THE ART OF SCIENTIFIC INVESTIGATION 



Take, for example, the testing of a new vaccine against a disease. 

 The vaccine may prove effective in several experiments but we 

 must still be cautious in saying it will be effective in future. 

 Influenza vaccine gave a considerable degree of protection in large 

 scale trials in U.S.A. in 1943 and 1945, but against the next 

 epidemic in 1947 it was of no value. Regarded as a problem in 

 logic the position is that by inductive inference from our data we 

 arrive at a generalisation (for instance, that the vaccine is effec- 

 tive). Then in future when we wish to guard against the disease we 

 use this generalisation deductively and apply it to the particular 

 practical problem of protecting certain people. The difficult 

 point in the reasoning is, of course, making the induction. Logic 

 has little to say here that is of help to us. All we can do is to 

 refrain from generalising until we have collected fairly extensive 

 data to provide a wide basis for the induction and regard as 

 tentative any conclusion based on induction or, as we more often 

 hear in everyday language, be cautious with generalisations. 

 Statistics help us in drawing conclusions from our data by ensur- 

 ing that our conclusions have a certain reliability, but even 

 statistical conclusions are strictly valid only for events which have 

 already occurred. 



Generalisations can never be proved. They can be tested by 

 seeing whether deductions made from them are in accord with 

 experimental and observational facts, and if the results are not 

 as predicted, the hypothesis or generalisation may be disproved. 

 But a favourable result does not prove the generalisation, because 

 the deduction made from it may be true without its being true. 

 Deductions, themselves correct, may be made from palpably 

 absurd generalisations. For instance, the truth of the hypothesis 

 that plague is due to evil spirits is not established by the correct- 

 ness of the deduction that you can avoid the disease by keeping 

 out of the reach of the evil spirits. In strict logic a generalisation 

 is never proved and remains on probation indefinitely, but if it 

 survives all attempts at disproof it is accepted in practice, 

 especially if it fits well into a wider theoretical scheme. 



If scientific logic shows we must be cautious in arriving at 

 generalisations ourselves, it shows for the same reasons that we 

 should not place excessive trust in any generalisation, even widely 

 accepted theories or laws. Newton did not regard the laws he 



88 



