F. Bandelier, published in the Proceedings of the American Antiquarian Society, at the meeting held at Worcester, October 21, 1880.

Bibliographies of the writings of Dickens and Thackeray; a list of the published writings of Herschel on astronomical subjects, by E. S. Holden, and many other interesting lists of books, pamphlets, and articles have appeared since our last meeting. Particulars regarding them may be found by consulting a recent file of the Library Journal. The paper, "Library Aids," which I read at the meeting of this Association in Baltimore last February, has been published as a circular of information by the Bureau of Education in Washington, and is probably in the hands of all librarians.

Lists of good books are appended to many of the articles in the Encyclopædia Britannica.

The Birmingham Free Libraries have just issued a preliminary list of Bibliography in the new reference library, Birmingham, 1881-82.

Very valuable bibliographical aid is to be had from the Bulletins of the Boston Public Library and of Harvard University. Twelve numbers have been issued of Bibliographical Contributions prepared in the Library of Harvard University, and edited by Justin Winsor.

Among the latest issues in this series are: 11. Samuel H. Scudder, The Entomological Libraries of the United States. 12. A List of the Publications of Harvard University and its officers, 1870-1880. 14. William H. Tillinghast, Notes on the Historical Hydrography of the Handkerchief Shoal in the Bahamas. 15. J. D. Whitney, List of American authors in Geology and Paleontology.

It is announced that No. 17 of this series will be: A List of the most useful Reference Books, by Justin Winsor.

Cornell University has begun the publication of "The Library." The first number was issued with the date January, 1882, and contains notes, list of additions, etc. This number has two bibliographical lists, namely: Works relating to architecture, in Cornell University Library, and Petrarch Bibliographies.

Twelve monthly bulletins a year are issued by the Cincinnati Public Library. When gathered into volumes, indexes are provided, namely, a subject-index and an index of authors, anonymous works and collections.

In closing this branch of my subject, I must call attention to an important work on anonyms and pseudonyms, the first volume of which bears the imprint of the present year, namely, Halkett, S: and Laing, J: Dictionary of the anonymous and pseudonymous literature of Great Britain, including the works by foreigners written in, or translated into, the English language. Vol. I. Edinburgh. £2. 2. (Boston agents, Lockwood, Brooks, & Co. $10.50.)

For current information in regard to pseudonyms and anonyms, reference is made to this heading of the Library Journal.

While finishing this portion of my report the welcome news comes of the publication, by Trübner & Co., of a second edition of their Catalogue of Dictionaries and Grammars of the principal languages and dialects of the world.

Miscellaneous.

In conclusion, I will mention the titles of five books recently published, which librarians will at once recognize as sources from which to obtain information very generally sought for.

Who wrote it? An index of the authorship of the more noted works in ancient and modern literature, by William A. Wheeler. Edited by Charles G. Wheeler. Boston: Lee & Shepard, 1881.

Familiar Allusions. A hand-book of miscellaneous information, by William A. Wheeler and Charles G. Wheeler. J. R. Osgood & Co., 1882.

The reader's hand-book of allusions, references, plots, facts, and stories, by Rev. E. Cobham Brewer. J. B. L. & Co.

Words, facts, and phrases. A dictionary of curious, quaint, and out-of-the-way matters, by Eliezer Edwards. J. B. L. & Co., 1881.

I

Ogilvie, J. Imperial dictionary of English. New edition, by C. Annandale. Vols. 1 and 2. Of this work, which is practically a new work, by Mr. Annandale, and which claims to contain many thousand more words than any other dictionary of our language, the London Spectator of November 26, 1881, says: "We have no hesitation in saying that it will prove a most thorough piece of workmanship, and that, among reference-books of its class, it will hold the first place, both as an authority and a source of instruction and entertainment."

