College and Research Libraries Letters To the Editor: Jessie Carney Smith, in her excellent ar- ticle "Special Collections of Black Litera- ture in the Traditionally Black College" (C&RL, September 1974), neglects to mention one important fact. That is her own role in making available to librarians and researchers the catalog of the distin- guished Negro Collection at Fisk where she is University Librarian. With the participation of Mrs. Smith and her staff, G. K. Hall and Co. is currently in- volved in publishing that card catalog in book form. It thereby joins the published catalogs of the Moorland Collection at Howard and the Schomburg Collection of the New York Public Library, both of which she describes (as well as other black collections not included in her study) and which are also published by G. K. Hall. "Overdue Policies" Richard Newman G. K. Hall & Co. Boston, Massachusetts Four letters have been received com- menting on the article "Overdue Policies: A Comparison of Alternatives" by ]an Baaske, Don L. Tolliver, and Judy Wester- berg which appeared in the September 197 4 issue of this journal ( p.354-59 ). Pertinent extracts from these letters are presented below with a response by one of the authors, Don L. Tolliver, executive di- rector of learning resources, University of Wisconsin- Whitewater. To the Editor: It is incredible that the authors would have concluded that "a search of the litera- ture produces scant statistical data" to sup- port the assumption that overdue notices COLLEGE Fr RESEARCH LIBRARIES are "a necessary part of library materials" (p.354). It is even more appalling to see such a statement in a learned journal while undergirded by a bibliography which in- cludes only six-count them-six references to the literature, none of which is more than five years old. . . . During my doctoral studies, I spent nearly three years in probing into the vari- ous aspects of the delinquent borrowers in academic libraries. In my dissertation, the results of these investigations were care- fully and fully summarized. In that particu- lar piece of "library literature," I described how I undertook to determine the differ-_ ences, if any, which occur in the return rate of materials of delinquent borrowers as related to various stimuli (overdue no- tices). . . . In order to test how these overdue notices influenced the response in a signifi- cant manner, data were gathered subse- quently regarding the rates of response. The findings showed that statistically sig- nificant differences in response existed the more direct the stimuli and in the predicted manner. The conclusions drawn by Baaske, Tol- liver, and Westerberg in a way corroborate the results of my earlier study, that is that overdue notices "appear to have an impor- tant reminder effect and improve the return rate of overdue books" (p.359). In fact, in my summary it was suggested that future research in this area might be undertaken by introducing other variables as treat- ments. The effort to test the effectiveness of the threat of encumbrances in urging stu- dents to return library materials is an excel- lent example of what I had in mind . By no means do I wish to denigrate the statement of the problem, the description I 71 72 I College & Research Libraries • January 1975 of the design and procedures, nor the re- sults of the research described by the Pur- due trio. No matter how sophisticated the design and implementation of a research project, however, there is nothing that su- persedes certain fundamentals of scholarly investigation. In this case, the answers to the basic questions of whether the problem is one which had never been solved; or had previous research on the subject been found and examined; or can the results of other research be used in solving the present problem-all seem to have been less than thoroughly explored. Response: Le Mayne W. Anderson Director of Libraries Colorado State University F art Collins Essentially, · Dr. Anderson stresses the need for thorough literature reviews of pre- vious research on the subject under study. I agree 100 percent with his criticism and regret that during the course of our work we did not find his 1970 Ph.D. thesis en- titled "Delinquent Borrowers in an Aca- demic Library." With his work in hand, we could have potentially made a more signifi- cant contribution to library literature. One word in our defense: In the real world of a library research unit, one works within the constraints of management needs, time, and cost factors per study. When a real problem is at hand, one does not always have the luxury of time necessary for an ex- tended literature review. Often alternatives must be suggested to management imme- diately in order to meet deadlines for policy formulation. Don L. Tolliver To the Editor: . . . There seems to be a tendency to as- sume that as long as statistical tests of va- lidity are met then the conclusions are sound. . . . An example of uncontrolled variables can be seen in the article by Baaske, Tolliver, and Westerberg .... One must assume that Purdue University has a published overdue policy which in- cludes fines and threat of encumbrance. Therefore, the fact that subjects did not re- ceive notices or formal threats of such ac- tion does not mean that they were ignorant of normal policy and perhaps affected by it. In other words, environmental con- straints may have been such that a true test of the effect of different notices, or lack of them, was impossible. The authors conclude that overdue no- tices "appear to have an important remind- er effect and improve the return rate of overdue books." Unless a more exhaustive study of the data exists which was not pub- lished, this seems unproved. We still lack conclusive proof that the length of the loan period is preeminent in determining book returns. We suspect that different types of borrowers, e.g., undergraduate vs. graduate students, have different book use period re- quirements. And it is entirely possible that some differentiation in use periods might be identified on the basis of subject field. Thus it would seem that a true test of the effect of overdue notices could only be con- ducted in a less contaminated environment, utilizing a more homogeneous group of borrowers, and focusing on book returns in a particular subject area. Robert L. Burr Director of Circulation Earl Gregg Swem Library The College of William and Mary in Virginia Williamsburg Response: Uncontrolled variables can always con- taminate results, especially if they are gen- erated in a systematic fashion. One way to minimize their effect is to employ a ran- dom sampling technique. Any contamina- tion effects which might have been present were probably randomly distributed across subjects, thus, not systematically affecting the data. In other words, subjects in each treatment group, in all probability, had an overall equal awareness of the library's ex- isting overdue policies. There could well have been some contamination from learn- ing effects, in that subjects depended on overdue notices as a reminder to return books. Yet, a review of the data generated by a small pool of subjects (who were first- time users of the library and therefore in all