12 



29. Here, then, we distinguish a different kind of force 

 belonging to two descriptions of evidence, viz. certainty, which 

 proceeds from the direct testimony of the senses; and probability, 

 which consists of the frequency of the connection between the 

 thing seen, and the thing unseen, which latter is inferred from 

 witnessing the former. 



30. Probability always implies some doubt, as an inference may 

 always be superseded by a perception : but the degree of doubt 

 belonging to probable evidence is proportioned to the compara- 

 tive frequency of the association inferred, and the absence of such 

 association. Thus a man who should conclude a luminous body, 

 looking like fire, to be hot, would be less likely to be wrong, than 

 one who should conclude such a person to be a Christian, merely 

 because he goes to church; heat being most commonly conjoined 

 with a luminous body, and perhaps not one person in a hundred 

 of those who go to church being a Christian, except just when 

 it suits his humour or his convenience. Experience, or the 

 testimony, in this case, of another sense, would confirm the former 

 inference ; and the latter inference would be liable to be super- 

 seded, in a proportionate number of instances, by the belief of 

 experience. 



3J. Upon this simple plan the immense fabric of human 

 reasoning is constructed. We appeal to the senses for wjequivo- 

 cal proof, or a testimony, the impression of which, while the 

 senses preserve common relations, cannot be superseded; and in 

 the absence of such proof we act upon probability, and we rely 

 upon it with a strength proportionate to the following gradations 

 of evidence: 1st, when the connection is universal, we on this 

 account are apt to suppose it necessary; 2nd, when the exceptions 

 bear but a trifling proportion to the frequency of the association; 

 3rd, when the exceptions are not so frequent as the association ; 

 4th, the judgment is suspended by a recollection of an equal 

 frequency of the associations and of the exceptions; 5th, the 

 inference is negatived, or probability inclines us to the opposite 

 belief, when the exceptions are more frequent than the associa- 

 tions. This gradation appears necessarily true, because we give 

 credit to the inference of a connection, from our past experience 

 of it; and the probability of its occurrence, under apparently 

 similar circumstances, must be in proportion to the times which it 

 does occur, or has occurred. This seems no more than a 

 definition of the common sense of mankind on this subject. 



32. In examining the truth of a proposition, our business is 

 to take account of the evidence, which we shall often find to be 

 most complicated; a knot which the understanding ties perhaps 

 involuntarily, and which the understanding by any effort can 

 scarcely again untie. We shall find the facts, authority, or basis 

 of the evidence to be sometimes doubtful; we shall find it 

 ecessary to examine on what the truth of the evidence depends: 



