Knowledge, Concepts, and Categories. Koen Lamberts and paradox motivating their analysis is that similarity is used to explain concepts, and concepts are used to explain similarity David Shanks, eds. Cambridge, MA: The MIT Press; 1997: 464 pp. Price: $30.00. (ISBN 0-262-62118-5.) (p. 44). Into this vicious circle falls information science because ''similarity'' is probably the ideal operationalization for the fuzzy concept of relevance. Information science is shifting from a systems focus to a What does it mean that two bibliographic documents are users focus. There are calls in the research literature for a new similar? Perhaps the answer is that they have many properties dynamic, user-centered approach to relevance, and new conferin common. But this does not get us too far since any two ences on the contexts of information seeking. Qualitative methbibliographic documents have an infinitude of common attriods suddenly have gained prominence. In this era of shifting butes (e.g.: Neither were published in Bulgaria in 1938, neither paradigms and methodological uncertainty, it is delightful to were found in the Australian Outback, neither have inlayed gold find a book that summarizes the latest empirical research on leaf, neither are on fire at this moment, etc.). We conclude that mental representation in contemporary cognitive psychology. I any two bibliographic documents are similar. But by the same found myself a student again with ideas for research projects token, any two documents have an infinitude of attributes that popping off each page. This collection contains state-of-the-art are different (e.g.: The third word in this one is ''dog,'' the reviews by leading cognitive researchers of mental representathird word in the other one is ''cat,'' this one has 67 sentences, tion, concepts, and categories. Translated into the coinage of that one has 68 sentences, this one is online and that one is information science, these are the fundamental concerns of inpaper, etc.) We conclude that any two bibliographic documents dexing, online searching, search strategies, relevance research, are dissimilar. Obviously, there is some work to be done here etc., summarized and presented with the supporting empirical discovering how and why two pieces of text are perceived to research. In effect, this is the hard science of a user-centered be similar. information science.
Hahn and Chater present avenues out of this circle by re-The volume falls into three parts, the first of which has the viewing the strategies of ''respects'' of similarity, surface and most immediate rapport with information science. Here one deep similarity. No doubt, we information scientists will trace finds Evan Heit's ''Knowledge and concept learning,'' Ulrike these same peregrinations when we finally comes to grips with Hahn and Nick Chater's ''Concepts and similarity,'' Gregory relevance. Among their comments are methods of modeling L. Murphy and Mary E. Lassaline's ''Hierarchical structure in similarity including the spatial model where an n-dimensional concepts and the basic level of categorization,'' James Hampspace is constructed out of the similarity relations among obton's ''Conceptual combination,'' and Linda B. Smith and Lajects. It is akin to building a city map from the distance among rissa K. Samuelson's ''Perceiving and remembering: Category buildings. Building such a document universe has been the goal stability, variability and development.'' The second third of the of information scientists where one could navigate among bibbook covers more technical aspects of cognitive neuroscience, liographic records. Hahn and Chater present devastating argubrain structures, and process models of categorization, while ments of the impossibility of using similarity as the basis for the last third of the book presents some specific models of this n-dimensional domain. Say goodbye to the ''docuverse'' categorization and function learning.
that would permit searchers to navigate in a three-dimensional The articles in the first third of the book make it essential world selecting relevance articles. reading for information scientists who want an overview of the Murphy and Lassaline focus on hierarchical structures. Inforfundamental psychological research on the human activities of mation science is replete with generic trees of descriptors and naming categories, placing objects in categories, creating hierarsubject classifications. An anchoring idea is the basic level of chies, creating multiword descriptors, and so on. These are all categorization. The information science analog would be topical fundamental information science activities. For example, Heit's relevance, i.e., the term that the indexer chose. They review the contribution points out that what a person learns about a new qualities of the basic level of categorization illustrating that it category is greatly influenced by what a person knows about is more than just cue validity. They discuss a number of probother, related categories. This resonates with the theorizing of lems of hierarchical structures, among them the problem of William Cooper (1971), largely forgotten today, that to search transitivity. For example, ''car seat'' is widely viewed as a online is to deduce. Online searching is a process that combines chair, and ''chair'' is widely viewed as a type of furniture, but the searcher's background and deductive mechanisms. Heit few regard car seat as a piece of furniture. How prevalence is comments: ''Perhaps the most fundamental generalization is this problem in the generic trees and subject classifications that that in learning about new categories, people act as if these we use in information science? categories will be consistent with previous knowledge. People
The theme of Smith and Samuelson's contribution is that no stem to act with economy, so that previous knowledge structures two ideas are ever exactly the same, that what is known in any are reused when possible'' (p. 10). From this insight flows the real moment depends on the context. Here is the empirical integration of prior knowledge with new observations. This is foundation for the analysis of online searching moves. They manifested in information science research in both understandexamine category variability-concepts like ''water'' and ing the moves of online searching, and the priming effects of ''mother'' that resist a single cohesive and coherent concept. reading through a set of retrieved records. This is the empirical foundation for the analysis of the act of Heit considers many issues including the exemplar model indexing, a practical art that has lacked a theoretical developof categorization, which suggests that categorization is based ment (Farrow, 1991). on similarity to category exemplars. Are ERIC major descriptors There are many more connections to be made. This book would exemplars of a category? Are the records retrieved by ERIC make an excellent textbook in a course on the cognitive aspects major descriptors exemplars of their category? These are uninof information use. I have added it to my personal collection. vestigated research questions. Heit also considers categorization based on feature interpretation effects. An information science analog would be the work of Park (1994), who interviewed Terrence A. Brooks searchers about the most compelling part of bibliographic records. This book acts as a research source book for these basic