264 DEVELOPMENT AND PURPOSE CHAP. 



6. But it may be asked, at what stage does corroboration 

 yield certainty ? In any system each constituent generalisa- 

 tion is open to the doubt whether all the relevant circum- 

 stances have been taken into account. At what stage and 

 on what grounds can we regard this suggestion as over- 

 thrown? To answer these questions we must, I think, 

 recast the form of statement of the principles on which 

 inference works. We have assumed so far that a relation 

 which is independent of any outside terms is universal. 

 This as it stands is an assertion about reality. Suppose we 

 say instead that an c observed relation of which no external 

 ground is in any way suggested by experience is a logical 

 ground of a universal assertion.' This is not only an asser- 

 tion about reality though it is indirectly an assertion 

 about reality but also a rule for our thought. Let us 

 take this as the principle of our reasoning and see what it 

 leads to. Clearly it would not be a sound rule if the above 

 assertion as to reality were untrue, but that assertion might 

 be true, and yet this principle might be unsound. In 

 assuming it, therefore, as sound, we are assuming some- 

 thing more than before. We are assuming, since it is a 

 rule for guidance, that there is some good to be got by its 

 means. That is not, as I shall presently show, the whole 

 basis of the assumption, but is at least one of its implica- 

 tions which is corroborated if good results. 



Assuming, however, that this is the implication of 

 rational thought, we see that any generalisation to which 

 exception can be taken on the ground that experience does 

 suggest the possibility of some external ground is to the 

 extent of the value of that suggestion doubtful. In point 

 of fact, we do hold generalisation in suspense as long as 

 experience does suggest any possibility of the kind. If no 

 specific interference of a disturbing condition is suggested, 

 there still remains a general doubt of the adequacy of 

 human observation based on the fact that with the utmost 

 care we sometimes err. Our ordinary inductions may be 

 regarded as open to a vague doubt of this kind. On the 

 other hand, if we have succeeded in establishing a system 

 of interconnected generalisations of a kind which has never 

 been shaken, of such a system we might say that experience 



