The Chromium-Plated Tabulator: Institutionalizing an Electronic

IEEE Annals of the History of Computing 1058-6180/01/$10.00 © 2001 IEEE 75

In many organizations, data processing is lookedupon as simply ‘the old tabulating operationwith chromium plating.’

—Richard G. Canning and Robert L. Sisson, The Management of Data Processing, 1967

The history of computing has had, as yet,remarkably little to say about the people whoordered and used computers, or of the purpos-es to which they were put. This has been par-ticularly true of the use of computers inbusiness, despite the insistence of historian andconsultant James Cortada that “a quick look athow computers were used suggests that the his-tory of the digital computer is every bit asmuch a business story as it is a tale of techno-logical evolution.”1 My aim here is to followthat quick look with a more considered exami-nation of the acquisition and usage of comput-ers to give a new perspective on an establishedtopic: the transition during the mid-1950s fromelectromechanical punched card technology tothe first generation of electronic computers.

Work by historians such as Martin Campbell-Kelly, JoAnne Yates, and William Aspray hasconsistently shown that the computer industrywas, more than anything else, a continuation ofthe pre-1945 office equipment industry—and inparticular of the punched card machine indus-try.2 Their careful exploration of computer tech-nology and the dynamics of the computerhardware industry leave little doubt that IBM’seventual dominance of the computer industryowes as much to the events of the 1930s as tothose of the 1960s. This is in itself a major depar-

ture from the perception, common during the1950s and common today, that each new gen-eration of computer equipment is a revolution-ary technology without historical roots, abreakthrough plucked fully formed from theforehead of (to mix a metaphor) Prometheus.

The next stage in our exploration of the his-tory of computing must take us beyond the sup-pliers of computer technology and into thefirms and occupations using it. By examiningthe crucial initial shift from punched card tocomputer, in the context of historian RuthSchwartz Cowan’s “consumption junction” (theplace where technology meets user), we findnew dimensions of continuity and discontinu-ity in usage to complement those in technolo-gy, distribution, and production alreadyexplored by historians.3

This article examines the early use of com-puters for routine clerical and accounting jobsby large American corporations—an activity Irefer to here as administrative computing, butwhich was firmly established by the late 1950sas data processing.4 For several decades, suchroutine administrative work had dominatedusage of the punched card machine, and, dur-ing the mid-1950s, this activity edged out sci-entific and technical computation as theprimary function of electronic computers.

The first managerially oriented discussion ofthe computer’s possibilities for business, a sub-ject that peaked about 1954, presented it as ascientific marvel of electronic technology,poised to spark a “second industrial revolution”that would transform the office much as the

The Chromium-Plated Tabulator:Institutionalizing an ElectronicRevolution, 1954–1958Thomas HaighColby College

The computer promised business of the 1950s an administrativerevolution. What it delivered was data processing—a hybrid of newtechnology and existing punched card machines, people, andattitudes. The author examines how first-generation computers weresold and purchased, and describes the occupations (analyst,programmer, and operator) and departments that emerged aroundthem. This illuminates claims of a more recent electronic revolution inbusiness.

first had transformed the industrial workshop.The postrevolutionary order would mark asharp discontinuity with the past. But while therhetoric of revolution continued to sweep thecomputer field at periodic intervals, it was rap-idly eclipsed in practice by the conservativereality of data processing. Promoted by IBM andembraced by punched card staff, the concept ofdata processing included both administrativecomputing and conventional punched cardwork. It served both well, by reinforcing thecontinuing importance of punched card tech-nology, and the organizational and profession-al continuities between the two kinds ofmachine. My emphasis is therefore on the con-struction of a new corporate institution—thedata processing department—from a combina-tion of the existing tabulating, and systems andprocedures, functions.5 By 1958, thousands ofcomputers had been installed, and the shape ofthe data processing department had been large-ly standardized. It would change only slowlyover the next 20 years. Alongside the newdepartment came the occupational identity ofthe data processing worker, an identity that wasitself a hybrid of new computer influences andold punched card culture.

