College and Research Libraries 92 College & Research Libraries niques in a few pages, but the professional literature has not yet adequately met this need. Chapter 7, "The Last Planning Steps" contains PERT and Grantt charts and an ''Operations Research Time Study, I I all of which can aid the staff in reducing the complexities of moving to manageable de- tails. Upon reading through this useful, compact volume of 132 pages of text, it sometimes seemed as though it were in- tended for the librarian who would be faced with doing everything entirely alone and at other times as though one might just be able to get a little outside help. For the novices who read this book, they may not realize just how much help is available nor how possible it might be for them to ask for help. This book is filled with many how-to details that help the read to under- stand the planning process, but not al- ways fully nor absolutely correctly. It is important for the librarians who read this book to recognize their own limitations in time and experience and to understand when they can do parts of the planning themselves and when they should seek outside professional help.-Gloria Novak, University Library, University of California, Berkeley. Swisher, Robert, and Charles R. Mc- Clure. Research for Decision Making: Methods for Librarians. Chicago: Ameri- can Library Assn., 1984. 209p. $25. LC 84-12381. ISBN 0-8389-0398-3. This is a difficult book to review because it addresses a significant gap in the litera- ture on research for librarians but does not, in my opinion, fill the most important part of that gap. The authors identify ''action research'' as research for decision making. This fits well into the more gen- eral trilogy: research for the sake of better understanding, research for action, and research for no good reason at all. In each category it is possible to do good research or bad research. In fact, research that is good for one category may be fair or even poor in another. The trouble with this book is that it does not tell us enough about how to do good action research, or even how to recognize (and thus avoid) bad action research. I am January 1986 left somewhat ill at ease by a book that tells me something is doable, and worth doing, but doesn't show me, through either good or bad examples, how to do it well. On some subtle questions, such as the meaning of confidence intervals, the au- thors do a very nice job. On other issues, dealing with logical relations, they have more trouble. An example is the use of flowcharts toil- lustrate the interrelation of concepts, processes, signals, or other entities. Flow- charts can be powerful aids to thought, or abominable. The worst are diagrams with eight or ten circles, all linked by lines, to suggest that all the parts have something to do with each other. At the other ex- treme we have the feedback diagrams of systems engineering, which can be so pre- cise that the diagram itself specifies a dif- ferential equation, up to a few undeter- mined constants. The flowcharts on pages 5, 7, and 9 are poor because they mix concepts, products and processes in a confusing fashion. The authors do not use the powerful analysis of the relation to environment presented in Churchman in "The Systems Ap- proach." He stresses the distinction be- tween resources, which may be used by the system, and constraints, which must be obeyed or satisfied. Churchman's book is accessible to a bright high school senior, and is worth reading. Again, the flowchart selected on page 36 to illustrate the use of flowcharts is not a good example, because it does not make clear where items enter the system, how they come out, and how many of them fol- low each path. The basic structure of the process is simpler than it appears here . Confusion about functional relation- ships is shown in the pair of graphs on page 15 that ought, by the labeling of their axes, to be symmetric to each other. None- theless, the dotted lines droop down in both graphs. This kind of carelessness is reflected throughout the text. Properties, . concepts, and objects are loosely inter- changed even within a single sentence. In longer discussions an implied equivalence is set up (for example, between research competency and research literacy, on page 15) among different concepts, mak- ing one feel the unseen presence of Humpty Dumpty. The discussion on page 6 adduces the properties of stable closed- loop systems to open library systems. On page 8 we are asked to believe that the in- puts to a library system are its goals. Else- where we are told that a library sets its goals, and given a flowchart implying that the setting of goals is a part of action re- search. Although Swisher and McClure don't make common mistakes in their discus- sion of statistical inference, there is still something to be desired. Most library re- searchers today will be presented with SPSS output or something like it. The au- thors could have shown us what that looks like, and have illustrated it with a reasonable set of ample data (perhaps fifty or one hundred data elements.) If their mission is to overcome the fright librarians may feel upon seeing this stuff, the book should display one or two tame examples, to ease that fright. Rather earlier, on page 16 they cite a hy- pothetical case in which a study estab- lishes "a statistically significant relation- ship . . . between women undergraduates and skills taught." I have no idea how the rows and columns of the cross tabulation would be labeled, and I submit that the reader won't either. If the authors do, they should have told us. If they don't, then how can we be confident of their seri- ousness? Furthermore, a key point about the'' use of statistics" is not brought out. The whole idea of confidence intervals is de- signed to prevent premature rejection of some natural hypothesis (usually called the null hypothesis, H 0 ) in favor of an al- ternative that may appear better through the action of chance alone. In very rough language, the 95 percent confidence inter- val is designed to make the odds against this particular mistake 19:1. HOWEVER! In action research we are usually not . ,, testing a new fertilizer'' (perhaps that is more the domain of the reader of type III research)-we are trying to "learn some- thing new.'' Most statistical packages build in the null hypothesis that variables are unrelated. That is absurd. What we usually want to ,know is: "How much are Recent Publications 93 they related?" Is the relation. of manage- rial or economic significance? If I am trying to estimate whether a par- ticular library is circulating as many books as it ought to, I have some idea that this is related to the number of students enrolled in the departments that it serves. To see whether it is "off the line" I assemble the relevant data and draw some kind of plot. If I hand the problem to a statistician she may do a regression analysis, and may tell me that the R-squared value is large, and that I can have high confidence in the re- gression. What that means is that the (ab- surd!!) null hypothesis built into com- puter program (namely that the two variables have nothing to do with each other) can be rejected. It does not mean that every branch ought ยท to lie on the curve. (This can be dealt with by calculat- ing the band of error, which some pro- grams do, but my point is that we are not interested in preserving the null hypothe- sis here-it is a straw woman.) To sum up, the authors know a great deal about research, and about statistics, but they have not shared the most impor- tant parts of that knowledge with their readers. The project planning chart (page 29) is a useful example for someone who has not done project research before. Chapter 4, on surveys and questionnaires, contains some good tips and pointers. Taken together, the book cannot be rec- ommended. It is not informed by a single critical intelligence, and in places it looks as if the authors shared a single sentence (many run to sixty and seventy words) making the same point twice. The impre- cision in the treatment of ideas will disturb experienced managers and experienced researchers alike. It would make a poor in- troduction to either subject for those with- out experience. In spite of some bright spots, this is rather more a book about the literature than about research. The impor- tant gap is still unfilled.-Paul B. Kantor, Tantalus Inc.~ Cleveland, Ohio . Evaluation of Reference Services. Ed. by Bill Katz and Ruth A. Fraley. New York: Haworth, 1984. 334p. $29.95. LC 84- 12898. ISBN 0-86656-377-6. (This work