C1'eation of compUH" "'.,---' 

llO 
in (9) presents me results 01 a comparison test 01 me first mree 

CREATION OF COMPUTER INPUT IN AN EXPANDED 
CHARACTER SET 

Donald V. BLACK: System Development Corporation, Santa Monica 
California (Formerly, University of California, Santa Cruz, Calif.) , 

Keypunching of an expanded character set for library catalog data is 
described. The set included 101 different characters. Source documents 
were shelf list cards, the master record at the University of California 
Library, Santa Cruz. At the end of February, 1967, some 50 million 
characters, 1'epresenting more than 110,000 separate titles, had been 
punched. Some of the considerations leading to the adoption of this 
method for the creation of machine readable input are given, and details 
on costs and production rates. 

For manipulation by a computer, data must be converted to machine 
readable fornl. There are still only a few reasonably flexible means of 
creating machine readable records, especially if the data include an. ex~ 
panded character set. Five possible methods utilize one of the fnUoWlllg. 
standard keypunch, paper tape-producing typewriter, optical character 
reader, keyboard device that encodes dh'ectly onto magnetic tap~, or f 
keyboard tenuinal that inputs directly into a computer. DescriptIOns. 0 

some of these methods are available in the literature. The Johns Hopkin! 
University (1) used optical character recognition which can handle a ft~_ > 
alphanumeric representation, whereas Southern Illinois (2) used mar. 
sense scanning to convert only a limited amount of information. Car~ 
wright (3) and IBM (4) discuss direct computer input from a keybo~r 
terminal. Buckland (5) discusses the use of the paper tape-produc:nf 
typewriter. Hammer (6) and Kilgour (7) discuss keypunching. PatflC 

t( 8) discusses several methods of conversion, but only in the abstrac · 

...Cbap\110ds above.does not discuSS the relative merlts 01 these methods, but. 

'fb paper ts the details of a system that has converted approximately 
es 

11resen . h 2 1ra"eris;fuon characters 01 library catalog data on more t an 0 anguag ,
1 


500 to et of 101 characters.
with • ~ iversity 01 california at Santa Cruz is one 01 three university 

'!'h. n recently established by the State. It opened lor business in the 


~Pls~~65 with a core collection 01 some 55,000 titles in approximately 

fw000 volun,es. Early in the operation 01 the Library, it was decided to 

SO, achin ro as much as possible; therelore me existing catalog 


emods
eos. as had to be converted il the original collection were to be a part 

'f,~e lutur machine system. The creation 01 the core collection lor the 


e
;)"ee neW campuses 01 the University 01 Calilornia has been described 

in the literature (10). 

METHODSBids were sought to convert the catalog records during the summer 01 
)965. The shell list record produced by me new campuses' project was 
the master record and was to be me source lor conversion. Unlortunately, 
the shell list consisted 01 both printed Library of Congress caIds and cards 
produced at me new campuses' project Irom typewritten multilim mas' 
ters. No editing was to be done on me shell list caIds. The only addition 
was the stamping 01 an arbitrary number using a five-digit automatic 
numbering machine, the purpose 01 the number being to keep individual 

punch cards together for each entry. Weighing me responses to me request lor bids was a disherutening 
experience. Only lour responses were received Irom a total 01 15 requests 
sent out. The bid request did not specify the method to be used to 
convert to machine readable form, but only the resulting machine read­
able record. Since the specifications had used punch cards as an example, 
P,:,haps this limited the minking 01 some 01 the organizations involved, 

With the result that they did not choose to bid, 
e Three .bids were based on keypunching. One was from Florida and me 
.ompleXlhes of the task made the choice of such a distant company 
:mpossible. II problems had arisen during the course 01 me conversion, 

ravel costs would have been excessive. 
cF'0ther .response estimated the cost to be about $1.50 per record. 

