Knowl. Org. 30(2003)No.2 87 B. Hjørland: Fundamentals of Knowledge Organization

Fundamentals of Knowledge Organization1

Birger Hjørland

Royal School of Library and Information Science, Copenhagen Email: [email protected]

Birger Hjørland is Professor in information science (IS) at the Royal School of Library and Informa- tion Science in Copenhagen, Denmark. He was formerly professor in IS at the College University of Borås, Sweden. He has a Ph.D. in information science from the University of Goethenburg, Sweden and a Master of Arts in psychology from the University of Copenhagen, Denmark. For twelve years, Professor Hjørland worked as a research librarian at the Royal Library in Copenhagen, where he was a subject specialist in psychology and the coordinator of the library's computer based reference serv- ices. He has taught the history and philosophy of psychology at the University of Copenhagen and also taught information retrieval in the Institute of Applied and Mathematical Linguistics in Copen- hagen. In spite of Professor Hjørland's experience in the field of psychology, his work is much influ- enced and oriented by philosophical and sociological perspectives. He is a member of ASIS&T and ISKO and is the author of Information Seeking and Subject Representation: An Activity-Theoretical Approach to Information Science.

Hjørland, Birger. (2003). Fundamentals of Knowledge Organization. Knowledge Organization, 30(2). 87-111. 74 refs.

ABSTRACT: This article is organized in 10 sections: (1) Knowledge Organization (KO) is a wide interdisciplinary field, much broader than Library and Information Science (LIS). (2) Inside LIS there have been many different approaches and traditions of KO with little mutual influence. These traditions have to a large extent been defined by new technology, for which reason the theoretical integration and underpinning has not been well considered. The most important technology-driven traditions are: a) Manual indexing and classification in libraries and reference works, b) Documentation and scientific communication, c) Informa- tion storage and retrieval by computers, d) Citation based KO and e) Full text, hypertext and Internet based approaches. These traditions taken together define very much the special LIS focus on KO. For KO as a field of research it is important to establish a fruitful theoretical frame of reference for this overall field. This paper provides some suggestions. (3) One important theoretical distinction to consider is the one between social and intellectual forms of KO. Social forms of KO are related to professional training, disciplines and social groups while intellectual organization is related to concepts and theories in the fields to be organ- ized. (4) The social perspective includes in addition the systems of genres and documents as well as the social system of knowl- edge producers, knowledge intermediaries and knowledge users. (5) This social system of documents, genres and agents makes available a very complicated structure of potential subject access points (SAPs), which may be used in information retrieval (IR). The basic aim of research in KO is to develop knowledge on how to optimise this system of SAPs and its utilization in IR. (6) SAPs may be seen as signs, and their production and use may be understood from a social semiotic point of view. (7) The con- cept of paradigms is also helpful because different groups and interests tend to be organized according to a paradigm and to de- velop different criteria of relevance, and thus different criteria of likeliness in KO. (8) The basic unit in KO is the semantic rela- tion between two concepts, and such relations are embedded in theories. (9) In classification like things are grouped together, but what is considered similar is not a trivial question. (10) The paper concludes with the considering of methods for KO. Basically the methods of any field are connected with epistemological theories. This is also the case with KO. The existing methods as de- scribed in the literature of KO fit into a classification of basic epistemological views. The debate about the methods of KO at the deepest level therefore implies an epistemological discussion.

1. What is Knowledge Organization (KO)? organization of information in bibliographical rec- ords, including citation indexes, full text records and In the Library and Information Science community (LIS) the Internet. Information Science (IS) is basically Knowledge Organization (KO) means especially the about the best way to construct such bibliographical 88 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization records (which is done in KO) as well as the optimal – Semantic relations way to utilize given records (which is done in infor- – Subjects mation retrieval IR). – Subject access points (“SAPs”), and so forth. KO is, however, a much broader concept. Knowl- edge is organized in, among other things: I also find that KO lacks serious explorations into the methods of KO and the methodological basis of KO. – The social division of labour (e.g., in disciplines) Developments in practice seem much more influ- – Social institutions (e.g., in universities) enced by progress in IT than by progress in KO – Languages and symbolic systems proper. Sometimes people seem even to expect that – Conceptual systems and theories IT in itself can replace conceptual and theoretical – Literatures and genres concerns. In my opinion we may have the best over- view of progress in KO if we consider five technol- Library and Information Science or just Information ogy-driven stages in the development of KO. If this is Science, (L)IS, has often ignored this broader meaning true, it is an expression of a crisis in the field: that the of KO and has thus failed to be based on such field is not driven by developments in its own re- broader theories; or any theories at all. My central search. It is lacking accumulated knowledge, which view is that such broader views cannot be ignored. can be transformed from one IT-platform to another. Any attempt to develop fruitful principles for KO in LIS must be based on broader theories of KO. 2. Technology Driven Phases in the Development The principal actors in IS are the knowledge pro- of Knowledge Organization (KO) ducers (e.g., authors), the users and the intermediar- ies. It is their interaction with bibliographical records In library and information science (LIS) the concept that is the focus of (L)IS. Each of those actors brings of KO is connected to the development of classifica- with him certain pre-understandings, views, concepts tion and indexing systems in libraries, bibliographies, and languages mainly acquired during socialization in and electronic databases. Hjørland, (2000; Hjørland & society. The success of the interaction depends on Kyllesbech Nielsen, 2001) described five technology these pre-understandings, concepts and languages. driven stages in the development of KO. Each stage is This applies not only to the “match” of concepts but not replaced by subsequent stages, but continues to especially to their ability to support users’ tasks. A influence the field. Because of this fact the technology theory of such broader forms of KO is therefore es- beyond those stages may also be seen as one among sential in order to construct efficient systems for KO other causes for the development of more or less sup- in LIS. plementary or competing approaches in KO. LIS has claimed status as a field of study from about 1876 when Melvin Dewey first published his The stages are: classification system and when schools of librarian- ship began to emerge. What are the accumulated re- a. Manual indexing and classification in libraries and reference sults of more than one century of research in KO? works Among the results are a number of standards and guidelines, some theoretical developments, for exam- Cataloguing in libraries and organizational principles ple, Cutter’s rules (1904), advances in facet analysis and in reference works go far back in time. One can trace of course important changes in information technol- the beginnings of the use of alphabetical order to an- ogy (IT). It is, however, difficult to sketch the more tiquity, although it was not generally adopted until theoretical and scientific progress in this field because well into the Middle Ages, and it was not the only it seems largely atheoretical and fragmented and be- method of classification possible (Daly, 1967). How- cause many different lines of thought seem to co- ever, in the Middle Ages and at the beginning of exist. In my opinion KO lacks theories about its most modern time catalogues were probably seen as inven- fundamental concepts such as: tories designed for controlling the collection and for information needed for new acquisitions, rather than – Concepts as tools for subject searches. – Criteria for class inclusion A professionalisation of the classification and in- – Meaning dexing of books in libraries appeared around 1876 –Indexing with the publication of the Dewey Decimal Classification, Knowl. Org. 30(2003)No.2 89 B. Hjørland: Fundamentals of Knowledge Organization and with the foundations of library schools (and thus played a central role in the establishing of interna- “library science”), and so forth. Charles A. Cutter tional abstracting journals3. They were less interested (1837-1903), Melvin Dewey (1851-1931), Henry E. in keeping books for their own sake or for broad cul- Bliss (1870-1955) and S. R. Ranganathan (1892-1972), tural purposes, and were highly interested in estab- among others, have been recognized as founding fa- lishing services which could stimulate the application thers. of knowledge to specific purposes. The foundation of It is generally recognized that this stage produced user studies (Bernal, 1948) and bibliometrics (e.g., principles for KO that are still valid and important. Bradford, 1948) is also part of this stage/tradition, Among them are “Cutter’s rules” and Ranganathan’s which may first and foremost be characterized by a approach to the classification of subject fields. Other more specific subject approach, a deeper level of in- principles have of course become obsolete, and some dexing, more emphasis on modern technology and a are controversial. Frohmann (1994), for example, more scientific attitude towards goals and problems. considered Dewey’s approach to be very harmful be- This stage laid the foundation for the later develop- cause it ignored the semiotic nature of classification ment of online bibliographic databases. and just proposed an empty formalism: “Dewey's subjects were elements of a semiological system of c. Information Storage and Retrieval by Computers (mainly 1950-) standardized, technobureaucratic administrative software for the library in its corporate, rather than Computer science4 has its own development, which high culture, incarnation” (pp. 112-113). It must be should not be confused with the development of li- added, however, that there is very little overall dis- brary, documentation and information science. As a cussion of the importance of these (and other) princi- theoretical discipline computer science was founded ples. They have been considered standards or canons in the 1930s, well before the appearance of the mod- rather than research-based principles subject to inve- ern computer. The founding fathers were the logi- stigation, modification and revision. cians Alonso Church (1903-1995), Kurt Gödel (1906- 1978), Stephen C. Kleene (1909-1994), Emil Post b. Documentation and Scientific Communication (1897-1954), and Alan Turing (1912-1954). This early work has had a profound influence on both the theo- The documentation movement developed from about 1895. retical and the applied development of computer sci- The founders were Paul Otlet (1868-1944) and Henri ence, but these persons are not directly connected to Lafontaine (1854-1943), who established The Interna- the history of LIS or KO, in my view. Computer sci- tional Institute of Bibliography2 and in 1905-1907 published ence has different educational programs, scientific the first edition of the Universal Decimal Classification, journals and conferences compared to LIS and it also UDC, which was an extension of the Dewey Decimal has somewhat different goals compared to LIS. Arbib, Classification designed for deeper subject analysis. Kfoury, and Moll (1981) write that “Computer science seeks The American Documentation Institute was founded in 1937 to provide a scientific basis for the study of information processing, the and by 1968, had changed its name to the American So- solution of problems by algorithms, and the design and programming ciety for Information Science (ASIS). It is difficult to de- of computers.” In my opinion the goal of LIS is to opti- scribe the exact differences between librarianship and mize the utilization of documented knowledge. The documentation. Documentalists were less interested primary purpose of libraries is to provide physical and in libraries and their collections; they were more in- intellectual access to information. The intellectual ac- terested in bibliographical control, in scientific com- cess is provided by the organization of either the munication and scientific documentation, in informa- physical documents themselves or by organizing tion services to industry and in the utilization of the document representations in catalogues, bibliographies knowledge in the documents. Implicitly they shared a and databases. LIS is not primarily focused on con- view of KO as based on subject knowledge and sub- structing algorithms, but on informing people about ject literature. Documentalists often regarded them- documents. LIS is supposed to be more open to dif- selves as more service-minded, more technology- ferent views, more reflective and meta-oriented and oriented and more advanced than librarians. Where demonstrate gaps and uncertainties in knowledge to traditional librarians often had an orientation to- users. This is very different from making, for exam- wards the humanities, documentalists were mostly af- ple, an expert system that performs optimally by re- filiated with science, technology and business. They flecting some generalized cognitive models. indexed single articles in journals and books and 90 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

