College and Research Libraries


68 I College & Research Libraries· January 1969 

braries. To date, Robert D. Desmond, its 
editor, and the Library of Congress have 
done an outstanding job of developing 
this tool. With the help of this survey they 
should be able to enhance the value of 
NST for the effective bibliographical con-
trol of serials in the future.-]os eph H. 
Treyz, University of Michigan. 

Bookplates for Libraries; Contemporary 
Designs for School, Public, College, and 
University Libraries. By Edward H . 
Shickell, with an introduction by Wil-
liam R. Holman. Austin, Texas: Roger 
Beacham , 1969. 69p. , illustrated. $12.95. 

The evolution of bookplates since the fif-
teenth century, and particularly their col-
lection , categorization, and admiration 
since the latter part of the nineteenth, oc-
cupies a substantial literature, much of it 
privately printed. Add to this a smattering 
of earlier books on bookplate design, and 
this handsome new volume of original de-
signs for libraries stands out as unusually 
fresh and attractive. 

It is to some degree complem entary to 
Mr. Holman's Library Publications, a 1965 
Beacham publication distributed b y John 
Howell Books, and is, like this larger and 
earlier volume, published to stimulate 
more interesting and imaginative printing 
for libraries and their clientele. 

Mr. Shickell's seventy-two specimen plates 
make use of a number of the better t ype-
faces and his own skillful calligraphy 
rendered in four colors suitable to library 
plates. Although their range of both color 
and form is limited by the fact that they 
are one man's work, he is both imaginative 
and eclectic, and his variety and taste can-
not but be stimulating to librarians seek-
ing to design bookplates. 

Mr. Holman's inb·oduction presents both 
encouragement and practical advice, in-
cluding the suggestion that if all else fails 
to produce a work of art the reader may 
violate Mr. Shickell' s copyright a little b y 
lifting a design direct from the book. The 
type faces used are carefully identified, 
and an index leads you to the plates in 
which they appear.-David H eron, Uni-
v ersity of Kansas. 

Evaluation of the MEDLARS Demand 
Search Service. By F. W. Lancaster, 
Bethesda: National Library of Medicine, 
1968. 276p. (available from NLM Office 
of Public Information) . 

Of all the automated information retrieval 
systems which are currently in operation, 
the MEDLARS (Medical Literature Anal-
ysis and Retrieval) System of the U.S. Na-
tional Library of Medicine has perhaps 
most captured the world's imagination and 
attention and has put both the United 
States and medicine as a subject discipline 
in the forefront in the use of computers as 
an aid in solving problems in information 
transfer. MEDLARS is a machine system 
designed to serve several purposes includ-
ing the monthly production and printing 
of Index Medicus, one of the world's pri-
mary medical indexing media. It has as 
well the capability to produce and print 
subsets of a large file of literature citations 
either on a continuing basis for special 
subject groups or on demand for individ- · · 
uals. The system inherently must, there-
fore, possess some of the trade-offs that 
are inevitable in any multi-purpose sys-
tem. 

This study is not an evaluation or descrip-
tion of the entire MEDLARS system; 
(such a description is being currently pub-
lished by the National Libra1y of Medi-
cine, under the title: Description and His-
tory of MEDLARS) . It is rather an at-
tempt to evaluate its " demand search 
module," a component designed to pro-
duce, by computer, comprehensive bibli-
ographies on many-faceted subjects on re-
quest. Nevertheless, in the process of 
studying this report, a reader can learn 
much about the construction and use of 
the entire MEDLARS system. In fact, 
some of the problems and prerequisites ex-
plored in the study have relevance to all 
kinds of literature searching, manual as 
well as machine. 

There do not seem to be any particularly 
new methodological approaches offered in 
this study. They are essentially modifica-
tions and refinements of those developed 
by Cleverdon and others. Nevertheless , 
the misgivings expressed by Alan M. Rees 



in his thoroughgoing review of the general 
subject of evaluation in the second volume 
of the Annual Review of Information 
Science and Technology ("Evaluation of 
information systems and services") in 
which he predicted that ". . . difficulty 
will probably be encountered in attribut-
ing recall and precision failures to the in-
dexing process, indexing language, search 
formulation and so on," do not seem to 
have been warranted. The basic problem, 
of course, is not resolved, and that is the 
central position of "relevance," the basis 
upon which both precision and recall are 
evaluated, and which in the end must re-
main largely a personal and a subjective 
judgment. Another problem is that to some 
extent the recall ratio (percentage of 
known relevant articles retrieved) and 
precision ratio (percentage of retrieved ar-
ticles which are relevant) is in effect not 
only an evaluation of the system under 
study but of the effectiveness of the paral-
lel search. 

Testing and evaluation is a responsibility 
of any management producing a product 
for consumers, in order both to control 
quality and to improve the efficiency of the 
system. Since this is one of the few large 
systems using controlled vocabulary tech-
niques, it has perhaps an unusually large 
onus of this kind of responsibility, in view 
of the proliferation of KWIC and other 
systems in which input is largely automat-
ed and do not depend as much on the 
human factor which although it can be 
more insightful can also be more erratic 

Recent Publications I 69 

than the machine. The findings that the 
MEDLARS demand search module is op-
erating on the average of about 58 per 
cent recall and 50 per cent precision do 
not provide any comparisons with other 
information retrieval systems either ma-
chine or manual because no other system 
to this reviewer's knowledge has been as 
rigorously and thoroughly examined. Nor 
do Lancaster's conclusions and recommen-
dations offer much promise that these rat-
ings can be substantially improved for per-
formance of the entire system. 

The configuration of the next generation 
MEDLARS system now under active de-
velopment is still not known to us, but if it 
is still committed to the use of a controlled 
vocabulary and human indexing, some of 
the cautionary and advisory conclusions of 
this study will still be valid, e.g., the ne-
cessity for continual quality control, the 
expansion in the use of sub-headings 
(which should also substantially benefit 
the manual system), and publication of 
the entire entry (cross-reference) vocabu-
lary. On the other hand, the recommenda-
tion on restricting foreign language mate-
rial seems rather parochial particularly 
since the study is based only on U.S. in-
vestigators, and in view of the develop-
ment of MEDLARS decentralized centers 
in other parts of the world. 

This study deserves detailed examination 
by anyone interested in information stor-
age and retrieval either as a producer or a 
consumer.-David A. Kronick, University 
of Texas.