College and Research Libraries Th~ Future of Generalized Systems of Classification IT IS FASHIONABLE at present to dismiss the future development of generalized classification as if there could be no such thing-as if no synthesis were ever pos- , sible again because no one can see any unifying factor or factors in the prolifer- ation of subjects with which we have to deal. This is equivalent to saying that the sky is less blue because the blind man does not see it. There will be new gen- eralized classification systems in the fu- ture for the simple reason that we have to have them. And anything we have to have sooner or later is found. George Gaylord Simpson, the verte- brate paleontologist, points out that clas- sification in zoology has proceeded alter- nately by stages of analysis and synthesis, carried out by men who might be cate- gorized as "lumpers" and "splitters."1 In library classification, by analogy, the older systematizers, Dewey, Bliss and oth- ers, would be "lumpers." Those who make faceted classifications and special subject analysis systems of all kinds are "splitters." When the "splitters" have finished analyzing the new alignment of parts that make up the sum total of human knowledge-an analysis which must be performed if we are to make any use at all of the mechanical and elec- tronic aids now available to us-there is no reason to believe that a new synthesis cannot or will not be made. It is his- torically true that the impasse of one age is solved in the next, or the one after the next. The only fault, a very human one present in every generation, is that impatient men grow hopeless or even • 1 George Gaylord Simpson, "The Principles of Classification and a Classification of Mammals," Bulletin of the American Museum of Natt,ral tHis - tory, LXXXV (October 1945), 22-24 . . SEPTEMBER 1963 Bv PHYLLIS A. RICHMOND Mrs. Richmond is Supervisor of River Campus Science Libraries in the University of Rochester. antagonistic because they cannot see the solution in their own lifetimes. In classification, as in any other disci- pline, one must lay groundwork for fu- ture development. Even if it is possible to be clairvoyant to the extent of pre- dicting correctly what form future de- velopment must take, it is still necessary to take steps to get there. An Einstein, for instance, will hit upon "some great unifying idea from which one can de- duce consequences that can ultimately be bought into agreement ... with ob- served and measurable phenomena." 2 But even an Einstein does not build on nothing. He takes account of all the abor- tive, inconclusive, tentative, incomplete, or even rejected solutions of his prede- cessors. Thus fortified with knowledge of the major blind alleys, at least, he can lay out his course. It is the duty of those now working in classification to discover and explore all possible approaches so that valuable negative as well as pqsitive evidence will be available for the ulti- mate synthesizer. This is not to say that all current work in classification will be fruitless or that things will necessarily proceed so slowly that a synthesis is impossible before the end of the century. It is to say that avenues not yet considered must be ex- plored and that work already begun must be carried much further. Ideas, no matter how bizarre, deserve .considerft- 2 Morris R. Cohen, American Tho"ttght, a Critical Sketch (Glencoe, Illinois : Free Press, 1954), P· : 81. 395 tion, because, with any research, we nev- er know exactly where we are heading. Nor do we know which line of develop- ment will bear the ultimate prize. Essentially classification is the process of taking identifiable entities and relat- ing them in one way or another to each other. These items inay be anything the mind distinguishes as an entity in itself. Since the mind is rather free in coalesc- ing raw material of all types into entities, this means that anything "thinkable" is classifiable. Past syntheses almost exclusively have been constructed in a dendritic pattern -an inverted tree-from general to spe- cific. The tree has been delineated as a plane figure and treated as if division were the only method of construction. Many of the writers who look with hor- ror upon classification as a tool for or- ganizirtg knowledge see only this type of deductive system and identify all classifi- cation with it. A different approach is to build new generalized classification systems from collections of particulars, gathered by observation, ordered by reasoned crite- ria, and formed by induction into gener- alities as an ascending scale. These gen- eralities can then be tested through hy- pothesis and deduction, and finally con- clusions derived from the process, which, when tested and verified by further ob- servation, will provide a solid basis for a more accurate representation of the world in which we live. This is essential- ly the scientific method, defined in very simple terms. Note that the main classes which will result from such a process are not predicted or predictable in advance. Such a system, however, does not pre- sent a perfect picture of knowledge be- cause it, too, usually ends in a dendritic pattern, although sometimes the net re- sults are expressed in chains or even in chains made up from intermediary ma- trix or lattice analysis patterns. Some- thing vital has been left out. This is