FISHERY BULLETIN: VOL. 70, NO. 3 



northern anchovy larvae were undersampled by 

 a factor of 3.4 relative to sardine larvae because 

 of differences in retention rate and rate of de- 

 cline in catch with increase in size. His esti- 

 mates of retention rates also were based on dif- 

 ferences between observed catch curves and hy- 

 pothetical curves based on a constant rate of 

 decline. 



METHODS 



The data for this study on fish larvae were 

 taken from CalCOFI plankton tows. From 1949 

 to 1968, CalCOFI used a single silk net (type 1) 

 of 0.55-mm mesh with a 1-m diameter mouth 

 opening. During 1966-68 this net was paired 

 with one of nylon of 0.333-mm mesh (type 2) 

 with a 1/2-m diameter mouth opening. In 1969 

 the netting of the larger net was replaced with 

 monofilament nylon netting of 0.505-mm (type 

 3), The variance of mesh width of the type 3 

 net is considerably less than that of the type 1 

 net. Other details of the net characteristics are 

 available from Smith". The nets were lowered 

 and raised obliquely at a rate of 1.5 knots to a 

 depth of 140 m in 1966-1968 and to 210 m during 

 1969. Since very few anchovy or sardine larvae 

 occur below a depth of 140 m, the change in depth 

 should make no diflference in the results of this 

 study. Other details of sampling are described 

 by Ahlstrom (1966). 



All sardine and anchovy larvae in each sam- 

 ple were identified and measured to the nearest 

 0.5 mm, standard length. The data revealed 

 evidence of varying degrees of personal bias 

 towards favoring measurements of whole milli- 

 meter rather than half millimeter. Sette (1950) 

 was aware of the potential for this type of bias 

 in measuring adult fish and stated: ". . . to avoid 

 personal bias in favor of whole or half centi- 

 meter marks, the measuring scale had uniform 

 graduation marks and they were serially num- 

 bered. In addition to avoiding bias, this had the 

 advantage of giving two digit numbers for all 

 listings and computations, the data being di- 

 vided by two for conversion to centimeters at 

 the final stage of work." Perhaps it would be 



prudent to follow the advice of Sette, if 0.5 mm 

 accuracy is desired. Evidence of personal bias 

 in measuring the smallest sizes of larvae was 

 also noted. Since the smallest larvae are often 

 distorted, it is difficult to make objective mea- 

 surements. Because of the above described bi- 

 ases, measurements of the larvae are grouped 

 into the intervals shown in Tables 1 and 2. 

 Larvae captured by the type 2 net were multi- 

 plied by the ratio of volume of water sampled 

 by the type 1 net to the volume of water sam- 

 pled by the type 2 net to adjust for the smaller 

 size of the type 2 net. The samples were chosen 

 on the basis of the presence of sardine larvae or 

 moderate numbers of anchovy larvae. 



The following equations, using the notation of 

 Regier and Robson (1966), were used to esti- 

 mate mesh retention of the type 1 net 



Wijfc = SijUoik + Cifc 



(1) 



' Paul E. Smith, National Marine Fisheries Service, 

 Southwest Fisheries Center, La JoUa, Calif. 



where nuk — number of larvae of size Lj caught 

 by type i net in Mh sample. 



Sij = mesh retention of type i net to 

 larvae of size Lj, i.e., 



Sij = riij/Nj 



where Nj = either absolute or relative num- 

 ber of larvae of size Lj in the 

 population. 



Cjic = error term that is assumed to be 

 normally distributed and inde- 

 pendent of W2jfc. 



The use of equation (1) to estimate mesh re- 

 tention of the type 1 net implies the assumption 

 that S2i is 1 for all ;, i.e., Nj :=:: n-zj. Preliminary 

 analysis of data obtained from a series of paired 

 samples of the type 2 net and a finer meshed 

 net indicates that this assumption is valid for 

 northern anchovy and Pacific sardine (P. E. 

 Smith, personal communication). This method 

 of estimating Sij from a known or estimated 

 Nj is noted as the "direct approach" in the term- 

 inology of Regier and Robson (1966). 



The least squares estimate of Sij is given by 

 Cochran (1963) as 



840 



