Duplicate entries versus see crossreferences

Transcription

1 Duplicate entries versus see crossreferences in back-of-book indexes Virgil Diodato When there is a choice, does a back-of-book indexer use a duplicate entry or a see reference? Guidelines suggest that it is preferable to use the duplicate entry if it would not add to the length or complexity of the index. The author studied 1,100 see references in 202 back-of-book indexes and concluded that 22% of the see references should have been replaced by duplicate entries. Failure to select a duplicate entry instead of a see reference occurs most frequently in science and technology books and in indexes with no subheadings. Introduction When should a back-of-book indexer use a duplicate entry (sometimes called a double posting) instead of a see reference? For example, if the author of a book writes about cats, the indexer may start with an entry like this: cats 15, 21-3, 42, 77-8, , 230. To help readers who look in the index under 'felines,' there is a choice. The indexer can provide a see reference: Example 1: cats 15, 21-3, 42, 77-8, , 230 felines see cats Or, the indexer could use duplicate entries: Example 2: cats 15, 21-3, 42, 77-8, , 230 felines 15, 21-3, 42, 77-8, , 230 Which is the better choice? Two advantages of the see reference in Example 1 are: (1) it may take up less space than a duplicate entry; (2) it implies to the reader that the word 'cats' is more likely to appear on the given pages than the word 'felines.' A disadvantage of the see reference is that the user has to turn from the 'f' section to the 'c' section of the index. The advantage of the duplicate entries in Example 2 is that the same page references (also known as loca tors) are listed with 'cats' as well as with 'felines.' The major disadvantage of a duplicate entry is that it may add to the size and complexity of the index. Guidelines The British indexing standard BS3700:1988 and draft 4.1 of the proposed revised American indexing standard ANSI/NISO Z X discuss duplicate entries. In back-of-book indexes:... a 'see' cross-reference should be replaced by an entry if there are few location references or if the entry does not occupy more lines of type than would the cross-reference.1... a duplicate entry under the alternative head ing should be made if a 'see reference' would occupy more space...? One of Wellisch's guidelines is: 'A double entry is the better choice if a heading has no more than three locators or only one subheading....'3 Taken together, these guidelines give us what I call The Guideline: Replace a see reference with a duplicate entry if the duplicate entry would: (1) contain the same number or fewer lines than the see reference; and (2) have three or fewer page references; and (3) have one or no subheading. Unless one or more of these three situations occurs, use a see reference instead of a duplicate entry. Previous research Gratch, Settel, and Atherton;4 Bishop, Liddy, and Settel;5 Liddy, Bishop, and Settel;6 Diodato and Gandt;7 Diodato;8 and Nwodo and Otokunefor9 together have studied cross-references in about 1,000 back-of-book indexes. Three of these studies also examined duplicate entries: Gratch, Settel, and Atherton; Bishop, Liddy, and Settel; and Liddy, Bishop, and Settel. They all used the terminology 'multiple entry,' which includes duplicate entries as well as any instance of the same textual matter being indexed under three or more alternative entries. Gratch, Settel, and Atherton found that 29% of humanities books and 59% of social sciences books each had at least one multiple entry in their index. Also, 'in many cases the use of multiple entries and The Indexer Vol. 19 No. 2 October

2 DUPLICATE ENTRIES VERSUS SEE CROSS-REFERENCES IN BACK-OF-BOOK INDEXES cross-references was found in the same index.'10 Bishop, Liddy, and Settel found at least one set of multiple entries in 33% of indexes. Multiple entries were most likely to occur in science and technology books, the field in which they found the lowest per centage of books with cross-references.11 Liddy, Bishop, and Settel discovered that indexes created by professional indexers were more likely to contain multiple entries than indexes made by authors of the books.12 Do users have a preference for either duplicate entries or see references? I am completing a small sur vey of librarians and college professors on several design issues in back-of-book indexes. They were shown examples of see references and duplicate entries like Examples 1 and 2 above. Librarians and profes sors disagree. According to the data collected so far, most librarians (58% 83 of 143) prefer see references. Most professors (58% 53 of 92) prefer duplicate entries.13 Methodology of this study The objective of the current research was to deter mine whether indexers, wittingly or not, follow The Guideline for choosing between a duplicate entry and a see reference. I searched 344 back-of-book subject indexes exam ined two years earlier.14 The books had come from a sample listed in the 1988 cumulation of the American book publishing records In my return to these books recently, I found 202 of them in our university library, and so they became the sources for this study. In each of the 202 back-of-book indexes I looked for duplicate entries and see references. To make the study feasible, I examined parts only of each index. Method of testing The Guideline To test whether the choice of a see reference obeyed The Guideline given above, in each index I examined: the first ten see references; if there were less