Last in the list of the necessary items of statistics comes that of readers or users of the library--the most interesting in some ways, and the most disappointingly vague. Presumably your users fill out some kind of blank form of application and have their names entered in a book. It is therefore easy to give, as is usually done, the total registration and its annual increase. But this is evidently not the number of actual users of the library. Who are the “actual users”? The expression itself is vague. To be complete you should have the numbers of those who have used the library within one, two, and three days, and so on back indefinitely. There is no place where the line may be drawn between “live” and “dead” cards. But such statistics are too elaborate to collect regularly, so that the ordinary library leaves this subject in its pristine mistiness. There are some pretty variations of it, however, which may be gone into if there is time. For instance, how are your users divided, according to occupation? This you can ascertain from your applications provided the applicant is required to state his occupation. Here again the result is for registered users, not actual users. Again, how are your users distributed topographically? The result of this inquiry may be shown graphically on a map, and it is particularly valuable when one is thinking of moving or of establishing a branch; but it takes more time than is at the disposal of most librarians. Here, I believe, ends the enumeration of necessary kinds of statistics. In each kind the collection may be reduced to a minimum; but the librarian must, if the library is to be maintained at all, keep a cash account, count the books, and make some kind of a list of them. Also, if at all possible she or he must be able to tell how many books are circulated and how many users’ names are on the books. This is the minimum; the maximum is fixed only by considerations of time and usefulness. First among the kinds of statistics that are not absolutely necessary, but interesting and often useful, is that of routine work done--letters written, visits made, cards written. This may easily be carried to excess. Then there is the enormous class in which the data are obtained not directly, but by comparison of other data. To this class belong the financial comparisons already noted. For instance, by comparing the circulation of separate classes with the total we get class percentages--a very useful type of statistics; by comparing circulation with books on shelves we get the average circulation of each book, etc. There is no end to the varieties of this class of statistics, and they may be rated all the way from “very valuable” to “useless” or even “nonsensical”. The whole class would require a separate paper to discuss.