2U=week periods (19U8) the figures for the 2U-week period include the 

 loss for the first 12 weeks, 'Aiis same procedure is applied when com- 

 puting the per cent gain. 



The conversion is derived from the total weight of food fed a 

 given lot divided by the total gain in weight made by the surviving 

 fish on the diet. Here, again, no compensation is made for mortality. 



Whenever a high mortality occurred in a lot, the fish were checked 

 for demonstrable deficiency symptoms „ At the conclusion of the ex- 

 perimental period, all lots of fish were examined for evidence of 

 dietary deficiencies. During 19ul; and 19U5, erythrocyte counts were 

 made on samples of fish taken from key diets. Because time did not 

 permit the sampling of more than four fish from each diet, this pro- 

 cedure was abandoned in favor of cursory examinations of large samples 

 of 50 to 100 fish for gill color as being more indicative of the true 

 condition of the entire group. 



Diets, which were discontinued, were abandoned due to an avitaminosis 

 or an acute anemia in the fish which precluded the possibility of 

 their survival until the conclusion of the experiment. Other di ts 

 in which deficiency symptoms were recognizable but were not acute 

 were continued throughout the experimental period. In some cases, 

 the deficiency symptoms of the fish on the various diets are listed 

 in the tables giving the time of their occurrence and the condition 

 of the fish at the conclusion of the experiment. Unquestionably, 

 some deficiency symptoms were present which were not recognized either 

 because the symptoms were not clear cut or they haH not been previously 

 described. The presence of pale spleens, which occurred in fish on 

 several diets, may be an example of the latter case. 



It will be noted that the per cent gains in weight varied tre- 

 mendously from year to year. The initial, average weight of the 'fish 

 at the start of the experimental periods showed a similar variation. 

 As larger fish gain at a reduced rate, the initial size variation in 

 the fish between years explains the differences that exist in the per- 

 centage gains of fish fed comparable diets in different years when 

 the water temperatures were similar. 



The least difference for significance at the 5 per cent confidence 

 interval was computed from the fiducial limits of the mean difference 

 as determined by analysis of variance for paired experiments (Snedecor 



19y6)„ The least difference is expressed as twice fiducial limits 

 of the mean difference. As the data presented in the tables, with 

 the exception of the conversion factors, are the means of the two 

 groups of fish fed each diet, it is possible to determine if the dif- 

 ferences i^hich exist between the fish fed any two diets is statistically 

 significant merely by noting whether the difference is greater or less 

 than the least difference. 



10 