a study of the very nature of the material to be described. In the final analysis, it is this nature which will determine the system of classification. Any system which ignores actual makeup of indi- vidual subjects, with all their ramifica- tions, is doomed from the start. The universe we live in is apparently open and genuinely infinite, both infi- nitely big and infinitely small. Data, laws, methods, theories in all fields are partially and imperfectly known. On one hand, the possibility of discovery seems unending. On the other hand, the use of creative imagination appears limitless.' While the idea of progress as a funda- mental pervading force has been some- what shaken as an accepted philosophic view in the last two decades, there is no corresponding diminution in either cre- ativity or discovery, both of which are highly individualistic matters and prob- ably not related to progress as an ideal. It is the job of classification to show the waxing and waning of ideals as well as ideas, since the spirit of the times, its Zeitgeist~ adds dimension to any aspect of the sum total of human knowledge. For reasons not entirely clear, this Zeit- geist has usually been ignored, especially. in classification involving the sciences, medicine and technology, although it exists in these areas just as much as in the humanities. The totality of the intel- lectual atmosphere in each era deter- mines to a large extent what shall be accomplished in that era. This totality, which is a dimension external to the in- dividual, can be both narrow, represent- ing the sum total of knowledge in a given field 1 of research, or it can be broad, re- flecting the generally-held attitudes, mo- tivation, philosophical outlook, mores, and such of a period of time. There is also an internal factor which adds dimension to classification. This is the individual's own world...outlook, his Weltanschauung~ which affects his per- sonal motivation, intuitive capacity, and 396 COLLEGE AND RESEARCH LIBRARIES imagination. This results from learning, influence of other people, inspiration, and experience. It is a process of accre- tion, that is to say, lifetime learning added up, and of inner maturation. The long process of growing up as a part of being alive is most noticeable in research in the humanities, where outstanding work is almost never done by young men. The addition of further dimensions to the material to be classified greatly broadens the base of classification. The necessity for describing a body of knowl- edge with no known boundaries, which is the real work of classification, calls for a type of description which also can have no boundaries. In a discussion of the future of physics, Albert Einstein wrote: The belief in an external world inde- pendent of the perceiving subject is the ba- sis of all natural science. Since, however, sense perception only gives information of this external world or of "physical reality" indirectly, we can only grasp the latter by speculative means. It follows from this that our notions of physical reality can never be final. We must always be ready to change these notions--that is to say, the axiomatic substructure of physics--in order to do jus- tice to perceived facts in the most logically perfect way. Actually a glance at the devel- opment of physics shows that it has under- gone far-reaching changes in the course of time.3 This is true of all subjects, and it is this truth that has made all ~lassifications ob- solete over a period of time. So far as generalized classification is concerned, we have not even discovered most of the main categories, to say nothing of the thousands of related categories which must exist. At best, our classifications are an approximation, and subject to con- stant revision as new information be- comes available. Does this mean that all classification is a waste of time, since it can never be finished? The answer is "yes" only if by 8 Albert Einstein, The World As I See It (New York: Covici, Friede, 1934), p. 60. SEPTEMBER 1963 classification we mean a process of as- signing a fixed and -relatively unalterable . "address" to each categorical item. An ideal classification should be entirely flexible. Until all research and scholarly efforts cease, there will be no way of avoiding major alterations in subject organization, and, in fact, there should be no question of having anything but constant change. In making systems, pro- vision for this change should be all the way through as required, not just at marginal spots where it will be the least upsetting. As the Indian classification theorist, Ranganathan, has ably put it: A classificatory language must be nimble enough to keep step with the field of knowl- edge. Its expectation of life is determined by the · degree of its self-perpetuating quality, i.e., by the smallness of the dependence of the classifier on the classificationist [maker of the classification] to seize the correct class numbers of the new formations. The degree of its self-perpetuating quality is determined by the notational devices with which it is armed.4 In other words, the classification must be flexible enough t<:> cope with unlimited additions to itself, and it must not be defeated by the notation used with it. The notation used in a classification is an extremely important feature. Most notations are inadequate and actually strangle the .classification system in the process of trying to reflect its internal structure. Something that will break this stranglehold and at the same time will broaden the class structure by freeing it from dependence upon limited criteria is a must for the future. It may even prove advisable to avoid notation in a classification or to use a random number system, with no effort to reflect the struc- ture. Classification is and must be a mirror of the intellectual as well as the external world it seeks to represent. If this com- 4 S. R. Ranganathau, "Self-Perpetuating Scheme of Classification," Journal of Documentation, IV (March 1949), 240. 