With the advent of computers in the 1950s, LIS and ans’/information specialists’ monopoly on subject ac- KO became influenced by this new technology, and cess, and established a direct competition between many people felt that the future of LIS must be as a SAPs produced by different agencies. part of computer science. There is an indication that An underlying philosophy at this stage has often “information scientists” did not regard themselves as been that the length of the searchable record in itself part of library science or as dealing with documents was the most important parameter in retrieval (Lan- and their representations (cf., Hjørland, 2000a). They caster, 1998, pp.6-8). SAPs were often seen as merely felt much more closely related to computer science. “semantic condensations” of the texts represented This may have confused the theoretical development (implying that the ultimate goal was full text repre- of KO, because theories related to “information” ver- sentation and nothing more). Research was domi- sus theories related to “documents” are related to quite nated by quantitative methodologies while little re- different kinds of outlooks. As a theoretical concept, search concerning qualitative differences (semantics “information” tends to move LIS and KO towards or meanings) between different kinds of SAPs was es- theories about control, feedback, coding and noise in tablished at this stage. The underlying philosophy has transmitting messages, while “document” tends to been empiricist in more than one sense of this word. move LIS towards theories about meaning, language, It was first and foremost empiricist in its attempt to knowledge, epistemology and sociology. Therefore, measure the efficiency of subject retrieval points em- in LIS there may be a whole paradigmatic conflict pirically, for example, by measuring “recall” and hidden in those words. “precision.” It was also empiricist in its avoidance of This third stage in the development of KO as a “metaphysically” based classifications, and in its fa- field has been influenced by experimental approaches vour of “atomist” SAPs such as the Uniterm system in which recall and precision are well-known meas- devised by Mortimer Taube in 1951 and similar sys- ures, by extensive use of statistical models of the tems depending on specific words from the document properties of the document representations, by ap- themselves. proaches that try to automate KO using Artificial In- One associated tendency at this stage or approach telligence (AI) and expert systems, by applying natu- was the attempt to formalize and automate retrieval ral language processing (NLP) techniques and the and to eliminate human interpretation and subject like. An offshoot of this stage is also the cognitive analysis. We must distinguish between the financial view that tries to model the users’ cognitive functions pressure to automate practical systems on the one in the computers. side and the scientific evaluation of the performance The establishing of computer based abstract serv- of various aspects of human based and mechanized re- ices such as Chemical Abstracts and MEDLINE in trieval systems on the other side. It is a legitimate and the 1960s is an important achievement at this stage. highly desirable goal to reduce costs and improve ef- The development of descriptor based and free text re- ficiency in information systems. Basic research trieval (mainly based on titles and abstracts), Boolean should, however, illuminate basic strengths and draw- logic, field specific subject access, measuring recall backs in different approaches, and not be blinded by and precision and other innovations were extremely the pressure to use automated or cheap solutions. Be- important developments in document retrieval. In- cause of such tendencies important approaches related formation retrieval as a research tradition started to interpretation and qualitative aspects have been with the ASTIA and Cranfield experiments in the much neglected to this stage, and the research has not 1950s, and today's TREC full text experiments may accumulated as satisfactory a body of knowledge as is be seen as a continuation of this tradition. This third needed. stage improved information services and research ef- forts in IS in a most important way. In my opinion d. Citation based retrieval and KO (1963-) these technological innovations and the goals they try to solve define the core area of what is termed infor- Eugene Garfield's introduction of the Science Citation mation science. Index in 1963 marks the fourth important stage in the Computer technology made it possible to use development of SAPs. The possibility of retrieving many kinds of Subject Access Points (SAPs), both of documents according to the citations they receive the traditional kinds produced by information spe- represents a real innovation in IR, and this technique cialists and of words from the documents themselves is able to supplement all forms of term-based retrieval (e.g., titles and abstracts). This removed the librari- in very important and qualitative new ways. This in- Knowl. Org. 30(2003)No.2 91 B. Hjørland: Fundamentals of Knowledge Organization novation has also contributed to research in motives study of citation behaviour can illuminate the real for citing other documents, in sociological patterns in problems of IR: That cited documents are not simply citing, in the relative role of terms and references as a set of documents sharing a fixed set of attributes SAPs and in the semantic relations between citing which are not represented in the non-selected items. and cited papers. Overall, this stage has brought the Documents which are similar from the point of view concept of networks of documents into the focus of KO. of retrieval algorithms do not need to be co-cited, In this way, citation based retrieval has changed whereas documents which are not similar are often our understanding not only of subject relatedness, but co-cited. Ordinary retrieval algorithms and citation also of the concept of subject matter and of the fun- practices seem simply to reflect different theories damental aim of IR itself. Since it may be relevant to about subject-relatedness. cite papers, which have no words in common with Since authors may cite other papers in order to the citing papers (or no simple semantic relation such flatter or to impress, the prediction of which refer- as narrower terms, broader terms, synonyms, etc.), ences a given author will select in the end for a given naïve conceptions of subject relatedness or subject paper cannot be used as a valid criterion in IR. Crite- matter can no longer persist. Semantic relations may ria for IR should not be based on social-psychological be implicit or latent. Semantic relations in science are motives, but on epistemological principles for the ad- determined by theoretical advances, which may vancement of public knowledge! In this way, our in- change the verbal description of the research phe- sight from citation indexes has, in a very profound nomena completely. This is the reason why statistical way, changed not only the methods of IR but also the patterns in vocabulary may sometimes turn out to be concept of subject relatedness itself and the basic aim a less efficient measure of subject relatedness than pat- of retrieving information. We can no longer regard terns in citations. the prediction of individual use as the ideal criterion Citation behaviour is extremely important as a for IR, nor can we regard IR as a value free technique. theoretical object because the goal of IR is to provide Instead, we have to face the insight that the goals of the references which are useful in solving a specific IR are deeply rooted in epistemological norms for problem. A scientific article is a documentation of the what should be regarded as good science and good ci- solving of a specific research problem. The problem is tation behaviour. formulated in the article, and this problem has de- Pioneers in the integration of bibliometrical meth- termined what kind of information was needed by ods with the more traditional methods of KO are the author in order to contribute to this problem. Mike M. Kessler (1965), Miranda Lee Pao and Dennis Based on the information need, information has been B.Worthen (1989), Pao (1993) and Lorna K. Rees- sought and selected, and the documents actually used Potter (1987, 1989,1991). I also see my own research are finally cited in the article. Each of the thousands as pioneering in this respect both theoretically and of articles produced weekly is in a way a case study in empirically (e.g., Hjørland, 1992, 1993, 1998a+b, IR. Not only does every article pose a definite IR 2002b). It has subsequently also been used by, among problem, but the list of references provided by the others, research students of Peter Ingwersen (e.g., author is the key to how that particular person has Schneider & Borlund, 2002). solved the problem. Thus it is possible to check theo- ries of IR against how they match the actual docu- e. Full text, Hypertext and Internet (mainly 1990-) ments cited! In the traditional (positivist) philosophy of science, studies should be able to “predict” future Full text retrieval marks the fifth and final step in the behaviour or events. In our case this principle would development of SAPs. Until this point space limits imply that theories and models of IR should be able were a major constraint in the development of subject to predict what citations will turn up in the single access systems because length of record in itself is an papers. Most research on relevance and on IR seems important parameter in retrieval. At this stage, every to have overlooked this fact. From what we do single word or every possible combination of words know, it seems extremely unlikely that an algorithm in full text documents is a potential SAP, as is every should be able to select references from electronic da- thinkable kind of value-added information provided tabases and end up with just the set of references rep- by authors, readers, or intermediaries. Given full-text resented in a given article. From this point of view, representations the first important theoretical prob- theories of IR seem naïve and unrealistic and the goal lem to arise is whether any kind of value-added infor- of prediction seems to be wrong. A more detailed mation is necessary? Can such extra information as 92 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization provided by abstracting and indexing – at least in It is important to realize, however, that the devel- principle – increase recall and/or precision? If this is opment has been very much technology-driven. This not the case, then it seems as if we have reached the is not a satisfactory circumstance for a discipline with end of the line, that no further contributions from re- the ambition of being a science and producing general search or practice in IS are needed any longer. theories and principles. Although those five stages de- The answer to this question is closely linked to fine a topic and a common goal, they do not, how- theoretical views on the concept of subject. In the ever, define a cumulated fund of findings, theories view of Claus Poulsen (1994) a subject is something and principles. On the contrary, there are often latent that is expressed in the literature in a transparent and conflicting views between those stages or traditions. self-evident way. By defining subjects this way it is Those traditions have only to a very limited extent impossible to even ask whether a given text always been in the form of a dialogue. This is why overall represents the optimal representation of itself. By de- concepts for thinking about KO have been missing or fining subjects as informative or epistemological po- at least strongly underdeveloped. In the next sections tentialities Hjørland (1992, 1997) establishes the pos- we shall examine some of those concepts, theories sibility that documents may be implicit or even and principles that seem most important for this task. wrong about their own subject matter and hence that The goal of LIS is to write a history of its theoreti- there still exists a need for information professionals. cal development abstracted from the concrete technolo- To take an extreme example: A document about Jews gies in which its principles have been studied. This is written by a Nazi author should not only be indexed difficult because the general knowledge base is not as being about Jews, as it claims to be. It is important well established. Also, the tendency has been that li- to make the Nazi view visible in the subject analysis brary and information science has passively used the (and, for example, index it as Nazi propaganda about technology without contributing much to its devel- Jews). Subjects are not objectively “given” but are in- opment. If LIS is to be able to contribute valuable fluenced by broader views which are important for knowledge, its focus must be abstracted from con- the information seeker to know and should therefore crete technologies. ideally be part of the subject analysis. Whether or not this is also a practical, economical and realistic solu- 3. Information Processing in the Social Division of tion is another question that must be illuminated by Labour: The Intellectual and the Social Organi- evaluating specific subject access systems. zation of Knowledge. The introduction of full text databases has empha- sized the theoretical problems concerning subject Hjørland (2002d) analysed the official definition of matter and meaning and has also made composition information science given by the American Society for In- studies highly relevant to IS. In this phase, qualitative formation Science and Technology (ASIST), which states: aspects become more important than ever before. It is not just the problem of making efficient algorithms Information science is concerned with the gen- for IR, but also the problem of identifying the under- eration, collection, organization, interpretation, lying values and goals that such algorithms are going storage, retrieval, dissemination, transformation to serve (see Hjørland, 2003b). and use of information, with particular empha- sis on the applications of modern technologies * in these areas (Borko, 1968; Griffith, 1980).