early, tllls was too costly, and since bids of this nature are apt to be
~::ervative in 'h. matter ~f ~Itimate tot~1 costs, we I~lt the choice 01 

eth an mgam7.abon to do tne lob would, mdeed, result m a target figure 

at would be too high. 
I ?nly one bid used optical scanning as the method 01 conversion. Un· 
orufately, the bid was for me scanning only, and Library staff members 
wou d have had to retype the records for the scanner. Since the cost 



the scanJling alone was close to 301 a title, that bid was also 

lJaSed"'ll~.~being ultUnately more costly, choice 01 a kelY"nching service in San Francisc~ was made m
1'b" fill . 01 its pr<>xirillty to Santa Cruz, on the enthuSlas 01 the 

'" tbO b""J, task to be undertaken, and on a reasonable cost estimate. 
lJidDor I<?r b':.re completed the task in slightly more than three months. , au'" Po , '!'he se~ge Hon was done on an IBM/056 Verifier during this mass con­

.<: '" .....+J CO rl ... ... 
~ <d , "i"? 
01-4 CO lJ"\ U"\ t,eY "erwca.o I I I ;e<Si;e eJ<Pan charact~r set employed is shown in Figure 1. It was» !> -"" .... N 
<: ' .... ..... dedco, col '" 0 because it was aVlUlable on an IBM/1401 computer at the Los.<: " 
>,. +J ... CO COIr\, P CO I I 

..... >, I -::t -::t cbg'j':. caIDpus 01 the University (UCLA). At that time, it was the only >< tdp (V") I IS 
'"' $,,; ~;:1~ ~with sucb a printer on the West Coast. The character set had been 

! 
g .& ~ .~ ~ ro...... ... 

z ..., Joe Joe 0) f CO CO I, ted by librarians at UCLA from characters offered by IBM in the 
..; ~~ Pi~ '}J~~ 

co al IJ) Joe 0 I • 
..-i .;.> ~~§IJ)'rlPi ... ~~ (J) ~er of 1964 for the 1403 printer.'" >r< 


A 
 'rle:;j~~g <rt ...... ar As Figure 1 shOWS, digits and 100ver-case letters are punched as usual;'" Po
'"' col<LI@l ~ 1-' U el~~::;~'5 -::tarC(~ er'case letters are preceded by the word-separator character (0·5·S). ;:;, <: <: " 

Q) Q) X""" J..4 '" 0 '" cO ' t-- N N 
U U Q} rd Q} 6'gg. ..... "'~ '?~..;..;«: () () r-I '-" > ~ 

...... ~g lor the special characters is descnbed in the tables. There are'-" d ,....-! ~ -rl" (V")

col col ~ 0 0 .;.> ~ Ul <LI ~ (J.): h ... CO «)
.;.>~ Q) Q} , o;:j Q.! 0) Pi 0) .c ~rl+J roOO"'CO I I 


.;.> > u ~ .... <d .;.> ..-i <d ..; "" 
 """ minus signs: the one printed from an ll·punch is below the center 
'd'rO >0 '" ~ a.g -a:ci~~ 0J,'f'cr~ ~ e ~ ~ ~ ~ ~ ~ ~ 

;I;«: ~ ~ E-o ~ (J) E-o U '" 0:l.!ll8c"l«:~~~~;:1~ 01 the character; obtaining a centered minus requires a multiple punch u ~ '" ~ 
-&3 (11. ). The underscore prints in a space by itself, just as do other char· 0:; 'B

Po< ~ acters. It requires special programming to overprint this character by 0UJ co co ..... 0 ~£, , ,
r-I C\J (V") -::t If'\ \0ffi '):lr-co~J,~ ..... ..... suppressing paper spacing. The virgule overprint requires two columns E-o o ;:1 

..... ~ ~ 
CJ It> punch. Sharp.eyed readers will notice that the virgule appears twice ~~ 

u in Figure 1, and it has been counted twice lor the total 01 101 char­
o;J '""l acters. The blank has also been counted as a character, but the black ~ , < o lZ -A '" ... > ....... "" 


Ed «: .'ifJ square, which was not used at Santa Cruz, was not counted.fu ~ All data elements were encoded in fixed card fields; that is, the field 
for each type of inforrnation had a fixed length, generally 300 characters. 

