356 



Fishery Bulletin 102(2) 



o f=0 

 + f = 0.35 

 A f= 1.50 



1.000 



School size 



T 2.000 



Figure 3 



Simulated dynamics of school-size distributions using m dl as the mortal- 

 ity model and w as the aggregation model. This simulation started with 

 aggregated initial conditions (S = 10 6 ). The aggregation parameter was 

 a=10~ 6 . The top panel shows school-size distributions lin log-log scale) at 

 selected times (t). The lower panel gives the mean school size Uibar) and 

 school abundance IN) versus time. 



size (m dk ) led to poorer survival with preferential attach- 

 ment (Fig. 8, C and D). 



Discussion 



Koslow (1992), Rickman et al. (2000), and others have 

 commented on the inherent variability in stock-recruit- 

 ment data and the limited predictive power of determin- 

 istic stock-recruitment models. Therefore, there is no 

 expectation that one could select the models described 

 here over other stock-recruitment models on the basis of 

 fits to data. Although the aggregation-mortality models 



may be fitted to stock-recruitment data, the real useful- 

 ness is as a guide to selection of stock-recruitment models 

 used in management, as a mechanism for integrating 

 research on recruitment processes, and as a tool for explor- 

 ing the structure of recruitment variability. 



The aggregation-mortality models were introduced 

 by using an analogy with evolving random networks 

 ( Barabasi and Albert, 1999 ) and were shown to be analyti- 

 cally equivalent (Appendix 2). Modeled fish are subjected 

 to competing forces of organization (aggregation) and decay 

 (mortality), as in a network in which links to nodes in the 

 network are created, destroyed, and rewired (Albert and 

 Barabasi. 2002). An important finding of Barabasi and 



