104 EXPERIMENT STATIOISr RECORD. 



tion. This is common to all experiments concerned with the func- 

 tional activity of living organisms, and is due to the uncertainty of 

 their following exact rules of uniformity, i. e., to what we designate 

 as individuality. 



Mitchell and Grindley have presented a most interesting and sug- 

 gestive study on this element of uncertainty in the interpretation of 

 feeding experiments. It is one of the most effective critical studies 

 of any branch of our station work, and should be very helpful in 

 directing attention to the improvement of experiments of this class 

 and their interpretation. The bulletin illustrates not only the dan- 

 ger to be guarded against from a scientific standpoint, but to an 

 even greater extent when deductions are to be made for the guidance 

 of the farmer, because the latter often can not impose the precise 

 experimental conditions required. 



Manifestly experiments of this class are crude and hence lacking 

 in absolute accuracy. This should be recognized to guard against 

 overconfidence and too broad generalizations; and at the same time 

 the effort should be put forth to improve the methods both in plan- 

 ning and execution. As a first step we need to know for our own 

 information the extent of the experimental error and its source. 

 Until we do know this the necessity of improvement is not apparent 

 and its means is uncertain. Kefinement of certain stages of the 

 feeding may be more than counterbalanced b}^ the inherent errors 

 due to poor selection of animals or some other defect. The extent 

 of the experimental error is an index to the degree to which deduc- 

 tions can be safely drawn, and will indicate caution in making broad 

 generalizations for the benefit of the practical feeder. 



The extensive review of experiment station literature in the United 

 States made by Mitchell and Grindley develops the nature and the 

 source of the experimental error and points to methods of reducing 

 it. They find an average coefficient of variation in gain of about 

 twenty-one per cent for similarly treated lots of sheep and of about 

 seventeen per cent for steers and swine. This points to the danger 

 of small lots of animals and of imeven selection of individuals. As 

 the authors say, " increasing the size of lots is no remedy for a poor 

 selection of experimental animals," and "can not eliminate indi- 

 viduality by merely reducing its effect on the average."' Further- 

 more, " the necessity of selecting homogeneous lots of animals is not 

 appreciably diminished by the balancing of heterogeneous lots.-' 



The critical analysis represented by this bulletin points out the 

 inherent weakness of such experiments, as commonly made, and the 

 need of more scientific and dependable methods in our present feed- 

 ing trials. They are not all that they should be or might be made, 

 and thev are not all that we have assumed them to be. Whether 