IN 1870 I was requested to examine the vari

ous systems of shelf-arrangement then in use, with the view of recommending one for adoption in the New York Apprentices' Library, which had been successively arranged on the numerical and alphabetical plans, and had found both unsatisfactory. None of the schemes examined seemed to be free from objection, and a careful study of them was made, with the idea of combining their best features in one system. It soon became evident to me that all the possible methods of shelf-arrangement might be reduced to three fundamental forms the Numerical, the Alphabetical, and the Classified. It then occurred to me that if a system could be devised that combined the three forms in one, it would approach nearer perfection than any other, and it is this leading idea that forms the basis of the "Combined system," devised by me in 1871, which has been in successful operation in the Apprentices' Library since that time. A brief account of the plan will be found in the preface to the catalogue of the Apprentices' Library, published in 1874, and a fuller explanation, written at the solicitation of the editor, appeared in the Library journal, Vol. 3, No. 1. Without any effort on my part, other than the above, the scheme has attracted considerable attention, and has been adopted, with more or less modification, in at least ten libraries in this country. Indeed, I may say that, either directly or indirectly, it has suggested all the systems of shelfarrangement, devised since 1871, that have come to my knowledge. Most of these plans have been published in the Library journal, and have been taken up for consideration at the annual conventions of the Library Association. In reading the proceedings of these meetings I have been unable to discover any reference to myself as the originator of the method of combining the three fundamental forms of shelf-arrangement, although elsewhere, and in private, most of the authors of these

modified systems have freely acknowledged their indebtedness to me. I make the above explanation not only in justice to myself, but to prevent any possible misconception in the minds of those who are unacquainted with the facts.

The system submitted in the present paper is substantially the same as the one devised by the author in 1871, but contains such improvements and modifications as have suggested themselves in an experience of ten years with the plan in its crude form. The changes introduced are considerable, but do not in any way affect the essential and fundamental principles. In its present dress I believe the system to be worthy of consideration and examination, and, if not accepted in all its details, it may still be of value, as it was in its original form, in suggesting improvements and modifications to others.

The points in the present system that seem to me to require special consideration and explanation, are: A, the classification; B, the author-number; C, the title-number; D, the treatment of duplicates and editions; and, E, the manner of indicating the size. I will consider each of the points in order.

A. The Classification.

There are 23 main departments, of which 20 are devoted to the 8° and smaller sizes, and the remaining three to the 4° and larger sizes. Twenty of these departments are designated by the initials of their names. In the remaining three the class letter designates the size, Q being used for the quartos, F for the usual folios, and X for the Xtraordinary sizes, such as elephant folios and the like, that have to be placed in cases, or specially constructed shelves.

Each of these 23 departments, except class N (novels), is divided into nine classes, which are designated by the Arabic figures 1 to 9. They are also arranged alphabetically. In the 22 departments arranged on this plan there are

A final division of these 220 classes is made by dividing each into four sub-classes, designated in the tables by the letters a, b, c, d. These letters are only for reference, and are not actually used in applying the scheme. How one division is distinguished from another will appear when we come to the explanation of the author-number.

As

I call the classification mnemonic because it is alphabetical and self-explanatory. The order of the alphabet is peculiarly adapted for mnemonic purposes, as it is universally known and understood. I could easily have thrown the tables into a logical form, but I have carefully refrained from so doing, as I am sure that, like its innumerable predecessors, it would have been satisfactory to no one but its maker. There are many libraries arranged in logical order; but no two of them are arranged alike, and the only thing that the classifiers from Aristotle to Messrs. 'Cutter and Perkins are agreed upon is to disagree. There must be something radically wrong in a method that results in discord. the basis of each new logical scheme depends upon some preconceived metaphysical idea in the mind of the classifier, I would call it the Subjective method, and would substitute for it an alphabetical or Objective method, in which the order of the classes is conditioned by something outside of the mind of the classifier, that is to say, by the names of the subjects themselves. Assuming that an alphabetical arrangement of classes is desirable, there would probably be very little difference of opinion as to the order and nomenclature here chosen, as I have endeavored, as far as possible, to select the names most generally associated with their respective subjects. Still, in this respect, the scheme is, to a great extent, only tentative, and is open to improvement. If space permitted, much additional argument might be advanced in favor of the proposed arrangement, but I will content myself, at present, with one that seems to me to have considerable weight. If the Alphabetico-classed method of classification, as exemplified in the catalogues of the Harvard, Congress, Brooklyn,

harmonizing the rival claims of the Systematic and Dictionary catalogues, it is not clear why the same method should not work just as well on the shelves. Precisely the same arguments will apply in the one case as in the other.

