We must, however, draw a very important distinction. There is one method by which uniformities of a certain kind can be detected in the behaviour of purposive intelligent beings, without insight into the nature of their individual purposes—the method of statistical averages. Thus, though it would be quite impossible to say with certainty of any individual man that he will shoot himself or will get married, except on the strength of insight into his individual character and interests, we find by experience that it is possible to say, within a certain narrow range of error, what percentage of Englishmen will shoot themselves or will get married in the year. The percentage is, of course, rarely or never precisely realised in any one year, but the longer the period of years we take for examination, the more exactly do the deviations from the average in individual years compensate one another. The explanation is, of course, that on the whole the incentives to marriage or suicide, in a reasonably stable state of society, remain constant from year to year, so that by taking an average of several years we can eliminate results which are due to individual peculiarities of temperament and situation, and obtain something like a measure of the degree in which the general conditions of social existence impose a certain common trend or character on the interests and purposes of individuals. Two things are at once noticeable in connection with all uniformities obtained by the method of averages. One is that the result formulated in the statistical law is always one to which the actual course of events may reasonably be expected to conform within certain limits of deviation, never one to which we have a right to expect absolute conformity. Not only is the actual number of marriages, _e.g._, in any one year, usually slightly above or below the average percentage computed, _e.g._, for a ten years’ period, but as we compare one longer period with others, the average percentage for the longer period itself fluctuates. It is only in the “long run,” that is, in the impossible case of the actual completion of an interminable series, that the computed average would be exactly realised. As every one who has to deal with averages in any form knows, precise realisation of the computed average within a finite series of cases would at once awaken suspicions of an error somewhere in our calculations. Thus the uniformities of this kind are never absolutely rigid; they are ideal limits to which the actual course of events is found to approximate within certain limits of divergence.