I begin with the efforts made by computersalesmen and other computer enthusiasts tomarket the computer as a revolutionary tech-nology in the world of industrial management,and examine how these ideas influenced thecorporate committees set up to justify substan-tial investment in an unproven technology. Ithen explore the applications for which com-puters were most commonly used and comparethe tasks performed by the “revolutionary” elec-tronic marvels with those of a more mundanetechnology, punched card machines. I pay par-ticular attention to the physical environmentsaccorded to the two kinds of machine, and tothe continuing reliance of computers on broad-er systems of punched card and other data pro-cessing technologies.

My focus then shifts to the origins of, andinitial expectations for, the four main data pro-cessing tasks of operation, analysis, program-ming, and supervision. In each case, I explorecontinuities between these tasks and thoseinvolved in earlier work of tabulating, officemanagement, and systems and procedures.Examination of various computer-related jobsshows how they were accommodated into thebroader occupational identity of data process-ing, an occupation in which managerial orien-tation rather than technical excellence markedthe professional. I conclude by reexamining theimplications of these findings for the history of

computing, and exploring the lessons wemight draw by comparing this particular elec-tronic revolution in business with the so-recently ballyhooed explosion of electroniccommerce technologies over the Internet.

Selling a revolution …Early managerial discussion of the comput-

er treated it as an electronic marvel about totransform the world of business. Althoughexperimental computers such as the ENIACreceived a reasonable amount of publicity dur-ing the 1940s, businessmen would at first readlittle about them on the pages of Fortune,Forbes, or Business Week. These “giant electron-ic brains” were presented as scientific curiosi-ties. From 1953 onward, however, the arrival ofcommercially available computers from Univacand IBM made the computer an object of pro-fessional concern for managers. At the sametime, the computer manufacturers began a con-certed push to “educate” management as to thevalue of their wares.

The object of discussion in the early 1950swas not so much the computer but the moregeneral topics of “electronics for the office” orof automation. As consultant John M. Thesisinformed an audience of cost accountants in1954, “The word ‘electronic’ has become a‘buzz word’ joining ‘atomic’ and ‘space flight’in conveying the impression of scientificmagic.”6 It was by no means apparent in theearly 1950s that the stored-program, electron-ic, digital computer was the key electronicproduct. Initially, it seemed merely a represen-tative of a much more general class of elec-tronic technologies, which would replacehumans in the performance of an ever-widercollection of tasks. Even in the managerialpress, considerable credence was given to theimminent arrival of totally automated factories,triggering massive technological unemploy-ment and fundamental social change. Thus thepotential of electronics in business was fre-quently claimed to lie in its corresponding abil-ity to automate many activities currentlyconducted by white-collar workers and man-agers. As Thesis remarked later in the same arti-cle, “Electronic data processing systems arenow the highest form of mechanization avail-able for business procedure applications.”6

Edmund Berkeley, an actuarial methodsexpert turned computer evangelist, set the paceearly on, in his 1949 classic Giant Brains orMachines That Think. Not only did Berkeleyclaim for the new machines a “power … verysimilar to the power of a brain,” but he alsolaunched a thousand confused misappropria-

76 IEEE Annals of the History of Computing

Chromium-Plated Tabulator

tions of technological history with his sugges-tion that it would “take a load off men’s mindsas great as the load that printing took off men’swriting.”7 A year later, the inaugural editorialof Systems and Procedures Quarterly suggestedthat future historians would view this “comingrevolution in paperwork” as equal in magni-tude to the earlier revolution in industrial pro-duction. The editorial writer, who was also thehead of Shell Oil’s Methods Department, con-tinued his argument in a 1951 article in whichhe claimed that “[p]aper work may be entirelyeliminated” by the automatic transmission ofelectronic impulses.8

Another article, published in 1953, extend-ed this theme in a way that affords us a valu-able insight into the role of revolutionaryrhetoric in selling an unproven technology. Itsauthor, W.B. Worthington (a systems expertthen working for Hughes Aircraft), trumpedeven Berkeley’s expectations with the claimthat “The changes ahead appear to be similarin character but far beyond those effected byprinting.” If, as he suggested, “the ominousrumble you sense is the future coming at us,” itcould be assumed that any firm not agileenough to jump on board this noisy juggernautwas liable to be crushed by it. He revealed thekind of thinking that led aerospace firms likeHughes to be among the first to order a largeelectronic computer:

It takes about five years for Mr. Management toget his feet on the ground in this field of appli-cation of electronics to administrative systems....The first competitor in each industry to operatein milliseconds, at a fraction of his former over-head, is going to run rings around his competi-tion. There aren’t many businesses that canafford to take a chance on giving this fellow afive-year lead. Therefore, most of us have to startnow, if we haven’t started already.9

