Stage III classification of Higham and Nicholson 

 (1964) were retained. They were blotted on paper 

 towels to remove excess moisture, weighed to the 

 nearest 0.1 g, split longitudinally and turned in- 

 side out, and placed in individual jars of Gilson's 

 fluid modified by Simpson (Bagenal 1967). The 

 jars were shaken to liberate all eggs. After the 

 Gilson's fluid was poured off, along with most pul- 

 verized ovarian tissue, the ova were washed and 

 decanted in water several times and forced 

 through a sieve to remove remaining fragments of 

 ovarian tissue, spread on large trays covered with 

 paper towels, and dried under incandescent lamps. 



Higham and Nicholson (1964) described four 

 stages in the maturation of ovaries. Ovaries in the 

 immature and intermediate stages contain only 

 undeveloped ova; ovaries in the maturing and ripe 

 stages contain developing as well as undeveloped 

 ova. They concluded that only maturing ova 

 ripened during each spawning period. Maturing 

 ova were described as being opaque and yellow 

 and between 0.35 and 0.78 mm in diameter. I fol- 

 lowed this description to separate immature from 

 maturing ova. I also measured fecundity by es- 

 timating the number of maturing ova in both 

 ovaries. Instead of counting ova in sample sections 

 of the wet ovary, however, I counted ova in two 

 replicate samples of the dried ova that had been 

 separated from connective tissues. Before being 

 weighed, eggs were allowed to equilibrate with air 

 humidity. Each sample was weighed to the 

 nearest 0.01 mg. If both ovaries weighed more 

 than 12 g, two samples, each weighing 1/350 of the 

 total weight, were taken. If the ovaries weighed 1 2 

 g or less, two replicate samples, each weighing 

 0.035 g, were taken, since fecundity would have 

 been difficult to estimate in samples smaller than 

 0.035 g. Proportional sampling tended to 

 minimize the counting error for a fixed amount of 

 counting effort. Ova in each sample were counted 

 under a stereoscope. The number of ova in both 

 ovaries, A^, was estimated by multiplying the 

 number of ova in the two samples, N,, by the ratio 

 of total dry weight of eggs, W, , to dry weight of eggs 

 in samples, W^ (N =A^sW,/W,). To minimize count- 

 ing error between samples, a coefficient of varia- 

 tion of 3.0^f or less was maintained. 



Preliminary calculations indicated that fecun- 

 dity could be estimated with a precision of about 

 159c if 30 fish were selected randomly from each 

 age-class. The ultimate number in each age-group 

 was age 1, 21; age 2, 34; age 3, 33; age 4, 12; and 

 age 5, 1 (Table 1). 



Table l. — Mean number of eggs (thousands! and mean ovary 

 weight (grams), by age. for Atlantic menhaden sampled from the 



'Coefficient of variation 



Fecundity 



The regressions of fecundity on ovary weight, F 

 = 6,908(OW) - 17.937(OW)2, and fecundity on 

 total fish weight, F = 293(TW) + O.'ZUiTW)^, were 

 curvilinear, but fecundity on body weight only,F 

 = 488(BW), was linear. Thefl^ values were 0.981, 

 0.675, and 0.916, respectively. Although the rela- 

 tive merits of predicting fecundity from different 

 variables are debatable (Bagenal 1967), these 

 three models seem less useful than fecundity on 

 age, which can be used to determine reproductive 

 potential and calculate life table estimates, and 

 fecundity on fish length, which can be used to 

 predict the number of eggs spawned by different 

 size classes. 



A statistical test failed to support the curvilinear 

 relation implied by a plot of fecundity on age, 

 perhaps because of the few fish in older age- 

 groups. Of the two linear models tested for es- 

 timating fecundity at age, I selected F = 

 92,592( Age ) as the better estimator (r^ = 0.879; SE 

 slope = 3,440; SE regression = 89,110). It had 

 tighter confidence limits and a higher r^ than the 

 model F = a + bL. 



A logarithmic model (log F = a + bL) was 

 selected to describe the curvilinear relation be- 

 tween fecundity and length and was fitted to both 

 my data and the data of Higham and Nicholson 

 ( 1964) (Figure 1). Values predicted by this model 

 fit observed values more closely over the entire 

 range than those predicted by the nonlogarithmic 

 model (F = fcjL -H 6.2L^). The difference in the slope 

 coefficients of the logarithmic model fitted to the 

 two sets of data was significant (P<0.001). Esti- 

 mated fecundities were in reasonable agreement 

 for fish up to 275 mm, but diverged for large fish. 

 For 3.50 mm fish the model fitted to Higham and 

 Nicholson data predicted about 1.75 as many ova 

 as the model fitted to my data. 



Differences in fish ages or in the time of year fish 

 were collected, or actual changes in fecundity 

 might account for differences in estimates of ova 



309 



