236 U. S. BUREAU OF FISHERIES 



almost entirely of fragments of the siphon. For feeding, it was 

 soaked in warm Avater for a short time and then mixed with th& 

 ground liver. Very good results were obtained with a mixture of 

 40 per cent sheep liver, 50 per cent soy-bean meal, and 10 per cent 

 shrimp bran, which produced more rapid growth than a mixture of 

 equal parts of sheep liver and soy-bean meal. A mixture of 75 per 

 cent sheep liver with 25 per cent wheat middlings produced sur- 

 prisingly rapid growth; in fact, the average percentage of daily 

 increase in weight in the lot fed this combination of foods was greater 

 than that of any of the other experimental lots. The fish showed a 

 tendency to be " pot bellied," however, and were fatter than those in 

 the other lots. It is not believed that such short, thick-bodied fish 

 are as desirable as those that approximate wild fish in shape. Furth- 

 ermore, the presence of a number of exceptionally large fish in this lot 

 at the end of the experiment indicates that considerable cannibalism 

 occurred. 



Straight sheep liver does not appear to be a more satisfactory food 

 for yearlings than for fingerlings, at least on the basis of growth, 

 alone, as in this respect the controls were much smaller than the 

 other experimental lots, with the single exception of the fish fed sheep 

 liver and Mexican beans. A mixture of equal parts of Mexican beans 

 and sheep liver produced less growth in the brook trout than any 

 other of the experimental diets ; but with rainbows the same mixture 

 produced much more rapid growth, exceeding sheep liver in effec- 

 tiveness. Unfortunately, no rainbow trout were fed the liver and 

 soy-bean meal mixture, so that we have no evidence as to whether the 

 relative value of soy-bean meal and Mexican beans would be reversed, 

 as in the fingerlings. 



An essential part of the experiments at Holden is the work on 

 selective breeding, by means of which, it is confidently believed, it 

 will be possib]e to produce a superior strain of trout. Owing to in- 

 adequate facilities and the short time the Holden station has been 

 used for experimental work, it has not been possible to make much 

 progress in this direction. Selective-breeding experiments require a 

 great deal of space, owing to the many small lots of fish that must 

 be kept separate from each other. 



During the season of 1927 six pairs of brook trout were mated and 

 the progeny reared in separate compartments. As the parents all 

 were selected fish, it is not surprising that the young made much bet- 

 ter growth than do average fingerlings from the common brood stock. 

 The comparatively slight variation in size in individual fish in each 

 lot was especially striking. As every fish culturist knows, ordinarily 

 there is considerable difference in the rate of growth of individual 

 fish in a mixed lot of fingerlings of the same age, and it is necessary 

 to grade them frequently to prevent cannibalism ; but in fish having 

 the same parents there was surprisingly little difference in size 

 throughout the summer; in fact, in some instances they were more 

 uniform than fingerlings that were graded carefully. 



Although the progeny of the selected pairs were held in crowded 

 quarters in hatchery troughs for several months, the losses in some 

 instances were remarkably small. In one lot the, combined loss of 

 eggs, fry, and fingerlings up to September 30 was only 9 per cent, and 

 in another lot it was 12 per cent. It is planned to select the best fish 



