DeMartini and Lau: Criteria for size at maturity in snappers 



451 



pending on ovary size, and including the ovarian 

 wall) was taken from the central one-third of either 

 the right or left ovary (random choice) and placed in 

 107c formalin sea-water. 



Laboratory processing 



Most ovary specimens were processed within 1-2 

 months after collection. Diameters of 25 of the larg- 

 est viable oocytes were then measured (random axis, 

 63x magnification) for each specimen with a dissect- 

 ing microscope. The median of 25 diameters provides 

 a cost-efficient estimator of average maximum oo- 

 cyte size for ehu (Lau and DeMartini, 1994). 



For each preserved ovary specimen, a subsample 

 of the ovary was dehydrated, imbedded in paraffin, 

 and a minimum of three sections was cut (at 6 |im) 

 and stained with Harris's hematoxylin, followed by 

 eosin counterstain (Hunter and Macewicz, 1985). 

 Slide sections were examined with a compound mi- 

 croscope at 40-500x for the presence and relative 

 quantity of eosinophilic yolk ( unyolked oocytes: class 

 1 of Murphy and Taylor, 1990; partly yolked: classes 

 2-3; yolked: classes >4). If oocytes were yolked, we 

 further noted the presence of postovulatory follicles 

 (POF: class 6), hydrating or hydrated oocytes (HYD: 

 class 7), and oocytes undergoing a or /3 atresia (class 

 8: Murphy and Taylor, 1990). Individuals were des- 

 ignated as "immature" if the most advanced oocytes 

 were unyolked or partly yolked without substantial 

 atresia, "ripening mature" if fully yolked but lacking 

 POF or HYD, "resting mature" if a majority of the 

 yolked oocytes were atretic, or "ripe mature" if POFs 

 or HYDs were present. Fish that might in the future 

 have resorbed yolked oocytes without spawning were 

 necessarily classified as mature. 



Statistical analyses 



We used oocyte volume (OV) to characterize egg size 

 because a volumetric measure was more directly pro- 

 portional to OW than was a linear measure such as 

 oocyte diameter. OV was calculated as the volume of 

 a sphere, 4/3 m-^. where r was one-half of the median 

 diameter of the largest oocytes (Lau and DeMartini, 

 1994). The sampling distributions of dependent and 

 regressor variables were first examined by using raw 

 and log-transformed data. Log-transformed data 

 were least skewed; logarithms (base 10) were there- 

 fore used in all subsequent analyses. 



We used analysis of covariance (ANCOVA) to de- 

 termine if OW was suitable for quantifying matura- 

 tion state of individual fish after adjusting OW for 

 body size, egg size, or both measures. For each spe- 

 cies, the analysis evaluated fish of three ovarian de- 

 velopmental stages (unripe immature, ripening or 

 resting mature, and ripe mature), identified histo- 

 logically. We were thereby able to evaluate the po- 

 tentially confounding effects of oocyte development 

 and body size on ovary weight ( Erickson et al., 1985b). 



We further used two classification techniques (dis- 

 criminant analysis and logistic regression) to distinguish 

 between mature and immature fish. These two techniques 

 were used because we recognized the general need, in 

 studies such as this, to classify the maturity status of 

 individual fish 1 ) based on both gross morphometries 

 and histological criteria and 2 ) with gross metrics alone, 

 without the added benefit of histological evidence. At a 

 minimum, ovary tissues of some of the fish examined 

 in any study need to be examined histologically as a 

 check on the accuracy of gross classification. 



Both logistic regression and discriminant analy- 

 sis can be used to classify dichotomous states such 



