Beside these two sorts of probability—derived from •imperfect experience and from •contrary outcomes—there is a third arising from •analogy, which differs from them in some significant respects. According to the account I have given, all kinds of reasoning from causes or effects are based on two things: •the constant conjunction of any two ·kinds of· objects in all past experience, and •the resemblance of a present object to one of the kinds. These have the effect that •the present object invigorates and enlivens the imagination, and •the resemblance together with •the constant union conveys this force and liveliness to the related idea, which we are therefore said to believe. If you weaken either the •union or the •resemblance, you weaken the force of transition and thereby weaken the belief that arises from it. The liveliness of the first impression can’t be fully transferred to the related idea unless •the conjunction of objects of their kinds has been constant and •the present impression perfectly resembles the past ones whose union we have been accustomed to observe. In probabilities of chance and of causes (discussed above) it is •the constancy of the union that is diminished; and in the probability derived from analogy it is only •the resemblance that is diminished. Without some degree of resemblance there can’t be any reasoning. But this resemblance can be greater or smaller, and the reasoning is proportionally more or less firm and certain. An experience loses some of its force when transferred to instances that don’t exactly resemble it; but as long as there is some resemblance remaining there is still a basis for probability.