116 Journal of Library Automation Vol. 14/2 June 1981 tions only. They do not list the individual works that may be contained in publica- tions. If an analytic catalog were to be built into a computerized system at some time in the future , the structure code would be a great help in the redesign, be- cause it makes it easy to spot items that need analytics, namely those that contain embedded works, or codes 2, 4, 5, 6, 8, 9, 10, 11, and 13. A searcher working with such an analyt- ic catalog could use the code to limit out- put to manageable stages-first all items of type c, for example; then broadening the search to include those of type d; and so forth, until enough relevant material has been found. The structure code would also be useful in the displayed output. If codes 5 or 8 appeared together with a bibliographic de- scription on the screen, this would tell the catalog user that the item retrieved is a set of many separately titled documents. A complete list of those titles can then be displayed to help the searcher decide which of the documents are relevant for him. In the card catalog this is done by means of contents notes . Not all libraries go to the trouble of making contents notes, though, and not all contents notes are complete and rtliable . The structure code would ensure consistency and com- pleteness of contents information at all times. Codes 10 and 13 in a search out- put, analogously, would tell the user that the item is a serial with individual issue ti- tles. There is no mechanism in the con- temporary card catalog to inform readers of those titles. Codes 4 and 7 would tell that the document is part of a finite set, and so forth. It has been the general ex- perience of database designers that a rec- ord cannot have too many searchable ele- ments built into its format. No sooner is one approach abandoned "because nobody needs it," than someone arrives on the scene with just that requirement. It can be anticipated, then, that once the struc- ture code is part of the standard record format, catalog users will find many other ways to work the code into search strategies. It can also be anticipated that the pro- posed structure code, by adding a factor of selectivity, will help catalogers because it strengthens the authority-control aspect of machine-readable catalog files. If two pub- lications bear identical titles, for example, and one is of structure 1, the other of structure 6, then it is clear that they can- not possibly be the same items. However, if they are of structures 1 and 7, respec- tively, extra care must be taken in catalog- ing, for they could be different versions of the same work. Determination of the structure of an item is a by-product of cataloging, for no librarian can catalog a book unless he understands what the structure of that book is-one or more works, one or more documents per item, open or closed set, and so forth . It would therefore be very cheap at cataloging time to document the already-performed structure analysis and express this structure in the form of a code. REFERENCES l. Herbert H. Hoffman, Descriptive Cataloging in a New Light: Polemical Chapters for Li- brarians (Newport Beach, Calif.: Headway Publications, 1976), p.43. Revisions to Contributed Cataloging in a Cooperative Cataloging Database Judith HUDSON: University Libraries , State University of New York at Albany. INTRODUCTION OCLC is the largest bibliographic utility in the United States. One of its greatest assets is its computerized database of standardized cataloging information . The database, which is built on the principle of shared cataloging, consists of cataloging records input from Library of Congress MARC tapes and records contributed by member libraries. OCLC STANDARDS ln. order to provide records contributed by member libraries that are as usable as those input from MARC tapes, it is im- perative that the records meet the stan- dards set by OCLC and that the catalog- ing and formatting of the records be free of errors. Member libraries are requested to follow the nationally accepted cataloging code (Anglo-American Cataloging Rules, North American Text, 1 • 2 for records input before December 12, 1980, and Anglo- American Cataloguing Rules, Second Edition, 3 for records input later), the Li- brary of Congress' application of the cata- loging code, and the various MARC for- mats in preparing records to be input. 4 • 5 The cataloging rules dictate what kind of bibliographic information should be in- cluded in the cataloging records, a pre- scribed system of punctuation that iden- tifies the various fields of the cataloging record (International Standard Biblio- graphic Description, ISBD), which access points should be provided, and what form the entries should take. The MARC for- mats provide a standardized method of identifying the various fields and subfields in a cataloging record and, through the use of indicators, information necessary to make the record easily manipulated by computers. In addition, fixed fields pro- vide coded information about the catalog- ing records. The form of main, added, and series en- tries can be verified in the National Union Catalog to ensure that member libraries are following the Library of Congress' ap- plication of the cataloging code . By the same token, subject entries can be verified in the appropriate subject heading list (e.g., Library of Congress subject head- ings, Sears subject headings, etc.). A STUDY OF OCLC MEMBER CATALOGING A major problem with the use of contributed cataloging is the amount of re- vision needed to bring the records up to the standards described above. In 1975, a study of the quality of a group of member- contributed catalog records was conducted by C. C . Ryans. 6 The first 700 mono- graphic records input into OCLC after September 1, 1975, to which Kent State University attached its holdings were examined. 