Labelle et al.: Determination of age and growth of South Pacific albacore 



657 



fisheries (Labelle, 1993a). Up to 13 rings, assumed to 

 be annuli, were visible in the vertebrae sampled. 

 Sample sizes for all relative age classes were greater 

 than 20, except for classes 1, 12,, and 13. Significant 

 differences in the standard deviation of length-at-age 

 were detected among the age classes (Bartlett's test, 

 X L> =64.683, P<0.001), and the standard deviation was 

 greatest for relative age classes 5-7. This trend differs 

 from the results obtained in the length-frequency analy- 

 sis, which indicated increasing standard deviation with 

 age (although only linear trends were possible in this 

 analysis). 



Differences in mean size between the successive age 

 classes were calculated to reveal other anomalies in 

 growth patterns. In theory, absolute growth should be 

 rapid at first and decrease progressively in later life 

 (see Ricker, 1975). The pattern observed for albacore 

 agrees with these theoretical expectations, except for 



the high growth rates observed be- 

 tween relative ages 8->9 and 

 11->12 (Table 2). No statistical or 

 biological criteria justified the 

 omission of the data associated 

 with ages 8— >9 from the analysis. 

 However, the large growth incre- 

 ment obtained for age 11-^12 could 

 simply be biased because of the 

 small sample size (2) for relative 

 age class 12. 



The VB model was fitted to the 

 aggregate vertebral-ring-count 

 data for relative age classes 2-11. 

 The data for relative age classes 

 1, 12, and 13 were excluded be- 

 cause of their small sample sizes. 

 The VB model provided a good fit 

 to the remaining data (Fig. 3). The 

 residuals were generally well cen- 

 tered around zero and the slope of 

 a linear regression of residuals 

 against relative age was not sig- 

 nificantly different from zero (F- 

 test=0.001;P=0.973). 



Of the albacore sexed, 59 were 

 females and 70 were males. As 

 these were distributed in a simi- 

 lar fashion with respect to rela- 

 tive age class as the pooled data 

 set (females, males, and unsexed), 

 the VB parameters were estimated 

 for each sex (Table 3). Likelihood 

 ratio tests (Kimura, 1980) were 

 used to test for differences in 

 growth parameters between sexes. No significant dif- 

 ferences were detected for any of the tests carried out 

 (Table 4). Therefore, the growth parameters estimated 

 from the pooled data set were considered to be repre- 

 sentative of growth patterns of male and female alba- 

 core within the size range sampled. 



Comparison of growth curves 



A direct comparison of the VB growth curves estimated 

 from length-frequencies and vertebral-ring-counts is 

 slightly complicated by the assumption of r„ = for the 

 length-frequency analysis and the estimation of t = 

 -1.922 year for the vertebral-ring model. This incon- 

 sistency would result in the growth curves being out of 

 phase by almost two years, apart from any other dif- 

 ferences that might be present. One way of comparing 

 the two curves is to compare the expected growth of 