..," It was not necessary, however, to use the entire field or to fill it with g ..-i -­ :a til zeros or other codes. No terminating characters were used to separate 
II the fields. Each type of information was included on one or more cards'"' ~ ~ ~ 
~ ~ U ~ ~ m m :arin ~ code which would tell the computer precisely what type 01UJ 1l ~ ~~ ~(Q" gA 

II '"' .<: a .., "" '"' ~ () as ..., 0) <LI () v ormation the card ( s) contained. All of this is illustrated in Table l.g; j ~ .....;:;, s:ll:! aUg ~'"'-ti~~.;.> <: U There are basically two ways that information can be encoded into ~ ,.a<>: 1-'~g~ ~~6 H~~P~" rd 0) '-" J..t ~1J)~!!~gbj'ga~p~,.... cards. This is discussed in references (3) and (6) especially. To use a.cH~MenO)~ 
en 0 EI ro ~ ,0 r::!30)O.rllQo)....-i~VJdCO.....-lA"II ~ ~: '" .~ ..; ... ~ ° '" °M ~ .;.> SO'"8 rl0 ~ '" h '" r-Id ~~og.;i§ n,",daIQo,"PodJl) .... p,..; u~etelY variable lonnat it is necessary to bave field delimiting codes. Cd 

~ A«:~~p., (J) 0:; U W ::0: Z ~ ~ ~ «: u UJ «: ~ Po< 000Z I' ~ xed sequence 01 data elements is established (e.g., author, title, pub­
co CO CO CO ""'-!cococoEd O?Cfo'?'?'f'?'? 'f'f<J?"T'9,-i~

'e~ (\J ~ ~ J.Po< nlS er, .etc.), one code will suffice to separate each field. Fixed sequence "" M I I I ~ 'f t-;- ~ ~ 1" If' ~ tt 9 ~ 1" u;' '? tj ~§
uu (V")-::tlJ"\\O 0000" ° o 0 ";;:1 ;:1 ;:1 ;:1 ;:1 ~ ~ ~ ~ ~ ~ ~~ :Sltate~, however, making provision lor establishing sequence lor@ 

UJ <d.<:-e ry pOSSIble data element that can occur in a catalog entry· and if one 
~ ~ § ~ Op e e ements do not occur lor some specific entry, then the keypunch08 ....... • l:C 
 Ormor I ' '"Ii'­~. ~ ~~J "'" @J "** everatrJr must remember to add the required field-delimiting code lor .~ ~l<:~ 

cr­U , ery data element not present, and to add the code in the proper se­... 
o~ .;.> '" a"fu . II a number 01 individual codes are to be used to delimit fields, H u

ell-<!)-@J 1\ V + ....... -II- , h ,,,,,,,,... 1* .. . '-" ...-- ....... ,......, 
 H ceEd en becomes difficult to find codes that are not otherwise used for fu .<:: " colU 



114 journal ot Library Au-tomation Vol. 1/2 June, 1968 


Table 1. Punch Card Input Format 'fable ):. (c ont.)
field IV Cals. Comments 
Field ID Gols. Comments . card, No. 200-224 (Limits, 1-5 cards per title; 1-5 titles) 

')'JIle'fitl 1-60 May be cont. on up to 4 additional cards Shelf Key Card, No. 000 
Call No. 1-20 As desired e 61-67 Unused 
Year 21-24 Year of Publication con . Indicator 68 Is '-' if continued on next card 
Copy 25-26 (Blank or 01-99) tspecial Code 69 See Table 3 
Series 27 (Either alphabetic or numeric O.I( ) card No. 70-72 200-224 

Volume No. 28-30 (Blank or 001-999) . Accession No. 73-80 Same as Shelf Key Card 


(Blank or 01-99)
Part No. 31-32 pUblisher/Source Card, No. 300-305 (Limit' 1-2 cards per publ./source, 
Donor No. 33-36 No. of donor on g~t list (may be blank) 3 publ./sources total) 
Date Rec'd 37-40 Month, year receIved at Library Publ./Source 1-60 May be continued on a second card 
Location 41-42 Alpha code designating location on 61-67 Unused 

campus 68 Is 'o' if publ./source is cont. on next card 
Type 43 O=book, l=serial, 2=reference Cont. Indicator 69 p=publisher, S=source 