7 The analysis included changes in or additions to main, added, or series Communications 117 entries, changes in descriptive cataloging, and changes in or additions to subject headings . The study dealt only with the revision of cataloging; revision of the for- matting of records was not noted. The Kent State study found that 393 revisions were necessary to 283 records. The re- maining 417 records were considered to be acceptable, i.e., they adhered to AACR and ISBD rules and to the OCLC stan- dards for input cataloging. RECENT DEVELOPMENTS RELATING TO QUALITY CONTROL Since these records were studied, the Internetwork Quality Control Council was formed in 1977 by the OCLC Board of Trustees. 8 Its primary purpose is to iden- tify problem areas regarding quality con- trol and distribute information to networks concerning problems and solutions. Its role is to promote quality control through education and by monitoring the imple- mentation of standards. In addition, OCLC' s documentation has steadily improved. The recent publication of the Books Format9 and the recent re- vision of the cataloging manual10 provide clear and specific information on OCLC' s formatting requirements. With these developments in mind, it would seem likely that the quality of the contributed cataloging has improved since 1975. In order to test this assumption, a number of cataloging records were ana- lyzed in an effort to replicate the Kent State study. The analysis of these records differed from the earlier study in that dif- ferences in the treatment of series were not noted because one library's treatment of series can reasonably be expected to differ from that of another . METHODOLOGY The records included in this study con- sist of 1,017 monographic catalog records to which the State University of New York at Albany (SUNYA) Library added its holding symbol during an eight-month period from November 1979 to July 1980. The records included only those that were entered into the OCLC database after 1976. Cataloging revisions that were noted 118 journal of Library Automation Vol. 14/2 Jun e 1981 consisted of changes in main and added entries to make them consistent with Li- brary of Congress form of entry, and the inclusion of other added entries that were deemed necessary to provide adequate ac- cess to the material. In addition, correc- tions or additions to the imprint and the collation· were noted, as were typograph- _ ical errors in all fields . Subject headings that were changed to make them consis- tent with Library of Congress subject headings and subject headings and/or sub- divisions added to provide better subject access to the material were also noted . ANALYSIS OF CATALOGING Cataloging revisions were required for 43 percent of the 1,017 records examined (596 changes or additions were made to 437 records). Changes or additions to sub- ject headings were made to 22.4 percent of all the records in the SUNYA sample, and represented the most common revi- sion . Changes in descriptive cataloging were made to 20 percent of the records, and changes or additions to main or added entries were made to approximately 16 percent of the records. Table 1 compares the results of this analysis with the findings of the e arlier study . It should be emphasized that the two studies are not exactly comparable be- cause the Kent State study included differ- ences in the treatment of series, while this study noted only typographical errors in series statements. The findings of this analysis do not bear out the hypothesis that the quality of member-contributed cataloging has im- proved since 1975. The overall percentage of records requiring cataloging revision is similar in both the Kent State and the SUNYA samples . The percentage of changes made in the various areas of the cataloging records was similar, with the exception of added entries and subject headings . In the SUNYA sample , more re- visions and additions were made to these two areas. This difference between the two samples may reflect variation in the cataloging policies of the two libraries rather than the presence or absence of more errors in member-contributed cata- log records . ANALYSIS OF OCLC REPORTABLE ERRORS AND ADDITIONS In the fall of 1979, OCLC distributed its revised cataloging manual, which includes a chapter dealing with quality control. 11 The chapter delineates the errors and changes that are to be reported to OCLC for correction or addition . The cataloging records examined in this study were also analyzed with these criteria in mind. This analysis (table 2) revealed that 661 reportable errors or changes were found on 486 records (47.8 percent of all the rec- ords). Reportable errors or changes in- cluded formatting errors or omissions such as incorrect assignment of tags, incorrect or missing indicators, subfield codes or fixed fields, and errors affecting retrieval or card printing . Other types of errors in- Table 1 . Comparison of Two Studies of Cataloging Revision Area Needing Kent State Sample* SUNYA Sample Revision or Addition Number Percentage Number Percentage Main Entry 44 6.2 46 4.5 Title Statement 28 4.0 76 7.5 Edition Statement 4 0.6 2 0.2 Imprint 29 4.4 64 6.3 Collation 111 15.9 58 5.7 Series 55 7.9 3 0.3 Subject Heading 88 12.6 228 22 .4 Added Entries 44 6.2 119 11.7 Total Records in Study 700 100.0 1017 100.0 Records Requiring Revision 283 40.4 437 43.0 Number of Revisions Made 393 596 *Source: Constance C . Ryans, "A Study of Errors Found in Non-MARC Cataloging in a Machine- Assisted System," journal of Library Automation 11 :128 (June 1978). Communications 119 Table 2 . Errors and Additions Reportable to OCLC Number Percentage of Total Records Percentage of Total Errors and Additions 19 6 13 17 59 Errors in Transcription of Data Incorrect Assignment of Tags Incorrect or Missing Subfield Codes Incorrect Assignment of 1st Indicator Incorrect Assignment of 2d Indicator Incorrect Fixed Fields Incorrect ISBD Incorrect Form of Entry (less than LC) Errors