In laying out the details of the classification the law of proportion has been strictly observed. If we divide human knowledge into the three well-defined groups, HISTORY, LITERATURE, and SCIENCE, it will be found that each has just seven departments in the present scheme. The same law is observed in the minor divisions, as far as practicable, so that although there are only 887 heads, every important subject about which books are written, or that is likely to be sufficiently represented in a library, is provided with a separate rubric. In some of the later schemes very little attention is paid to this law. We find, for instance, minute subdivisions of Philosophy, Photography, and Engraving, the three classes embracing 120 out of 1,000 classes, or 8 per cent. of the whole; whereas Geography and Travels, which in most libraries have ten times as many works as all three combined, have no more sections than Photography! On the other hand, Fiction, which is the most largely represented class, in circulating libraries at least, is relegated to an obscure corner of Literature. The natural consequence of this unphilosophical proceeding is a multiplication of numbers where economy is most desirable, and the evil is sought to be overcome by either omitting the class symbols altogether, or by substituting some arbitrary mark in their place.

B. The Author Number.

Having a system of alphabetically arranged classes from A.o to Z.9, the problem is to unite with them a series of numbers sufficiently large to provide for the probable acquisitions in each class. I have selected 999 as the lowest admissible number. The usual method of numbering the separate works in each class, in the "Movable" system, is to take them in the order of their acquisition and call the first No. 1, the second No. 2, and so on. This is essentially arbitrary, as there is no reason whatever, aside from the mere accident of purchase,

why a book should have one number rather than another. As the subject treated of determines the place of each book in a scheme of classification, it would be more logical and consistent to have the number conditioned by something in the book itself. It was this consideration which led me, in 1871, to devise my system of alphabetical notation, which forms one of the essential and peculiar features of my original plan, and which has been adopted with more or less variation in most of the schemes devised since that time. The 999 numbers in our table might be arranged in one series of alphabetical combinations; but in that case only 230 classes could be numbered, and our tables have 880 exclusive of Fiction. If we want to number more than 230 classes, we must divide the 999 numbers into as many series as there are subdivisions in each of the classes A.o to Z.9. I have selected four divisions as sufficient. This gives three series of 300 each and one of 100 numbers. In the numbering table these four series are headed a, b, c, d, and correspond to, and are to be used in numbering, the similarly designated sub-classes in the mnemonic classification. Care has been taken that the series of 100 numbers is always used for the less important classes.

If we take a number of alphabetically arranged works, such as Directories, Cyclopedias, and Catalogues, and average the space occupied by the several letters, we shall find that we can make nine nearly equal divisions with the following letters: No. 1 beginning with A, 2 with B, 3 with D, 4 with G, 5 with I, 6 with M, 7 with O, 8 with S, and 9 with T. This scheme of division is easily remembered, as the vowels A, E, I, O, and U, have the odd numbers 1, 3, 5, 7, and 9. It not only forms the basis of the several numbering tables, but has been applied in numbering the classes. It will, therefore, serve as a mnemonic key to the whole system, and will enable one to not only give the class number of every important subject, but to make a pretty shrewd guess as to the author number. Thus British kallography is K. 2; British history, E. 2; Music, D. 6; Chemistry, C. 2; Biblical theology, T. 2; and Engineering, U. 3. In the few cases where this idea could not be

:

entirely carried out, it will be found that there is a variation of only one number.

In applying the key to the first table of 100 numbers, we get 9 divisions of 11 numbers each; consequently the key letters will be numbered as follows: A. o; Br. 113; D. 223; G. 338; I. 44; M. 55§; O. 663; S. 777; T. 888; and the letters will be found so numbered in the table, the fractions being of course suppressed. In table b, of 300 numbers, the key letters are numbered A. 100; Br. (11 × 3+100) 133; D. 166; G. 200; I. 233; M. 266; O. 300; S. 333; T. 366. In table these numbers are increased by 300 each, and in table d by 600 numbers each.