The above description of the five technology-driven This definition does not discuss how professions developmental stages in KO provides, in my opinion, other than information scientists are concerned with a clear picture of what KO means with LIS. In view- “the generation, collection, organization, interpreta- ing those five stages together, we have a rather well- tion, storage, retrieval, dissemination, transformation defined picture of the overall topic with which we are and use of information”. In fact, we know that many dealing. Behind those different traditions is a certain professions could claim to be equally concerned. By general goal that can be expressed as the optimisation neglecting this issue, the definition fails to define in- of KO in libraries, databases, reference works and on formation science since a definition of IS should spec- the Internet. This goal is important to keep in mind ify the special role of information professionals in because it is this goal that gives KO in LIS its impor- studying or handling information. tance for practice. Knowl. Org. 30(2003)No.2 93 B. Hjørland: Fundamentals of Knowledge Organization

Astronomers, for example, can be seen as experts KO involves two kinds of organization: The intel- who identify, process and interpret information from lectual organization of knowledge may also be the universe. Astronomers “read” both nature and termed the cognitive organization of knowledge. This books, but nature is mostly considered the key in- is basically the organization of knowledge in con- formation in the sciences. As products of their activi- cepts, conceptual systems and theories. If we use the ties they may publish their empirical and theoretical periodical system of chemistry or the zoological tax- findings. The library, documentation and informa- onomy of biology as the basis of indexing systems, tion profession, although interested in all kinds of we are using intellectual or “cognitive” systems of documents, has a core interest and expertise con- KO. The social organization of knowledge, on the nected to communication of published documents. other hand, is basically the organizations in profes- Information scientists are not experts in interpreting sions, trades and disciplines. If we refer to disciplines information from the stars, but are, for the most part, in our knowledge representations, we are using “so- experts in information documented by, for example, cial” systems of KO. It should be mentioned that this astronomers (e.g., indexing and retrieving astronomi- social organization principle has been the basic one in cal documents). In this example information has been traditional library classification such as the Dewey defined in a broader sense than usually implied in in- Decimal Classification (DDC): formation science. Just as astronomers can be said to handle informa- A work on water may be classed with many disciplines, tion professionally, all other professions can be said such as metaphysics, religion, economics, commerce, physics, to do so as well. Publishers, researchers, historians, chemistry, geology, oceanography, meteorology, and history. lawyers, teachers and anybody else can be said to be No other feature of the DDC is more basic than this: professionals in handling information in some way or that it scatters subjects by discipline (Dewey, 1979, p. another. The role of information specialists may be xxxi). relatively clear when the target group is, for example, astronomers: Information specialists are experts on There is very little serious debate about this principle forms of publications, databases, reference tools, and in the literature, supporting the view that many li- so forth. In the case of, for example, historians and brarians and information specialists seem to be un- lawyers, borderlines are much less clear because the comfortable with it. Whether or not this principle is key information that these professions seek, interpret approved it should be clear, however, that the social and use is itself contained in publications and other organization of knowledge in disciplines is very im- documents. The historian, not the librarian or infor- portant for KO and for optimising information seek- mation specialist, is the expert in seeking, organizing, ing. When it has been established which discipline a interpreting and utilizing the documents (mostly un- given query belongs to, the hardest part of the re- published) needed in his or her professional work, trieval task may well have been completed. The social but an information professional is more professional organization of knowledge is indeed important to regarding some specific problems such as databases, study, and apart from some research in bibliometrics cataloguing, and so forth. A basic difference in (see Hargens, 2000), this study has hardly begun in knowledge about information sources is typically LIS in spite of the significance expressed in the quota- that the subject specialist starts from a narrow point tion given above. Today, the “empirical” basis for and works bottom-up to more general information classifications such as DDC is mainly the literature sources, while the information specialist starts from a classified by the system. One way to expand the so- broad overview and works top-down to more specific cial dimension could be to study the development of information sources. In this way their competencies occupations in society, for example, as displayed in are supplementary. the Standard Occupational Classification System (U.S. De- The special focus of LIS is on documented knowl- partment of Labor, Bureau of Labor Statistics, 2002). edge produced by human beings in some kind of There are of course different theories or concep- documents of potential use to other human beings. tual frameworks on both the intellectual and the so- Light from the stars is not information for the LIS- cial organization of knowledge (as well as about their community, but astronomical information as pro- interrelatedness). One view sees scientific concepts, duced and used by astronomers is. This distinction theories and fields as reflecting a neutral and objective may seem subtle, but is important in order to con- reality. This might be termed science as a mirror meta- struct a firm theoretical basis for KO. phor, and it is related to rationalism. Scientists may 94 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization organize themselves according to such pre-established should be evaluated according to the job they are fields. Basically, however, the intellectual organiza- supposed to do for us. As such they are parts of theo- tion is not connected to – let alone formed by – the ries that have been developed in order to make us act social division of labour. Another view might see sci- in relation to specific problems. entific concepts, theories and fields as useful tools The organization of knowledge in traditions, ide- constructed in order for human beings to accommo- ologies and paradigms may be seen as the combining date the demands of life. This is the science as a map concepts between the intellectual and the social or- metaphor, which is related to pragmatism.5 Even if a ganization. They are cognitive organizations based on map is a reflection or representation of a reality, a social influences. map is still not a mirror. A map is first and foremost a tool for certain human activities. The kind of activi- 4. Documents and the Sociological Perspective of ties and interests the map is going to serve has a major Knowledge Production, Knowledge Interme- influence on how the map is made. From a pragmatic diating and Knowledge Consumption perspective the intellectual organisation of knowledge is deeply rooted in and connected to the social or- It is important for information science to provide ganization of knowledge. models of actors, institutions and information serv- One influential view today is social constructivism. ices in different discourse communities. The actors This view is related to the pragmatic view presented are knowledge producers, intermediaries, and users of above. Often, however, social constructivism and knowledge. The institutions are research institutions, pragmatism are opposed to kinds of realism such as publishers, libraries, and so forth. Information serv- scientific realism. One of the modern pragmatic phi- ices may be classified in primary services (e.g., pub- losophers is Richard M. Rorty. According to Rorty, lishing houses and journals), secondary services (e.g., scientific realism and pragmatism are two views that bibliographical databases) and tertiary services (e.g., cannot be combined; pragmatism is seen as an anti- professional encyclopaedias and literature reviews). I realist position. According to other philosophers have found the UNISIST (1971) useful. Figure 1 be- (e.g., Dewey, 1929; Ellis, 1990), realism must indeed be low is a modified version of the UNISIST-model based on pragmatism. This corresponds to my own from Trine Fjordback Søndergaard, Jack Andersen, view, which may thus be termed pragmatic realism. Scien- & Birger Hjørland (2003, p. 303), in which the stip- tific concepts and fields tend to represent parts of real- pled ellipse symbolises a knowledge domain. Figure 2 ity in a way which is functional for human activity. To shows the same model (not previously published) in the degree that social constructivism is opposed to real- which the Internet resources are highlighted as inte- ism, I disagree with that view. One should avoid the grated with the printed sources. pitfall of sociological reductionism. However, I have At this point, I want to emphasize that such mod- found much research done under the banner of social els provide one specific perspective on knowledge or- constructivism deeply relevant for the understanding of ganization. By introducing the actors and the divi- the structure of many knowledge fields. For example, sions of labour between different organizations and in order to illuminate the classification of mental ill- services, such models are sociological by nature. They ness it is relevant to do both experimental research help us to understand the different functions of kinds and to do “discourse analysis.” Experimental research of literature, and they enable us to establish a mean- may show that schizophrenia is not a fruitful aeti- ingful typology of scholarly literature, including the ological concept because all major psychoses may well known differentiating differentiation between have a common genetic basis (see Kringlen, 1994). primary, secondary and tertiary literature. “Discourse analysis” on the other hand may uncover the fact that most of our theories and concepts in this field are “social constructions” influenced by specific interests. In particular it may draw our attention to different cultural experiences and avoid tendencies to universalistic explanations. In this way it may open up for classifications that are better suited for a dia- logue between different perspectives. In the end, our scientific concepts and classifications (e.g., “schizo- phrenia” versus “Einheitspsychose”) are tools that Knowl. Org. 30(2003)No.2 95 B. Hjørland: Fundamentals of Knowledge Organization

Figure 1: The revised UNISIST-model modified for the domain analytic approach 96 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Figure 2: The revised UNISIST-model integrating printed and Internet resources and modified according to the domain analytical approach Knowl. Org. 30(2003)No.2 97 B. Hjørland: Fundamentals of Knowledge Organization

