78 
Fishery Bulletin 95(1), 1997 
decrease over time provided a better fit to double- 
tagging data for southern bluefin tuna, Thunnus 
maccoyii, than the model used in this study. They 
reasoned that tags might become more securely fixed 
over time, and thus less likely to be shed, as the fish 
grows and tissue is built up around the tag shaft. I 
fitted the three-parameter variable-rate shedding 
model (model 4 in Hampton and Kirkwood [1990]) to 
the pooled data set and to the three species-specific 
data sets and found that the improvement in fit over 
the constant shedding-rate model was negligible in 
each case and did not warrant the addition of the 
extra parameter. There is thus little evidence of a 
decline in shedding rates over time in these data. 
This may in part be due to the relatively short peri- 
ods at liberty (maximum of 2 years) of the double- 
tagged tuna in this study compared with those for 
the southern bluefin tuna (up to 18 yr) analyzed by 
Hampton and Kirkwood (1990). 
Given compliance with the assumptions of the ex- 
periment and the appropriateness of the model, it 
can be concluded that losses of tags through shed- 
ding are relatively modest (about 11% after two years) 
for the RTTP. This shedding rate is comparable to 
those reported by Hampton and Kirkwood ( 1990) for 
the more recent southern bluefin tuna double-tag- 
ging experiments (16% and 12% after two years for 
experiments 7 and 8, respectively), where comparable 
tags and techniques to those used in this experiment 
were used. Other tuna tagging experiments have 
reported substantially higher tag losses after two 
years, e.g. 30%-50% for the early southern bluefin 
tuna experiments (Hampton and Kirkwood, 1990), 
43% for eastern Pacific yellowfin tuna (Bayliff and 
Mobrand, 1972), and 35% for Atlantic bluefin tuna 
(Lenarz et al., 1973; Baglin et al., 1980). It is pos- 
sible that the higher shedding rates observed in some 
of these experiments were due to inferior tags, in 
which the streamers were prone to detach from the 
tag head. The streamers of tags used in this experi- 
ment and the recent southern bluefin tuna experi- 
ments were heat fused to the tag heads, making de- 
tachment impossible under normal conditions. 
The analysis of tag-seeding and associated data 
indicated that, despite extensive publicity and attrac- 
tive rewards for tag finders, failure to report tags 
was a significant source of tag loss in the RTTP. Given 
the diverse nature of the fishery, its spatial extent, 
and the methods of processing large quantities of fish 
caught by purse seiners in particular, this is hardly 
surprising. The estimated overall reporting rate in 
fact compares more than favorably with those for 
some tagging experiments carried out on more local 
scales (e.g. Campbell et al., 1992 for coastal shrimp 
in the Gulf of Mexico ). My estimates of reporting rates 
based on tag-seeding data may, if anything, err on 
the pessimistic side. It is suspected that one cause of 
failure to report purse-seine— caught tagged tuna may 
be the detachment of tags (through abrasion) from 
fish while they are held in the vessels’ wells. If this 
occurs, detached tags would likely be flushed out of 
the wells into the sea, after which detection would 
be highly improbable. It is possible that tags placed 
in dead tuna by observers were more prone to de- 
tachment in the well than tags placed in live tuna 
that had been at liberty for some time. The tag head 
and the portion of the tag shaft imbedded in the 
musculature of live tuna were frequently observed 
to be encased in a fibrous capsule, which would tend 
to fix the tags more securely than tags placed in dead 
tuna. “Shedding” of seeded tags could conceivably 
result in losses of seeded tags of the same order as, 
or greater than, the immediate tag-shedding rates 
estimated from the double-tagging experiment on live 
tuna (about 6%). It may be possible to estimate the 
extent of this problem by conducting a double-tag- 
ging experiment for seeded tags. 
The main purpose of estimating tag-shedding and 
reporting rates is to allow these processes to be in- 
corporated into analyses of the tagging data for the 
purpose of estimating mortality rates. Typically, this 
would involve substitution of the point estimates of 
the parameters into equations such as Equation 1; 
mortality rates that are free of the effects of these 
tag losses could then be estimated from the tagging 
data (e.g. Kleiber et al., 1987). However, where the 
ultimate objective of the analysis (mortality rate es- 
timation) is stock assessment related, it is impor- 
tant to have not only estimates of the mean rates 
but also estimates of their precision that are uncon- 
ditional on estimates of nuisance parameters such 
as tag-shedding and reporting rates. 
In this study, estimates of precision (expressed as 
95% confidence intervals ) of tag-shedding and report- 
ing rates were obtained by using the percentile 
method applied to the bootstrap distributions of the 
parameter estimates. For the tag-shedding analysis, 
I confined this to estimates of precision of Q 2yr , al- 
though confidence intervals for the model parameters 
could be similarly derived. The advantage of the boot- 
strap approach as applied to the analysis of tag re- 
porting is that it allowed the precision of the overall 
reporting rate to be easily determined given some 
knowledge, or reasonable assumptions, regarding the 
reporting-rate probability distributions for differing 
components (in this case, based on unloading loca- 
tion) of the data set. The approach also provided a 
convenient means of integrating uncertainties in tag- 
shedding and reporting rates (via the individual boot- 
strap values) into a similarly structured bootstrap 
