320 



Fishery Bulletin 88(2|. 1990 



result in a higher rate of shedding. Tags used in the 

 latest experiments do not appear to have suffered from 

 such defects. In experiments 7 and 8, the use of a spe- 

 cially designed tagging cradle may have resulted in 

 more effective and less traumatic tagging than in 

 earlier experiments, where fish were tagged on a mea- 

 suring board. Also in experiments 7 and 8, a deliberate 

 effort was made to anchor the tags behind the second 

 dorsal fin ray extensions. Tags successfully attached 

 in this way should have a very small probabDity of shed- 

 ding. Unfortunately, little quantitative information is 

 available on the efficacy of the technique of tag attach- 

 ment adopted for experiments 1-6, but it is known that 

 greater care was taken in tag attachment in the later 

 experiments. If the proportion of ideally attached tags 

 in experiments 1-6 differed markedly from those for 

 experiments 7 and 8, then commensurate differences 

 in shedding-rate estimates would result. 



While the lower tag shedding rates in experiments 

 7 and 8 are obvious on first inspection, the situation 

 is less clear for the earlier experiments. In general 

 terms, the precision of estimates of shedding rates, par- 

 ticularly long-term rates, will increase with the size of 

 the database. On these grounds, pooling of the data for 

 experiments 1-6 would appear to be an attractive op- 

 tion. However, it would be appropriate to do so only 

 if the assumptions underlying the analysis of the pooled 

 data remained valid. Critical among these assumptions 

 is that all tags have identical and independent shed- 

 ding probabilities. The primary reason for classifying 

 the tag releases in the 1960s and 1970s as six separate 

 experiments was a suspicion that these probabilities 

 may well have been heterogeneous. As mentioned 

 earlier, with this classification we were attempting to 

 minimize within-experiment differences in geograph- 

 ical area, fish size, capture and tagging methods, and 

 tagging personnel. While in principle further subdivi- 

 sion is possible, for example by tagging vessel in order 

 to take account of different tagging teams, we felt this 

 may leave too few data in each subset to obtain reliable 

 estimates of shedding rates. Also, no assertion is made 

 that the classification chosen is an optimal one; indeed 

 it is by no means obvious that suitable criteria for op- 

 timality could be defined. A statistical way of examin- 

 ing whether or not it is appropriate to pool data for 

 experiments 1-6 is to carry out a likelihood ratio test 

 of the hypotheses that the tag shedding parameters for 

 each experiment are equal. This hypothesis proved to 

 be resoundingly rejected (P<0.001). It appears that 

 pooling is not a viable option for these experiments. 



One of the subsequent analyses that we had envis- 

 aged for recovery data adjusted for tag shedding in- 

 volved application of the method of Hearn et al. (1987) 

 to obtain estimates of natural and fishing mortality 

 rates. This method, which is essentially a cohort anal- 



ysis of the tagging data, gives particular weight to long- 

 term recaptures and therefore requires accurate esti- 

 mates of long-term shedding rates. Despite the large 

 numbers of recoveries in the experiments described in 

 this paper, precise estimation of long-term shedding 

 rates has not been possible. This is best exemplified in 

 the results of experiment 1, where no single best fit- 

 ting model was obtained, and for which the point esti- 

 mates of long-term shedding rates differ markedly. 

 Even where a single best-fitting model was available, 

 considerable uncertainty still remained about the long- 

 term shedding rates. It therefore seems essential to 

 take account of this uncertainty in any subsequent anal- 

 yses using methods such as that of Hearn et al. (1987). 

 This is the subject of further research. 



Acknowledgments 



Dr. G. Eckert, Dr. V. Mawson, and Ms. S. Wayte 

 reviewed and provided helpful comments on an earlier 

 draft of this manuscript. We also thank Dr. W.S. Hearn 

 for useful comments and discussion, particularly in rela- 

 tion to the assumption of independent and identical 

 shedding rates. 



Citations 



Baglin, K.E.. Jr.. M.I. Farber, W.H. Lenarz, and J.M. Mason, Jr. 



1980 Siieddiiig rates of [ilastic and metal dart tags from Atlan- 

 tic bluefin tuna. Thunniu^ thywius. Fish. Bull., U.S. 78: 

 179-185. 

 Bard. Y. 



1974 Nonlinear parameter estimation. Academic Press, NY, 

 341 p. 

 Bayliff, W.H., and L.M. Mobrand 



1972 Estimates of the rates of shedding of dart tags from 

 yellowfin tuna. Inter-Am. Trop. Tuna Comm. Bull. 1.5:465- 

 .503 (Engl, and Span.]. 

 Beverton, R.H.J.. and S.J. Holt 



1957 (_)n the dynamics of exploited fish populations. Fish. In- 

 vest. Ser. II, Mar. Fish. G.B. Minist. Agric. Fish. Food 19, 

 533 p. 

 Hearn, W.S.. R.L. Sandland, and J. Hampton 



1987 Robust estimation of the natural mortality rate in a com- 

 pleted tagging experiment with variable fishing intensity. J. 

 Cons. Int. Explor. Mer 43:1(17-117. 

 Hynd, J.S. 



1969 New evidence on southern bluefin stocks and migra- 

 tions. Aust. Fish. 28(5):26-30. 

 Hynd, J.S., B.B. Jones, G.L. Kesteven, and N. Sproston 



1967 Tunas. /« CSIRO Division of Fisheries and Oceanog- 

 raphy Annual Report 1966-67, p. 20-24. CSIRO Div. Fish. 

 Oceanogr., Cronulla, Australia. 

 Kendall, M.G., and A. Stuart 



1961 The advanced theory of statistics. Vol. 2: Inference and 

 relationship. Charles Griffin, London, 676 p. 