5. Subject Access Points in Electronic Retrieval. ling “universalistic,” “reductionist” or “individualist” approaches. The purpose of studies is to explicate and In electronic retrieval, words, phrases and texts are represent the implicit meanings that are lost in mer- taken out of contexts and merged. Words and sym- ging. bols from many different journals are merged; words from different sections and parts of texts are merged, Three major approaches to IR and KO are: and so forth. Words are taken out of their context, a context that contributes to their meaning. By using a) Traditional IR based on term frequencies in the advanced technology it is possible to re-establish parts texts themselves and in the total database. of this lost context (e.g., by means of proximity ope- b) Citation bases’ organization and retrieval based on rators). Any string of characters and any symbol networks of citations between documents. from the texts themselves or from so-called value ad- c) Traditional library classifications based on disci- ded information become a possible subject access plinary divisions and disciplinary criteria. Those point. The art and science of information retrieval disciplinary criteria are by principle external to the and KO is to utilize all of these possibilities in an op- documents themselves and are NOT derivable timal way and to add further value to the representa- from either full text databases or hyper linked tion and retrieval of documents. networks. You cannot, for example, construct a What we must emphasize in IS, is to study the map of Spain from a collection of documents ways in which words and symbols are given meaning about Spain, whether you use co-citation analysis by their specific contexts. Of importance are the way or any other techniques. Maps are produced by in which different disciplines construct their mea- geographers, and their conceptual structures are nings, the way in which document composition pro- subsequently applied in library classification. vides meaning to words and symbols, and the way in which different controlled vocabularies construct All these approaches are often used concurrently and meanings. If a word is employed, for example, in the bibliographical records often contain a mixture of all title, in the abstracts, in the methodology section or of them. We should develop more (explicit) know- in the conclusion, we might (sometimes) be able to ledge about their relative strengths and weaknesses attribute different priorities and meanings to the for different kinds of queries. word, and this may help us to give different probabi- KO involves two kinds of decisions: To make a litistic evaluations of whether documents should be classification system (or a thesaurus, etc.) and to deci- retrieved or not. People, who know certain databases de where to classify a given document in that system. very well, including the subject literature and subject Both decisions are related and involve epistemological language they cover, probably employ this kind of implications. Both decisions are made at the end, ul- knowledge in tacit ways. The job for information timately based on what is considered relevant: What studies is to help to explicate the underlying princi- classes are relevant? What kinds of connections bet- ples. Just as ordinary people can speak a language and ween classes are relevant? What kinds of documents use grammar, linguistic experts have to explicate what are relevant for given users or purposes to put to- competent users do. gether in a given class? In Birger Hjørland and Lykke Kyllesbech Nielsen Different kinds of SAPs have different informatio- (2001) we summarized research on the relative infor- nal values in information retrieval. SAPs are more or mational value of different kinds of subject access less functional in revealing “the subject” of the docu- points (SAPs). It is important to realize that the rela- ments that they represent. It is important to realize, tive value of, for example, terms from a text versus ci- however, that which documents are relevant to a gi- tation searching (based on the bibliographical refe- ven query depends, among other things, on the theo- rences) is not constant. It varies according to discipli- retical perspective in which the query is embedded. A nary norms among other things. This implies that it description of “the subject” is not a neutral, objective is important to uncover such different disciplinary activity, but is influenced by different theoretical norms: In our above mentioned review we tried to views and interests. The consequence of this view is integrate such knowledge from many different sour- that “the subject itself” (the something that the SAPs ces. Our approach implies that the rich flora of do- are supposed to identify) cannot be regarded as being cuments and domains should be investigated by in- objective in the positivist sense. Based on such consi- formation science. This is in contrast to the prevai- derations, Hjørland (1992) defined the subject of a 98 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization document as the document’s epistemological (or in- however, take this basic structuralist claim for gran- formative) potentialities. A document has given poten- ted: that given languages affect our semantic structu- tialities, and the job for indexers (and other SAPs) is res and thus our classification of the world. to identify those potentialities. This is easily under- For human informative activities, the proper per- stood if we compare this with citation indexes. There spective of the meaning of “meaning” is very impor- may be many reasons to cite a given document, but tant. This is a difficult concept for which activity some of these reasons may be rather widely varied. theory provides a fruitful understanding. The produc- For the group of authors citing a given document for tion of books, texts and other documents is a special the same reason this document has the same function development in literate cultures. Documents are tools and the same “subject.” Indexing is a process in the having specific kinds of functional values in those cul- service of helping future authors to identify docu- tures. The view of AT and social semiotic6 theories is ments to be cited. This is done by identifying exactly that meanings, signs and documents are developed to those attributes that make it citable. The connections function in relation to standardized practices in between cited and citing papers are relevance connec- communities. We use, for example, the Bible and the tions, and so also are (or should be) the connections Hymn Book in our standardized religious practices, between indexing terms and the documents indexed. textbooks in our standardized teaching practices, law books in standardized legal practices, and so forth. 6. Social Semiotics and Activity Theory: The When we name something, we facilitate the use of Teleological Nature of KO that object. By naming churches (or books on churches) “tourist attractions,” we facilitate a certain The structuralist view of how our concepts are use of churches (or documents). We perform an act formed by our languages is shown in Figure 3 below. by means of language. Given names (or SAPs) will It shows that there is no one-to-one relation between always facilitate some uses of a document and make meanings in different languages. When individuals other uses relatively more difficult. No terminology learn a language, they learn the concepts of that lan- or technology or kind of SAP can ever be neutral. guage, and consequently they classify the world in the Concepts and documents have more or less stable way that is given by that particular language. For ex- functional values in relation to such standardized ample, Germans and Danes classify “tree” in different practices. There are of course always different views ways. By implication languages may affect the way of whether such standardized practices should be we conceptualise the world. According to Louis changed or remain unchanged, and there are always Hjelmslev (1943) each language puts arbitrary borders different kinds of possible changes of those practices. on reality, while activity theory (AT) finds that our There are always dynamic developments in meaning. symbolic systems tend to capture functional aspects The most important cause of this is the purpose for or affordances in the things we perceive (see Al- which we use the concepts, which is connected to the brechtsen, Andersen, Bødker, & Pejtersen, 2001). practices we want to change. Often scientific and technological development English *German *Danish *French Italian Spanish changes standardized practices in a rather uncontro- Tree Baum Arbre albero Árbol versial way. In other cases, however, changes in prac- Træ Leña tices are related to different political interests, to dif- Wood Holz legno ferent theories or “paradigms.” Different paradigms Bois Madera tend to influence given practices in different ways and Woods Skov bosco Bosque Wald by doing so they also tend to change our symbolic sy- Forest Forêt foresta Selva stems as well as our production of documents and the Figure 3: Cultural relativity in word meanings form, content, meaning and use of those documents. * Originally presented by the Danish structural linguist Louis The proper study of symbols and documents is thus Hjelmslev (1943). based on the study of the functions and interests tho- Extendet by information from Buckley (2001) se documents are serving. Any given tool (including a sign, concept, theory, Hjelmslev’s figure may be criticized; for example, it finding or document) has given potentialities in relati- has been mentioned that he omits the Danish term on to new social practices and paradigms. Users of li- “lund” (small forest) and that this omission may wea- braries may discover that previously neglected docu- ken this particular example. In this place we shall, ments may be very useful to develop a new social Knowl. Org. 30(2003)No.2 99 B. Hjørland: Fundamentals of Knowledge Organization practice or to defend an existing practice. Or they Robert de Beaugrande (1997) in linguistics. “I would may find that well-known documents may be re- certainly be happier if my findings had turned out to interpreted and used in new ways. Especially the kinds be less sobering or disturbing; but I can only report of documents known as “classics” have such abilities what I have in fact found” (§89). to be reused and re-interpreted by each generation. In the view of pragmatism (and Activity Theory) But even the most humble documents may turn out languages are tools that are adapted culturally to suit to be of value. This does not, of course, imply relati- the needs of their users. General languages may be vism regarding the relevance of documents: tools also seen as adaptations that suit the needs of major parts quickly become obsolete and “only of historical inter- of the populations, while languages for special purpo- est.” There may of course also be a discrepancy bet- ses are adaptations that suit the needs of specific ween the potential value of some documents and the groups such as chemists, lawyers, musicians, and so cost of storing, organizing and using them. forth. Both languages, (or more broadly symbolic sy- The meaning of any sign is its potential quality of stems) as also media, documents and information sy- for referring to some objects or states of affairs (Kar- stems, are teleological or goal-directed. They are op- patschof, 2000, p.197). A tool is something that has a timised to do certain functions, while relatively igno- functional value for some human (sub)culture. Lan- ring other purposes and goals. They should be seen as guages (and sublanguages) are also tools with functio- specializations in the social division of labour. This nal values. In languages, there are terms for tools. The implies that no system of KO can be optimised to do meaning of a word for a certain kind of tool is the all kind of tasks equally well. Although modern functional value of referring to a certain functional computer based retrieval systems are very flexible and value, defining the quality of the tool (Karpatschof, seem efficient for almost all tasks, it is important to 2000, p.197). A hammer may, for example, be termed consider the limitations of each kind of system for “hammer” (denotation) or “murder-weapon” (conno- different kinds of tasks and different domains. The tation). The word we use about a tool facilitates one way knowledge is organized in information systems or another use of it. Words may be more or less ap- must be relevant to the specific purpose of that parti- propriately used for a specific object in relation to a cular system. Relevance must, in my opinion, always given task, activity or discourse. In other words, to be regarded in relation to a goal. Birger Hjørland and use words is a kind of act (verbal acts) often used to Frank Sejer Christensen (2002) defined relevance this accomplish something extra-verbal. The meanings of way: “Something (A) is relevant to a task (T) if it increases the the words we use may be more or less suitable for our likelihood of accomplishing the goal (G), which is implied by T” [p. purposes, and in those cases we try to develop new 964]. Our conception of relevance is thus connected words or change the meaning of some old words. to meaning and to different views and traditions in Such changes in meanings are visible in the study of society. It is a basic view in the pragmatic tradition different fields, traditions and paradigms. that the nature of knowledge is to fulfil goals for the It is very naive and reductionistic to disregard organism or system that possess the knowledge. All these kinds of cultural inter-mediating factors in peo- kinds of knowledge or information processing pro- ple's relationship with information. The dominant cesses or institutions are basically telelogical by na- traditions in both information science and in behav- ture. They have to fulfil certain goals. Their relevance ioural and cognitive sciences have, however, ne- criteria are constructed backwards from those goals, glected these cultural aspects and just tried to study and people learn those relevance criteria by being so- generalized human relations to something termed “in- cialized and educated within a particular context and formation.” This dominant approach may be broadly tradition. termed “behavioural” in spite of different attitudes to versions of behaviourism. In this tradition people are 7. Paradigms, Epistemologies and KO expected to react to something in a specific, mechani- cal way without considering the culturally- The concept of “paradigm” became influential determined meanings and without considering the through Thomas Kuhn’s (1962, 1970, 1996) book. different goals and values in the meanings and in the Kuhn did not, however, recognize the social sciences documents. This has, in my opinion, brought about a as sciences or as paradigmatic. He considered the so- situation in which we have inherited very little useful cial sciences to be preparadigmatic. In his view a pa- knowledge from these areas on which to advance our radigm should be a set of common assumptions. As sta- field and practice. That view is also expressed by ted by Dogan (2001): 100 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Within a formal discipline, several major theo- tion science (see section 9 and 10 below); it is also the ries may cohabit, but there is a paradigm only case that the knowledge that we are going to organize when one testable theory alone dominates all is already conceptualised and organized according to other theories and is accepted by the entire sci- more or less invisible structures, determined by diffe- entific community. When Pasteur discovered rent paradigmatic influences. This is done in many the microbe, the theory of spontaneous genera- “layers” in ordinary languages, in subject languages, tion collapsed: contagion became the new para- in documentary forms, in networks of citing papers, digm. In the social sciences, however, we see at in informal networks and so on. Different paradigms best a confrontation between several non- test- usually imply different values and goals. KO in in- able theories. Most of the time there is not even formation science must know those different values a confrontation but careful mutual avoidance, and be able to fulfil the different goals. superb disregard on all sides. This is relatively easy, owing to the size of scientific disciplines, 8. Concepts and Conceptual Relations as Units of KO and their division into schools; it is true for all countries, big or small. (p. 11024) Concepts are normally seen as the units of thought and knowledge. This implies that the fundamental Nonetheless, his theory has been very influential in units of KO are the (semantic) relations between con- the social sciences, and the concept of paradigm is of- cepts. If knowledge is defined as justified true belief ten used in regard to the different schools, approa- (as in the Platonic tradition) then real knowledge is ches, systems and so on. Thus, the concept of para- hard or impossible to identify and to classify. It is digm is usually applied in a broader way. “Paradigm” more fruitful to speak of knowledge claims, rather may, in the philosophy of science, mean more or less than knowledge itself. To speak of knowledge claims as implicit background assumptions concerning the ob- the things represented in the literature and the thing ject of research, concerning research methods, con- to be classified is a more careful way of speaking, and cerning the usefulness of research, and so forth. In an there is no real loss by this way of speaking.7 even broader meaning the concept may be used with The basic function of concepts is to provide a basis more or less coherent views concerning a collective for dealing with the world. Concepts provide borders practice, such as teaching, divine service, administra- and classes in a continuous world. “Blue” delimits tion of justice, politics, and so forth. Kuhn’s 1970 some wavelengths, “castle” delimits some kind of postscript in the second edition defines among other buildings, and “music” delimits some sounds. Our things a paradigm as a constellation of beliefs, values, theories, conceptualisations and paradigms tend to techniques and so on shared by members of a given classify things and knowledge about things according community. to the same basic principles. Anders Ørom (2003, in Håkan Tjörnebohm, in1974, (as cited in Andersen, press) has demonstrated how, for example, pictures in 1999, p. 89) defined paradigms as systems of (explicit museums, the literature about art and the classificati- or implicit) basic assumptions and epistemic ideals in ons used in libraries and bibliographies all tend to be scientific disciplines. A paradigm is a super individual influenced by the same paradigm (e.g. the stylistic pa- structure of meaning, which is formed and reproduced radigm, the cultural historical paradigm, the semiotic in disciplinary socialization, teaching and scientific paradigm, the social historical paradigm, the feminist communication. Components of paradigms include: paradigm etc.). A descriptor or a class represents a concept8, and 1) Ideals and beliefs about science (epistemic information retrieval is essentially concerned with goals, methods and criteria in the production semantic relations between queries, document repre- and evaluation of scientific results inside the dis- sentations and texts. The smallest unit is the relations cipline); between two concepts. The theory of concepts (mea- 2) World-view hypotheses, basic ontological as- ning, semantics) is, however, probably one of the sumptions; and most difficult and muddled research fields today. It is 3) Ideals concerning significance for society and important to understand the different views between culture, for practical use and for enlightenment. traditional theories based on logical positivism and al- “Paradigm” in the broadest meaning of the word is a ternative views based on pragmatic theories. Tem- very central concept for KO. Not only do scientific merman (1997) describes some of these differences. paradigms determine how we study KO in informa- Knowl. Org. 30(2003)No.2 101 B. Hjørland: Fundamentals of Knowledge Organization