3=gov't pub., 4=see auth., , Special Code 70-72 300-305 
5=see subj., 6=see also subj. Card No. 73-80 Same as Shelf Key Card 


Language 44-47 From 1 to 4 one-letter codes indicating Accession No. 

lang. Collation Card, No. 400 
 As desired

Suppress 48 H "S", entry appears on shelf list only , Collation 1-40 
Unused

49-69 Unused . 41-69 400 (1 card only) 
Card No. 70-72 Must be '000' . Card No. 70-72 Same as Shelf Key Card 
Accession No. 73-80 8-digit No. which sequences a batch of Accession No. 73-80 

(Limit: 1-5 cards per comment; accessions in Call No. sequence (gener. Commentary Card, No. 500-509 2 commentaries, (1 of each type) (I,ally only final five digits are used) 
May be cont. on up to 4 more cards 

Personal Author, No. 100-104 (Limit: 0-5 ca.rds) (I, Commentary 1-60 
Unused61-67Author 1-60 Name of author, left justified Is '_' if commentary cont. on next card 

61-68 Unused Cont. Indicator 68 's' for commentary to appear on shelf
69Special Code 69 See Table 2 Special Code 

list only Card No. 70-72 100 through 104 
Card No. 70-72 500-509Accession No. 73-80 Same as Shelf Key Card 
Accession No. 73-80 Same as Shelf Key Card 

Corporate Author, No. 110-119 (Limit: 1 or 2 cards per author; 
Subject Card, No. 600-604 (Limit: 1 card per subject; 5 subjects) 0-5 authors) (I, 


Subject 1-60 As desired
Corporate Author 1-60 May be continued on second card 
61-68 Unused61-67 Unused 69 May be used to indicate level of 

Cont. Indicator 68 Is '-' if author is cont. on second card Special Code 
subjectl) 4 

Special Code 69 See Table 2 
600-60470-72Card No. 70-72 110-119 Card No. Same as Shelf Key Card 

Accession No. 73-80 Same as Shelf Key Card Accession No. 73-80 

I): H 2 commentary entries are used, at least 1 must have an '5' code. 
The first author to be processed by the computer is considered th~(I, 

The Special Code is ignored by the system at the present time. 
main author. The main author appears on all catalogs and is represente 
in the title by "00". 



.l.lQ 	 jotlrnm oT Lwrary L-1tlromanon VOL l/ ~ June, HJOO 

legitimate data. It seemed easier to input with fixed fields and s' 
waste a few card columns. That is, if a particular field ends anYWhllnP1y

ere'the body of the card before the final column (for example, 60), the 0 ll\ 
tor simply stops and feeds in a new card. It is possible that a Per:\. 
may have only one character of data on it, in addition to a card-s cardequ
number and item-type number. In practice it would seem that the 10el\:e 
time in card feeding is not significant, and blank Hollerith cards are SS III 

. 	 ~ry
cheap mdeed. 

Training for the mass conversion effort at the keypunch service in S 
Francisco proved relatively easy. An operator's guide was produ~~ 
showing the codes and conventions for each data element found on th 
typical catalog card. Only the shelf list card was used for the conversio e 
Tables 2 and 3 show the various elements that were coded. There we~' 
two forms of shelf list cards, as mentioned above: Library of Congres~ 
cards and cards produced by the new campuses' program at San Diego. 
On Library of Congress cards everything was encoded except Roman 
numeral pagination and size information, and the information at the 
bottom of the card: the call number used in the Library of Congress itself, 
the Dewey number, the LC card number, and the name of the originat­
ing library if any. On the home-made cards, Roman numeral pagination 
and size were not encoded. Everything in Roman characters was 
punched. Cards with only a small amount of information in Roman char­
acters had a legend punched, "for complete entry see shelf list." It is 
simply not possible in a short article to give all the fine points of conver­
sion. Rules for all contingencies were devised and most proved easy to 
follow. Twenty operators, working in two shifts of ten each, converted 
the 55,000 titles that existed in June, 1965, in about three months' working 
time. All data elements to be used later for sorting purposes on the 
computer were key verified, but for the first month of the conversion the 
entire record for each title was verified. 