than ten, I examined all the see references; each entry (called a 'target' entry) to which the ten see references referred. I compared the size of each see reference (number of lines) to the size and complexity of its target entry (number of lines, page references, and subheadings). I assumed that a duplicate entry would take up the same number of lines, page references, and subhead ings as the target entry (or part of the target entry) to which the see reference pointed. Consider these two examples: Example 3: See reference: ABM (antiballistic missile) see Nixon, Richard Target entry: Nixon, Richard ABM program and 111 Example 4: See reference: CBA software see cost benefit analysis Target entry: cost benefit analysis case study 102 software Apple 30 IBM 32-3 mainframe 33 Example 3 fails to follow The Guideline. Its see ref erence contains two lines of type. The target entry contains two lines, one page reference, and one sub heading. If we assume a duplicate entry would have the same characteristics as the target entry, then all three points of The Guideline are satisfied. The see reference should have been replaced by a duplicate entry, perhaps one like this: ABM (antiballistic missile) Nixon, Richard and 111 In Example 4, the two-line see reference points specifically to the bottom four lines of the target entry. The 'software' subheading includes three page refer ences and three subheadings of its own. Therefore, I assumed that a duplicate entry would take up four lines and have three page references and three sub headings. This would violate points 1 and 3 of The Guideline. Using the see reference in Example 4 obeys The Guideline. Method of counting the frequency of duplicate entries Each time an indexer uses a duplicate entry, a choice, conscious or not, has been made not to use a see reference. How often does this happen? Answering that turned out to be an almost impossible job. To find all the duplicate entries in an index, one would have to compare each entry with every other entry. Instead, in each of the 202 indexes I examined: the first 25 entries; all the entries that had any of the pages between 50 and 59, inclusive, as page references. For each such entry, I looked through the index for a duplicate entry listed under some reordering of the words in the original entry. For example here is a pair of duplicate entries I found: Example 5: Biblian, weavers of 130-3, 142-8, weavers of Biblian 130-3, 142-8, The Indexer Vol. 19 No. 2 October 1994

3 DUPLICATE ENTRIES VERSUS SEE CROSS-REFERENCES IN BACK.-OF-BOOK INDEXES My search for duplicate entries was limited. To make it feasible, I looked only under words found in the original entry. In the above example, I did not look under synonyms for 'weavers' nor for other names of 'Biblian'. Also, for feasibility, I stopped counting when I determined that an index had at least ten pairs of duplicate entries. These methods of look ing for duplicate entries led me to undercount the actual number of duplicate entries. Results: general information The 202 back-of-book indexes contained 1,100 see references. According to their Library of Congress (LC) classifications, most of the books (51%) were in the social sciences; 28% in science and technology; 20% in arts and humanities; and 1% in A and Z classes of LC. Because only two indexes were in the A and Z classes, they will not be mentioned in the rest of this paper. See Table 1. TABLE 1 Number and characteristics of indexes See references Too long/complex for duplicate entry Should replace with duplicate entry Results: adherence to the guideline Seventy-two per cent (790 of 1,100) of the see refer ences indeed should have been see references rather than duplicate entries, according to The Guideline. Each of the 790 had a target entry that was too long or complex for transforming the see reference into a duplicate entry. As explained below, this list of good see references was -increased from 790 to 859 out of 1,100. Results: frequency of indexes with duplicate entries Fifty-eight per cent of the indexes (117 of 202) had at least one duplicate entry each. Indexes in science and technology were most likely to have duplicate entries, as 70% of these had one or more; 63% of social sciences indexes and only 32% of arts and humanities indexes had duplicate entries. See Table 1. Failure analysis According to The Guideline, 28% (310 of 1,100) of the see references should not have been see references. They should have been replaced by duplicate entries, because doing so would not have increased the size or complexity of the indexes. However, for 69 of these it was understandable that they were not duplicate entries. Each of these 69 see references was like the see reference and target entry in Example 6: Example 6: See reference: Apple CBA software see cost benefit analysis Target entry: cost benefit analysis case study 102 Proposed duplicate entry: software Apple 30 IBM 32-3 mainframe 33 Apple software CBA (cost benefit analysis) 30 One can imagine that the index could be more com plex if the indexer had replaced the see reference with the proposed duplicate entry. The indexer who creates this duplicate entry may also make duplicate entries for the other subheadings under 'software.' Wellisch supports the use of such duplicate entries under cer tain conditions.16 Nevertheless, I gave the benefit of the doubt to the 69 cases like this and believed that the indexers retained the see reference for fear of mak ing the indexes too complex. That leaves = 241 (or 22% of 1,100) see ref erences for which