There is a very useful theorem for calculating the certainty of a matter, such as the evidence for a crime. When the pieces of evidence for some matter are interdependent, that is, when the pieces of evidence cannot be tested except against each other, then, the more evidence is adduced, the less credible is the matter in question, because anything which would make the earlier parts fail will make the later parts fail too. {{When all the pieces of evidence for some matter depend equally on a single piece, the number of pieces neither increases nor decreases the probability of the matter, because their joint value as evidence is included in the value of the piece on which they all depend.}} When the pieces of evidence are independent of each other, that is, when the evidence can be tested other than by each other, then, the more evidence is adduced, the more credible is the matter in question, because the falsity of one piece of evidence does not affect the validity of the others.

It may seem odd that I talk of probability in relation to crimes, which have to be certain if they are to call for punishment. But the paradoxicality here will disappear if we see that moral certainty is, strictly speaking, nothing but a probability, though a probability of such a sort as to be called certainty because every reasonable man necessarily assents to it out of force of habit born of the need to act and antecedent to any theorising.