Note the distinction between on the one hand subtitle “International Journal. Devoted to Concept logical and ontological classification, which in tra- Theory, Classification, Indexing, and Knowledge Re- ditional [theory of] Terminology is supposed to presentation”. In spite of this, the contributions be possible in the mind without considering or about concepts have hitherto been very limited in using language and before the naming of the number, and the concept of concepts is much better concepts takes place, and on the other hand cate- explored in other fields like philosophy, cognitive gorisation which is a result of the interaction bet- and developmental psychology, artificial intelligence ween language and the mind (p. 55). 9 and linguistics. Different theories of concepts are deeply linked The importance of concepts has been recognized in with epistemological theories. Empiricism, rationa- the KO community. Dahlberg (1991) is one promi- lism, historicism and pragmatism have their different nent example. The journal Knowledge Organization has as conceptions of concepts.

“Paradigms” In philosophy In psychology

Empiricism Simple concepts correspond to simple sensations. Classical associationism. Behaviourism. Connec- tionism (neo-associationism or parallel distributed There are no necessary relations between simple processing, which works with inductive, self- concepts. Simple concepts may be combined to organizing, “neural” networks for the processing complex concepts. Nominalism: General concepts of sensory input) are names, which we put on classes of things (empirical generalizations).

Rationalism Simple are concepts, which cannot in fruitful Classical Artificial Intelligence. Cognitivism. ways be defined by other concepts. They are not Works with deductive, rule-governed algorithms experienced, but inborn or maturated. Complex for the processing of sensory input. “Classical” or concepts are defined from simple concepts. Fun- “Aristotelian” concepts exhaustively defined by damental concepts are concepts, which are neces- sets of necessary and sufficient attributes. sary in order to describe a field. Simple, complex and fundamental concepts enter into certain ne- cessary mutual relations. The differentiation bet- Modern theory (Rosch, Lakoff) Prototypetheory. ween simple and complex concepts is absolute Concepts are more or less prototypical instances (independent of domain, interests, points of view of things. etc).

Criticism (Kant) Concepts represent knowledge of the world me- Jean Piaget’s “genetic epistemology”. (Empirico-rationalism) diated by our forms of reason or categories like “Concepts mature in the individual. They grow space, time, thing and cause. out like the teeth”

Historicism and Concepts are formed in a historical process on Psychoanalytical interpretation. A satisfactory hermeneutics the basis of pre-understanding and holistic per- analysis of the functions of cognition cannot just ception. There is circularity between the forming explore sensation, memory and thinking in isola- of simple and complex concepts. The relations tion, but must involve the whole person, and his between simple and complex concepts are relative or her developmental history both individually in relation to interests. Traditions and social and collectively. Concepts are thus formed by in- communities play important roles for the for- fluence of personal characteristics such as sex and ming of concepts. social class.

Pragmaticism and Knowledge and concepts are formed by people’s Activity theory: Our concepts are not just de- critical realism practical activity in relation to the objects of the termined by attributes in physical objects. They activity. have also “historical depth”. They are formed in the historical contexts of the objects, which have meaning.

Fig. 4: Basic Conceptions of “Concept” 102 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Frank C. Keil outlines some very important devel- In almost all these cases the discoveries fol- opments in theories about concepts: low a similar course. Certain entities are ini- tially classified as members of a kind because ...one thing has increasingly emerged. Coher- they share many salient properties with other ent belief systems, or “theories,” are critical to bona fida members of that kind and because understanding the nature of concepts and how their membership is in accordance with current they develop. Although philosophers and some theories. This classification may be accepted for psychologists of a more classical bent have often centuries until some new insight leads to a reali- argued about the interdependence of theories zation that the entities share other, more fun- and concepts, for the most part cognitive psy- damentally important properties with a differ- chologists in the last 20 years have not. As a ent kind not with their apparent kind. consequence of the work done in the 1960s on Sometimes it is discovered that although the concept learning and the later “anticlassical” re- fundamental properties of the entities are not sponse to that paradigm, it became the norm to those of their apparent kind, they do not seem talk about concepts as consisting of combina- to be those of any other familiar kind either. In tions of features or as being the result of sum- such cases a new theoretical structure must de- marizing operations on exemplars and/or di- velop that provides a meaningful system of clas- mensions. Although the classical and newer sification. probabilistic views clashed on the issue of There are many profound questions about whether features for natural concepts were nec- when a discovery will have a major impact on a essary and sufficient, their debate masked a scheme of classification, but certainly a major shared set of assumptions that the meanings of factor is whether that discovery is made in the concepts could be fully described by lists of fea- context of a coherent causal theory in which the tures and simple probabilistic or correlational discovered properties are not only meaningful operations on those features. but central (Keil, 1989, p. 159). As researchers began to look at the pheno- mena associated with concepts more closely, This long citation is important, I believe, because it however, it became increasingly clear from se- demonstrates the deeper reality of kinds and concepts veral different lines of evidence that something that science discovers. It has important implications was missing... . . Part of the meaning of a con- for the methodology of KO. First and foremost it cept may be the apprehension of the theoretical challenges many user-oriented and empiristic ap- relations that explain its internal structure... . proaches. such relations have always been fundamentally To the degree that this view is correct, the relati- important in the philosophy of science. (Keil, ons between two concepts are thus relative to the 1989, p. 267) theoretical systems (or paradigms) in which they are embedded. We cannot say objectively and once and The history of all natural sciences documents for all, for example, that “schizophrenia” and “manio- the discovery that certain entities that share depressive disorders” are narrower terms (NT) in re- immediate properties nonetheless belong to dif- lation to “psychosis”. It has been suggested by Krin- ferent kinds. Biology offers a great many exam- glen (1994) that aetiologically schizophrenia might ples, such as the discoveries that dolphins and not be a useful concept, while one should probably whales are not fish but mammals, that the bat is alternatively operate with the old German concept not a kind of bird, that the glass “snake” is in “Einheitspsychose”. It follows that schizophrenia is fact a kind of lizard with only vestigial limbs not NT in relation to psychosis. To the degree that beneath its skin. In the plant kingdom it has this is correct the semantic relations between “schi- been found, for example, that some “vegetables” zophrenia,” “manio-depressive disorders” and “psy- are really fruits and that some “leaves” are not chosis” are synonym relations, not generic relations. really leaves. From the realm of minerals and We may conclude that the basic units in KO, the elements have come the discoveries, among oth- semantic relations between two concepts, may be re- ers, that mercury is a metal and that water is a lative to the perspective and the theory from which compound ... they are considered. Because of this fact, KO cannot be done just from successive combinations of ele- Knowl. Org. 30(2003)No.2 103 B. Hjørland: Fundamentals of Knowledge Organization ments, but must reflect broader perspectives and ficial attributes – e.g., “the big book of trains” which theories. considers most aspects of the “railway system”. Scien- tific classifications, on the other hand, reflect some 9. Criteria for Likeliness in KO. deeper properties, such as the classification of chemi- cal substances in organic and inorganic compounds, A fundamental principle in KO is that like things precious metals, etc., based on atomic theory. When should be brought together, while different things scientific principles of classification are applied, see- should be separated. “Likeliness” is a concept that mingly related objects may be separated and seemin- may also be expressed by other terms such as “simi- gly different objects may be grouped together. For larity”, “sameness” (used by James, 1890), “resem- example, whales were once classified as fish, but are blance” or “equivalence”. Many writers in LIS have today – influenced by evolutionary theory - classified defined classification and KO by using the concepts as mammals.10 These examples demonstrate that na- of likeliness. For example: ïve realism is not adequate as a method to classify do- cuments. – Ernest Cushing Richardson (1964, p. 1) defined If we consider different examples, we will see that classification as the “putting together of like conceptual relations have different kinds of motiva- things, or more fully described, it is the arranging tions. of things according to likeness and unlikeness. It may also be described as the sorting and grouping – “Institutions for Information Science” are generi- of things”. cally subordinate to “institutions”. This relation- – Henry E. Bliss (1935, p. 3) wrote: “In dealing with ship seems to be motivated by purely logical rela- the multiplicity of particular things, actualities, tions (or by relations inherent in a given language). and specific kinds, we find that some are alike, in – “Copenhagen” is part of “Denmark” (Whereas general characters and in specific characteristics; Malmö in Sweden is not. This last example is mo- and we may consequently relate them in a class, or tivated by the fact that Denmark lost this part of classes, that is classify them”. territory to Sweden in 1658). The parts of a coun- – Frederick Wilfrid Lancaster (born 1933) defined try are thus defined by social arrangements. This classification as “sorting items into ‘conceptual type of semantic relations is in other words based classes’” and “forming classes of objects on the ba- on human conventions. sis of their subject matter”. (Lancaster, 1998, p. 17). – Whales are today classified as mammals. The ex- – Arlene G. Taylor: “The placing of subjects into ca- planation of this semantic (generic) relation is due tegories; in organizing of information, classificati- to evolutionary theory.11 on is the process of determining where an informa- – Psychology (and also psychopathology) may be tion package fits into a given hierarchy and then classified as part of neuroscience (natural science), assigning the notation associated with the appro- as part of the social sciences or as part of the hu- priate level of the hierarchy to the information manities (or otherwise). Such differences are, for package and its surrogate.” (Taylor, 1999, p. 237). example, visible in how such fields are placed in the organizational structures in universities. Such How do we decide what things are alike? How do we classifications (and semantic relations) often in- develop our criteria for likeness? The literature on volve professional interests. It is, for example, KO in LIS seems to ignore this core problem. Is this partly a question of professional power whether a because the problem is seen as obvious? Is this based field (e.g. psychopharmacology) is monopolized by on a kind of naïve realism: that things are what they a profession. At the deepest level, the question of look like, and that people’s immediate sense of like- whether psychology is a human science or a natu- ness is adequate as the basis for KO? ral science is a scientific question related to theo- But when are two things alike? Naïve realism con- retical questions within psychology. It is well fuses seeming similarity with similarity in an objec- known that psychology is divided over this ques- tive sense. Some metals might look like gold, but tion. Different paradigms in psychology have dif- might not be precious metal. Chemical analysis (not ferent answers. For behaviorism psychology is common sense) has to make a distinction between clearly a part of the natural sciences (cf. Watson, “what looks like gold” and “what is gold”. Classifica- 1913). For humanistic psychology it is clearly a tion made for children may be based on more super- part of the humanities. In classifying psychology as 104 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