I have no support. These 241 refer ences may have failed the index users. Failure analysis: a caveat To pass final judgement on each of the 241 cases would require finding out why the indexer used the see reference. I did not investigate these reasons. Surely some of the indexers would give convincing arguments for not replacing the see reference with a duplicate entry. Failure analysis: subject field The failure to replace a see reference with a dupli cate entry was much less likely in the social sciences indexes than in any other field. Indexers in the social sciences books that I examined did a good job: only 9% (52 of 598) of see references in the social sciences should have been replaced by duplicate entries. See Table 2. It was not surprising that one-third of the arts and humanities see references were failures. There is some justification for the lack of duplicate entries in many of these subject indexes that included names of people, The Indexer Vol. 19 No. 2 October

4 DUPLICATE ENTRIES VERSUS SEE CROSS-REFERENCES IN BACK-OF-BOOK. INDEXES TABLE see references that should have been replaced by duplicate entries. Subject field: Arts/humanities Science/technology Social sciences 59 of 177 (33%) 130 of 305 (43%) 52 of 598 (9%) Subheading format: Set-out Run-on Hybrid No subheadings 124 of 569 (22%) 80 of 378 (21%) 11 of 89 (12%) 26 of 64 (41%) Note: Each per cent is the number of see references that should have been replaced by duplicate entries compared to all see references in the given category. places, and works. The indexers may have felt that names do not deserve the same kind of treatment that a subject does. I do not always agree with that, how ever. Instead of sending the reader from 'Hueffer, James' to two page-references under 'van Hueffer, James,' why not duplicate the page-references under both versions of the name? Failure of so many see references (43%) in science and technology was surprising. How could there be so much failure to use duplicate entries in the field whose indexes commonly contain duplicate entries? Why was it so common to give the user an instruction like, 'energy of motion see kinetic energy' when there was only one page-reference in the target entry? The answer may be that such a see reference makes the preferred terminology clear to the user. Precise termi nology is a hallmark of science and technology; index ers may have felt that some duplicate entries teach the user improper terminology. Failure analysis: subheading format Many (41%) of the see references examined in index es that had no subheadings should have been replaced by duplicate entries. Having no subheadings may be a failure in itself, and so it was not surprising that the producers of such poor indexes also had few duplicate entries. Only a few (12%) of the see references in indexes that employed hybrid subheadings should have been replaced. Hybrid subheadings combine the features of set-out (or line by line) subheadings with those of runon (or paragraph style) subheadings. One would expect that indexers who can handle such a fine point as the hybrid subheading would be careful enough to make good decisions about using duplicate entries and see references. Twenty-two per cent of the see references in indexes that used only set-out subheadings and 21% of see ref erences in indexes using only run-on subheadings should have been duplicate entries. Perhaps this means that these indexes, as a group, are typical indexes and that we can expect about one-fifth of see references to be inappropriate in a given index. See Table 2. Conclusion We should require ourselves as indexers to be aware of all the tools available to us and when to select the appropriate tool. See references and duplicate entries both can help index users, but sometimes one of these tools is more helpful than the other. The guidelines noted above can help us decide when to use which tool, but the choice depends on each individual situa tion: the book being indexed, its audience, author, indexer, editor, publisher, subject field, the indexer's experience in periodical indexing (in which duplicate entries may be dangerous as the indexer may forget about them in subsequent issues of the index), and so on. In this study it seemed as if the indexer made the correct choice between see reference and duplicate entry 78% of the time and failed in 22% of the deci sions. But we can never be sure about this without knowing more about the ways in which people use indexes and more about how indexers make their design choices. Acknowledgments The author thanks: Georgianna Henry and June Woodbridge for their assistance in analyzing the data; James D. Anderson for providing information about the proposed American indexing guidelines; and Hazel Bell for her com ments on an earlier draft of this paper. References 1. British Standards Organization. British standard recom mendations for preparing indexes to books, periodicals and other documents. Standard BS3700:1988. London: British Standards Institution, 1988, National Information Standards Organization. Committee to Revise ANSI/NISO Z39.4 American National Standard Guidelines for Indexes. Proposed American national stan dard guidelines for indexes and related information retrieval devices. Standard ANSI/NISO Z X, draft 4.1. Bethesda, MD: National Information Standards Organization, 15 September 1993, section Wellisch, Hans H. Indexing from A to Z. Bronx (NY): Wilson, 1991, Gratch, Bonnie; Settel, Barbara; Atherton, Pauline. Characteristics of book indexes for subject retrieval in the humanities and social sciences. The Indexer 11(1), Apr. 1978, Bishop, Ann P.; Liddy, Elizabeth D.; Settel, Barbara. Index quality study, part i: quantitative description of back-of-the-book indexes. In Indexing tradition and inno vation: proceedings of the 22nd annual conference of the American Society of Indexers. Port Aransas (TX): American Society of Indexers, 1991, Liddy, Elizabeth D.; Bishop, Ann P.; Settel, Barbara. Index quality study, part ii: publishers' survey and quali tative assessment. In Indexing tradition and innovation: proceedings of the 22nd annual conference of the American Society of Indexers. Port Aransas (TX): American Society oflndexers, 1991, Diodato, Virgil and Gandt, Gretchen. Back of book 86 The Indexer Vol. 19 No. 2 October 1994