This argument illustrates one of the mostremarkable continuities in the discussion of cor-porate computing. For a half century, enthusiastshave used the dazzle of microseconds and mega-hertz, joined more recently by the exponentialcurves of Moore’s law, to argue that the unprece-dented power of the latest revolutionary com-puter technologies is about to dictate acorresponding and almost effortless revolutionin business. The only thing to do is to get out ofits way, and maybe to employ their own servic-es. Yet on the human level—the level of organi-zational structure, productivity statistics, andmanagerial practices—change has been at bestevolutionary and invariably painful. Worthing-

ton’s claims closely mirror those used by a morerecent cohort of technology pushers to persuadefirms to sink vast sums into unproven andimmature technologies for electronic business. Itis a timelessly powerful and seductive claim: Thenew technology will rapidly and fundamentallyreshape your firm’s competitive position, it willtake years to fully implement, and it will conferinsurmountable advantages on the first of yourcompetitors to embrace it. As a result, there is nochoice but to invest massively in this “disruptivetechnology”—anyone waiting to see how and ifit works will be swept into the dustbin of busi-ness history by the insurgent competitors thatreally “get it.”10

The computer moved rapidly from a subjectof speculation among systems experts into anactual tool of corporate administration—bring-ing rhetoric of revolution into the mainstreamof managerial discussion along with it. TheHarvard Business Review had little time forunderstatement in 1954 when it published itsfirst detailed article on the application of elec-tronic computers to business. The article, ashameless piece of corporate self-promotion byGeneral Electric, boasted of the firm’s successin applying a Univac to automate its payrollproduction. GE’s Univac was the first purchasedby a private corporation and the first Americancomputer of any kind to be installed primarilyfor corporate administration. This move, theeditors opined, “may eventually be recorded byhistorians as the foundation of the secondindustrial revolution ….” According to the arti-cle, the computer could pay for itself by pro-cessing payroll for just 5,000 employees andrunning for two hours a day. Anything accom-plished by the computer beyond that would gostraight to GE’s bottom line. The followingyear, another GE manager laid down the gaunt-let to management in impassioned terms.Urging them to embrace the power of opera-tions research and the computer to automatetheir decision making, he asked, “Isn’t there adanger that our thought processes will be leftin the horse-and-buggy stage while our opera-tions are being run in the age of nucleonics,electronics, and jet propulsion?”11

The claim of revolution thus came in twoconceptually distinct forms. One was the oper-ations research or management science posi-tion that the computer would automate,optimize, and therefore revolutionize the orga-nizational decision-making process. Thesearguments received early and powerful expres-sion at a 1955 Harvard University conference,at which Herbert Simon, Russell Ackoff, andothers explored the computer’s potential as a

October–December 2001 77

scientific tool to fundamentally reshape thepractice of management. The title of a 1956article, “Can You Afford the ‘Practical’Approach to Electronics,” captured the claimsmade for this approach. Its author, a consult-ant, argued against the apparently low-riskapproach of using computers to gradually auto-mate current processes while leaving the over-all management structure intact. Instead, hewarned, the potential of the “visionary”approach was so great that the true risk lay inevolutionary thinking.12

The other claim of revolution was in clericalcost reduction through the direct substitutionof capital and electronics for clerks andmechanical devices. This more conservativeidea of revolution required its adherents toargue that the associated cost savings were sohuge that any hesitation in this area would beirresponsible. As one consultant put it, “Theestimated savings have sounded almost unre-alistic … if your company is not presentlyengaged in an electronics study program, isyour reason good enough?” This view wasbacked by the influential and generally conser-vative Controllers Institute in its first evalua-tion of computing’s economics. Its author,Frank Wallace, a partner with consulting firmPeat, Marwick, Mitchell & Co., explained to hispenny-wise fellows that “unwarranted cautioncan deny a company a major instrument ofcompetitive and financial leverage.”13

These two distinct claims (clerical and man-agerial) of benefits that were too good to miss,and too urgent to defer, came together in apowerful way. Order a computer now, save amillion dollars a year on clerical costs, and useits spare capacity to revolutionize manage-ment. Or prevaricate, and so sit out the secondindustrial revolution and be crushed by yourcompetitors.14