belonging to science, or social science or humani- – Qualitative methods ties, one is actually involved in a theoretical battle – Text based methods between paradigms. (Which most people find – People based methods rather uncomfortable). – Institution based methods (e.g. studies of univer- – Classifications and semantic relations may be es- sity organisations) tablished by empirical generalizations. One exam- – Bibliometrical methods ple is Berlin & Kay (1969), who established the – Word frequency based methods empirical generalization that human languages re- – Sociological methods flect the classification of colour sensations in essen- – Historical methods tially the same ways universally, regardless of his- – Pragmatic, epistemological and critical methods torical and cultural differences. Such empirical generalizations avoid the uncomfortable theoreti- It is important to realize that an important quality of cal issues addressed above. There are, however, a system of KO may be its function as a standard. Of- other kinds of problems with this approach, which ten classifications and other systems are made as are closely connected with the basic assumptions compromises between experts in a field and declared in empiricism. a standard. Scientific arguments between researchers – Classifications and semantic relations may also be in the field are not published and discussed in the lite- purely accidental or ad hoc. Such classifications rature as a part of this process. This method is epi- may serve some purposes very well. In general, stemologically a kind of “authoritanism”: the quality however, classifications that reflect essential char- of the system is claimed by reference to its authority acteristics in the objects are the most valuable. as a standard. However, as researchers in KO we have (This “essentialism” should not be confused with to consider other qualities of KO too and should not an objectivism that ignores human activity). accept “authoritanism” as a criterion of scientific knowledge. We may conclude that there exist many different One distinction, one that may be encountered in kinds of criteria for likeliness. They may be conven- the literature, is between computer based methods tional, logical, psychological and so on. Regarding and “manual” methods of KO. The term “manual” is natural kinds, however, they should especially be a little peculiar, because what is meant is rather intel- seen as domain-specific criteria which are discovered lectual methods. This distinction is not in itself a fun- by science. They are not just something that can be damental distinction. If it is possible to develop some extracted from users or from statistical investigations. explicit rules for how to classify documents, then such rules can be used by humans or by computers. 10. Methods of Knowledge Organization In other words, sometimes humans are not perform- ing better than computers, which follow rather sim- By methods of KO we mean the methods of con- ple instructions. At other times, however, humans structing systems of KO such as classifications and apply deep knowledge, which is not available to thesauri. Also the The processes of indexing and clas- computers. And sometimes computers process huge sification are also considered methods of KO. amounts of data that are not possible for humans to What are the methods for classification and KO? process. We should thus not consider computer based We may first make a distinction between classifica- versus manual methods as a basic methodological dis- tion in the sciences (such as biology or archaeology) tinction. We should rather classify the methods of and in LIS. This paper is about KO in LIS. However, KO in other ways asking what kind of rules, what we shall argue that basically the methods of KO in kind of knowledge and what kind of cognitive proc- LIS are connected to the same fundamental paradigms esses are involved in the process? In general our in epistemology as the methods of classification in knowledge of how humans classify is limited.12 sciences and other fields. First, however, let us Another distinction is between quantitative versus shortly consider some methods: qualitative methods. Quantitative methods may, for example, be based on word frequencies. Qualitative – Standardization methods may, on the other hand, be based on inter- – Computer based KO pretations of meaning. Again, I do not consider this a – “Manual” or “intellectual” methods fundamental distinction. It is well known that any – Quantitative methods quantitative investigation presupposes an adequate Knowl. Org. 30(2003)No.2 105 B. Hjørland: Fundamentals of Knowledge Organization qualitative analysis. Qualitative methods, on the perts, but something that might be constructed by other hand, may result in a number of categories and working with the literature (not just as a kind of thus approximate the quantitative approach. The so- opinion poll). The question of cognitive authority is ciologists Pierre Bourdieu uses many statistics (quan- of course extremely important and difficult. Scholarly titative data) in his research; however, he explicitly work is not just “subjective”, choosing authorities ac- denies doing this in a positivist way. There are thus cording to personal preferences. The more qualified, important ways in which to use data that go deeper the less “subjective” in this sense of the word. The than the distinction between quantitative and qualita- “objectivity” of the decision of scholarly work is tive approaches. partly based on the knowledge of different traditions Systems of KO may use data from the literature, and metaperspectives and on the cumulated historical from people, from organizations, etc., in many differ- evidence. On could perhaps say that a fundamental ent ways and in many combinations. Bibliometrical difference between empiricism/rationalism and his- approaches are mainly based on literatures, and so are toricism/pragmatism is the emphasis on reading in many conventional and “manual” methods (including the research process. the well known principle of identifying literary warrant Basically, the methods of KO are related to fun- for a given class or descriptor). The use of historical damental theories of epistemology. All researchers in knowledge about the development of knowledge any field are always more or less influenced by cer- fields (as exemplified in Wallerstein, 1996 and Hjør- tain ideals on how to obtain knowledge. This is also land, 2000c) is a different way to use literature as data the case with regard to constructing systems of KO. If for KO. This last approach also uses the development students of LIS have taken a course in the philosophy of the organization of universities as a basis for con- of science, they should be able to identify the domi- structing KO. nant epistemic approach in any paper on KO (and People may also be used as sources on which to also in any system of KO). Some papers and systems base the development of KO. In cognitive science and are mostly based on empirical generalizations. Sys- artificial intelligence the methods of knowledge elicitation tems based on word-frequency measures may be the are well known (e.g. Cooke, 1994). This is, in my best example. Other papers and systems are mostly opinion, an indirect method because we have to know the based on rational rules and deductions (while often methods by which the experts, from whom we ob- ignoring empirical issues). Facet-analytic systems in tain the knowledge, got this in the first place. In the tradition of Ranganathan may provide the best other words we must be able to argue which knowl- examples. A third kind of system is based on the edge claims are best substantiated. If we do not ad- study of the evolution of knowledge fields. To a cer- dress this issue, we are only indirectly qualified to tain degree the disciplinary based systems such as construct and evaluate systems of KO. People may, Dewey (DDC) can be said to conform to the ideal. however, also be used as sources in other ways. They On the other hand the updating of this system is may, for example, be seen as members of discourse com- more is and more influenced by the principles of the munities and be studied as such. We may study the so- facet-analytic tradition (as discussed by Miksa, 1998). cial division of labour and the dependency between And in some sense the DDC is also empirically based different people and groups of people (compare on the basis of the literature, which is classified by the Whitley, 1984). system. DDC is thus not as clear in its methodological When we are going to decide what kind of sources principles as the former examples. The fourth basic to build on (literature, people, organizations), we ideal is the pragmatic epistemology. In this case the sys- need to know something about how the relevant tems are developed on the basis of the analysis of goals, knowledge is distributed. Methods of knowledge values and consequences. For example, female research- elicitation seem to be built on the assumption that ers may regard the goal of women’s emancipation as the needed knowledge is ready at hand within a the main goal of feminist scholarship. They may thus group of experts or other people. This may be more regard any relations that support this goal as important. or less the case. However, more scholarly approaches “Similarity” is thus identified by classes of documents build on other assumptions. The meaning of words is sharing similar functions in relation to this goal. Prag- traced in great detail and registered in historical dic- matic classification may thus also be termed critical or tionaries with quotations and citations from authori- political classification. It may at first seem strange and tative authors. In this case, the relevant information is opposed to common scientific norms. not supposed to be ready at hand in the mind of ex- 106 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Pragmatic epistemology and pragmatic KO do not Concrete things have ways [emphasis in original] mean that a person (or a whole field) can just do things of acting, as many ways of acting as they have the way that suits his personal interests (or the interests points of interaction with other things. One of the researchers in the field). If this is done, if research thing is callous, unresponsive, inert in the pres- just produces “social constructions” then reality will ence of some other things; it is alert, eager, and make those constructions incoherent. They will be op- on the aggressive with respect to other things; in posed by empirical and theoretical arguments. The a third case, it is receptive, docile. Now different production of incoherent “knowledge” is not valuable ways of behaving, in spite of their endless diver- and cannot be a serious goal. Therefore pragmatic phi- sity, may be classed together in view of common losophy is bound up with a form of realism. The prag- relationship to an end. No sensible person tries matic method is not opposed to aspects of empiricism, to do everything. He has certain main interests rationalism and historicism. It claims, however, that no and leading aims by which he makes his behavior isolated evidence is enough. The final criteria of truth coherent and effective. To have an aim is to limit, are connected to human goals and activities. You can- select, concentrate, group. Thus a basis is fur- not avoid considering such issues although they may nished for selecting and organizing things accord- seem uncomfortable. ing as their ways of acting are related to carrying The founder of pragmatism, Charles Sanders Peirce forward pursuit. Cherry trees will be differently (1839-1914), wrote: grouped by woodworkers, orchardists, artists, scientists and merry-makers. To the execution of The rational meaning of every proposition lies different purposes different ways of acting and re- in the future. How so? The meaning of a propo- acting on the part of trees are important. Each sition [its logical interpretant] is itself a proposi- classification may be equally sound when the dif- tion. Indeed, it is no other than the very propo- ference of ends is borne in mind. sition of which it is the meaning: It is a transla- Nevertheless there is a genuine objective stan- tion of it. But of the myriads of forms into dard for the goodness of special classifications. which a proposition may be translated, what is One will further the cabinetmaker in reaching that one which is to be called its very meaning? his end while another will hamper him. One It is, according to the pragmaticist, that form in classification will assist the botanist in carrying which the proposition becomes applicable to on fruitfully his work of inquiry, and another human conduct, ... that form which is most di- will retard and confuse him. The teleological rectly applicable to self-control under every theory of classification does not therefore com- situation and to every purpose. This is why he mit us to the notion that classes are purely verbal locates the meaning in future time; for future or purely mental. Organization is no more conduct is the only conduct that is subject to merely nominal or mental in any art, including self-control. (Peirce, 1905) the art of inquiry, than it is in a department store or railway system. The necessity of execution In this way the pragmatic epistemology sees the supplies objective criteria. Things have to be meaning of words as connected to speech acts and to sorted out and arranged so that their grouping the goals that humans try to satisfy through their will promote successful action for ends. Conven- acts. In the same way it regards semantic relations and ience, economy and efficiency are the bases of classifications as determined by their functions as classification, but these things are not restricted tools for human goals. As the pragmatic philosopher to verbal communication with others nor to in- John Dewey stated: ner consciousness; they concern objective action. They must take effect in the world. So far nominalism and conceptualism - the the- At the same time, a classification is not a bare ory that kinds exist only in words or in ideas - transcript or duplicate of some finished and done- was on the right track. It emphasized the teleo- for arrangement pre-existing in nature. It is rather logical character of systems and classifications, a repertory of weapons for attack upon the fu- that they exist for the sake of economy and effi- ture and the unknown. For success, the details of ciency in reaching ends. But this truth was per- past knowledge must be reduced from bare facts verted into a false notion, because the active and to meanings, the fewer, simpler and more exten- doing side of experience was denied or ignored. sive the better... (Dewey, 1948, p. 151-154). Knowl. Org. 30(2003)No.2 107 B. Hjørland: Fundamentals of Knowledge Organization

