Yee, Martha, Information Technology and Libraries
Editor's Note: The two items that appear here constitute an exchange of ideas relating to two communications published in recent issues of ITAL. The original article prompting this exchange was published by Joan Cherry in volume 17, number 3 (September 1998, available at www.lita. org/ital/1703_cherry.html). In volume 18, number 2 (June 1999, available at www. lita.org/ital/1802_crawford.html), we carried a response to Cherry written by Walt Crawford. That brings us to the two items here: a critique of Crawford's piece by Martha Yee, followed by a reply from Craw-ford.
Response to "Webcats and Checklists: Some Cautionary Notes"
I have some concerns about some assertions made about the IFLA guidelines on OPAC displays by Walt Crawford in his article "Webcats and Checklists: Some Cautionary Notes," in the June 1999 ITAL (vol. 18, no. 2).
First of all, he asserts that Principle 3--Assume large retrievals--"doesn't appear to be based on real-world testing." He briefly mentions some research on RLG's Eureka, asserting that "at least 75 percent of author, title, and subject retrievals yield twenty-five records or less." The findings are again briefly mentioned in Crawford's book, Being Analog, where he states that "If users got any results at all, the results fit on a single screen almost 70 percent of the time." In neither place is the research design more than vaguely described; thus it is difficult to examine the assumptions underlying the research, the research design, any flaws that might have occurred in the research design or statistical analysis, etc. It is not clear, for example, how many of the "users" he studied might have been librarians. It is also not clear whether the "results" discussed above consisted of headings (each with multiple postings) or bibliographic records. Also, it appears (from an e-mail communication from Mr. Crawford) that "subject" was defined as one subject heading, such that a main heading and all of its subdivisions counted as multiple "subjects." The alternative way to define "subject" would be to define it as all of the subject headings that comprise a main heading and all of its subdivisions. The definition can make quite a difference when doing research on the frequency with which users need to deal with large retrievals. I would assert that it is likely that the user's definition of "subject" would often be the latter; for example, "Racism" and "Racism--History" would be considered to be the same "subject" by most users, I suspect. The user who wants to select from everything available on racism should be shown the histories and allowed to decide if he or she is interested in that aspect of the subject or not.
It is well known that about 80 percent of authors in any given catalog publish only one book. However, it is probable that those authors are not the most likely to be sought by users, or they would have been encouraged to publish more books ...
Crawford also ignores the fact that there has been considerable research that indicates users frequently have problems due to large retrievals. Average retrievals of 91 to 247 bibliographic records and of 350 headings are reported in the published research. This research is described and cited in detail in the book by Sara Shatford Layne and myself, Improving Online Public Access Catalogs, published by the American Library Association in 1998; see p. 89 and p. 166-67. Problems presented by poorly displayed large retrievals are likely to get worse, as software is designed to let users retrieve records from multiple OPACs over the Internet. (See, for example, the prototype Z39.50-based Bradford OPAC2 (BOPAC2) Project, described at www.bopac2.comp. brad.ac.uk/~bopac2/report/.)
Secondly, he asserts that "The recommendations absolutely call for unlabeled displays as choices, and are lukewarm about labeled displays. The assertion that users `recognize the information when they see it' goes against almost all catalog research. …