38 



In addition to this qualitative work, Dr. Bristol Roach has in- 

 troduced exact methods. She has studied quantitatively the growth 

 of a single species in nutrient solutions differing only in the 

 nature of the carbohydrate present, the substances tested being 

 the sugars (pentoses, hexoses, disaccharoses), also mannite and 

 glycerol. The rate of growth of the algae in culture, as measured 

 by the increase in bulk, is constant under uniform favourable 

 conditions for about the first ten days after inoculation, and 

 parallel cultures have equal growth rates within the limits of 

 experimental error. It has therefore been possible to devise a 

 method for growing the alga under constant conditions of tem- 

 perature, light and aeration, and by taking daily measurements 

 of its bulk to ascertain the rate of growth in the presence of the 

 various compounds under investigation. In this way figures 

 have been obtained for a number of the sugars which can be 

 regarded as representing their relative values as energy sources 

 for the organism concerned. Without this physiological work 

 it is impossible to ascertain with certainty the part played by 

 the algae in the important changes going on in the soil. 



STATISTICAL CONTROL OF THE FIELD AND 

 LABORATORY OBSERVATIONS. 



It is one of the distinguishing characteristics of the recent 

 Rothamsted work that the field and laboratory observations are, 

 wherever possible, subjected to close scrutiny in the Statistical 

 Department, with the view of estimating the degree of prob- 

 ability attaching to the results, and of indicating modifications 

 in the plan of the experiments that may increase their 

 accuracy. The field data are examined in order to trace 

 correlations between weather, crop growth and other of the 

 quantities measured, the mass of the data being so great that 

 no other procedure gives equally useful results. 



As a preliminary, Mr. Fisher found it necessary to develop 

 adequate statistical methods for the study of field data. This work 

 has now progressed considerably. 



The methods of experimentation have been closely examined 

 and improvements introduced which allow of a far higher degree 

 of accuracy than could previously be attained. 



The difficulties of the older methods of field experimentation 

 arose from uncontrollable variations in the weather and the 

 soil. Experiments repeated on the same soil in different years 

 give discrepant results owing to the variation of the weather; 

 while experiments repeated on different land in the same season 

 give equally discrepant results owing to the variation of the soil. 

 In consequence, even well conducted field experiments suffered 

 from errors of the order of 5 or 10 per cent, a range of inac- 

 curacy too large to meet the requirements of the practical farmer, 

 to whom a difference of 5 per cent, in his average gross yield 

 may make the whole difference between profitable and unprofit- 

 able farming. In order to eliminate these errors, three types 

 of procedure have been adopted by experimenters : — 



1. To repeat an experiment for a long sequence of years, 

 so that the average yield may be taken to indicate the 



