430 SCIENCE PROGRESS 



time of weaning ; after that there was no significant difference 

 and at the time of slaughter the differences between the lots 

 was so small as to be within the limits of experimental error. 

 So far as the evidence of these experiments goes, it shows that 

 cotton seed oil is as suitable as cod liver oil as a substitute 

 for butter fat in feeding calves. 



I have long recognised that mere chemical analysis and 

 energy value or starch value do not tell all that is required 

 in order to enable us to determine the position and value of a 

 feeding stuff. At one time and that not so long ago energy 

 values, albuminoid ratios and chemical analyses were looked 

 upon as almost the whole gospel of the nutrition of farm 

 animals ; this period of development is still the one represented 

 in the text-books of agriculture and agricultural chemistry. 

 Now it seems that the pendulum is swinging strongly over to 

 the other side and it is desirable to utter just one note of 

 warning. In the reaction against the overgrown claims of an 

 old school, do not let us go to the other extreme and lose hold 

 of what was true and right in their work. Although our 

 methods of food analysis are very imperfect and all our work 

 is vitiated by this and by the great individual variations which 

 occur in experiments with animals, still if there be one solid 

 basis of well-established fact which we hold on to as scienti- 

 fically sound and unassailable it is the energy values of food- 

 stuffs and nutrients. Moreover we are on sure ground in 

 maintaining that energy values for the animal are the same 

 as for the inanimate machine, making due allowance for the 

 products of combustion obtained in each case. 



THE MAGNITUDE OF THE ERROR IN NUTRITION 



EXPERIMENTS 



(Prof. R. A. Berry) 



It is desirable to direct attention to the magnitude of the 

 experimental error in nutrition experiments 'on animals, dealing 

 especially with the case of pigs. 



In an experiment carried out at the experiment station of 

 the West of Scotland Agricultural College in 191 1, in which 

 seventy-six large white pigs were used with an average initial 

 live weight of yy6 lb. equally divided as to sex and all fed on 

 the same ration during fourteen weeks, the probable error for 



