lib-MOCS-KMC364-20131012122710 292 Journal of Library Automation Vol. 14/4 December 1981 We need a format which is consistent, easily maintainable without being uncontrollably disruptive, and responsive to changing needs which are likely to accelerate as we gain experience with online systems. Rather than recommending or support- ing the implementation of specific changes to the MARC format, it is essential that the library community begin to establish the framework and benchmarks necessary to maintain the MARC formats over the long term as well as to guide short-term consider- ations. ARL and others can play an impor- tant role in undertaking and encouraging a broader approach to this pressing problem. Such an approach will not only reduce the risk of decision making, but will also assist in the development of the cost/benefit data needed to enhance consideration of format changes. REFERENCES l. D. Kaye Capen, Simplification of the MARC Format: Feasibility, Benefits, Disadvantages, Consequences (Washington, D.C.: Associa- tion of Research Libraries, 1981), 22p. 2. "Principles of MARC Format Content Desig- nation," draft (Washington, D.C.: Library of Congress, 1981), 66p. 3. IchikoT. Morita and D. Kaye Capen, "A Cost Analysis of the Ohio College Library Center On-Line Shared Cataloging System in the Ohio State University Libraries," Library Re- sources & Technical Services 21:286-302 (Summer 1977). 4. Council on Library Resources Bibliographic Interchange Committee, Bibliographic Inter- change Report, no.1 (Washington, D.C.: The Council, 1981). Comparing Fiche and Film: A Test of Speed Terence CROWLEY: Division of Library Sci- ence, San Jose State University, San Jose, Cal- ifornia. INTRODUCTION For more than a decade librarians have been responding to budget pressures by al- tering the format of their library catalogs from labor-intensive card formats to computer-produced book and micro- formats. Studies at Bath, 1 Toronto, 2 Texas, 3 Eugene, 4 Los Angeles, 5 and Berkeley, 6 have compared the forms of catalogs in a variety of ways ranging from broad-scale user sur- veys to circumscribed estimates of the speed of searching and the incidence of queuing. The American Library Association pub- lished a state-of-the-art reporf as well as a guide to commercial computer-output mi- crofilm (COM) catalogs pragmatically sub- titled How to Choose; When to Buy. 8 In general, COM catalogs are shown to be more economical and faster to produce and to keep current, to require less space, and to be suitable for distribution to multi- ple locations. Primary disadvantages cited are hardware malfunctions, increased need for patron instruction, user resistance (par- ticularly due to eyestrain), and some ma- chine queuing. The most common types of library COM catalogs today are motorized reel microfilm and microfiche, each with advantages and disadvantages. Microfilm offers file- sequence integrity and thus is less subject to user abuse, i.e., theft, misfiling, and dam- age; in motorized readers with "captive" reels it is said to be easier to use. Disadvan- tages include substantially greater initial cost for motorized readers; limits on the ca- pacity of captive reels necessitating multi- ple units for large files; inexact indexing in the most widespread commercial reader, and eyestrain resulting from high speed film movement. Microfiche offers a more nearly random retrieval, much less expensive and more versatile readers, and unlimited file size. Conversely, the file integrity of fiche is lower and the need for patron assistance in use of machines is said to be greater than for self-contained motorized film readers. THE PROBLEM One of the important considerations not fully researched is that of speed of search- ing. The Toronto study included a self- timed "look-up" test of thirty-two items "not in alphabetical order" given to thirty- six volunteers, of whom thirty finished the test. The researchers found the results "in- conclusive" but noted that seven of the ten librarians found film searching the fastest method. "Average" time reported for searching in card catalogs was 37.3 min- -- utes, in film catalogs 41.6 minutes, and for fiche catalogs 4I. 7 minutes. A reanalysis of the original data shows a stronger advan- tage of fiche over film (45.3 minutes versus 51.7 minutes) when all times except dupli- cates are totaled, but that difference is al- most entirely due to one extreme score (203 minutes). 