EXPERIMENTATION 



matter how objectively minded the scientist may believe him- 

 self to be, it is very difficult to be sure that his judgment 

 may not be subconsciously biased if he knows to which group 

 the cases belong when he is judging them. The conscientious 

 experimenter, being aware of the danger, may even err by 

 biasing his judgment in the direction contrary to the expected 

 result. Complete intellectual honesty is, of course, a first essential 

 in experimental work. 



When the experiment is complete and the results have been 

 assessed, if necessary with the aid of biometrics, they are 

 interpreted by relating them to all that is already known about 

 the subject. 



Planning and assessing experiments 



Biometrics, or biostatistics, the application of the methods of 

 mathematical statistics to biology, is a comparatively new branch 

 of science and its importance in research has only lately won 

 general recognition. Books dealing with this subject have been 

 mentioned in Chapter One and I do not intend to do more here 

 than call attention to a few generalities and stress the need for 

 the research worker to be acquainted at least with the general 

 principles. Some knowledge of statistical methods is necessary 

 for any form of experimental or observational research where 

 numbers are involved, but especially for the more complex 

 experiments where there is more than one variable. 



One of the first things which the beginner must grasp is that 

 statistics need to be taken into account when the experiment is 

 being planned, or else the results may not be worth treating 

 statistically. Therefore biometrics is concerned not only with the 

 interpretations of results but also with the planning of experi- 

 ments. It is now usually taken as including, besides the purely 

 statistical techniques, also the wider issues involved in their appli- 

 cation to experimentation such as the general principles of the 

 design of experiments and the logical issues concerned. Sir 

 Ronald Fisher, who has done so much to develop biometrical 

 methods, discusses these topics in his book, The Design of 

 Experiments.^^ 



In selecting control and test groups, logic and common sense 

 have first to be satisfied. A common fallacy, for instance, is to 

 compare groups separated by time — the data of one year being 



19 