… Buying a computerFor many companies this was an easy choice

to make. Hundreds of computer installationswere ordered long before the computer’s eco-nomic value could be demonstrated.Contemporary estimates valued the hardwarein a typical installation centered on a singlelarge computer at $2 million—around $13 mil-lion in today’s dollars. In 1955, IBM hadinstalled only about 25 of its large 700-seriescomputers, chiefly the 701 and 704 modelsintended for scientific and engineering calcu-lations. Although the first of these had beensupplied in 1953, the majority of these instal-lations were still in the experimental stages.GE’s pioneering Univac installation had gar-

nered a spate of publicity in 1954. Yet, as PeterB. Laubach, part of a Harvard Business Schoolgroup investigating administrative computing,admitted in a 1955 Harvard Business Review arti-cle, the “revolution … appears to be off to a fal-tering start. Too much was promised too fast,with the result that many businessmen havegrown skeptical of the entire electronic data-processing field.” Of the dozen or fewer com-puters at work on administration, accounting,and statistics for business, he estimated that nomore than two or three were in full operation.15

An internal IBM market analysis, producedduring February 1955, indicates the company’sprogress in introducing large computers forbusiness administration. As the report’s fore-word noted, “Industry acceptance has beenever increasing, as evidenced by the many let-ters of interest received ….” The orders, it con-tinued, “seem to indicate a substantial marketfor machines of this capacity.” By this point,IBM had already taken orders for 99 of its large,administratively oriented computers. The situ-ation brought some comfort. But a questionmark remained because so few of thesemachines had yet been delivered, almost nonedoing useful work. As the report noted, therewas “very little actual experience, particularlyon commercial accounting applications wherethe bulk of the potential lies, to substantiatetheir worth to the eventual user.” It concludedthat administrative computing had reached aturning point, at which “[t]he success of theentire program will undoubtedly rest on thesuccess of the first few installations.”16

Without any proven savings, and few func-tional installations, to order a computer was anact of pure faith in technology’s transformativepower. Even the bullish consultant Thesis hadbeen obliged to concede that a manager look-ing for a demonstration would find “few exam-ples to be seen. You probably cannot see acomputer performing an operation similar toyour own.” Yet 1955 was a banner year for IBM,despite the firm’s caution. As well as workingits way through the fat order book for its 700-series machines, it began to deliver its medium-sized 650 machines. An average 650 systemcost about $3,750 a month to rent (equivalentto a purchase price of about $200,000).Although it was a true electronic, programma-ble computer, it processed information muchmore slowly than the large machines of the 700series and was designed to work closely withexisting punched card machinery. It proved tobe “computing’s Model T”17 —more than 1,100were in use for business applications by 1958.Both the 700s and the 650 were first-generation



machines, reliant on bulky and unreliable vac-uum tubes for their electronic capabilities.Other firms produced large computers (notablyRCA and Univac) and small ones (Burroughsand Univac), but these fit the same general pat-tern as the IBM models and offered the samecapabilities. Not until the end of the decadewere all these first-generation machines ren-dered obsolete by a second computer genera-tion of transistorized machines.18

In 1954 the computer was a revolutionarynovelty; by 1958, several thousand had beeninstalled. During the first four years of admin-istrative computing, estimated annual ship-ments of computer hardware rose from $10million to $250 million. Companies wereordering the machines faster than IBM couldbuild them. New machines were normallyannounced a year or two before the first modelswere delivered. Even in 1958, a companywould typically spend a year or two waiting forits newly ordered computer to arrive. Scarcitywas an incentive for companies to gain a placein line by rushing in an order that could laterbe canceled or delayed. IBM granted preferen-tial treatment to those firms that had placed anorder, both in admission to its programmertraining courses, and in the allocation of pre-cious practice sessions on its own computers.This practice gave these firms a sporting chanceof having some programs in a state of advanceddevelopment by the time the computer arrived,but it also made it hard to complete a thoroughstudy without placing an order.19

