Now it does not seem that such a process as this is likely to be free from error: and, indeed, no one pretends that it is. In fact there is scarcely any point upon which moralisers have dwelt with more emphasis than this, that man’s forecast of pleasure is continually erroneous. Each of us frequently recognises his own mistakes: and each still more often attributes to others errors unseen by themselves, arising either from misinterpretation of their own experience, or from ignorance or neglect of that of others. How then are these errors to be eliminated? The obvious answer is that we must substitute for the instinctive, largely implicit, inference just described a more scientific process of reasoning: by deducing the probable degree of our future pleasure or pain in any given circumstances from inductive generalisations based on a sufficient number of careful observations of our own and others’ experience. We have then to ask, first, how far can each of us estimate accurately his own past experience of pleasures and pains? secondly, how far can this knowledge of the past enable him to forecast, with any certainty, the greatest happiness within his reach in the future? thirdly, how far can he appropriate, for the purposes of such forecasts, the past experience of others? As regards the first of these questions, it must be remembered that it is not sufficient to know generally that we derive pleasures and pains from such and such sources; we require to know approximately the positive or negative degree of each feeling; unless we can form some quantitative estimate of them, it is futile to try to attain our _greatest possible_ happiness--at least by an empirical method. We have therefore to compare quantitatively each pleasure as it occurs, or as recalled in imagination, with other imagined pleasures: and the question is, how far such comparisons can be regarded as trustworthy.