82 
Fishery Bulletin 95(1 ), 1997 
histological analyses. According to the results of the 
histological analyses, swordfish were considered to 
be in spawning condition, i.e. spawning was in 
progress or imminent in the region in which the fish 
was captured, if they were either “gravid” or “spawn- 
ing or partially spent” (classes 5 and 6 of Taylor and 
Murphy [1992] ). Individuals in these conditions were 
classified as in an active (A) reproductive status. All 
others were classified as quiescent (Q). 
In our study the results of histological analyses 
represent the known reproductive status of indi- 
vidual swordfish, and the H(i) are various hypoth- 
esized classification methods (Table 2) for placing 
swordfish into categories A or Q. In some cases, these 
H(i) indicate minimum-length criteria used to deter- 
mine which data from individual swordfish should 
be included in estimates of average values of GI. To 
facilitate the examination of impacts of minimum- 
length criteria on classification methods, these cri- 
teria were treated as a component of the H(i) in our 
analyses. We do not attempt to define minimum- 
length criteria for size at maturity in this study (cf. 
Taylor and Murphy [1992]); our concern was to de- 
termine if classification methods based on GI were 
independent of such criteria. The hypotheses identi- 
fied in Table 2 as “Present study” are representative 
of a multitude of hypotheses we examined. 
The optimum value (OV) and confidence intervals 
for the value of GI to be used as criteria, GI*, to esti- 
mate the reproductive condition of individual female 
swordfish were determined by using maximum-like- 
lihood estimation procedures and the following 
model: 
erwise R = 0. Then for individual fish, i, selected at 
random, 
P (a swordfish is reproductively given its gonad in- 
dex) = KiGlY* x (1 - n(GI)) (1 ~ R \ 
Maximum-likelihood estimates of niGI ) are given by 
the number of successes in the series of trials deter- 
mined by the value of GI* in the data, divided by the 
number of trials. Thus, for our data set: [Gl v ..., GI k *, 
GI n ] and [i? p ..., R k , I? ;j ] = [the observed values 
of GI] and [the respective measure of reproductive 
status determined by histological analyses] for indi- 
vidual fish, these estimates are given by 
7T{GI) = 
(number of individuals with 
R = 1 and GI <GI* k ) 
k-1 
(number of individuals with 
.fi = 1 and GZ > GI^) 
n-k + 1 
for individuals 
with GI < GI* k 
for all others 
It follows that the optimum value, GI * , is that which 
maximizes the following log-likelihood function 
( LKLHD ): 
LKLHD = II* x ln(7r( GI)) + ( 1 - R) x ln( 1 - /r(G/))] 
Results and discussion 
Let R - 1 if a swordfish is reproductively active 
(classes 5 and 6 of Taylor and Murphy [1992]), oth- 
Table 2 
Levels of gonad indices (GI) used to classify the reproduc- 
tive activity of female swordfish and minimum eye fork 
length (EFL) criteria used to standardize statistics for com- 
parison among areas and times, as employed by various 
researchers. Formulations for GI(1) and GI(2) are given in 
the text. 
H(i) 
Classification method 
Author 
1 
GI(1) >3.0 
Kume and Joseph 
2 
GI( 1) > 7.0 and EFL > 150 cm 
Miyabe and Bayliff 
3 
GIG) > 7.0 and EFL > 160 cm 
Sosa-Nishizaki 
4 
GIG) > 4.0 and EFL >131 cm' 
Arocha and Lee 
5 
GIG) >6.0 
Present study 
6 
GI(2)> 1.37 
Present study 
1 Authors used lower-jaw fork length >150 cm. 
Results of classifying individuals as either A or Q 
based on the results of the histological analyses and 
from application of the various H(i) (treated in each 
test as the null hypothesis) are given in Table 3. It is 
clear that H(2) and H(3) fail to classify individuals 
correctly according to reproductive status as deter- 
mined by histological analyses. Under H(2) and H(3), 
individuals below the minimum-length criteria are 
not classified. About 75% of the individuals whose 
lengths were above the size restrictions stated in 
these hypotheses were classified correctly, but only 
48% of the individuals in this group that were repro- 
ductively active were correctly classified which rep- 
resents a significant type-1 error that may be ex- 
tremely costly in terms of loss of information on the 
distributions of reproductively active swordfish. How- 
ever, this is a result of the value of GI included in 
the hypotheses and not a result of restrictions placed 
on lengths of individuals included in the analyses, 
as is evidenced by the results for the other H(i). We 
also note that length was not found to be a signifl- 
