538 



Fishery Bulletin 104(4) 



most hauls, were fairly abundant in bycatch, an(i were 

 each represented by a range of sizes. Tailor iPomato- 

 mus saltatrix) were grouped into five length classes 

 with approximately equal numbers of fish in each class 

 (140-145, 146-170, 171-195, 196-220, 221-280 mm); 

 the snapper Pagrus aiiratus were also so grouped into 

 five length classes (70-105, 106-120, 121-135, 136-155, 

 156-230 mm). Data were analyzed as the proportion 

 of fish in each size class, to remove differences due to 

 different numbers of fish caught in the various types of 

 gear. Size classes were then used as variables in multi- 

 variate analyses to compare the various types of gear. 

 Using data from experiment 1, we detected no dif- 

 ferences among types of gear in the proportional size 

 frequencies of either tailor or snapper. The mean Bray- 

 Curtis dissimilarities between pairs of samples were 

 all very similar to the mean measures within samples 

 (i.e., among replicates); therefore there was no sugges- 

 tion of differences in size frequencies among types of 

 gear for these species. Multidimensional scaling plots 

 (not included, Clark and Gorley^) supported this result 



because the centroids of the size classes of both species 

 were not clustered by type of gear. 



Composition of bycatch 



The types of species and relative abundances of spe- 

 cies in the bycatch were analyzed from several times 

 of sampling. Analyses used data without squid because 

 numbers of squid are irrelevant to the bycatch. Typi- 

 cally, there were no detected differences in assemblages 

 of species in the bycatch. One exception, in Experiment 

 2 is described. In our study, analyses of Bray-Curtis 

 dissimilarities showed some differences among types of 

 gear, where the square-mesh codend caught a different 

 assemblage. It seemed that fewer whitebait (Hyperlophus 

 vittatus) were being caught when the square mesh was 

 used (this was confirmed by SIMPERS analysis, which 

 showed that numbers of whitebait were an important 

 variable for discriminating between samples). To test 

 the validity of this conclusion, we performed analyses, 

 omitting numbers of whitebait. This time, all differences 



3 Clarke, K. R., and R. N. Gorley. 2001. Primer version 5: 

 user manual/tutorial, 91 p. Primer-E Ltd, Plymouth, U.K. 



■• See footnote 3. 