9 The Berkeley report of fiche/film compa- rability addressed the issue of retrieval speed directly. By constructing a series of look-up tests composed of items selected from a large public library COM catalog, the researchers were able to compare mi- crofiche and microfilm formats while hold- ing other variables constant. In one test in- volving thirty-six paid users and 252 trials, microfilm was determined to be faster by 7.6 percent (±2.5 percent). In a second test, forty volunteer users were timed in 240 trials and the advantage of film over fiche dropped to 5. 7 percent ( ± 2.5 percent) .1° Although rigorous in design and execution, the Berkeley experimenters used in their look-up tests questions that naive users might misinterpret, e.g., "You want a book about Paul Robeson, written by Eloise Greenfield. Find the listing and give the call number"; and some which could be confus- ing, e.g., "Does the library have any joke books? If so, give the call number for one. "11 Such questions potentially pose an element of uncertainty for subjects: Should I look under Robeson or Greenfield? under joke books or humor? In addition, questions were selected by "browsing the file for tar- get items," a procedure which could result in an uneven distribution of items which in turn could bias the results. Since the num- ber of observations is relatively large the reliability of the results is not questioned; the validity may be. The study reported here was executed by a class in research methods taught by the author during the same time as the Berkeley study; we used the same two formats of the same catalog, and attempted to answer the same question: Using the best available equipment, which microformat is faster to search? ASSUMPTIONS We assumed (I) the two forms of the cat- alog were identical; (2) the quality of the image was not significantly different; (3) a Communications 293 search for items selected randomly from the file and arranged randomly was a fair test of retrieval speed; and ( 4) graduate students in library science were reasonably repre- sentative users for a test of speed. METHODOLOGY We used a dictionary catalog from a pub- lic library system with 436, 79I entries, of which 5,63I were author, Ill,l58 were ti- tle or added entries, and 320,002 were sub- ject entries. Using a random number table, we selected from the catalog I6 entries which were reproduced and randomly ar- ranged to form the test. Of the I6 items, 3 were author entries, 8 were title or added entries, 5 were subject entries. The se- quence, which presumably would affect the speed of retrieval more in the film format because of the necessity to scroll from one letter to another, wasACWNS KCB WM H L P PAL. The test was then administered to thirty-seven volunteer graduate students randomly assigned to a Micro-Design 4020 fiche reader or an Information Design ROM 3 film reader. The two readers were located in the same room. The 86 fiche were held and displayed by a Ring King binder. All times were measured by a stopwatch. Questionnaires administered before and af- ter the test established that the two groups did not differ significantly in age or in self- perceived mechanical ability. Of the film users, 64 percent used micro-formats "occa- sionally" or "frequently" compared with 35 percent of the fiche users. Of the total group, 73 percent wore glasses and 62 per- cent reported prior physical problems with both film and fiche readers used before the test. RESULTS Table 1 shows that the mean speed of the film users was I6. 7 minutes, significantly faster than the 25.3 minutes recorded by the fiche users; the range of speed for the film users was less than V3 that of the fiche users. Even the slowest film user was faster than 70 percent of the fiche users. However, the fastest fiche user was faster than 70 percent of the film users. The range of fiche scores is more than 3 times that of the film scores (Figure I). The standard statistical test shows the difference of means to be signifi- cant at the .Oilevel. 294 journal of Library Automation Vol. 14/4 December 1981 Table I. Speed of Retrieval (in Minutes) Format Low Microfilm (n = 17) 12.3 Microfiche(c = 20) 14.6 t = 4.8,p< .01 DISCUSSION Searching motorized microfilm appears to be significantly faster than searching mi- crofiche, on the average, for relatively inex- perienced users. Even the slowest time on the film was faster than most fiche times. The wide range of fiche scores suggests the possibility that frequent users could im- prove their searching times; very experi- enced users may be able to search fiche faster than film. • Because of the relatively small numbers of subjects and observations •The author, an experienced fiche user, was timed at 11.6 minutes; this was the fastest time recorded by either fiche or film users.