METHODS 



A review of the literature on the nutrition of salmon and trout 

 such as was conducted by Karrick ( 191*8 ) reveals the many divergent 

 results attained by different investigators and often reversals in 

 different years by the same investigators. These contradictory re- 

 sults may be due to several factors but, principally, to variations 

 in methods of diet preparation and diet presentation. As pointed 

 out by Phillips and Hewitt (19U5), otherwise adequate diets may be 

 prepared and presented in such manner as to be onl3 r partially avail- 

 able to the fish and, as a co.nsequence 5 a reduction in growth or a 

 nutritional deficiency may result. Another reason for a misinter- 

 pretation of results is faulty experimental design in which it is 

 impossible to segregate the differences due to biological variation 

 and experimental error from the differences due to diet. The ex- 

 perimental techniques employed in the feeding trials at the Leaven- 

 worth Laboratory were designed to eliminate as :»any of these as 

 possible and to measure those which were unavoidable. 



The fish used in each year's experiments were selected from a 

 single day's take. During I9UI1 and 19U5 the thoroughly mixed pop- 

 ulation of fingerlings was distributed into the troughs by the se- 

 lection of random numbers. In this manner two adjacent troughs could 

 be used for a single diet while still retaining a random selection. 

 Previous experiments had determined that no significant difference 

 existed due to the placement of the troughs in the experimental sec- 

 tion. In 19J47 and 191*8 the use of the Leavenworth sampler (Hewitt 

 and Burrows, 19U8), in which all four pockets were closed, made pos- 

 sible the elimination of the above procedure.. The use of the sampler 

 insured a random distribution from a homogeneous population into all 

 the experimental troughs. In this procedure the required poundage 

 of fish for the entire experimental group was divided by the sampler 

 into four equal portions. These divisions were further subdivided 

 until the approximate weight of each individual trough of fish was 

 attained, at x-jhich point the exact distribution was made by weight,, 



In order to retain a representative population in each trough 

 throughout the experimental period, which is necessary for the use 

 of analysis of variance, the initial stocking of the individual 

 troughs was at 1,000 grams (2.2 pounds) for the 12 or ll* week ex- 

 periments (19UU, 19l&, and 19U7) and 500 grams (1.1 pounds) for the 

 experiments of 2 It- weeks duration (19U8). When experimental groups 

 are arbitrarily reduced to a fixed weight at intervals within an 

 experiment and a portion of the fish discarded, it is the opinion 

 of the authors that the fish retained on experiment are no longer 

 representative of the original population and, therefore, are not 

 comparable to groups fed other diets in which reductions in the 

 original population have been made at more or less frequent in- 

 tervals and by different amounts. 