Beginning in December, 1965, the Library at Santa Cruz began key­
punching operations. Mter a training period of a week and operational 
experience of four months, the local operators achieved a rate of 7,000 to 
8,000 keystrokes per hour, with a net error rate of only 12 errors in ap­
proximately 24,000 keystrokes; That is, the operators recognized a number 
of errors and corrected them at the time of initial punching. The 12 
remaining errors should be caught during proofreading, which we sub­
stituted for key verification in the ongoing production system. It was felt 
desirable to combine proofreading for transcription accuracy with the 
typical library practice known as "revision," which implies that the cat~­
log copy be reviewed for content as well as accuracy. This is true even 
for text taken from Library of Congress catalog copy. Elements such as. the 
form of entry, the form of series note if any, number of subject headl1lf 
and form, etc., are all reviewed by a cataloger other than the one w 0 
initially prepared the copy. Proofreading and revision was done from !l 

special Codes for Authors 
'table i,. Meaning Notation 

code 

6 
To create added notation on author catalog: 

type 1: Joint author Joint Auth. 
COIllpUer Compo 
Editor Ed. 


E Joint editor Joint Ed. 

G 
 Illustrator Illus. 
I publisher Publ. 
p Translator Trans. 
T 

Type 	2: To specify a substitute sort key: 
X 	 Use this author as a substitute sort key for previouS author. 

Previous author wUl appear on appropriate catalog but this 

author wUl not. 

Table 	3. Special Codes for Titles 

Code Meaning 
Suppress listing this title in title catalog 

X Title 	is a transliterated title 
T 

Title 	is a series title S 
Partial title P

D 	 Standard Title or conventional title 

In all cases, the first title encountered when processing a given entry 
will be the only title which appears in the author and subject catalogs. 

printout on a line printer having only 64 characters available in the char­
acter set. This number of characters suffices, however, since there are only 
~ usable card codes, i.e., the pattern of holes in each column of a Holler­
Ith card. There are only 64 valid combinations which can be read by the 
~mputer equipment. As illustrated in Figure 1, some characters with
dlacri~cals r~quire three punched columns to produce one character in 
the ultimate printout. 

RESULTS 
thFor the mass conversion which took approximately three months during 

e ~ummer of 1965, the total cost was slightly less than $34,000, or ap­
p:oxunately 60¢ per title. In a discussion of the project, after its conclu­
SIon, with the two supervisors of the service bureau in San Francisco, it 
Was agreed that in all likelihood the service bureau operators had just 



118 Journal of Library Automation Vol. 1/2 June, 1968 
"d the programs written to process the catalog card data. This 

,,ailable' statement to make, yet the road to the production of either 
~reached peak efficiency about the time the project terminated. 1'h 

had the project continued, the cost per title would have decreased. ;t is, 
work records maintained by the service bureau, it was apparent that rOIll 
:first two months of ~he project :vas a learnin.g period? as the output of the 
operators rose contmually dunng that penod of tune. During the ~e 
month the productivity curve leveled off considerably. ast 

Table 4 shows the cost of the production operation established in S 
Cruz. The costs used are somewhat arbitrary. For example, the ,~ta 
punch operator" classification at the University had six steps. In prodey. 
ing an average cost should the actual rate being earned by the keyPU:h
operators be used, or the beginning rate, or some other? The amount uS~d 
in Table 4 represents an average of the pay being received by the 2.3 
full-time equivalent operators, rounded upward to an even amount. Holl. 
erith card costs can also vary slightly. Table 4 uses $1.00 per thousand as 
a reasonable price and one which could probably be obtained anY'Nhere in 
the country. Costs are based on rates obtaining in February, 1967. 

