water temperatures rose rapidly. The temperatures for the second 12- 

 week period averaged !p[|.°F<,, and, differing from the relatively con-= 

 stant temperatures of the first period^ varied from 50 to 62° F„ 



The results brought about by the water temperatures were easily 

 distinguishable. During the first half of the experiment it can be 

 demonstrated that with a few exceptions high mortalities vere the 

 rule. The exceptions were a result of those diets without meal of any 

 kind. In addition to the high mortalities suffered by most of the 

 lots during this period of cold water, the (growth rates were low. As 

 soon as the temperatures started to rise, however, the growth rates 

 increased markedly in most of the lots. Later, after the residual 

 effects of the cold-water period had worn off, the mortality rates de- 

 clined and i\fi th but a few exceptions remained low. 



Since the water temperatures were definitely of two sharply de= 

 fined patterns of cold water and warm water, and the differing effect 

 of these two periods can be easily demonstrated by differences in mor- 

 tality and growth rates, the splitting of the experiment into two 

 halves serves as a most useful tool in explaining the outcome of these 

 trials. 



RESULTS OF EXPERIMENT 



The results of feeding the various diecs are summarized in Table 1. 

 The most reliable of these measurements were based on the mortalities 

 and the total weight of the fish in each of the two troughs in a lot 

 since these two measures lend themselves readily to statistical analyses. 

 The mean lot weights and mean mortalities, the average of the two troughs 

 assigned to a diet or lot, are based directly on Lhe aaily mortalities 

 and bi-weekly weights. The same basis was used for the per cent gains 

 and the per cent mortalities which were computed by dividing the 12= 

 week or 2lt-week total by the original weitht or nuuiber stocked. To 

 measure the differences in weights and mortalities produced by the 

 various diets, analysis of variance was used in the manner prescribed 

 by Snedecor (19ii6), and only the least difference at the 5 per cent 

 confidence level was employed as a basis of (Bomparison. In this same 

 table will be found two other measures, the conversions and deficiency 

 symptoms. The conversions were computed by dividing the total amount 

 of food fed by the total weis-.ht gain. The presence of anemia was 

 determined bj'- checking the gill color of 90 to l50 fish per lot in each 

 of the 20 lots. 



The numerous small tables in this paper which deal only with the 

 weights and mortalities were used to emphasize the differences achieved 

 with diets having a common base and one variable component. Thsse tables 

 are all taken from Table 1; therefore any mortality or weight from one 



