448 



Fishery Bulletin 90(3), 1992 



statements about whether the relationship is, in real- 

 ity, weak or strong cannot be made because the 95% 

 confidence limits around the sample correlation are 

 wide (Fig. 5). Published correlations between early life 

 stages and recruitment vary greatly in strength (e.g., 

 Peterman et al. 1988, Stevenson et al. 1989, Gushing 

 1990); unfortunately with short data series, true dif- 

 ferences in the biology of these species cannot be 

 distinguished from sampling error. The correlations 

 between early larval abundances and recruitment com- 

 piled by Peterman et al. (1988) also illustrate this point: 

 in only 1 of 7 cases do the 95% confidence limits around 

 R^ not include both 0.2 and 0.8. This problem of low 

 precision is less serious when the true correlation is 

 likely to be fairly high (Fig. 5, metamorphs). 



The precision of correlations is also relevant to anal- 

 yses involving oceanographic or climatic variables and 

 recruitment. These studies usually invoke hypotheses 

 that the environmental variables are agents of larval 

 mortality, either through transport or their effects on 

 the production or concentration of larval food (Shep- 

 herd et al. 1984, Hollowed and Bailey 1989); therefore, 

 their true correlations with recruitment can be no 

 stronger than the correlations for mortality rates 

 directly (Fig. 4). Yet the sampling variability of R'^ for 

 a short series of data suggests that there will be a good 

 chance of finding at least one strong sample correla- 

 tion among a group of 4-5 predictors that may be, in 

 reality, only weakly related to recruitment. Adding 

 more data will result in the sample correlation declin- 

 ing towards p ; this frequently results in the sample cor- 

 relation becoming nonsignificant (Koslow et al. 1987, 

 Walters and Collie 1988, Prager and Hoenig 1989). My 

 model results suggest environmental variables will be 

 strongly correlated with recruitment only if the en- 

 vironmental factor is related to mortality across all 

 prerecruit stages (e.g.. Fig. 4; covariance models). 



In summary, my analysis indicates that it is unlikely 

 that estimates of abundance of survival rates of the egg 

 and early-larval stages of marine fish will lead to useful 

 predictions of recruitment. Although mortality in the 

 earliest life stages is a major source of recruitment 

 variability, the late-larval and juvenile periods are also 

 important. Peterman et al. (1988), Fritz et al. (1990), 

 and Pepin and Myers (1991) argue for the need for coor- 

 dinated research on all prerecruit stages, rather than 

 focusing only on the early stages, and my results sup- 

 port this view. The modeling approach I have developed 

 here can be easily modified for any particular species 

 to estimate a priori the likelihood of success of pro- 

 posed recruitment research and to suggest particular- 

 ly fruitful avenues of investigation. 



Acknowledgments 



This analysis would not have been possible without 

 the efforts of many scientists in estimating the vital 

 rates of fish populations in multiyear studies. I thank 

 G. Gabana for discussions of variability and correlations 

 and for help collecting data. This paper has been im- 

 proved through the comments of G. Gabana. M. La- 

 Pointe, B. McKenzie, P.M. Peterman and D. Roff, and 

 two reviewers. Partial support was provided by post- 

 graduate fellowships from the National Science and 

 Engineering Research Gouncil of Ganada (NSERG) and 

 the Max Bell Foundation, and NSERG operating 

 grants to D. Roff. 



Citations 



Ahlstrom. E.H. 



1954 Distribution and abundance of egg and larval populations 

 of the Pacific sardine. Fish. Bull.. U.S. 56:83-140. 

 Alexander, G.R., and D.S. Shatter 



1969 Trout production and angling success from matched 

 plantings of brook trout and rainbow trout in East Fish Lake, 

 Michigan. J. Wildl. Manage. 33:682-692. 

 Al-Hossaini, M., Q. Lui, and T.J. Pitcher 



1989 Otolith microstructure indicating growth and mortality 

 among plaice, Pleurmiectes platessa, post-larval sub-cohorts. J. 

 Fish. Biol. 35(Suppl. A):81-90. 

 Allen, R.L. 



1977 Status of the upper Columbia river salmon and steelhead 

 runs. In Schwiebert, E. (ed.), Columbia river salmon and 

 steelhead, p. 23-30. Am. Fish. Soc. Spec. Ptibl. 10, Bethesda. 



Anderson, J.T. 



1984 Early life history of redfish {SebaMes spp.) on Flemish 



Cap. Can. J. Fish. Aquat. Sci. 41:1106-1116. 

 1988 A review of size dependent survival during pre-recruit 

 stages of fishes in relation to recruitment. J. Northwest Atl. 

 Fish. Sci. 8:55-66. 

 Bannister, R.C.A. 



1978 Changes in plaice stocks and plaice fisheries in the North 

 Sea. Rapp. P.-V. Reun. Cons. Int. Explor. Mer 172:86-101. 



Bannister, R.C.A. , D. Harding, and S.J. Lockwood 



1974 Larval mortality and subsequent year-class strength in 

 the plaice {Pleunmfdes platessa). In Blaxter, J.H.S. (ed.), The 

 early life history of fish, p. 21-37. Springer-Verlag. NY. 

 Barnaby, J.T. 



1944 Fluctuations in abundance of red salmon, Oncorh i/nch us 

 nerka. of the Karluk River, Alaska. Fish. Bull., U.S. 50: 

 237-295. 

 Bradford. M.J. 



1991 Effects of ageing errors on recruitment time series esti- 

 mated from sequential population analysis. Can. J. Fish. 

 Aquat. Sci. 48:555-558. 

 Buckley, L.J., and R.G. Lough 



1987 Recent growth, biochemical composition, and prey field 

 on larval haddock {Melanogrammus aeglefinus) and Atlantic 

 cod {Gadus morhua) of Georges Bank. Can. J. Fish. Aquat. 

 Sci. 44:14-25. 