How to sell a revolution? At least some ofthe eager new computer salesmen sought tobypass their traditional contacts in office man-agement, punched card, or systems depart-ments. In 1954, Dun’s Review and ModernIndustry hosted a roundtable discussion amongrepresentatives of office machine companies—whose candor was solicited by anonymity. Thenew technology, they confided, demanded anew approach, one to which their existingsalesmen could not easily adapt. The salesmenhad “to sell above normal channels. We’ve gotto go to the top.” Claimed one of the comput-er company executives: “We just will not talkto anybody below controller.” Office managers,a traditional audience for the salesmen, had lit-tle clout. Even time spent cultivating the high-er powered administrative “systems man”would likely prove wasted on discovery that“he has no power to purchase.” Instead,

[W]e try to use him as a bridgehead within anorganization to get an opportunity to study andmake a survey of that particular company’s

needs. Then we go back and make a proposal andtry to see to it that the systems man, either as ourambassador or going with him, presents that pro-posal to the person who has authority to buy.

Along with this new audience went a new kindof pitch. Rather than selling individualmachines, they asked instead “[d]on’t youthink that the office should be sold to manage-ment as a production unit like the machineshop?” To this end, one firm boasted that “Wehave stopped our salesmen using the word‘machines’; it’s out of the vocabulary entirely…. That’s shown last and talked about last, nomatter what the application is …. If you sell theidea, then you’ve got the sale.”20

How, and why, did so many companieschoose to order such expensive and unprovenmachines? Clever salesmanship could not, initself, explain why executives were willing totalk to computer salesmen, still less why theyagreed to order a computer. Were the potentialsavings gained from replacing clerks andpunched card machines with million-dollarcomputers so compelling as to render the wait-and-see approach more dangerous than thefools-rush-in philosophy? It seems unlikely.There is little evidence that companies thatwaited until, say, 1960 to install a computersuffered any negative results. So why did somany large, well-run companies manage toorder so many computers before either thecosts or the benefits were known? While schol-ars of the “new institutionalism” in organiza-tional analysis have drawn attention to fadsand herd behavior as powerful factors inspreading organizational features, this does notexplain how momentum first gathered. Afterall, our impression of 1950s corporate manage-ment is one of conformity, conservatism, andinflexibility than a group willing to gamble mil-lions of dollars on a whim.21

The answer lay in a ritual known as the fea-sibility study, used by businesses from the earlydays of administrative computing to investi-gate the computer’s potential. This studywrapped a host of unknowns, unknowables,and hopeful guesses in the apparently rationallanguage of financial analysis. Some of the ear-liest books and articles on the use of computersin business devoted themselves to examininghow such a study should be conducted. Myanalysis here relies primarily on five book-length guides to the study of computing. Twowere published by the Controllers Institute andwritten by consultants (one by Wallace, one bya Price Waterhouse team). Two were written byRichard G. Canning (author, consultant, and


publisher); the final one was written by theHarvard group’s Peter B. Laubach.22

These authors agreed that the feasibilitystudy should be rigorous; they included thecomparison of machines from several manu-facturers and identified potential areas forcomputer application. In a good study, theysuggested, sufficient analysis and preliminaryprogramming would be conducted to accurate-ly judge how efficiently each possible comput-er model would run a given job, and how manyhours each month it would take. Armed withthis knowledge, a company could calculatehow many clerks and punched card machinesthe computer would save once it was fullyoperational. This exercise would yield an esti-mate of the computer’s benefits. The costsseemed easier to estimate—rental for the com-puter, the one-time costs of installation, analy-sis, and programming, plus the recurring costsof consumables (tapes and cards) and of com-puter department operators and supervisors. Ifthe discounted benefits exceeded the costs,then the machine should be ordered.

On its face, this exercise appeared to placethe installation of a computer on the samerational basis as the decision to invest in a newset of lathes or to build a warehouse. In practice,however, it was deeply flawed. Although somecompanies ordering these computers spentmany man-years of effort learning about com-puter technology, the information needed tojudge the computer’s economic potential sim-ply did not exist. Nobody knew what the costof programming might turn out to be, whetherobsolescence would be a serious problem, howto quantify the so-called “intangible benefits”of improved management, or whether the com-puter could cope with nonroutine activities.Pioneers had already discovered that the stud-ies made by eager representatives of the com-puter manufacturers could not be takenseriously—these systems studies tended tounderestimate everything from the floor spacerequired to the complexity of the programsneeded to undertake a task to the cost of con-verting data from manual methods.23

