558 



Fishery Bulletin 89(4), 1991 



Daily bias 



L A - 



(L A N c ) + (L M N M ) 



Nr + Nm 



mm/day, 



(1) 



where N c was the number of fish in the treatment 

 group at the end of each day, N M was the number of 

 dead fish that day, and L M was the mean standard 

 length of the dead fish that day. The L A was calcu- 

 lated from a regression of sample mean lengths on 

 elapsed time over the course of the treatment. The 

 slope of this regression (L A on elapsed time) was the 

 apparent growth (Table 2), which was corrected by the 

 average of the daily biases to estimate the actual in- 

 crease in fish length. 



Fish mass was estimated from corrected length using 

 log-log regressions of dry mass on standard length for 

 each sample. The regression estimate for each sample 

 was calculated by assuming that length increased at 

 the corrected rate for that treatment. 



From each sample, a subset of four fish, chosen to 

 best match the estimated length and dry mass of the 

 sample, was analyzed for fat content by using Soxhlet 

 extraction with chloroform-methanol (2:1). Fat content 

 of food removed from stomachs was also measured. 

 Total energy content of fish and food was estimated 

 from published values (Brett and Groves 1979, Hunter 



and Leong 1981) for heat of combustion. The rate of 

 energy loss or gain in each treatment was calculated 

 as the regression slope of fish energy content in each 

 sample versus elapsed time. 



Results 



Rations 



Anchovy stomach contents from 12 samples (Table 2) 

 were weighed immediately after feeding. The average 

 amount of food in the stomachs was less than (but 

 correlated with) the net amount of food that was dis- 

 pensed per fish (Fig. 2). The difference was too large 

 to be due to the rate of gastric evacuation. A portion 

 of the food passed through gaps in the bottom of the 

 flume sections and into the surrounding tank. Recir- 

 culation of this food was prevented by screens. Regres- 

 sions of stomach contents on food dispensed (Fig. 2) 

 suggested that a greater proportion of the food dis- 

 pensed was eaten in the high-ration, fast-speed treat- 

 ment than in the other treatments. Although the re- 

 gressions were not significantly different, they were 

 used to estimate stomach contents separately for each 

 speed on days when stomach contents were not sam- 

 pled. The average of the resulting daily estimates was 

 calculated for each treatment (Table 1). 



