1 10 
Fishery Bulletin 11 7(1-2) 
translucent band and one opaque band on the corpus 
calcareum when the thin-sectioning method was used. 
We counted the number of convex structures for the 
burn method and translucent bands for the section¬ 
ing method. A single reader (reader 1: senior author) 
twice counted bands at 2 different times without prior 
knowledge of specimen length. A third count was made 
if the first and second counts did not coincide. If the 
third count was the same as either the first or second 
count, the duplicated measure was used in analysis; if 
the third count did not agree, a sample was excluded 
from analysis. A random subsample of 200 individuals 
was read by a separate reader (reader 2: S. Tanaka) to 
ensure consensus in interpretation of growth bands. To 
evaluate inter- (both readers) and intrareader (reader 
1) aging precision, an index of average percentage error 
(IAPE) (Beamish and Fournier, 1981) and mean coef¬ 
ficient of variation (CV) (Chang, 1982) were calculated. 
An age-bias plot also was constructed to test inter- and 
intrareader counts (Campana et ah, 1995). 
For age estimation, we assumed a tentative birth 
date of 1 June on the basis of the birth season esti¬ 
mated by Fujinami et al. (2017). The first band (birth 
band) was considered to be formed after parturition 
on the basis of the observation of vertebral centra of 
near-term embryos and neonates (see the “Results” 
section). In addition, we assumed subsequent growth 
bands formed annually, on 1 December (see the “Re¬ 
sults” section); therefore, the age of each specimen was 
calculated as follows: 
Age = (a - l) + ^^(a > 1,1 < [3 < 12), (1) 
12 
where a = the number of convex structures (translucent 
bands) deposited after the birth band; and 
(3 = the month when the individual was caught. 
Age verification 
To verify periodicity of growth-band formation, the 
most peripheral structure on each centrum was classi¬ 
fied as either convex (translucent) or concave (opaque). 
We analyzed monthly changes in frequency of each 
band on the centrum edge throughout the year. The 
periodicity of growth-band pairs was verified by using 
a statistical model developed by Okamura and Semba 
(2009). Three models were constructed according to dif¬ 
ferent periodicities of growth-band formation: an an¬ 
nual cycle, a biannual cycle, or no seasonal cycle. Oka¬ 
mura and Semba (2009) suggested that the model with 
the lowest Akaike’s information criterion (Burnham 
and Anderson, 2002) is preferred because it is estimat¬ 
ed to be closest to the unknown reality that generated 
the data. Vertebral centra with only one band (birth 
band) were excluded from analysis. 
Growth analysis 
The von Bertalanffy growth function (von Bertalanffy, 
1938) was fitted to observed length-at-age data by us¬ 
ing the maximum likelihood approach with the op- 
tim function in R (vers. 3.3.0; R Core Team, 2016), as 
follows: 
L t = LJl-e- k(t - t » > ), (2) 
where L t = the predicted length at age t (in years); 
L„- the theoretical asymptotic length (in 
centimeters); 
k = the growth coefficient (per year); and 
£ 0 = the theoretical age at zero length. 
We used Kimura’s likelihood ratio to test for a signifi¬ 
cant difference in the growth parameters of males and 
females (Kimura, 1980). We tested the null hypothesis 
(H 0 , all parameters are different between sexes) ver¬ 
sus the alternative hypothesis (Hi-t^, the sex-specif¬ 
ic growth model, in which one of the parameters is 
shared for each sex, and H 4 , all parameters are shared 
between sexes). The 95% confidence intervals (95% CIs) 
of parameter estimates were derived from 2000 resa¬ 
mpled data sets by using the bootstrap method. 
Theoretical longevity U max ) was estimated following 
methods of Taylor (1958) and Fabens (1965): 
t max =t o- ln(Q '° 5) (Taylor, 1958) and (3) 
k 
/ max = 5^“7^ (Fabens, 1965). (4) 
K 
Age at maturity and maternity 
According to criteria described by Fujinami et al. (2017), 
sexual maturation in males was classified into 3 stages 
on the basis of calcification of the claspers and testis 
development (for details, see Suppl. Table 1) (online only): 
1) immature juvenile, 2) maturing juvenile, and 3) ma¬ 
ture adult. For females, sexual maturity was assessed 
on the basis of uterine width, ovarian development, 
and the presence of embryos or fertilized eggs, with 5 
stages recognized (for details, see Suppl. Table 1) (on¬ 
line only): 1) immature juvenile, 2) maturing juvenile, 3) 
mature adult, 4) mature pregnant, and 5) mature post¬ 
partum. The maturation stage of each specimen was 
converted into binary data (immature=0, mature=l) at 
age intervals of 1 year for statistical analysis. A logistic 
regression model was fitted to the binomial maturity 
data, to determine ages at 50% maturity for both sexes, 
as follows: 
Y = 1 / [1 + exp{ -(a + (3/)}], (5) 
where Y - the proportion of mature individuals of each 
age; 
X = age; and 
a and (3 = coefficients. 
A generalized linear model with a binomial error struc¬ 
ture and logit-link function was used to estimate the a 
and (3 coefficients by using R. Female maternity ogive 
also was determined by using maternal condition ac¬ 
cording to criteria described by Fujinami et al. (2017) 
(for details, see Suppl. Table 1) (online only). Data for 
maternal condition (pregnant or postpartum) or non- 
maternal condition (immature or non-pregnant mature) 
