60 AN INTRODUCTION TO SCIENCE 



regulating judicial procedure. "He talks as a Lord 

 Chancellor," said Hobbes. According to Lord Chancellor 

 Verulam, Science must progress from step to step, never 

 committing itself to any hypothesis which is not the nec- 

 essary inference from observation. The true scientific 

 method is always to be strictly inductive a most useful 

 restriction, and especially necessary in Bacon's day. 



Descartes' richer imagination took longer flights. In 

 certain matters he even asked of his inner consciousness 

 how he himself felt that things ought to be? How would 

 he have made them had he had the making of the world ? 

 Then he collected evidence to show that they are as he 

 supposed ti priori that they would be. This is deduction ; 

 building downwards, although the process by which Des- 

 cartes tested his evidence was as strictly inductive as 

 Bacon could exact. 



After all, the difference between induction and deduction 

 is a question of name. We know nothing of the universe 

 but that which we have learnt by experience or that which 

 our predecessors have learnt by experience and have re- 

 corded for us. , \Vhen the imagination takes a long flight, 

 when it seeks an a priori explanation, it is but appealing 

 to experience, although it is unable to trace the steps 

 along* which the reason marches in seeking so distant a 

 cause for effects which are near at hand. And \vhen we 

 come back to experience for proof of the applicability of 

 far-fetched explanations the reason moves towards it by 

 processes of induction. Every hypothesis is by definition 

 an advance on knowledge. It is in the nature of a deduc- 

 tion that reason goes on before observation. Observations 

 are then built up to support reason. The difference between 

 induction and deduction is but a difference in degree. 



It is characteristic of science to proceed with the utmost 

 caution to build a pyramid of inductions, each tier of 

 which contains a smaller number of generalisations than 

 the tier upon which it rests, until the apex is a compre- 



