THE ART OF SCIENTIFIC INVESTIGATION 



determine. The groups are usually formed by ' randomisation ', 

 that is to say, by assigning individuals to one group or the other 

 by drawing lots or by some other means that does not involve 

 human discrimination. The traditional method of experimenta- 

 tion is to have the groups as similar as possible in all respects 

 except in the one variable factor under investigation, and to 

 keep the experiment simple. " Vary one thing at a time and make 

 a note of all you do." This principle is still widely followed, 

 especially in animal experiments, but with the aid of modem 

 statistical techniques it is now possible to plan experiments to test 

 a number of variables at the same time. 



As early as possible in an investigation, a simple crucial experi- 

 ment should be carried out in order to determine whether or not 

 the main hypothesis under consideration is true. The details 

 can be worked out later. Thus it is usually advisable to test the 

 whole before the parts. For example, before you try to reproduce 

 a disease with a pure culture of bacteria it is usually wise to 

 attempt transmission with diseased tissue. Before testing chemical 

 fractions for toxicity, antigenicity or some other effect, first test 

 a crude extract. Simple and obvious as this principle appears, 

 it is not infrequently overlooked and consequently time is wasted. 

 Another application of the same principle is that in making 

 a first test of the effect of some quantitative factor it is usually 

 advisable to determine at the outset whether any effect is pro- 

 duced under extreme conditions, for example, with a massive dose. 



Another general principle of a rather similar kind is the process 

 of systematic elimination. This method is well exemplified in the 

 guessing game where a series of questions such as " animal, 

 vegetable or mineral " is asked. One can often find the unknown 

 more quickly by systematically narrowing down the possibiUties 

 than by making direct but blind guesses. This principle is used 

 in weighing, when weights that are too heavy and too light 

 are tried, and then the two extremes are gradually brought 

 together. The method is especially useful in seeking an unknown 

 substance by chemical means, but it also has many applications 

 in various branches of biology. In investigating the cause of a 

 disease, for instance, sometimes one eliminates the various 

 alternatives until at last a narrow field is left for one to 

 concentrate on. 



