E. H. Richards 303 



It appears that a grass diet without any corn ration reduces the 

 value of horse faeces regarded as a nitrogen fixing medium to about the 

 level of grass and cake fed bullock droppings. Bullocks and cows on 

 grass alone seem to produce nothing in their excrement which can 

 function as food for nitrogen fixers — under the conditions of these 

 experiments at any rate — but the relatively poor digestion of the horse 

 leaves sufficient carbohydrate unassimilated to show an amount of 

 fixation well beyond the experimental error. 



In the next experiment horse, bullock, and cow faeces, all, however, 

 on different diets, were run in parallel for 112 days. To each flask was 

 added 0-1 gram calcium carbonate, 0-1 gram of garden soil, and 25 c.c. 

 tap- water. Each variety of dung was tested in triplicate and the mean 

 results are given below. 



Comparative fixation of Nitrogen in Horse, Bullock 

 and Cow faeces. 



matter in faeces ... ... ... + 3-0 -0-1 -0-0 



The sample of horse faeces was evidently not as effective as the 

 previous one and neither the bullocks nor cows were getting any cake 

 when these samples were taken (August, 1915). In this experiment 

 the bullock faeces did not fix even the small amount of nitrogen found 

 in the last experiment. The cow dung actually lost a little nitrogen, 

 thus falling into its natural position on the hypothesis that fixation is 

 a function of undigested residues from concentrated foods. 



To complete this part of the work it was necessary to examine the 

 droppings of a cow fed on a cake diet. By the kindness of Mr C. F. 

 Benson, manager of the Hon. Rupert Guinness' farm at Woking Park, 

 a suitable sample was obtained from an animal specially fed for the 

 purpose of this experiment which was run for 28 days with the results 

 set out below. 



Even when the animal was fed on an abnormally rich diet cow 

 faeces fixed practically no nitrogen, but it is significant that both the 

 duplicate experiments show an increase close to the experimental error, 