Estimates of how much computer time aprogram would need to run were sometimes offby a factor of ten. This meant that the comput-er would run fewer tasks and generate fewerbenefits. The lack of available computer capac-ity made it almost impossible to write and runprograms before making the commitment toorder a computer. In addition, the expense ofprogramming was massively underestimated.As late as 1956, it was often viewed as a one-time expense to be amortized over the life of

the computer, rather than an ongoing and con-stantly increasing black hole in the budget.24

Another problematic aspect of any calcula-tion was the computer’s assumed life span,which determined the period over which thestart-up costs of programming, conversion,installation, and training could be spread. Inhis well-researched 1956 book, Office Work andAutomation, Howard S. Levin presented somereturn-on-investment figures, using estimatesthat represented the prevailing consensus.Levin showed that, while purchase (as opposedto rental) could be justified using accepted costestimates and a 10-year assumed life span, “nei-ther rental nor purchase is supported if weassume a five-year useful life for the computersystem.” Thus, even the optimistically low esti-mates of costs and high estimates of benefits inthe mid-1950s could justify a computer’s acqui-sition only if the computer were assumed toremain useful for substantially longer than fiveyears. Punched card machinery, after all, wasdepreciated over 16 to 20 years according to IRSguidelines of the period—and none had yetbeen established for computer equipment.25

Benefits were just as hard to predict. TheControllers Institute study outlined a reason-able start to estimating clerical cost savings. Afirm should chart its most important clericalcosts and show the amount incurred perform-ing each task. Then came the leap of faith—how much of this could the computer save? “Itis obvious,” wrote Wallace,

that a computer could not replace all clericalcosts. Therefore, the clerical costs must be high-er than the cost of operating a computer. The artis not sufficiently advanced to give any rule ofthumb indications of how much higher presentclerical costs must be.26

Estimates of high cost savings were premisedon the idea that all or most of the clerks couldbe eliminated after conversion was complete—what else could automation mean? Later stud-ies found the actual potential for eliminationto lie between zero and 25 percent of the exist-ing clerical workforce in the area automated,although many firms claimed to have reducedthe rate at which their clerical staff expanded.Existing clerical work was much less routinethan had been assumed.26

The longer a company spent on its study,the more momentum built up behind the com-puter. Before it was ordered, still less installed,the computer gave office managers, and cleri-cal systems and procedures experts, a means toincrease their status in the eyes of their superi-



ors. The Harvard team, for example, docu-mented several firms in which their role in thestudy led to a formal upgrading of such groupswhile the study was still in progress. Expertswith a strong systems and procedures back-ground were adamant that the study shouldproduce figures on different options, includingimproved efficiency through better proceduresor the improved use of conventional punchedcard equipment. All agreed, however, that thestudy team was far more likely to be chargedwith a “yes” or “no” answer to the question ofcomputer acquisition than with an open-endedstudy of all possible technological, procedural,and organizational avenues for administrativeimprovement. Given this, the study was liableto become an exercise in the rationalization ofa decision to acquire.

Case studies of processes used by actualcompanies showed that the ordering decisionwas usually made without detailed estimates ofcosts and savings, and certainly without thetrial programming work that experts recom-mended. Canning, for example, confided that“Too often, the phrase is heard, ‘We are goingto use the XYZ machine but we’re not just surehow we are going to use it.’” In those firmswhere initial interest came from top manage-ment—rather than from lower level account-ing, systems and procedures, or tabulatingpersonnel—purchase was still more likely to beauthorized without detailed planning. Expertsalso accused firms of relying too closely on theestimates and studies provided by computermanufacturers.27

Study teams acquired a particular kind ofbias. Their self-selected members had spentmonths or years immersing themselves in theexciting new world of computing, learningabout hardware, writing sample programs,attending conferences, bonding with manu-facturers’ representatives, and hiring consult-ants. By the end of this process, they had begunto reorient their careers away from accountan-cy or office management and toward theemerging field of electronic data processing.Moreover, as a firm’s computer experts, theycould expect to wield considerable power with-in a new department if they recommended theacquisition of a computer. As John Dearden, aprominent skeptic, later observed:

Management should recognize that the recom-mendation of the feasibility study will almostcertainly be to acquire a computer and that itwill be difficult, at that time, to override this rec-ommendation. In fact, in many instances theonly decision management really makes is to

authorize an initial study. From the moment ofauthorization, the project develops momentumthat is just about impossible to stop.28

