for good; and that no process, however laborious, which is in the least likely to help us in avoiding this should be evaded. But it is at least possible that one source of error lies in mistaking one kind of reason for another--in supposing that, because there is, in one sense, a reason for a given conclusion, there is also a reason in another, or that because there is, in one sense, no reason for a given conclusion, there is, therefore, no reason at all. I believe myself that this _is_ a very frequent source of error: but it is at least a possible one. And where, as disagreements show, there certainly is error on one side or the other, and reason, too, to suppose that the error is not easy to detect, I think we should spare no pains in investigating any source, from which it is even possible that the error may arise. For these reasons I think I am perhaps doing right in trying to explain as clearly as possible not only what reasons we have for believing in an external world, but also in what sense I take them to be reasons. I proceed, then with my explanation. And there is one thing, which, I think my illustration has shown that I do _not_ mean. I have defined a reason for a belief as a true proposition, which would not be true unless the belief itself--what is believed--were also true; and I have used, as synonymous with this form of words, the expressions: A reason for a belief is a true proposition from which the truth of the belief _follows_ from which it _could_ be _validly inferred._ Now these expressions might suggest the idea that I mean to restrict the word "reason," to what, in the strictest sense, might be called a _logical_ reason--to propositions from which the belief in question _follows,_ according to the rules of inference accepted by Formal Logic. But I am _not_ using the words "follow," "validly inferred," in this narrow sense; I do _not_ mean to restrict the words "reason for a belief" to propositions from which the laws of Formal Logic state that the belief could be deduced. The illustration which I gave is inconsistent with this restricted meaning. I said that the fact that a statement appeared in the _Times_ might be a good reason for believing that that statement was true. And I am using the word "reason" in the wide and popular sense, in which it really might be. If, for instance, the _Times_ stated that the King was dead, we should think that was a good reason for believing that the King was dead; we should think that the _Times_ would not have made such a statement as that unless the King really were dead. We should, indeed, not think that the statement in the _Times_ rendered it absolutely _certain_ that the King was dead. But it _is_ extremely unlikely that the _Times_ would make a statement of this kind unless it were true; and, in that sense, the fact of the statement appearing in the _Times_ would render it _highly probable_--much more likely than not--that the King was dead.