 Effects of core and marginal brood reduc- competitive asymmetry imposed by hatching the final paragraph contained some ambiguity. A tions and enlargements on core and marginal chick asynchrony. corrected version follows: mortality. Numbers in square brackets in the centre Hatching asynchrony can serve several The dates provide a simple test of earthquake represent the average mortality and sample size of functions simultaneously — protection size. Suppose that the dates excluded January 1700. core and marginal chicks in unmanipulated broods against whole brood loss, uncertain food In that case, the tsunami in Japan could not repre- with no hatching failure, used as a baseline for all supplies, or sex-ratio bias — probably sent a Cascadia rupture longer than the 650-km dis- 1–3,7–9 comparisons. Arrow width represents the point esti- reflecting a balance among these . How- tance between the region of the dated snags and mate of the effect of an experimental enlargement ever, insurance is the only general advantage the far (California) end of the subduction zone. or reduction on mortality and associated numbers of hatching asynchrony in birds, being the Because a 650-km-long rupture at Cascadia would give the exact difference between unmanipulated only viable explanation for obligate brood be too small for magnitude 9 (ref. 7), dates exclud- 2,4,6,9 broods with no hatching failure and experimental reduction . All birds, including facultative ing January 1700 would strengthen geophysical broods. Numbers in parentheses are sample sizes brood-reducers, face the risk of early off- arguments8 against magnitude 9 earthquakes at 5,12 for experimental broods. Asterisks indicate signifi- spring mortality . Parents can protect Cascadia. By instead converging on January 1700, cant (P*0.05) differences between unmanipulated themselves against this contingency, and the dates mean that the northwestern United States control and experimental broods in approximate enjoy the benefits of insurance, by holding a and adjacent Canada are plausibly subject to earth- randomization tests of differences in means. low-cost replacement in reserve. Hatching quakes of magnitude 9. 352 Nature © Macmillan Publishers Ltd 1997 NATURE | VOL 390 | 27 NOVEMBER 1997 scientific correspondence Perls et al.3 of a relationship between longevi- 1. Perls, T. T., Alpert, L. & Fretts, R. C. Nature 389, 133 (1997). ty in women and the age at which they bore 2. Brawer, J. R. et al. Biol. Reprod. 49, 647–652 (1993). Mixed blessings for 3. MacMahon, B. et al. J. Natl Cancer Inst. 50, 21–42 (1973). children. Perls et al. showed that women sur- 4. Brinton, L. A. et al. Am. J. Obst. Gynecol. 167, 1317–1325 (1992). viving to at least age 100 were four times 5. Berstein, L. M. Oncoendocrinology of Smoking (Nauka, St middle-aged mothers more likely to have had children while in Petersburg, 1995). 6. Hytten, F. E. & Leitch, I. The Physiology of Human Pregnancy their 40s than women who survived to age (Blackwell Scientific, Oxford, 1971). 1 Perls et al. in Scientific Correspondence 73. But it is not possible to argue from these 7. Gavrilov, L. A. et al. Proc. Acad. Sci. (Russia) 354, 569–572 present intriguing data suggesting that data that “the driving selective force of hu- (1997). bearing children in middle age may be a man lifespan is maximizing the period of harbinger for extreme longevity. We have time during which women can bear child- reanalysed another study, which does not ren” without a comparison of the total num- lead to the same conclusion. ber of offspring produced (lifetime repro- Tree-ring dating the 1700 Perls et al. report a significant fourfold duction, LR) of the two groups of women. excess of births in middle age (N419 wo- Natural selection favours genes that Cascadia earthquake men age 40 or older) among 78 female cente- increase LR2,4, and consequently analyses narians when compared with a group of 54 must be based on LR if they are to conclude Geological evidence shows that an earth- women in the same birth cohort (born in that the length of the reproductive period in quake attended by a tsunami, or a series of Massachusetts in 1896) who survived only to women is a driving force in the evolution of such earthquakes, ruptured at least 900 km the age of 73. We re-examined this question human lifespan. The interpretation of Perls of the Cascadia subduction zone along the using data derived from a large case-control et al. would be supported if women in the west coast of North America between the of postmenopausal breast cancer in Massa- centenarian group had a larger LR than years 1700 and 17201. Satake et al.2 found chusetts, New Hampshire and Wisconsin2.