FISHERY BULLETIN; VOL. 87. NO. 3, 1989 



100 m^. Other statistics were calculated accord- 

 ing to Zar (1984). 



The analysis to derive correction factors for 

 day, night, and twilight catchability and to 

 derive estimates of larval mortality for 26 

 species involved three steps. The larval length- 

 interval used was calculated from the modal 

 length, which was always greater than the min- 

 imum length captured, to the maximum length 

 that was captured in all three light regimes. I 

 assumed that the modal length was the minimum 

 length fully retained by the 505 |jl mesh net. If 

 extrusion of larvae at modal lengths or gi-eater is 

 significant, length-dependent mortalities will be 

 underestimated and will have no effect on catch 

 ratios. The exponential decay regression model 

 was fitted to the larval length (A',) and catch per 

 100m^(F,)as 



Yi = aexp(-pZ,) -I- e, 



(2) 



for each species by day, night, and twihght 

 catches. The expected catches from these re- 

 gressions were then used to estimate the ratios 

 of catch by length for night:day and twihght:day. 

 The predictive nonlinear second-order poly- 

 nomial regression model: 



R, 



^,X,+ PiZf + e, 



(3) 



was fitted to the ratios by length. (The predicted 

 ratio (Ri) of night:twilight catches for any length 

 (Xj) is, of course, the ratio of night:day at X, 

 divided by the twilight:day ratio at Xj.) The 

 catches at each length per 100 m" were then 

 corrected using the predicted ratios that maxi- 



mized the catches. The exponential decay model 

 was again fitted to the corrected catches and the 

 slope of the line was used as an estimate of 

 length-dependent mortality. 



RESULTS 

 All Larvae 



The initial analysis examined the catches for 

 all taxa combined for all 8,312 stations, of which 

 6,530 contained fish larvae. The mean catch per 

 10 m" was calculated for 3,578 day stations, 3,332 

 night stations, and 1,402 twilight stations (Table 

 1). A significant difference (P < 0.05) was found 

 between night-versus-day and night-versus- 

 twihght catches with the ratios of the mean 

 catches of 1.62:1 and 1.44:1, respectively. The 

 combining of all data raises the question of 

 whether one or more year's data might have 

 been anomalous and produced the significant dif- 

 ference in catches. Each year was analyzed sep- 

 arately and is shown in Table 1. For each year 

 the highest mean catches always occurred at 

 night and in six of the eight years twilight 

 catches exceeded day catches. In four of the 

 eight years a significant difference (P < 0.05) 

 was found between night and day catches (1977, 

 1982-84). The consistent interannual relation- 

 ships in catches indicate that the combined 

 years' ratios are good indicators of average 

 catchability of all larvae during the 8 yr study. 



The analysis of the combined data-set inte- 

 grates the areal and temporal heterogeneity in 

 the distribution and abundance inherent within 

 the larval fish community. To determine if the 



Table 1. — The mean catch per tow (#/10 m^) by day, night, and twilight of fish larvae collected off the northeast 

 United States for years 1977-84 and combined. Numbers in parenthesis are standard error of the mean and 

 sample size. Values of Student Most for differences in mean catches. 



420 