397 bined world is a reality of four or five or more dimensions, the classification must have four or more dimensions, with some kind of spatial representation for visual- ization purposes. Such a spatial representation is a sine qua non of adequate symbolic descrip- tion of anything approaching reality. To use a letter-number string can be un- wieldy as well as distorting. What this spatial representation should be-a frac- tion, lattice, or matrix in depth, or some- thing better-cannot now be deter- mined. The forms suggested seem too rigid. One needs a spatial "notation" that is more amorphous. Such a spatial representation is needed for another reason. The human mind can cross subject boundary lines without much difficulty. Ideas in one field are applied or adapted to another with com- parative ease. It is commonplace for new area studies to be set up in academic cir- cles by drawing upon specialists from several fields to bring their different backgrounds to bear on a set of problems common to all of them. Thus we have new programs in space research, brain research, biomedical engineering, non- Western civilization and so on, all inter- d~J:>artmental affairs. Old geographic di- VISIOns are replaced by new ones: Sub- Saharan Africa, Southwest Asia, Middle Amer~ca. ~he late Robert C. Binkley, in teaching history, broke down ·the vertical strea~ of t~me applied to political-geo- graphic regwns (England since 1485) in- to horizontal units (all of Europe in ~ecembe~ 1587). The next step logically Is to realize that there is no reason for not treating space and time as a unit because that is exactly what they are. Space-time is a dimension common to everything. Even the apparent "here and now" of contemporary science has retro- spective features. For example, scientists never cease to remind us that, so far as they are concerned, the printed an- nouncement of new research is already out-of-date on the day it appears. This means that even as the announcement is :ead, it is describing something already In retrospect. This has significance-and not just in priority struggles. Even if a new generalized classification system is ~o mo:e than an ephemeral conglomera- tiOn, It has to take into account the space-time factor. Other dimensions are the Zeitgeist and Weltanschauung mentioned earlier. Just as no scholar or creative artist func- ti?ns ~n a ~acuum, so no output, be it a soenti~c discovery or a piece of poetry or a musical composition, comes from a void. 5 All of them have antecedents, and sooner or later most also have descend- ents. There are branches of scholarship ~hich specialize in locating and study- Ing connections and relationships among things which apparently have sprung up d~ novo . . A .classification which only descnbes a limited number of qualities pertaining to entities and which de- liberately stultifies itself because it must fit into a preconceived code of very lim- ited proportions has small chance of lasting. First a classification must be free to cross subject boundary lines as freely as our minds cross them. Then the notation, if there is one, must be so free ~hat . th~ classification can expand into Infinity In any direction. . Another point to be considered in the future development of generalized classi- fication, in addition to taking a clear look at the nature of the material to be classified and making a flexible system while freeing this system from depend- ence upon notation, is the need to be able to show relationships between ele- ments of a classification system in a com- pletely unhampered way. This is ex~ tremely difficult. Up to now, few classi- fication schemes, no matter how well designed or modified, have been able to show area relationships or cross-connec- 5 John Living.ston Lowes, The Road to Xanadu (New York: Vmtage Books, 1959) is an interesting example . · 398 C 0 L L E G E A N D R ESE A R C H L I B ·R A R IE S tions between subjects which pertain to two or more fields simultaneously. The only solutions attempted have been to enter such things in each related spot as a recognition of each separate emphasis, or to lock the codes for the individual related topics together in a linear chain or a series of more or less connected chains. In this respect, a system which can build its own class descriptions, as with Uniterms, has advantages. With present trends in research tending more and more to break down existing barriers between subjects, and the content of in- dividual fields being realigned in con- sequence of this, the barriers between subjects, as classically outlined, are a positive menace. If growth continues to be nonlinear, in the mathematical sense,6 and expansion continues in revised pat- terns, added