5 indexes and the characteristics of author and nonauthor indexing: report of an exploratory study. Journal of the American Society for Information Science 42(5), June 1991, Diodato, Virgil. Cross-references in back-of-book indexes. The Indexer 17(3), Apr. 1991, Nwodo, C. O. and Otokunefor, H. C. Indexing of books in Nigeria: some observations. The Indexer 16(4), Oct. 1989, Gratch, Settel, and Atherton, 17, Bishop, Liddy, Settel, Liddy, Bishop, Settel, Diodato, Virgil. [Preferences of index users.] Unpublished raw data Diodato, Virgil. Cross-references, American book publishing record cumulative New York: Bowker, Wellisch, 104. Virgil Diodato is Associate Professor at the University of Wisconsin-Milwaukee School of Library and Information Science. Bibliographical standardization The Times of 30 May, under the heading, 'What's in a catalog', reflects on the forthcoming amalgamation of the catalog(ue)s of the British Library and the Library of Congress, which between them hold more than 100 million works by more than a million authors. Many names thus recur multiply: 36 John Smiths have their works held in BL, 64 in LC; and 'when authors' names are transliterated from non-roman sources, the alphabet soup becomes thicker'. The two libraries hold 36 variant spellings of Tchaikovsky, while Mao Tsetung occurs in at least 16 forms.' The Anglo-American Authority file will resolve these difficulties by giving each author in the two libraries a unique identity and a standard name. The Times greets this initiative as 'the first successful attempt for four centuries to standardise the spelling of the two old English-speaking countries', seeing the orthographic independence hitherto claimed by each as 'shibboleths of linguistic nationalism'. Now, though, 'information technology is not bound by such petty nationalism... computerese is one jargon where American spelling sets the international standard'. Nevertheless, The Times deplores the prospect of 'that Cloud-Cuckoo-Land where everybody speaks and writes English according to the same rules', declaring, 'The only sort of English that can be stan dardized is trivial', and rejoicing that 'while the British Museum is blowing its own trumpet over the joint cat alogue, the Library of Congress still prefers already to blow its own horn over the joint catalog'. Benchmarking for indexing The National Federation of Abstracting and Infor mation Services (NFAIS), based in Philadelphia, has undertaken a Benchmarking Project, established by the Quality Task Force of its Information Policy and Copyright Committee. The processes and measures of indexing were taken as the benchmark, 'because measures are not known, processes vary, and a great deal of new technology is being used in indexing processes'. The project is reported by Pam Weaver in the NFAIS Newsletter, 36 (6) June Questionnaires were sent to information providers and users: 70 members of NFAIS and 1,318 of the Special Libraries Association, to determine what they consider to be important indexing concepts and which information providers would be identified as most respected in the industry for their indexing quality. 36% of the NFAIS questionnaires and 14% of SLA's were returned. Both groups rated accuracy/relevance, completeness, and consistency as the three most important criteria of good/quality indexing. The NFAIS members indexed most frequently by major terms, subject categories and author names, fol lowed by minor terms, descriptors, bibliographic infor mation and identifiers. SLA members searched most often by free-text, index terms/descriptors, title words and author names. Other fields included in the responses were date/publication, geographic locations, corporate source, and document types. Other charac teristics of good-quality indexing cited were: up-todate vocabulary; indexing appropriate to the audience of the product; distinguishing between major and minor descriptors; use of a good thesaurus, crossreferences, and hierarchical structure; knowledgeability; subject specialism; and quality control on spelling, typographical errors, etc. The most frequently cited products and their providers were: MEDLINE//m/e,v Medicus National Library of Medicine; PTS PROMT Information Access Company; ABI/INFORM UMI; Chemical Abstracts Chemical Abstracts Service/American Chemical Society; and Compendex Plus Engineering Information (Ei). Indexes on CD-ROMs Palmer's Index to The Times, covering the period 1790 to 1850, taken from 244 quarterly volumes, and two segments of Periodical Contents Index, a bibliographic database containing so far the complete title page of every issue of 474 humanities journals published in Europe and the US from their beginnings to 1960, have both now been issued by Chadwyck-Healey on CD-ROM. The Indexer Vol. 19 No. 2 October