FISHERY BULLETIN; VOL. 85, NO. 3 



Pr(s) = 1 - [SS(s)9 + SSis)s/SS{s)^ + SS(s)s 



+ SSin)^ +SSin)j] 



The classification of the 1984 data was exam- 

 ined to see whether the northern and southern 

 cells fell into the same geographic areas as those 

 which originally defined the growth types de- 

 scribed by Lynde et al. (fn. 2) and the growth 

 curves of Francis and Hollowed (i.e., the north- 

 west slope and Aleutian Basin areas or the south- 

 east shelf and slope areas). 



Analysis of Fecundity 



Collections of walleye pollock ovaries for fecun- 

 dity analysis were subsampled by geographic 

 area to provide at least four ovaries in each 5 cm 

 length interval over the entire available length 

 range. Ovaries were collected from walleye pol- 

 lock in all areas except the northwest shelf, where 

 observer coverage was inadequate. The oocytes 

 were separated from the ovarian tissue by wash- 

 ing them through successively finer meshed 

 sieves; and a volumetric subsampling method, 

 similar to that described in Gunderson (1977), 

 was used to estimate the total number of oocytes 

 in each ovary. In all ovaries examined, there was 

 a clear separation in the sizes of unyolked (un- 

 counted) and yolked (counted) oocytes. The histo- 

 logical analysis (described below) indicated that 

 ovaries showing this separation in egg size modes 

 were at the stage of maturity suitable for fecun- 

 dity counts. A minimum of five subsamples were 

 counted for each ovary; if the coefficient of varia- 

 tion between subsample counts exceeded 10%, 

 more counts were done, to a maximum of 10 

 counts. It was discovered that recounts of individ- 

 ual subsamples accounted for very little of the 

 overall variation between counts, so these were 

 not done on a regular basis. 



The average coefficient of variation between 

 subsample counts for each ovary was 5.20% 

 (range, 0.87 to 10.30%). The number of oocytes 

 was estimated for a total of 115 ovaries. 



Fork length, ovary-free weight, and the mean 

 of the subsample estimates of fecundity for each 

 fish were used to derive three relationships for 

 each area within the Bering Sea: length- 

 fecundity, length-weight, and weight-fecundity. 

 Nonlinear least-squares regression methods 

 (Dixon 1983) were used to estimate parameters 

 for each relationship. Unequal error variances 

 were observed in the dependent variables of 



weight and fecundity. This was accounted for by 

 the use of weighting factors in the nonlinear re- 

 gression procedure. The weighting factor used 

 was the inverse of the variance of the dependent 

 variable. 



Because comparison of linear regressions is the 

 most direct method and because the linear and 

 nonlinear regressions showed the same relative 

 positions for each area, Newman-Keuls multiple 

 range tests (Zar 1974) were done on the slopes and 

 intercepts of the equations resulting from the lin- 

 ear regressions, to examine differences in the re- 

 lationships by area. The regressions were fitted to 

 a length range (38 to 60 cm) common to all areas 

 before comparison. 



Tests performed included an overall test for co- 

 incidental regression, Newman-Keuls multiple 

 range tests on the slopes of the lines, and multiple 

 range tests for equality of the intercepts of those 

 relationships found to have equal slopes. The pro- 

 cedures outlined in Zar (1974) were followed for 

 all of these tests. If the tests indicated that the 

 relationship under examination did not differ sig- 

 nificantly between two or more areas, the data 

 from the areas were combined and a curve fitted 

 to the new set of data using the nonlinear proce- 

 dure. 



Histological Analysis 



The ovaries of walleye pollock collected for his- 

 tological analysis were classified by area, follow- 

 ing the same geographic scheme used for the 

 length-at-age and fecundity analyses. The collec- 

 tion was subsampled to obtain ovaries in a com- 

 plete range of development from each of the five 

 geographic areas. Sections were removed from 

 the central portion of the ovaries. Tanino et al. 

 (1959) have shown that there is no difference in 

 the size composition of oocytes throughout 

 walleye pollock ovaries. All sections were 6 to 10 

 |jLm in thickness and were stained with Mayer's 

 haemotoxylin and eosin. In all, 122 ovaries were 

 examined histologically. 



Oocytes were classified into 12 categories of de- 

 velopment (Hinckley 1986). The overall maturity 

 of each ovary was based on the most advanced 

 oocytes present, and each ovary was assigned to 

 one of 10 maturity classes (Hinckley 1986). 



Egg-stage frequency counts, as determined 

 from the histological slides, were used to examine 

 the process of oocyte development. A transect grid 

 was drawn on each slide, and oocyte counts for 

 each stage of development were made at each in- 

 tersection on the grid. Stage 8 (tertiary yolk) and 



484 



