count detached tags. Any tags fouled and detached 

 on deck during setting were retained by the fisher- 

 men, returned to us, and counted. Thus our effort 

 between tagging longline sets consisted only of 

 counting detached tags and replacing them with 

 new tags. 



The results of experimentally setting tagged long- 

 lines (i.e., the variation in proportions of detached 

 tags in relation to various attachment strengths, hook 

 types, tagging locations, and series) were tested by 

 analysis of variance (ANOVA) using the Statistical 

 Analysis System (Barr et al. 1976). 



Results and Discussion 



A total of 1,156 detachable tags (687 of one-series 

 and 469 of ten-series) were set with various hook 

 types and attachment strengths on two separate oc- 

 casions near Hudson Canyon (Table 1). Following 

 retrieval of the longlines we determined that 384 

 detachable tags had been lost, 96 at the east Hudson 

 location and 288 at the west Hudson site (Table 

 1). 



ANOVA of the proportions of tags detached 

 showed significant variations in detachment rate 

 (Table 2). Significant or near- significant probability 

 levels were calculated for variations in proportions of 

 detached tags in relation to the following sources of 

 variation: Tagging location (east or west Hudson); 

 series (one or ten); tagging location-series interac- 

 tion; attachment strength; hook type; and hook type- 

 tagging location interaction (Table 2). 



We know that some accidental tag loss occurred due 

 to fouling, which was observed at the east Hudson 

 site as the gear was being set. However, we believe 

 that detachment rate data actually reflect the relative 



abundance of tilefish tagged. This is supported by 

 observations made from a submersible at the same 

 time and location ( Grimes et al. 1 982 a) ; in a transect of 

 commercial longline gear, tilefish were hooked on 42 

 of 227 hooks (0.19 hooking rate). This hooking rate 

 was nearly identical to the 0.19 loss rate for all tags 

 set at the east Hudson location (Table 3). Evidently, 

 tag loss from fouling was a random event that oc- 

 curred irrespective of hook type or attachment 

 strength and thus did not affect the analysis, 

 although it could be logically reasoned that weaker 

 attachment strengths and curved hooks would foul 

 most readily. 



We know no obvious reason why 1) higher propor- 

 tions of one-series than ten-series tags were 

 detached, or 2) higher proportions of ten-series than 

 one-series tags were detached at the east Hudson site 

 (Table 3), causing the significance in the ANOVA of 

 the series and tagging location-series interaction 

 (Table 2). If tilefish were contagiously distributed, 

 one might expect these results from the ANOVA and 

 also expect overall tag loss to be contagiously dis- 

 tributed along longlines. A runs test (Sokal and Rohlf 

 1969) failed to demonstrate contagion in tag loss, and 

 Grimes et al. (1982a) failed to demonstrate contagion 

 for longline catches using the same statistical pro- 

 cedure. The significantly greater tag loss at the west 

 Hudson site (Tables 2, 3) presumably reflects greater 

 tilefish abundance there. 



Attachment strength was deemed significant by the 

 ANOVA because increasing proportions of tags were 

 lost with decreasing attachment strength (Table 3). 

 Apparently tilefish were able to detach most easily 

 those tags with 0.9 kg monofilament, followed by 1.8 

 and 2.7 kg. Among tags returned, four were attached 

 with 1.8 kg and two with 2.7 kg monofilament (Table 



Table 1. — Numbers of detachable tags of various hook types and 

 attachment strengths set, detached, and returned at east and west 

 Hudson Canyon tagging locations, August and September 1979. 



TABLE 2. — Analysis of variance of the proportions of tags detached 

 at east and west Hudson Canyon locations, August and Septem- 



664 