Figure 5 below summarizes the four fundamental transmitted through language and cultural products. methods of KO for which I have argued. It is impor- And any kind of pragmatism is limited by constrains tant to realize that they are idealizations. The do not set by the real world through empirical evidence. and they cannot exist in pure forms. Any empirical By knowing the principal strengths and weak- procedure must involve kinds of logic and non- nesses of basic epistemological positions, the informa- empirical evidence. All rules must be applied to some tion specialist is equipped with some general under- empirical reality. All empirical and rational methods standing of the strong and weak sides of different must be based on pre-understandings and meanings methods of KO.

“Scientific Classification” “Bibliographic Classification”

Empiricism Classification provided by statistical generali- Documents clustered on the basis of some zations (e.g. factor analysis) based in “simila- kind of similarity, e.g. common terms in tra- (Observations and inductions) rity”. ditional IR or bibliographical coupling. Examples: Classifications of mental illness in Examples: “Atlas of science” and visualizati- psychiatry (DSMIV) kinds of intelligence in ons (White & McCain, 1998). “Research psychology based on statistical analysis of test Fronts” I SCI and algorithms for information scores. retrieval.

Rationalism Classification based on logical, universal divi- Facet analysis built on logical divisions and sions. “eternal and unchangeable categories” (Principles of pure reason. Deductions) Examples: Frame based systems in Artificial Examples: Ranganathan, Bliss II & Langridge. Intelligence. Chomsky’s analysis of the deep Semantic networks. According to Miksa structure in language- (1996) the DDC has increasingly used this approach.

Historicism Classification based on historical or evolutio- Systems based on the study of the develop- nary development. ment of knowledge and knowledge produ- (Study of context and cing communities (the social division of development- Explicating Examples: Biological taxonomies based on (scientific) labour). pre-understanding) evolutionary theory. Classification of the sciences on the basis of Examples: Wallerstein (1996) (See also their history and organizational structures. Hjørland, 2000c). The feature of the DDC that it distributes subjects by discipline.

Pragmatism Classifications based on specific values, poli- Systems based on “cultural warrant” or “cri- cies and goals, e.g. feminist epistemology. tical classification”. Examples: The French (Analysis of goals, values and encyclopaedists, the Marxists, Classifications consequences in both subject serving feminist collections. and object)

Figure 5: Fundamental Methods of Classification

11. Conclusion

This paper has outlined the fundamental aspects of and the methods of KO in LIS are at the deepest level KO as a field of study. The basic assumptions and at- based on the same philosophical assumptions as the titudes have been that the basic units of KO are se- methods of science and scholarship. This implies that mantic relations between concepts. Such semantic re- the fundamental discussion of the basis of KO in LIS lations cannot primarily be established by universali- is strongly connected to the discussion of different stic assumptions, but much primarily be understood theories of epistemology. Epistemological studies ha- as domain specific, as uncovered by (and constructed ve been rare in LIS, and it seems urgent for our field by) scientific disciplines. KO in LIS cannot ignore to upgrade our qualifications in this area. concepts, theories and findings in specific disciplines, 108 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Notes position in a concept system which visualises logi- cal and ontological relationships. On the basis of 1 This article is a slightly modified version of a pa- the position in the concept system a definition per presented in Salamanca, Spain (Hjørland, will be formulated” (Temmermann, 1997, p. 53). 2003a). 10 Aristotle noticed that dolphins and porpoises had 2 From 1937 named Féderation Internationale de Docu- lungs. They did not lay eggs, and they nourished mentation, FID, and from 1986 International Fédéra- tion their unborn young via a placenta. He grouped for Information and Documentation, FID. them with the quadrupeds rather than fish. (Ari- 3 The history of the abstract journal goes, however, stotle was a lone voice. For over 2,000 years af- back to 1665, cf. Manzer (1977). terwards, his idea was lost, and whales were fish. 4 Computer science deals with “information tech- Only in relatively modern times were whales put nology”, IT, or with data processing. Computer back with mammals.) science is in Denmark called “datalogi”, the sci- 11 Reveal (1999) describes different paradigms, ence of data. methods and tools used in plant taxonomy and 5 I write this well aware that the pragmatic philo- mentions among other things the use of (1) anat- sopher John Dewey used the mirror metaphor: omy: 1800s, (2) chromosome numbers: 1920’s, (3) “A classification of books to be effective on the comparative garden studies: 1940’s, (4) palynol- practical side must correspond to the relationships ogy: 1950’s, (5) phenetics: 1960’s, (6) biochemical of subject-matters, and this correspondence can be systematics: 1970’s, (7) cladistics: 1980’s and (8) secured only as the intellectual, or conceptual, or- chloroplast DNA and RNA: 1990’s. ganization is based upon the order inherent in the 12 Klumpner (1993, p. 1) reports: “These problems fields of knowledge, which in turn mirrors the led the Indexing Study Group of the American order of nature.” (Dewey, 1929, p. viii) Psychoanalytic Association to an experiment. 6 Whereas traditional semiotics tends to be formali- About a dozen seasoned analysts independently stic and to abstract signs from their contexts of indexed a passage from the Standard Edition [the use; social semiotics examines semiotic practices, works of Sigmund Freud]. When they compared specific to a culture and community, for the ma- what they had done, all agreed that the failure to king of various kinds of texts and meanings in va- agree about which terms to index was humbling rious situational contexts and contexts of cultural- and impressive. The group did not even agree on ly meaningful activity. For an introduction to so- which words required see or see also directives, or cial semiotics see, for example, Hodge & Kress on the words that should follow those directives. (1988). The formulations in this paper have been A dead end?”. Klumpner himself tried to solve the inspired by Karpatschof (2000). problem by a quantitative study of actual produ- 7 Sometimes a differentiation is made between the ced indexes. What I would like to emphaze is that organization of knowledge and the organization in order to progress we need those kinds of disa- of other “things” (including physical objects, pro- greements recorded and discussed qualitatively in cesses, phenomena such as pain etc.). Other things the literature. They are extremely rare, for which than knowledge (claims) do not, however, classify reason I claim that we know little about “manual” themselves. They are also classified by our con- KO. ceptualization of the world. 8 In the literature of cognitive science (e.g. Keil, References 1989) a distinction is made between word mea- nings and concepts, or between concept structure Albrechtsen, H.; Andersen, H.H.K.; Bødker, S.; Pe- and semantic structure. People are described as jtersen, A.M.,(2001). Affordances in Activity Theory And possessing two different systems: a semantic sy- Cognitive Systems Engineering. Denmark. Risø: Avail- stem of word meanings and a concept system. able at: http://www.risoe.dk/rispubl/SYS/syspdf/ This distinction is not made in the present paper. ris-r-1287.pdf 9 “Saussurian structuralist semantics believe that the Andersen, H. (1999). Political Attitudes and Cogni- best way to describe meaning is to describe the tive Convictions among Danish Social Science Re- mutual deliminating of concepts (semantic relati- searchers. Scientometrics, 46(1), 87-108. ons). Traditional Terminology believes that the Arbib, M. A.; Kfoury, A.J. & Moll, R. N. (1981). Basis best way to describe concepts is to determine their For Theoretical Computer Science. Berlin: Springer-Verlag. Knowl. Org. 30(2003)No.2 109 B. Hjørland: Fundamentals of Knowledge Organization