Affecting Retrieval or Card Printing Bibliographic Information Missing Addition of Access Points 313 8 87 3 1 135 Total Number of Records Containing Reportable Errors or Additions Total Number of Reportable Errors or Additions 486 661 eluded incorrect or omitted access points (added or subject entries, ISBN, LC card numbers, etc.), errors in transcription of data, incorrect ISBN, and the omission of needed bibliographic information. Approximately 40 percent (408) of the records contained formatting errors, with over 29 percent (300) of the records con- taining incomplete or incorrect fixed fields. The apparent unconcern with fixed fields may stem from a lack of understand- ing of the value of correct fixed-field in- formation. The recent addition of date and type of material as qualifiers in a search of the database is one example of the use of fixed fields. In order to underscore their importance, it might be useful for OCLC to highlight this use of fixed fields and further explain to its members how other fixed fields might be used in online search strategies in the future. Errors in or omission of access points were found in 222 records (21.8 percent). These errors were also noted in the study of cataloging revisions discussed above, as were errors in transcription of data, in ISBD, and in omission of necessary bib- liographic information. SUMMARY OF FINDINGS Although the quality of the SUNYA sample seems equivalent to that of the Kent State sample, an analysis by date of input of the records examined indicates a slight decrease in the percentage of rec- 1.9 0.6 1.3 1.7 5.8 30.8 0.8 8.6 0.3 0.1 13.3 47.8 2.9 0.9 2.0 2.6 8.9 47.4 1.2 13.2 0.5 0.2 20 .4 100.0 ords needing correction for those records input in 1979 and 1980 (table 3). Perhaps this is the beginning of a trend toward more careful cataloging and formatting of records input by members. In summary, 589 of the 1,017 member- contributed records studied were found to require revision. Of these, 486 records contained er.rors or omissions that may be reported to OCLC, and 437 required cata- loging revision. It is discouraging to real- ize that approximately 60 percent of the member records used required revision. Such a high percentage of records needing revision necessitates the review of all member records .used if a library wishes to adhere to OCLC standards for cataloging. This leads to tremendous duplication of effort and negates, in part, the purpose of shared cataloging. Table 3. Yearly Breakdown of Catalog Records Total Records Percentage Year Number Needing Needing of Input of Records Correction Correction 1977 186 115 61.8 1978 332 202 60.8 1979 339 184 54.3 1980 160 88 55.0 INFLUENCES FOR CHANGE The implementation of AACR2 in 1981 provides the impetus for greater adher- ence to standards. Since all catalogers 120 Journal of Library Automation Vol. 14/2 June 1981 have had to learn the new cataloging re- quirements, greater care may be used in the formulation of records by member li- braries. The publication of clear and specific guidelines for reportable errors may help to alleviate the situation in two ways . First, the careful articulation of errors or desirable additions may impel member li- braries to place more emphasis on the quality control of input. Second, member libraries may report more errors, thus allowing OCLC to correct the master rec- ords. A change in the method of correcting errors and the rate at which they are cor- rected might be beneficial. Presently, errors on the master records can only be corrected by OCLC or by the inputting li- brary if it is the only library that has used the record. Such an arrangement is clum- sy and time-consuming. If other member libraries were trained and authorized to correct errors on master records, errors might be corrected as often as they are detected. In the long run, however, the responsi- bility for inputting catalog records that meet the standards for cataloging and for- matting rests with the member libraries. OCLC and the networks must develop methods of encouraging libraries to input records that are correctly formatted and cataloged . One way of alleviating the problem might be to develop training programs conducted by OCLC or by net- work staff that are aimed at those libraries identified as having high error rates. Another approach might be to give public recognition to libraries that contribute cataloging of high quality to the database. One example of this approach is the Pitts- burgh Regional Library Council's Fred Award, which annually honors the library with the lowest error rate in the PRLC network. 12 Through the use of peer pres- sure the member libraries and networks of OCLC can encourage adherence to the standards. In addition, they must continue to insist that OCLC address this annoying, expensive, and seemingly perennial problem. REFERENCES l. Anglo-American Cataloging Rules, North American Text (Chicago: American Library Assn., 1967), 409p. 2. Anglo-American Cataloging Rules, Chapter 6 (Rev. ed.; Chicago: American Library Assn., 1974), 122p. 3. Anglo-American Cataloguing Rules, Second Edition (Chicago: American Library Assn., 1978), 620p. 4. OCLC, Inc . , Cataloging: User Manual (Columbus: OCLC, 1979), 1v. (looseleaf). 5. OCLC Level I and Level K Input Standards (Columbus: Ohio College Library Center, 1977), 1 v. (looseleaf). 6. Constance C. Ryans, "A Study of Errors Found in Non-MARC Cataloging in a Machine-Assisted System," Journal of Li- brary Automation 11:125-32 Oune 1978). 7. Ibid., p . 127. 8. Frederick G. Kilgour, "Establishment of In- ter-Network Quality Control Council" (Un- published document, Ohio College Library Center, 1977), 2p. 9. OCLC, Inc., Books Format (Columbus: OCLC, 1980), 1v. (looseleaf). 10. OCLC, Inc., Cataloging: User Manual, 1v. (looseleaf) . 11. Ibid. 12. "PRLC Peer Council Cites Pittsburgh Theological Seminary Library for High Cataloging Standards," OCLC Newsletter 131:4 (Sept. 1980).