probability had not learned to depend on overdue notices) revealed the same re- sults as presented in the study. In a sense, the subjects assigned to group A served as a control group to which one can make comparisons regarding the effects of the other treatments. This study did not address the issue of varying loan periods nor were we particu- larly interested in differences between dif- ferent types of borrowers or differences in fields of study. Such an approach would have served to limit the degree to which we could generalize our findings. In summary, we needed to know the effects of overdue notices, threats, etc., as related to the user population in general. Don L. Tolliver To the Editor: The experimental design of the Baaske, Tolliver, and Westerberg study of overdue policies reported on page 355 of the Sep- tember issue calls for three observations on each subject. Presumably, the subjects are people, borrowers .... It will be noted that the criterion mea- sure, the observation, was in terms of the percentage of books returned. What values can this percentage assume? Only two pos- sibilities, as I see it: a borrower in any treatment group either has returned his book on a given day, or he hasn't. The per- centage is either 100 or 0. No other values are possible. In other words, we have data of nominal quality. This is the question asked to obtain 0 21 , for example: on day 28, has borrower 1 in group A returned the book, yes or no? This raises two serious questions about the study. ( 1) Were the observations really pre- measures and postmeasures, as the authors claim on page 356? I say they were not. 0 1 , 0 2 , and 0 3, for example, are not three successive observations on the same subject. Instead, as soon as an observation takes on the value of "yes" that sub- ject is eliminated from the study. (2) More important, were the observa- tions of high enough quality for arithmetic treatment? Again I say, no I The answers were in terms of "yes" and "no." If you add a "yes" and a "no" and divide by two, what is the result? A mean of "maybe"? Since the analysis of variance design em- ployed by the authors requires data of at Letters I 73 least interval quality, I have concluded, for the moment, two things: ( 1) This study to determine the differ- ential effect of overdue warning al- ternatives on return rates has mis- carried. ( 2) Our profession needs more concern with methodology, not less. And don't listen to the change for change's sake people! Response: Herbert H. Hoffman Catalog Librarian Santa Ana College Santa Ana, California Mr. Hoffman's comments concerning the design are technically correct. A true Campbell & Stanley design was not em- ployed, for subjects did "drop out" along the way. Perhaps if the study were done again, a chi-square ( X2 ) test would be em- ployed. However, the same results would likely be found and similar conclusions drawn. Another statistical test which is equally effective is the test of differences between proportions or percentages. Thus, the findings remain as follows: At the time measures were taken, the percentages of books returned clearly were not the same under the three treatments. Mr. Hoffman's statement that the data aren't worth analysis is erroneous. Also his statement that analysis of variance requires interval data is also wrong. It is quite com- mon to perform ANOV A on ranked data, which is ordinal, not interval. Indeed, our profession needs to be con- cerned with methodology, and constructive criticism is helpful; yet let us not lose sight of a more important issue, namely, know- ing which questions merit the energy nec- essary to complete a study. Don L. Tolliver To the Editor: . . . The authors are to be commended for the relatively complete description of the methodology they employed. However, some questions need to be raised regarding this methodology as well as the final conclu- sions reached by the authors. ( 1) A total of 4361 transactions were "randomly assigned to either treatment group A, B, or C." But the resulting assign- 74 I College & Research Libraries • January 1975 ment of 969, 1524, and 1868 transactions, respectively, is so unlikely as to defy belief. (A chi-square test of the hypothesis of equal likelihood is rejected at an exceed- ingly low level: p < < .0001.) Is there an ex- planation for this phenomenon? (2) The authors' statement that all pairs of means are significant at the .05 level ap- pears to be contradicted by another state- ment appearing later in the same para- graph: "no significant difference in return rate was found between Ss in Group A (overdue notice and threat of encum- brance) and Group B (overdue notice only)" (p.358). And in their conclusion, the authors write, "The threat of encum- brance is effective in urging students to re- turn library materials near the due date" (p.359). When the threat of encumbrance is accompanied by an overdue notice, this conclusion also seems to be contradicted by the first-quoted statement above. Which of these statements accurately reflects the au- thors' findings? ( 3) Finally, a major conclusion of the authors is that "the encumbrance system does not appear to have the cumulative and deterring effect of a fine system" (p.359). This conclusion appears to be entirely un- supported by the study, in which the eflect of fines on book return rates is not an ex- amined subject. Response: Stephen P. Harter Library Science/ AV Program College of Education University of South Florida Tampa Mr. Harter's comments concerning un- equal Ns is important. In this study, the as- signment of subjects to one of the three treatment conditions was determined by the last digit on each checkout card. Unfortu- nately, from the pool of transaction cards used, more cards happened to have last digits which, based on the instructions giv- en circulation personnel, provided for auto- matic assignment to Group C rather than Groups A or B. Unequal Ns do not diminish the quality of a study, although they can be difficult to interpret or can be misleading. The analysis used did allow for these very large unequal Ns. Mr. Harter is correct in indicating that the effect of fines on book return rates is not directly examined in this study. As indi- cated in the study report, no statistically significant difference in return rate was found between subjects in Group A (over- due notice and threat of encumbrance) and Group B (overdue notice only) . Thus, it was incorrectly reported that all pairs of means were significant at the .05 level. This was the only comparison of means that was not significant at the .05 level. However, the trends as illustrated in Figure 2 indicat- ed that threat of encumbrance has some ef- fect in encouraging students to return li- brary materials nearer the due date. In this study, this effect was not statistically differ- ent from receiving an overdue notice only. Yet, the trends still pointed in that direc- tion. Perhaps an observation is worth noting at this time. While results of field research may not be perfect, such results (especial- ly when studies are replicated) certainly can provide library management with need- ed information and are far better than no research at all. Don L. Tolliver