EXPERIMENTATION 



necessary in drawing conclusions as to how widely applicable 

 are results obtained under necessarily limited sets of circum- 

 stances. 



Darwin once said half seriously, " Nature will tell you a direct 

 lie if she can." Bancroft points out that all scientists know from 

 experience how difficult it often is to make an experiment come 

 out correctly even when it is known how it ought to go. There- 

 fore, he says, too much trust should not be put in an experiment 

 done with the object of getting information.^" 



The examples quoted are experiments which gave results that 

 were actually " wrong " or misleading. Fortunately they are 

 exceptional. Commoner, however, is the failure of an experiment 

 to demonstrate something because the exact conditions necessary 

 are not known, such as Faraday's early repeated failures to obtain 

 an electric current by means of a magnet. Such experiments 

 demonstrate the well-known difficulty of proving a negative 

 proposition, and the folly of drawing definite conclusions from 

 them is usually appreciated by scientists. It is said that some 

 research institutes deliberately destroy records of " negative 

 experiments ", and it is a commendable custom usually not to 

 publish investigations which merely fail to substantiate the hypo- 

 thesis they were designed to test. 



SUMMARY 



The basis of most biological experimentation is the controlled 

 experiment, in which groups, to which individuals are assigned at 

 random, are comparable in all respects except the treatment under 

 investigation, allowance being made for the inherent variability 

 of biological material. Two useful principles are to test the whole 

 before the part, and to ehminate various possibilities systemati- 

 cally. In the execution of an experiment close attention to detail, 

 careful note-taking and objectivity in the reading of results are 

 important. 



Biometrics is concerned with the planning of experiments 

 as well as the interpretation of results. A basic concept in 

 biometrics is that there is an infinitely large, hypothetical popula- 

 tion of which the experimental group or data are a random 

 sample. The difficulty presented by the inherent variability of 



25 



