456 



Fishery Bulletin 94(3), 1996 



ranges sampled, two different-size cast nets were 

 used. Of the 15 casts per substation, 10 were made 

 with a 5-m diameter, 10-mm mesh net, and five casts 

 were made with a 3-m diameter, 6-mm mesh net. 

 The smaller net was more effective in narrow up- 

 stream habitats. 



We used stratified-random sampling; cast nets 

 were thrown over schools of mullet juveniles, rather 

 than thrown randomly. Random sampling yielded few 

 wild mullet and very few tagged individuals. Cul- 

 tured and wild striped mullet schooled together in 

 fairly low densities within these clear-water nurs- 

 ery habitats, and our collections targeted these 

 schools. Nevertheless, data used to determine recap- 

 ture rates and proportions of tagged mullet in 

 samples were randomly distributed because there 

 was no a-priori indication that schools, once sighted, 

 contained tagged individuals. 



Mullet sampled in these collections were measured 

 and checked for tags with a portable tag detector 

 (Northwest Marine Technology, Inc.). All tagged 

 mullet were placed on ice and returned to the labo- 

 ratory for size measurement and tag analysis. Wild 

 fish were counted, measured, and released at the 

 sample site. After extraction, the binary codes were 

 read by a technician with a binocular microscope 

 ( 40x ). All tag codes were verified with a second ( blind ) 

 reading by a another technician. 



Data were analyzed with SYSTAT software 

 (Wilkinson, 1990). A randomized-block design 

 ANOVA was used to compare means. Treatments 

 (SAR, release microhabitat) were blocked over time 

 (3 release lots) within season (spring and summer). 

 Proportions were arcsine transformed. Orthogonal 

 contrasts were used to compare means (Sokal and 

 Rohlf, 1981). SYSTAT Basic was used to write tag 

 decoding algorithms. For each recaptured fish, the 

 algorithms identified SAR, release microhabitat, re- 

 lease date, release lot, and number of fish released 

 per treatment-lot combination on the basis of the 

 binary tag codes. An error-check algorithm was also 

 written. Variance estimates are expressed through- 

 out as standard errors. 



Results 



Recapture summary 



A total of 2,985 tagged cultured mullet were recap- 

 tured during the 11-month period of this study (Table 

 2). Of these, 2,642 were cultured fish from the present 

 study (1992 releases). None of the fish released in 

 1992 were recaptured at Kaneohe Stream. From the 

 tag codes it was determined that 304 (10.2%) of the 

 2,985 tagged fish collected had been released in 