Table 4. Cost Per Title to Produce Machine Readable Catalog Data 
Keypunch rental, $65.00/mo. $ .026 
(one-shift operation) 
Keypunch operator, $2.10jhr.+20% overhead .168 
Blank Hollerith cards, $1.00/thousand .009 
Machine listing of cards for proofreading .002 
(printing at 390 cards per minute) 
Proofreading, $5.00/hr., 120 titlesjhr. .042 
Correction of errors .020 

Total $ .267 

DISCUSSION 

There are, of course, hidden costs in the ongoing production at Santa 
Cruz that are difficult to fix because the University does not charge for 
them. For example, there is the cost of space occupied by keyp~ch 
operators and by equipment, the cost of air conditioning and electncal 
supply, the cost of adding internal partitions, doors, etc. Spread over ~ 
yearly total of some 30,000 new titles, these unknown costs and the addl­
tional costs of supervision could not be very great per title, assuming that 
the rate of keypunching production remains relatively constant. However, 
labor costs may prove to be a key factor in some geographic areas. In, 
Santa Cruz good keypunch operators were available at a reasonable cost, 
but in large metropolitan areas this may not be true. Since the ope~~tor 
cost is over 60 percent of the total per title, it obviously can be a CrItical 
factor. 

What happens after catalog copy is converted to machine readable forI1l 
by punch cards or any other method, depends on computer equiprnent 

.. eoS)' rds or book-fonn catalogs is not easy. Even the data them­
.,..,wg ca cause problems. For example tbe reader will note in Figure 1 
seI"" ':'cha'a used to indicate an up-sMt for capital \etters has a 

cter...t Ih de of 0-5-8 fixed by the manufacturer, IBM, and neces~ary to print 
~d ~e exp",ded cbaracter set cham. In retrospect ,t would have been 
.,rth t convert the firuJ code configurations as part of a computer 
""'"' ~g step than to punch them from the beginning. The 0-5-8 sMt
P'f~s a special code used within the 1400 series computers, and is known 
10 e IS ore! mark or wore! separator. In normal operation of the 1401 com­
asu~these marks are used to delimit fields within the mexnory of the ~chble. certain program commands use these marks to detect when the 
b bluing or the end of a field has been reached. Use of such a code 

;ift,e data can raise havOC with a program unless the programmer is con­

stantly alert to the problem and takes great pains to circumvent it. Some 

other cOde such as the $ sign might have been used and then converted, 

prior to the final run on the )401, as part of the computer processing to 


theWhilecode needed for printing purposes. the number of articles on catalog conversion has not yet be.ml 
overwhehnblg, it is apparent that there is a great deal of interest in the 
'field. One might ask, 'Should every library proceed to convert its own 
catalog?" DeGennaro has addressed this problem (11) and the reader 
is referred to that discussion. Perhaps the question has no ideal answer. 
It does seern unfortunate, however, that tbe pre-1955 National Union 
Catalog is not to be published from a record that is machine readable. 

It would seem possible, however, that in the future methods could be s 
devised to use macbine readable records produced by the larger b'br.,ie 
and some procedure whereby tbe smaller libraries could check their 
holdings against those of the larger libraries. By some fairly simple 
method, a subset of the master machine records could be selected for use 

in the catalogs of smaller libraries., ~ 
The Santa Cruz project began before the Library of Congress had an­

nounced results of preliminary plans for the MARC project (12). To a 
fu'am extent the catalog record at Santa Cruz could be converted into 

e MARC format, although MARC goes far . deeper in coding discrete 
elements of data within the catalog record than does Santa Cruz. How­
fiver, to the extent that discrete dahl elements are encoded and identi­
IZ proflCrly in the machine record, any catalog format can be trans-

drmoo .mto any other catalog format by the computer. The key is the 
proper Identification of each data element. 

ACKNOWLEDGMENToTh~ author wishes to thank his colleagues at System Development Cor­
p ration, ill particular Mrs. Ann Luke, for helpful comments on this paper. 



121 

__ ~ JVUIIUU or Ltbrary Automation Vol. 1/ 2 June, 1968 

REFERENCES 
1. 	 The Johns Hopkins University. The Milton S. Eisenhower Lib 


Progress Report on an Operations Research and Systems En i r~: 

Study ofa University Library (Baltimore: Johns Hopkins, 196i)ne~ 


2. 	 Southern Illinois University. Office of Systems and Procedur~. 

Automated Circulation Control System for the Delyte W. Morris ~ 

brary; the System and Its Progress in Brief (Carbondale, ill.: SouthLj. 