C. Title-Numbers.

Where a class contains several works by the same author, it is necessary to add a character of some kind to distinguish one individual work from another. I have chosen, as the simplest and most expressive, the 9 Arabic figures, and the 26 letters of the alphabet. As an author number may also be used alone, we have (9+26+1) 36 possible variations for each alphabetical combination. The separate works of an author can be arranged either alphabetically, or in the order of their acquisition. An alphabetical arrangement is more convenient and useful in Fiction, Biography, and Literature, but may be applied to other classes, if thought desirable, although it is of very little account outside of the classes just named. To secure an alphabetical arrangement the titlenumbers should be used as follows:

In the case of popular authors of fiction, biographies of celebrated characters, such as Napoleon, Washington, etc., and in voluminous and noted authors in other classes, the whole 26 letters can be utilized, the 9 Arabic figures being reserved for less important authors, the initial of the title (where letters are selected as title-numbers), and the figures corresponding to the key-letters (where the 9 figures are selected), being used to maintain the sub-alphabetical arrangement of the individual books. Where the exact initial or figure is already occupied the next vacant one can be taken.

The foregoing method of applying the titlenumbers, if carried out in all the classes, would

give a uniform average of two authors for each number. But 26 title-numbers is much too large for one author, except in Fiction and the other cases enumerated, and it will be found that 9 title-numbers are quite sufficient for most authors. Our 36 title-numbers will, therefore, generally enable us to number, alphabetically, four authors with each combination. By writing the figures and letters under the 9 key letters, we shall have four series, beginning with figure 1, and letters a, j, and r. If more than four authors in the same combination are to be provided with numbers, any of the unoccupied title-figures or numbers can be used for the purpose.

As our 36 title-numbers are not intended to be used exclusively for one author, but may accommodate two, four, or more, and as 36 is about the maximum number of works for a shelf, it would be better, perhaps, to call each of our alphabetical combinations an ideal shelf. When all the title-numbers of a combination are used the ideal will correspond with the real shelf. Bearing this in mind, there is no difficulty in providing for certain exceptional cases that may arise. There may be, for example, authors who will need more than the 9 or 26 numbers reserved for them in our plan. The remedy is simply to continue the same series of title-numbers in the nearest vacant combination, either preceding or succeeding. As such cases can only arise when the proper ideal shelf is filled, the new series will, as a matter of course, be continued on the next shelf, and all the separate works with the same initial or number will be immediately underneath those in the first series. It is, therefore, easy to provide for the most voluminous authors, and keep up the sub-alphabetical arrangement of their works. Taking, for example, an extreme case where an author has written 75 works in the same class, three shelves will accommodate them all, and any separate work can be found with ease, as all the a's and b's, etc., will be found together, only instead of being side by side they will be placed immediately underneath each other.

If the 36 title-numbers and 300 ideal-shelfnumbers should prove insufficient to provide for all the works in a class, the numbering capacity

can be increased to any desirable extent by subdividing the class by means of the 26 letters, added to the department letter. I do not think such division will be found necessary or even desirable, as the number of special heads is large enough for a library of 2,000,000 volumes, and the number of libraries that exceed this is limited. I merely indicate the possibilities of the system to provide for any contingency that may arise. As our 887 classes can be increased to 23,062, and each of these can accommodate 10,800 works, it is difficult to imagine a case where the system would not be able to provide every work with a special and distinct number, and still use no more than seven characters in any case.

In class N, Novels, where the number of individual works of most authors is very large, and where a satisfactory division into classes is hardly feasible, the system has been modified as follows: the series b, c, d, of author-numbers are extended to 3,000 each by allowing ten variations for each combination. The first series of 3,000 is devoted to English, the second to French, Spanish, and Italian, and the third to German and Teutonic fiction. The series of 100 combinations headed "a" is similarly extended to 1,000 numbers, and is divided into four sub-classes, as shown in the tables. In each of the series of 3,000 numbers, each number is uniformly divided among two authors, the title-numbers 1 to 9 being used for the less voluminous authors, and the 26 title-letters being reserved for the more popular writers. Each of our three grand divisions of Novels will, therefore, provide for 6,000 separate authors, and if these should be insufficient they can be increased to 156,000 by adding one letter to the class letter.

Our system of notation consists, then, of six characters only for the largest classes, viz., a department letter, four figures, and a titlenumber of one character. The ordinary shelf systems, using Arabic figures only, have just as many characters, and it would not be possible to number a library of 100,000 works with less, if restricted to Arabic figures; but our system is capable of marking 8,000,000 works without using more.

As I use 35 characters in my title-numbers,