beyond a doubt, and that the rest—all subordinate truth—is subject to error in various degrees. Any finite truth, to be made quite true, must more or less be modified; and it may require modification to such an extent that we must call it utterly transformed. Now, in Chapter xxiv., we have already shown that this account holds good, but I will once more insist on our fallibility in finite matters. And the general consideration which I would begin by urging, is this. With every finite truth there is an external world of unknown extent. Where there is such an indefinite outside, there must be an uncertain world of possible conditions. But this means that any finite truth may be conditioned so as to be made really quite otherwise. I will go on briefly to apply this.
Wherever a truth depends, as we say, upon observation, clearly in this case you cannot tell how much is left out, and what you have not observed may be, for all you know, the larger part of the matter. But, if so, your truth—it makes no difference whether it is called “particular” or “general”—may be indefinitely mistaken. The accidental may have been set down as if it were the essence; and this error may be present to an extent which cannot be limited. You cannot prove that subject and predicate have not been conjoined by the invisible interposition of unknown factors. And there is no way in which this possibility can be excluded.
But the chance of error vanishes, we may be told, where genuine abstraction is possible. It is not present at least, for example, in the world of mathematical truth. Such an objection to our general view cannot stand. Certainly there are spheres where abstraction in a special sense is possible, and where we are able, as we may say, to proceed a priori. And for other purposes this difference, I agree, may be very important; but I am not concerned here with its importance or generally with its
