CLASSIFICATION OF FALLACIES. 307 



tyranny is distressed, which, however true, proves nothing, 

 the middle term being undistributed.&quot; The former would be 

 ranked, in our distribution, among fallacies of generalization, 

 the latter among those of ratiocination. &quot; Which are we to 

 suppose the speaker meant us to understand ? Surely&quot; (if he 

 understood himself) &quot;just whichever each of his hearers might 

 happen to prefer : some might assent to the false premise ; 

 others allow the unsound syllogism.&quot; 



Almost all fallacies, therefore, might in strictness be 

 brought under our fifth class, Fallacies of Confusion. A fallacy 

 can seldom be absolutely referred to any of the other classes ; 

 we can only say, that if all the links were filled up which 

 should be capable of being supplied in a valid argument, it 

 would either stand thus (forming a fallacy of one class), or 

 thus (a fallacy of another) ; or at furthest we may say, that 

 the conclusion is most likely to have originated in a fallacy of 

 such and such a class. Thus in the illustration just quoted, 

 the error committed may be traced with most probability to 

 a fallacy of generalization ; that of mistaking an uncertain 

 mark, or piece of evidence, for a certain one ; concluding 

 from an effect to some one of its possible causes, when there 

 are others which would have been equally capable of pro 

 ducing it. 



Yet, though the five classes run into each other, and a 

 particular error often seems to be arbitrarily assigned to one of 

 them rather than to any of the rest, there is considerable use 

 in so distinguishing them. We shall find it convenient to set 

 apart, as Fallacies of Confusion, those of which confusion is 

 the most obvious characteristic; in which no other cause can 

 be assigned for the mistake committed, than neglect or inabi 

 lity to state the question properly, and to apprehend the 

 evidence with definiteness and precision. In the remaining 

 four classes I shall place not only the cases in which the evi 

 dence is clearly seen to be what it is, and yet a wrong conclu 

 sion drawn from it, but also those in which, although there be 

 confusion, the confusion is not the sole cause of the error, but 

 there is some shadow of a ground for it in the nature of the 

 evidence itself. And in distributing these cases of partial 



202 



