they represent the disability of past experience to furnish 'laws' for their calculation, and the practical importance which attaches to these incalculable or qualitative changes is very considerable.Though they may be comparatively infrequent and may appear on first inspection almost negligible breaks in the otherwise calculable continuity of the evolutionary process, their determinant importance is receiving ever greater recognition.In human conduct, individual or social, these mutations seem to play a larger part, chiefly by reason of the operation of the so-called 'freedom' of the human will.For whatever philosophic view be held regarding the determination of the acts of the will, its operation scatters mutations thickly over the realm of human conduct.Hence it remains true that science can do so much less in explaining and predicting human history than in any other department of nature.No doubt here, as elsewhere, science hopes to apply quantitative analysis of such increasing accuracy as to enable it to determine and predict a larger number of such mutations.Since there doubtless exist quantitative conditions for every qualitative change, it may seem theoretically possible for science some day to catch up with 'the art of creation.' This supposition, however, assumes that the number of permutations and combinations in 'nature' is limited, and that, therefore, in some extensive run history does repeat itself.The final victory of science thus seems to depend upon the adoption of a cyclical view of the history of the universe.But, for all present practical purposes of social processes, science is so far removed from this perfection that the economist and the sociologist are continually compelled to allow for unpredictable changes of such frequency and of such determinant importance that their claim to direct 'the general will' and to mould the conscious policy of a society must be very modestly expressed.

Such laws of causation as they derive from past observation and experiment must usually be conceived as laws of tendencies, seldom endowed with any rigorous authority of close determination, and still more seldom with accuracy of quantitative prediction.

§9.It is sometimes supposed that this hampering effect of the uniqueness, irregularity, novelty and freedom of the individual and social organisms can be got rid of by a process of multiplication in which particular eccentricities will cancel.To economists, in particular, there is a strong temptation to fall back upon the average man, in the belief that scientific determinism justifies itself through averages.Now the radical defect of measurement by averages, as a mode of social valuation, has already been disclosed.The ascertained fact that the average money income, or even the average real income, of the British people may have risen 10% within the last decade, disables itself, by the very process of averaging, from informing us as to the effect of this increase of national wealth upon national welfare.For this effect depends upon the distribution of the increase, and the process of averaging consists in ignoring this vital fact of distribution.

This defect of averages for purposes of interpretation, of course, involves a consequent defect for purposes of guidance in economic conduct.

The calculation that a given course of national conduct, e.g., the expenditure of so many millions upon improved transport, will raise the national or average income by so much, loses all the worth of its superficial exactitude unless we know how much of the increase is going to the landlord in rising rent, how much to the labourer in rising wages.

This, of course, involves no repudiation of the true utility of averages, but only of the spurious accuracy which their forms suggest.The exact statement that the average income of an English family has risen 10% in the last decade does imply a reasonable probability that an increase of total national welfare has taken place.8 But it gives no information as to the amount of that increase, and is consistent with the fact that there may have been a decrease, owing to a worsening of the distribution of the growing income, or of the labour and other costs involved in its production.

§10.So far upon the supposition that welfare is a quantity.It will occur to statisticians that the information to be got from averages of income may be justified by nicer discrimination.If, in addition to learning that the average income of all families has risen 10%, we discovered the different percentages which had been added to rent, interest, profits and wages, or, better still, the ratio of increase for the different income levels, we should surely then, by this extended use of averages, get nearer towards a quantitative estimate of the increase of welfare that had been achieved!