freedom to cross boundaries in all directions is a vital necessity. A further point is related to the pre- ceding ones. In a classification, it is not only the basic categorical concept that is significant, but also the modification thereof. The multidimensional quality of knowledge is no less complex than the multidimensional descriptive possibili- ties for each individual item in a classi- fication schedule. Even the descriptive terms themselves can be permuted to widen the scope of the system. A class is by nature an exclusive generalization. It is also inclusive, and in this respect covers considerable variety, according to its criteria. There are as many potential classification entities as there are terms applicable to the description of an ele- ment. A classification element, in turn, is only a temporary convenience, since it is potentially a class in itself, in the light of later knowledge. It may also be elim- inated in the same light. A classification system, on the other hand, may never be fully described in the terse terms of a 6 Ladis D. K ova ch. " Life Can Be SO Nonlinear" Ameri can Scientist, XLVIII (June 1960), 2i8-25. ' SEPTEMBER 1963 classification schedule. Actually, the tra- ditional classification systems have been made by arbitrarily selecting some de- scriptive factors and deliberately ignor- ing the rest. Indexes to the systems pick up some of the alternatives, but only in a hit-or-miss fashion. The faceted classi- fications, by their very nature, attempt to allow for a multitude of possible de- scriptive factors. In sorting out the multidescriptive .possibilities for classification elements, the criteria of the classification system are important. In the past, these criteria have been unwritten as a rule; in fact, it appears that in most cases they have been intuitively assigned. Sometimes the intuition has been quite realistic, judg- ing by the state of knowledge at the time the systems were made. The science sec- tions in the Dewey Classification, for instance, with the emphasis on paleon- tology, fit American science in the later 19th century like a shoe. Unfortunately, the creature wearing the shoe turned out to be a millipede-a situation with which Dewey's successors were never able to cope. Ranganathan has departed from intuitive practices to the extent of trying to spell out the criteria for each class. Others are carrying the process farther, but as yet there is no general agreement on what these criteria should be or even on how they should be reached. 7 The criteria of classification might serve as mental diffraction gratings. Something is needed to split concepts systematically the way such gratings work with rays of light or other rays. It would seem that current research on de- scriptor language8 is a fine beginning, but only a beginning. The argumenta- tive leap which produces a generaliza- tion from a collection of particulars 7 F or a. summary of the work of the English Classi - fica tion Research Group , see D. ] . Foskett, "The Classification Research Group, 1952-1962," L ibri, XII (1962), 127-38. 8 B . C. Vickery, On Retrieval System Theory (London : Butterworths, 1961), pp. 23-55. 399 must be made and its nature noted. No attempt to do this has come at all, per- haps because of lack of awareness among classification makers that such a leap exists. Those who work with machines may be more alert to this possibility. The advent of the giant calculating machine has inspired documentalists to see if a use cannot be made of it to or- ganize the tremendous body of knowl- edge 'which threatens to overwhelm us with the products of our own brains. Success both in the reduction of mathe- matical calculations and in handling huge masses of factual data in business has led to considerable experimentation with such machines for. literature search- ing and information retrieval. Three features, among others, stand out as a net result of this endeavor. First, those concerned with the or- ganization of knowledge, in attempt to make use of the machines, have tended to pay more attention to the capability of the machine than to the needs of the person using the know ledge or of the internal makeup of the knowledge itself. Early meetings between "hard- ware" men and librarians were unpro- ductive because · the librarians asked "What can you do?" and the engineers said "What do you want done?" Neither side understood the answers given by the other. Machines with positive talent in searching certain kinds of data were de- signed, but their effort was negated by the erroneous assumption that input was a simple matter, suitable for untrained clerks. After a succession of failures, the realization dawned that input is the ma- jor key to successful output. Here again, there was no rapport between the "hard- ware" men and librarians, who could have predicted this result. A productive combination of vast experience with clas- sification and subject analysis on one side and brilliant inventiveness of the other has not been realized because of lack of communication between the two groups. This gap has existed because there was no common frame of refer- ence between the two until the "hard- ware" men tried out their systems and in the process made the same mistakes that the librarians had made fifty years ago. It is to the lasting shame of the catalog librarians