180 



Fishery Bulletin 103(1) 



fled. We used cultured individuals to determine which 

 opaque or translucent zone represented the first growth 

 increment, although the accuracy of age validation with 

 cultured individuals has been questioned by Campana 

 (2001). In our study, the close correspondence between 

 the growth of cultured and wild fish over a period of 

 several years gives us confidence in using this approach 

 to validate first increment position. The slightly slower 

 growth rate observed in cultured striped trumpeter 

 can be attributed to jaw malformation — a phenomenon 

 that has been shown to affect feeding ability (Cobcroft 

 et al., 2001). 



Modal progression of the 1993 cohort through time 

 provided indirect validation for annual periodicity in in- 

 crement formation up until age seven. Validation across 

 all age classes was not possible in our study, although 

 validation after the age of five years was significant. 

 That is, validation was achieved past the average age 

 at which fish moved offshore into deeper water, and 

 past the age at which there was a significant reduction 

 in growth rate. 



The second consideration to address when studying 

 animal growth is model selection. Akaike's information 

 criterion is a standard method for model selection that 

 provides an implementation of Occam's razor, in which 

 parsimony or simplicity is balanced against goodness-of- 

 fit (Forster, 2000). However, model selection should not 

 rely on statistical fit alone; it should also provide a bio- 

 logically sensible interpretation across the entire range 

 of ages in the sampled population (Haddon, 2001). In 

 the case of striped trumpeter, the standard von Berta- 

 lanffy function provided a poor representation of growth 

 in older individuals, resulting in an unrealistically low 

 L r . This problem was largely overcome by the applica- 

 tion of a two-phase growth function. Similar to that 

 used on large pelagics, such as Thunnus maccoyii (Bay- 

 liff et al., 1991; Hearn and Polacheck, 2003). In their 

 application of the model, Hearn and Polacheck (2003) 

 considered biological traits when discussing the justi- 

 fication for age at transference, namely the reduction 

 in growth rate, and inshore to offshore migration. In 

 the present study we have considered analogous traits 

 to seed the age of transference for striped trumpeter. 



In this species there is a marked transition in size 

 structure between shallow and deeper reefs that occurs 

 at around 450 mm or between 4 and 5 years (Fig. 8). 

 In addition, a visual assessment of the length-at-age 

 data highlighted a marked decrease in growth rate at 

 a similar age. 



Solving for the age at transference produced a point 

 estimate that results in a sharp discontinuity in the 

 growth curve; an observation that Hearn and Polacheck 

 (2003) highlighted as biologically unrealistic. The range 

 of low negative log likelihood values described by the 

 age at transference profile is due to the patchiness of 

 data around these ages, creating uncertainty in the 

 model. We have assumed in this case that the converged 

 value of 4.4 years is accurate and that the variability 

 around this point is normally distributed with a stan- 

 dard deviation equal to one. By including the normal 

 probability distribution function we have effectively 

 created a smooth transition between growth phases. 

 This function implies that age at transference has some 

 level of inherent variability, which is likely to be more 

 biologically plausible than knife-edge transition. 



A further extension of the two-phase model was test- 

 ed by applying the seasonal growth version of the VBGF 

 (described in Eq. 3) to the first phase and a standard 

 VBGF to the second phase, but was disregarded because 

 of the effect of over parameterization on parsimony. 

 However, this approach did highlight the flexibility of 

 the two-phase model to allow for a more dynamic rep- 

 resentation of population growth characteristics. 



This study supports the assertion by Hearn and Po- 

 lacheck (2003) that discontinuity in growth rate may 

 be a more common phenomenon in fish than implied 

 by growth models reported in the literature. Such a 

 two-phase growth model, where age at transference 

 coincides with the transition phase from one fishery to 

 another, has proven useful. It allows separate growth 

 parameters to be tracked to each fishery, and as such, 

 provides a precursor to developing a more biologically 

 robust production model with dynamic parameters at 

 age and for fishing method. 



The predictive regression developed by Pauly (1980) 

 that estimates natural mortality is based on the direct 



