508 THE AMERICAN JOURNAL OF SOCIOLOGY
evolution would not fit the facts of another stage. Associations must therefore be classified functionally ', and, more than that, our working test of all functional classifications must be our teleological concepts. That is, we are bound to schedule associations in accordance with our judg- ment of their relation to the scale of ends that are at issue in the par- ticular situation in which those associations function.
In view, then, of the many previous efforts to classify social phenomena, and of their unfortunate failure to reach results at all satisfactory, it behooves everyone who ventures to renew the undertaking to do so with caution and modesty. It is probable that each type of classification thus far proposed has certain per- manent uses. Much more has been accomplished in the aggre- gate by previous attempts, abortive though we may judge each of them to be, than is likely to be added by any single organizer of data. In all probability each of the attempts at classification named above, with many others of which they are types, con- tains something of value. At all events, the limitations which the attempts encounter are full of meaning, if we can learn how to interpret them.
It does not follow that the task of classification which remains is a problem of arranging harmony and synthesis between the various previous classifications. We are rather in search of a harmony and synthesis of the different portions of subject-matter with which these attempts have been concerned. Nothing is plainer than that the different schemes which we have reviewed have been concerned primarily with one of several dif- ferent orders of facts. Some of them have seen society as it appears from the ethnologists' point of view; others, as it appears from the economists', the technologists', the legists', or the moralists' point of view. Accordingly, as we have found to be the case throughout the social sciences, the makers of the respective classifications have each unconsciously based their calculations upon a very narrow abstraction from the social whole, and have tried to make this abstraction a secure nucleus about which the rest of the facts might gather.
The program now to be proposed for a more adequate method begins with our conception of the whole in and of which
