Hampton. Estimates of tag-reporting and tag-shedding rates for tuna in the tropical Pacific Ocean 
71 
would be detected during or after unloading of the 
catch in port offered the opportunity for tagged tuna 
to be planted in the catches before these detection 
processes began. Furthermore, the layout of purse- 
seine vessels and the method of onboard handling of 
the catch facilitated the opportunity for planting 
tagged tuna surreptitiously, out of sight of the vessel’s 
crew. Such tag-seeding operations would be more 
difficult on other types of vessels, e.g. pole-and-lin- 
ers and longliners, operating in the fishery. 
Field operations Selected observers on purse sein- 
ers were asked to plant up to five tagged tuna in the 
catch during a voyage. The number of tagged tuna 
was limited to five so as not to attract undue atten- 
tion during unloading; it was not unusual during the 
RTTP for five (and sometimes more) tagged tuna to 
be recovered from a single unloading. The exact tim- 
ing of tagging individual fish depended on the cir- 
cumstances encountered during a cruise, particularly 
the frequency of successful sets. Therefore, the pe- 
riod over which the five tags were seeded ranged from 
a few days to several weeks. 
Fish were tagged discretely, usually on the well 
deck (one level below the work deck where the fish 
are landed), as they passed down the chute just prior 
to entering the well. The tags and manner of attach- 
ment were identical to those used in the tagging pro- 
gram proper. Tag numbers, dates, species, sizes, and 
well numbers were recorded and the information sent 
to SPC at the completion of the voyage. Upon recov- 
ery, seeded tags were processed in the same fashion 
as genuine tag recoveries. Tag finders were paid the 
standard reward for seeded tags and were not informed 
that the tags were part of a seeding experiment. 
Estimation of return rates of seeded tags Return 
rates of seeded tags were calculated for the overall 
data set, for the three species (skipjack, yellowfin, 
and bigeye tuna ) and for the seven unloading loca- 
tions represented in the data (American Samoa, Ja- 
pan, Korea, Philippines, Puerto Rico, Solomon Is- 
lands, and Thailand). For one unloading location 
(American Samoa), there were sufficient returns to 
estimate reporting rates by time period (year). Dif- 
ferences in seeded tag-return rates among species, 
unloading locations, and time periods were assessed 
by using chi-square tests (Sokal and Rohlf, 1981). 
Return rates were estimated by assuming that the 
number of returns, r, in a given category was a bino- 
mial variate. Given the number of tags seeded, N, the 
estimated return rate is given by p-r/N. Under these 
conditions, 95% confidence limits for return rates were 
also obtained. Lower and upper confidence limits, p A 
and p B , for p were determined by solving the equations 
where l-2a is the confidence level (0.95 in this in- 
stance). Solutions for p A and p B can be easily obtained 
using an optimization program, such as the Microsoft 
Excel Solver. 
Estimation of overall reporting rate for the 
RTTP An unbiased estimate of the overall return 
rate of recovered tags (i.e. the total number of tags 
returned divided by the total number of tags recap- 
tured) is required for the estimation of fishing and 
natural mortality rates from the RTTP data. The 
return rates of seeded tags can be considered as 
sample means of the overall (population) mean re- 
porting rate. It transpired that seeded tag-return 
rates varied greatly by unloading location, requir- 
ing that the data be stratified by unloading location 
in the estimation procedure. The parametric boot- 
strap (or Monte Carlo) approach was used to obtain 
approximate 95% confidence intervals for the over- 
all reporting rate and its components (with the per- 
centile method), taking account of the different prob- 
ability distributions of reporting rate by unloading 
location. One thousand simulations (or bootstrap 
replicates) were run. In each, the weighted average 
reporting rate across locations is given by 
where R is the number of tags returned from loca- 
tion j and p'j is the bootstrap (or pseudo) reporting 
rate for location j. 
For each replicate, the p', were randomly sampled 
from probability distributions. For recoveries in lo- 
cations covered by tag-seeding experiments, beta dis- 
tributions B(x , y , a jt b) were used to represent the 
probability distributions of the true reporting rates. 
These continuous distributions are related to the bi- 
nomial distributions defined by the tag-seeding data 
by x =r and y^N-r + 1 (Mendenhall and Scheaffer, 
197^). 'the limits of the distributions, a and b, would 
normally be 0 and 1, respectively. In this case, we 
assumed b - 1 and set the lower limit of reporting rate 
for location j, a-, to the local tag-return rate (i.e. num- 
ber of local returns divided by the number of local 
releases), so as to avoid the possibility of estimated 
recoveries out-numbering releases for any replicate. 
For two locations, Solomon Islands and Philippines, 
