Abstract. - To test the hypoth- 

 esis that year-class strength in ma- 

 rine fishes is determined in the early- 

 larval stages, and that these stages 

 can be used to predict recruitment, 

 I modeled the recruitment process 

 using a modified form of key-factor 

 analysis. Using data compiled from 

 the fish literature, I found a signifi- 

 cant relationship {K" 0.90, P« 0.001, 

 n 97) between the mean and interan- 

 nual variance of stage-specific mor- 

 tality rates that provided variance 

 estimates for the model. The R'~ 

 values for the true correlation be- 

 tween abundances of small larvae 

 and subsequent recruitment for four 

 example species of marine fish were 

 predicted to lie between 0.10 and 

 0.57, depending on the assumptions 

 of the model. I therefore suggest 

 that recruitment levels are 'ixed 

 after the early-larval period. How- 

 ever, the precision of sample correla- 

 tions are too low (10-yr data series) 

 to empirically test whether abun- 

 dances or mortality rates of early lar- 

 vae are in reality strongly or poorly 

 correlated with recruitment. After 

 metamorphosis, the strength of the 

 true relationship and the precision of 

 sample correlations increase suffi- 

 ciently to permit precise forecasting 

 of recruitment. Recruitment is a 

 complex process in which variation 

 in all life stages contributes substan- 

 tially to the variability in final abun- 

 dance; therefore, researchers should 

 recognize the importance of the later 

 prerecruit stages and the interac- 

 tions among all stages. 



Precision of recruitment 

 predictions from early 

 life stages of marine fishes 



Michael J. Bradford 



Department of Biology, McGill University 

 1205 Ave. Dr Penfield, Montreal, Quebec H3A-1BI, Canada 

 Present address: Department of Fisheries and Oceans, West Vancouver Laboratory 

 4160 Marine Drive, West Vancouver, BC V7V IN6, Canada 



Manuscript accepted 6 May 1992. 

 Fishery Bulletin, U.S. 90:439-453 (1992). 



A major problem in the management 

 of marine fisheries is the unpredict- 

 able fluctuations in stock size result- 

 ing from variable recruitment. Hjort 

 (1913) first recognized this recruit- 

 ment variability and proposed a num- 

 ber of hypotheses that linked the 

 survival of small first-feeding larvae 

 and subsequent year-class strength 

 (reviewed by Wooster and Bailey 

 1989). These hypotheses have formed 

 the basis of much research on the 

 early life history of fishes, research 

 which has been largely focused on the 

 'critical period' during the transition 

 from endogenous to exogenous 

 modes of feeding. While much has 

 been learned about the biology of 

 larval fish, the evidence for a critical 

 period of increased mortality and a 

 link between the larval stage and 

 recruitment remains equivocal (May 

 1974, Ware and Lambert 1985, 

 Peterman et al. 1988, Campana et al. 

 1989). Explanations for this failure 

 have ranged from sampling and 

 technical difficulties, such as inap- 

 propriate scales of sampling (Leggett 

 1986, Taggart and Leggett 1987, 

 McGurk 1989), to the suggestion that 

 no such critical period exists, and 

 that all prerecruit stages contribute 

 to some degree to variability in year- 

 class strength (Sissenwine 1984, An- 

 derson 1988, Peterman et al. 1988). 

 There have been few attempts to 

 model the recruitment process to 

 assess the likelihood that early-larval 

 mortality is a dominant feature of 

 year-class variability. Manipulations 



of life tables have shown that small 

 changes in larval mortality have the 

 potential to cause great variation 

 in recruitment (Smith 1985, Houde 

 1987 and 1989, Pepin and Myers 

 1991); however, the influence of the 

 larval stages in a fully dynamic model 

 incorporating variability in all stages 

 has not been investigated. In par- 

 ticular, the role of postlarval mortal- 

 ity in causing recruitment variability 

 is unclear and has been the cause of 

 some controversy (Sissenwine 1984, 

 Peterman et al. 1988, Taggart and 

 Frank 1990, Wooster and Bailey 

 1989). 



An often-stated justification for 

 research on the early life history of 

 fish is to provide short-term forecasts 

 of recruitment, thereby allowing 

 managers to adjust fishing regula- 

 tions in response to changes in stock 

 size (Gulland 1989). While it is obvi- 

 ous that the stages very close to re- 

 cruitment will give the most accurate 

 predictions, sampling these stages is 

 often difficult and expensive (Smith 

 1985), unless they are caught inciden- 

 tally in other fisheries. Rather, ef- 

 forts have usually been concentrated 

 on finding a predictive relationship 

 between recruitment and the abun- 

 dance or some measure of survival of 

 larvae and recruits, on the working 

 assumption that Hjort's hypothesis of 

 year-class determination at this early 

 stage is valid (Peterman et al. 1988, 

 Gushing 1990). 



The utility of short-term (i.e., an- 

 nual) predictors of recruitment in the 



439 