Illinois University, 1963). ern 


3. 	 Cartwright, Kelley L.; Shoffner, Ralph M.: Catalogs in Book FOrm 
(Berkeley: Institute of Library Research, 1967). . ... 

4. 	 International Business Machines. Federal Systems Division: Report on 
Pilot Project for Converting the Pre-1952 National Union Catalog t 
a Machine Readable Record (Rockville, Maryland: IBM, 1965) . 0 

5. 	 Buckland, L. F.: Recording of Library of Congress Bibliographical Data 
in Machine Form rev. ed. (Washington, D. c.: Council on Library 
Resources, 1965). 

6. 	 Hammer, Donald P.: "Problems in the Conversion of Bibliographical 
Data-A Keypunching Experiment," American Documentation, 19 (Jan­
uary 1968), 12-17. 

7. 	 Kilgour, Frederick G.: "Development of Computerization of Catalogs 
in Medical and Scientific Libraries," Clinic on Library Applications 
of Data Processing, University of Illinois, 2nd, 1964, Proceedings 
(Champaign: Illini Union Bookstore, 1965), p. 25-35. 

8. 	 Patrick, Robert L.; Black, Donald V.: "Index Files; Their Loading and 
Organization for Use," Libraries and Automation; Proceedings of the 
Conference on Libraries and Automation, Airlie Foundation, Warren­
ton, Virginia, May 26-30,1963 (Washington, D. C.: Library of Congress, 
1964 ), p. 29-53. 

9. 	 Chapin, Richard . E.; Pretzer, Dale H.: "Comparative Costs of Co~­
verting Shelf List Records to Machine-Readable Form," Journal of Lt- . 
brary Automation, 1 (March 1968), 66-74. 

10. 	Voigt, Melvin J.; Treyz, Joseph H.: "New Campuses PrograIll 
(UCSD, UCI, and UCSC)," Library Journal, 90 (May 15, 1965) , p. 
2204-08. 

1l. 	 DeGennaro, R. A.: "A Strategy for the Conversion of Research Library 
Catalogs to Machine Readable Form," College & Research Libraries, 28 
(July 1967), p. 253-257. 

12. 	 U. S. Library of Congress, Information Systems Office: A Prelimi~ry 
Report on the MARC (Machine-Readable Catalog) Pilot ProJect . 
(Washington, D. C.; Library of Congress, 1966). 

COSTS OF LIBRARY CATALOG CARDS 
PRODUCED BY COMPUTER 

Frederick G. KILGOUR, Ohio College Library Center, Columbus, Ohio 

.PffJductW costs of 79,831 cards are analyzed. Cards were produced by 
ntout ".,.;ants of the Columbia.Harvard-Yale procedure employing an 

IBM ({l0 Document Writer and an IBM 1401 computer. Costs per card 
ranged from 8.8 to 9.8 cents for completed cards. 

Early in September, 1964, the Yale Medical Library put into routine oper­
ation the Columbia-Harvard-Yale computerized technique for catalog 
card manufacture (1), and during the following three years Yale pro­
duced over 87,000 cards. The principal objective of the CHY project 
was an on-line, computerized, bibliographie lnformation retrieval system. 
However, the route selected for attaining the objective included manu­
facture of cards from machine readable data to keep up the manual 
catalog while machine readable records were being inexpensively ac­
cumulated for computerized subject retrieval. Catalog cards were only:iii .product of the system, but their production was designed to be as 

clent as possible within constraints of the system. Nevertheless, this 
Pjper will examine CHY card production costs as though this segment 
? the system were an isolated procedure, yielding but one product, as
hethe case in classical library procedures. Costing will disregard other 
Ii nellis, such as accession lists and machine readable data produced lor 

ttIe, or no, additional expense. ~ Columbia Medical Library and Harvard Medical Library also in· 
S ed IBM 870 Document Writers and tested the programs for card 
prodUction, but neither library routinely produced cards. However, Co­