Beaugrande,R. de (1997). On history and historicity Frohmann, B. (1994). The Social Construction of in modern linguistics. Formalism versus function- Knowledge Organization: The Case of Melvin alism revisited. Functions of Language, 4/2, 169-213. Dewey. Advances in Knowledge Organization. 4, 109-117. http://www.beaugrande.com/ Griffith, B. C. (Ed.). (1980). Key Papers in Information Sci- Berlin, B. & Kay, P. (1969). Basic Color Terms: Their Uni- ence. New York: Knowledge Industry Publications. versality and Evolution. Berkeley: University of Cali- Hargens, L. L. (2000). Using the literature: reference fornia Press. networks, reference contexts, and the social struc- Bernal, J. D. (1948). Preliminary Analysis of Pilot ture of scholarship. American Sociological Review, 65 Questionnaire on the Use of Scientific Literature. (December), 846-865. In: Royal Society [Great Britain]. The Royal Soci- Hjelmslev, L. (1943) Omkring sprogteoriens grundlæggelse. ety Scientific Information Conference: Report and København: B. Lunos bogtrykkeri a/s. (Many la- Papers Submitted. 1948 21 June – 2 July; London, ter editions and translations, e.g. Prolegomena to a England. London, England: Royal Society. Pp. Theory Of Language. translated by Francis J. Whit- 589-637. field. Baltimore : Waverly Press, 1953). Bliss, H. E. (1935). A System of Bibliographical Classification. Hjørland, B. (1988). Information Retrieval in Psy- New York: The H. W. Wilson Company. chology. Behavioral and Social Sciences Librarian, Vol 6 Borko, H. (1968). Information science: What is it? (3/4), 1988, 39-64. American Documentation, 19(1), 3­5. Hjørland, B. (1992). The concept of “subject” in In- Bradford, S. C. (1950). Documentation. London: C. formation Science. Journal of Documentation, vol. 48, Lockwood. no. 2, 172-200. Buckley, G. (2001). Semantics. http://www.ling. Hjørland, B. (1993). Emnerepræsentation og informationssøgning. upenn.edu/courses/Spring_2001/ling001/ Bidrag til en teori på kundskabsteoretisk grundlag. Göteborg: semantics.html [visited 2003-01-08] Valfrid, Cooke, N. J. (1994). Varieties of knowledge elicita- Hjørland, B. (1997): Information Seeking and Subject Repre- tion techniques. International Journal of Human- sentation. An Activity-theoretical approach to Information Sci- Computer Studies , 41(6) , 801-849. ence. Westport & London: Greenwood Press. Dahlberg, I. (1991). Philosophical foundations of Hjørland, B. (1998a). Information retrieval, text conceptual ordering systems. Advances in Knowledge composition, and semantics. Knowledge Organization, Organization, 3, 102-119. 25(1/2), 16-31. Daly, L. W. (1967). Contributions to a History of Alphabeti- Hjørland, B. (1998b). The Classification of Psychology: zation in Antiquity and the Middle Ages, Brussels: Lato- A Case Study in the Classification of a Knowledge mus. Field. Knowledge Organization, 24(4), 162-201. Dewey, J. (1929). Introduction. IN: H. E. Bliss: The Hjørland, B. (2000a). Documents, Memory Institutions, Organization Of Knowledge And The System Of The Sci- and Information Science. Journal of Documentation, vol. ences. New York, Holt (pp. vii-ix). 56(1), 27-41 Dewey, J. (1948). Reconstruction in Philosophy. Enlarged edi- Hjørland, B. (2000b). Library and Information Sci- tion. New York: Beacon. (Original work published ence: Practice, theory, and philosophical basis. In- 1920) formation Processing and Management, 36(3), 501-531. Dewey, M. (1979): Dewey Decimal Classification and Relative Hjørland, B. (2000c). Review of Wallerstein et al. Index. Edition 19. Albany, N.Y.: Forest Press, (1996). Open the Social Sciences, report of the 1979. Vol. 1. Gulbenkian Commission on the Restructuring of Dogan, M. (2001). Paradigms in the Social Sciences. the Social Sciences. Stanford, Calif.: Stanford IN: International Encyclopedia of the Social & Behavioral University Press. I: Knowledge Organization, 27(4), Sciences, (vol. 16, pp. 11023-11027). Ed. By N. J. 238-241 Smeltser & P. B. Baltes. Amsterdam: Elsevier. Hjørland, B. (2001). Towards a theory of aboutness, Ellis, B. (1990). Truth and Objectivity. Oxford: Basil Black- subject, topicality, theme, domain, field, content... well. and relevance. Journal of the American Society for Infor- Fjordback Søndergaard, T.; Andersen, J. & Hjørland, mation Science and Technology. 52(9):774–778. B. (2003). Documents and the Communication of Hjørland, B. (2002a). Domain analysis in information Scientific and Scholarly Information. Revising and science. Eleven approaches - traditional as well as Updating the UNISIST Model. Journal of Documenta- innovative. Journal of Documentation, 58(4), 422-462 tion, 59(3), 278-320. 110 Knowl. Org. 30(2003)No.2 B. Hjørland: Fundamentals of Knowledge Organization

Hjørland, B. (2002b), Epistemology and the Socio- Karpatschof, B. (2000). Human activity. Contributions to the Cognitive Perspective in Information Science. Anthropological Sciences from a Perspective of Activity Theory. Journal of the American Society for Information Science and Copenhagen: Dansk Psykologisk Forlag. ISBN: Technology, 53(4), 257-270. 87-7706-311-2 Hjørland, B. (2002c) The Methodology of Construct- Keil, F. C. (1989). Concepts, kinds, and cognitive development. ing Classification Schemes. A Discussion of the Cambridge, Massachusetts: The MIT Press. State-of-the-Art. IN: Proceedings of the Seventh Interna- Kessler, M. M. (1965). Comparison of the results of tional ISKO Conference 10.-13. July 2002, Granada, Spain. bibliographic coupling and analytic subject index- Ed. By María J. López-Huertas. Würzburg, Ger- ing. American Documentation vol. 16, no. 3, 223-233. many: Ergon Verlag, pp. 450-456. (Advances in Klumpner, G. H. (1993). A Guide To The Language Of Knowledge Organization, Vol. 8). Psychoanalysis. An Empirical Study Of The Relationships Hjørland, B. (2002d). Principia Informatica. Founda- Among Psychoanalytic Terms And Concepts. Madison, tional Theory of Information and Principles of In- Connecticut: International University Press, Inc. formation Services. IN: Emerging Frameworks and Kringlen, E. (1994). Is the concept of schizophrenia Methods. Proceedings of the Fourth International Conference on useful from an aetiological point of view? A selec- Conceptions of Library and Information Science (CoLIS4). tive review of findings and paradoxes. Acta Psy- Ed. By Harry Bruce, Raya Fidel, Peter Ingwersen, chiatrica Scandinavica, 90 (suppl. 384), 17-25. and Pertti Vakkari. Greenwood Village, Colo- Kuhn, T. S. (1962). The Structure of Scientific Revolutions. Chi- rado, USA: Libraries Unlimited. (Pp. 109-121). cago: The University of Chicago Press. 2rd. eition Hjørland, B. (2003a). Fundamentals of knowledge or- 1970. 3rd edition 1996. ganization. IN: Tendencias de investigación en organización Lancaster, F. W. (1998). Indexing And Abstracting In The- del conocimiento / Trends in Knowledge Organization Re- ory And Practice. 2nd ed. London: Library Associa- search. (Pp. 83-116; +English abstract p. 51 + Spa- tion. nish abstract p. 19). José Antonio Frias & Críspu- Manzer, B. M. (1977). The Abstract Journal, 1790-1920: lo Travieso (Eds.). Salamanca, Spain: Ediciones Origin, Development and Diffusion. Metuchen, N.J.: The Universidad de Salamanca. Scarecrow Press. Presentation available at: http://www.db.dk/bh/ Miksa, F. L. (1998). The DDC, The Universe of Knowledge, and publikationer/Filer/Presentation.ppt the Post-Modern Library. Albany, NY: Forest Press. Hjørland, B. (2003b), Social And Cultural Awareness Pao, M. L. (1993). Term and citation retrieval: A field and Responsibility in Library, Information and study. Information Processing & Management, 29(1), Documentation Studies. IN: Aware and Responsi- 95­112. ble:Papers of the Nordic-International Colloquium on Social Pao, M. L., & Worthen, D. B. (1989). Retrieval effec- And Cultural Awareness and Responsibility in Library, In- tiveness by semantic and pragmatic relevance. Jour- formation and Documentation Studies (SCARLID). Edited nal of the American Society for Information Science, 40(4), by W. Boyd Rayward, Joacim Hansson, Vesa 226­235. Suominen. Lanham, MD: Scarecrow Press, (in Peirce, C. S. (1905). What pragmaticism is. The Monist, press). 15, 161-181. Hjørland, B. & Albrechtsen, H. (1995). Toward A Poulsen, C. (1994). Informationens skygge og foran: informa- New Horizon in Information Science: Domain tionskvalitet, informationseksplosion og online kataloger. Ro- Analysis. Journal of the American Society for Information skilde, Denmark: Institut for Datalogi, Kommu- Science, 46(6), 400-425. nikation og Uddannelsesforskning, Roskilde Uni- Hjørland, B. & Kyllesbech Nielsen, L. (2001). Subject versitetscenter. Access Points in Electronic Retrieval. Annual Re- Rees-Potter, L. K. (1987). A bibliometric analysis of termi- view of Information Science and technology, vol. 35, 249-298. nological and conceptual change in sociology and economics with Hjørland, B. & Sejer Christensen, F. (2002). Work the application to the design of dynamic thesaural systems. Vol. tasks and socio-cognitive relevance. A specific Ex- 1-2. Ph.D dissertation. Dissertation Abstracts Interna- ample. Journal of the American Society for Information Sci- tional, 48(06), SecA, p.1344. ence and Technology, 53(11), 960-965. Rees-Potter, L. K. (1989). Dynamic thesaural systems: Hodge, R. & Kress, G. (1988). Social Semiotics. New a bibliometric study of terminological and con- York; Cornell University Press. ceptual change in sociology and economics with James, W. (1890). The Principles of Psychology. (2 vols.). application to the design of dynamic thesaural sys- New York: Henry Holt. Knowl. Org. 30(2003)No.2 111 B. Hjørland: Fundamentals of Knowledge Organization

tems. Information Processing & Management, 25(6), 677- Temmerman, R. (1997). Questioning the univocity 691. ideal. The difference between socio-cognitive Rees-Potter, L. (1991). Dynamic thesauri: the cogni- Terminology and traditional Terminology. Hermes. tive function. Tools for knowledge organisation Journal of Linguistics, 18, 51-93. Available at: and the human interface. Proceedings of the 1st Interna- http://hermes.asb.dk/archive/FreeH/H18_04.pdf tional ISKO Conference, Darmstadt, 14-17 August 1990. (Visited 2003-01-11). Part 2, 1991, 145-150. Törnebohm, H. (1974). Paradigm i vetenskapernas värld och Reveal, J. L. (1999). PBIO 250 Lecture Notes. Princi- i vetenskapsteorien. Göteborg: University of Göte- ples of Plant Taxonomy. http://www.inform. borg. umd.edu/PBIO/pb250/prin.html (hentet 10-02- Unisist (1971), Study Report on the feasibility of a World Sci- 2003) ence Information System. By the United Nations Edu- Richardson, E. C. (1964). Classification. Theoretical and cational, Scientific and Cultural Organization and practical. Third edition. Hamden, Connecticut: The the International Council of Scientific Unions. Shoe String Press, Inc. (Reprinted unaltered from Paris, Unesco. 1930 edition). U.S. Department of Labor. Bureau of Labor Statistics Rorty, R. (1979) Philosophy and the Mirror of Nature, (2002). Standard Occupational Classification (SOC) Sys- Princeton, NJ: Princeton University Press. tem. Washington. D.C. http://www.bls.gov/soc/ Schneider, J. W. & Borlund, P. (2002). Preliminary (Visited December 30, 2002). study of the potentiality of bibliometric methods Wallerstein, I. (1996). Open the Social Sciences, report of the for the construction of thesauri. IN: Emerging Gulbenkian Commission on the Restructuring of the Social Frameworks and Methods. Proceedings of the Fourth Interna- Sciences. Stanford, Calif.: Stanford University Press. tional Conference on Conceptions of Library and Information Watson, J. B. (1913) Psychology as the behaviorist Science (CoLIS4). Ed. By Harry Bruce, Raya Fidel, views it. Psychological Review, 20, 158-177. Peter Ingwersen, and Pertti Vakkari. Greenwood Whitley, R.(1984). The Intellectual and Social Organizations Village, Colorado, USA: Libraries Unlimited. (Pp. of the Sciences. Oxford University Press. 151-165). Ørom, A. (2003) The organization of knowledge Taylor, A. G. (1999). The Organization of Information. about art. Paper to appear in Knowledge Organization Englewood, Colerado: Libraries Unlimited, Inc. 2003.