that so few of them have made any attempt to mechanize the experienced approach which they alone possess. The secoQd feature is one that gives a false sense of capability to the ma- chine. On the surface, finding a piece of factual data in a mass of other factual data appears not unlike finding a box of rotary switches in a warehouse full of electronic supplies. The joker here is that the warehouse is a selective situa- tion, limited to a homogeneous collec- tion of highly specialized items, and serving a small segment of the total buy- ing public. If a collection of information has the equivalent qualities, homoge- neity and limited access, then the ma- chine solution to the problem of infor- mation retrieval looks pretty good. How- ever, under such restricted circumstanc- es, practically anything, including a hier- archical classification of the most rigid type, works well. The difficulty comes when the collection becomes hetero- geneous. Even given the same special situation, if the information to be retrieved is theoretical instead of factual-an idea, a line of thought, an argument for or against a point, a suggestion for future research-the machine falls even flatter on its transistorized face than the pld classification. The classification at least has some way of indicating relationships and can give a hint to where to look for an answer. The machine is an aU-or- nothing proposition. The third factor is a temptation which came with the machine. The very suc- cess with mathematics, the relatively easy manner in which a computer can be C 0 L L E .G E A N D R ESE A R C H LIB R A R IE S programmed to deal with a formula, 'the facility with which symbolic logic can be handled, all suggested that the thing . to do was to find a mathematical model, or a series of mathematical models, to express the essence of organization of knowledge. Fit the data to the model and the model to the machine and the problem of information retrieval was as good as solved. Unfortunately this was equivalent to putting the cart before the horse, as even the most elementary grasp of the scientific method should have warned. Quantification-the for- mula for the law, the shorthand for an accepted generalization-is usually the last step, not the first, in the whole process of organization. Even when one begins with a great unifying id'ea, sub- ject to verification by later observation, quantification comes slowly. For infor- mation retrieval, the mathematical mod- el, to be satisfactory, has to . wait until answers are found to questions that have not yet been asked. There are many questions to be asked and answered before new generalized classification systems of any stature can be expected. A few of the basic ones have been raised in this study: How do we represent a multidimensional reality, presuming such exists, on paper? How do we cross departmental lines in sub- ject fields with a classification system the way the human mind crosses them? How do we show relationships between parts of a system in a completely unhampered way? How do we indicate the multidi- mensional descriptive possibilities for ~ach individual item being classified? flow do we orient our criteria for classi- ficati?n in sue~ a way that concepts may be ~lffract.ed Into elements in a syste- matic fashwn? Should we dispense with notation altogether? If not, can we use something more graphic than a nota- tion made up of letters and numbers? I~ it Rossible to get better spatial percep- u~n Into our classification systems? It SEPTEMBER 1963 would . be convenient to replace the plane and linear by something with ~epth, perhaps analogous to proper mo- tiOn and radial velocity in astronomy. Can .we a~oid the pitfall of adapting our classificatiOn systems to the capability of the computer instead of vice versa? Can we make a machine that will han- dle the infinite variety that composes the sum total of human knowledge? Will our final ~lassification synthesis be capa- ble of being represented by mathemati- cal models? Th~ lines of development to be pur- sued In the future make it imperative that we also ask questions pertaining to methodology: Where can classification borr~w ideas, techniques, and philo- sophic approaches from other disci- plines? Do we have to stoop to some degree of fantasy in our classification composition before we can conquer with cold, hard facts? In other words, how much creative imagination can we put into a classification system without los- ing touch with reality? Is it possible that w_e have ~ade .a mistake in tending to view classificatiOn as a science rather than a art? Is some combination of the two, perhaps achieving the rapproche- ment between science and the humani- ties that has eluded us, the ultimate answer? The future of generalized classifica- ~ion d~pends in large part upon man's Ingenuity. So far, there has been no limit to the capabilities of the human mind, and there seems, therefore, to be no justification for the . view that classifi- ~ation as a way or organizing knowledge Is dead merely because the philosophic approaches used · so far have led to blind alleys. It is time to look for new ap- proaches. •• AcKNOWLEDGEMENT The author is grateful to Mrs. Pauline Ather- ton, American Institute of Physics, and Gertrude Oellrich, Rutgers University, for reading and commenting upon this article in manuscript. 401