Going back for a moment to the question of experts, probably the most difficult advice to procure, with any degree of satisfaction, is regarding fiction, whether in English or in foreign languages. It has been said that one may approve a book simply on the author’s name, or even on that of the publisher, and this is still true in isolated cases, but in these days, when both author and publisher are continually trying experiments, continually varying standards and style, each book must be dealt with individually. I do not see how one can decide whether a given novel should or should not be bought for a library without reading it through from cover to cover or hearing a report from someone who has so read it and who understands the wants and limitations of the American public library. This is a line, it seems to me, along which great improvement in our selection is possible; but I confess I do not see my way to an immediate solution of the problem. Possibly this is a good opportunity to say a word for a method of testing the adequacy of one’s collection which has scarcely been used as it deserves. One of the most difficult things for a librarian to ascertain is whether his collection is properly distributed among the different classes, and by this I mean, as before, distributed in accordance with the legitimate requirements of the community. It is not possible to find by a statistical method exactly what people need, but it is possible to find out what they want, as indicated by the kind of books that they read. The statistical record of this will be found in the class percentages of circulation. Whether or not the library is equipped to supply this need is indicated by the class percentages of books on the shelves. A comparison of these two percentage tables is always most interesting to the book selector. It does not enable him automatically to select books, but it does indicate points for fruitful investigation. To take some actual cases, I find a library with four per cent of history and six per cent of literature on the shelves, whereas the corresponding circulation percentages are five and seven. This is prima facie evidence that the collections in those two subjects are used rather more than the others and could well be increased. In cases where it is not desirable to encourage circulation in a given class, such an indication should evidently meet with no response. The circulation of fiction always runs far beyond its proportion, and it is neither proper nor desirable for the library to try to keep up. Thus in three libraries where the percentage of adult fiction on the shelves is 20, 19 and 17, respectively, I find the corresponding circulation percentages to be 34, 35 and 27. What, let us ask ourselves, are library statistics for? Is all the labor concerned in their collection and assemblage to result simply in a table that is to be glanced at for a moment with more or less interested curiosity, or do