It is important to insist that this question, limited as it is, is the utmost, to which, with any knowledge we have at present or are likely to have for a long time to come, Practical Ethics can hope to give an answer. I have already pointed out that we cannot hope to discover which is the _best_ possible alternative in any given circumstances, but only which, among a few, is better than the others. And I have also pointed out that there is certainly no more than a probability, even if we are entitled to assert so much, that what is better in regard to its immediate effects will also be better on the whole. It now remains to insist that, even with regard to these immediate effects, we can only hope to discover which, among a few alternatives, will _generally_ produce the greatest balance of good in the immediate future. We can secure no title to assert that obedience to such commands as ‘Thou shalt not lie,’ or even ‘Thou shalt do no murder,’ is _universally_ better than the alternatives of lying and murder. Reasons why no more than a _general_ knowledge is possible have been already given in Chap. I. (§ 16); but they may be recapitulated here. In the first place, of the effects, which principally concern us in ethical discussions, as having intrinsic value, we know the causes so little, that we can scarcely claim, with regard to any single one, to have obtained even a _hypothetical_ universal law, such as has been obtained in the exact sciences. We cannot even say: If this action is performed, under exactly these circumstances, and if no others interfere, this important effect, at least, will _always_ be produced. But, in the second place, an ethical law is not merely hypothetical. If we are to know that it will always be better to act in a certain way, under certain circumstances, we must know not merely what effects such actions will produce, _provided_ no other circumstances interfere, but also that no other circumstances will interfere. And this it is obviously impossible to know with more than probability. An ethical law has the nature not of a scientific law but of a scientific _prediction_: and the latter is always merely probable, although the probability may be very great. An engineer is entitled to assert that, if a bridge be built in a certain way, it will probably bear certain loads for a certain time; but he can never be absolutely certain that it has been built in the way required, nor that, even if it has, some accident will not intervene to falsify his prediction. With any ethical law, the same must be the case; it can be no more than a generalisation: and here, owing to the comparative absence of accurate hypothetical knowledge, on which the prediction should be based, the probability is comparatively small. But finally, for an ethical generalisation, we require to know not only what effects will be produced, but also what are the comparative values of those effects; and on this question too, it must be admitted, considering