Doing the same things, fasterWhatever the stated reason for ordering it, as

the delivery date for a computer approached,the recipient had to choose specific tasks for itand undertake the necessary programming andconversion work. At this point, the wide-eyedenthusiasm, such as evinced by methods expertBerkeley and others intent on the computer’sundergirding a new approach to managerialdecision-making, was often displaced by theurgent need for the installed computer to douseful work immediately. This more pragmaticmind-set took hold, as word of difficulties spreadand earlier language came to seem embarrassing.As the Harvard group concluded, “the so-calledgiant brains cannot think. Looked at in properperspective, automatic data processing methodsare merely an extension of present punched-card data-processing methods ….”29

To cause a revolution, the computer wouldhave to offer something businesses had beenunable to achieve without it, whether an enor-mous reduction in clerical costs or a transfor-mation of its operations. The evidence is clearthat, typically, the administrative computers ofthe 1950s merely supplied what was already per-fectly attainable with punched card machines ormanual methods. The computer may have pro-vided reports, totals, and other output more rap-idly—and almost certainly at higher cost—thanprevious methods, but it did not substantiallyalter the domain of possible work.

In 1957, a survey conducted by the NationalOffice Management Association found that halfof the largest firms examined (those with morethan 5,000 office workers) had already installedat least one large computer, such as the IBM 700series. An additional 14 percent had orderedtheir first large machine but not yet received it.All these firms were still running conventionalpunched card installations alongside the newmachine.30 At this point, more computers wererunning engineering computations of one kindor another (56 percent) than payroll or inven-tory control (38 percent each), reflecting theearly dominance of scientific over administra-tive computing. However, companies still await-ing delivery of their first computer were muchless likely than the pioneers to have earmarkedit for technical calculations, as the dominantcorporate applications of large computers werealready shifting from scientific and technicalcomputation to administrative tasks. Whenasked about future intentions, 98 percent of the


firms already using computers either had inven-tory control programs running or planned todeploy them in the future, making this by farthe most widely considered application.31

The purposes to which the new million-dollar electronic computers were applied borea striking similarity to those already performedby punched card machines. One third of theoverall sample were using punched cardmachines, in contrast to the 0.4 percent usinglarge computers. The average punched cardinstallation ran six different jobs on itsmachines. The three leading punched cardapplications were (in descending order) salesstatistics, payroll, and inventory calculations—the same applications that dominated the com-puter’s administrative use.

Why payroll? The payroll run had a numberof attractive characteristics. It took place everyweek and was handled in the same manner onevery occasion. It applied to the whole companyyet seemed reasonably straightforward and rou-tine. A large company had tens of thousands ofpeople on its weekly payroll, ensuring arespectable volume, and the job required enoughcalculations to be highly time-consuming bymanual methods. Tax, overtime, union dues,retirement benefits, vacations, and bonuses allhad to be considered. Much of the complexity inthe process came from legislative requirements,making it a process that would be around forsome time to come and could not be eliminatedby procedures improvements or organizationalstreamlining. But, in principle at least, it was notso complex and full of special cases requiringhuman judgment as to be impossible to auto-mate by computer. Many companies had stan-dardized and centralized payroll operations,reducing the chances of conflict with divisionalmanagers. As we have seen, many firms werealready using punched card machines in theirpayroll runs, meaning that much of the requireddata had already been coded into machine-readable form. Payroll stood as good a chance orbetter as any potential application to pay foritself through clerical savings.

The publicity given to GE’s choice of payrollas the first application for its administrativeUnivac also directed attention of subsequentcomputer installations toward payroll. This ledthe Harvard report’s authors to worry thatinfluential “leader” firms might lead followerfirms astray through blindly emulating theirchoices. Some observers, especially those withan orientation toward operations research ormanagement theory, complained that to usesuch a powerful machine on such a mundanetask was to squander its potential in making a

significant difference to company manage-ment. But it was the similarity of payroll andother early computer applications to routineclerical tasks already performed by punchedcard machines (together with their potentialfor tangible job savings and the lack ofupheaval to managerial culture) that madethem so attractive. The Harvard team foundthat one of the firms it examined had madethis choice quite deliberately. On the advice ofa consultant with a punched card background,the firm had chosen to “mechanize the exist-ing system. In his opinion, it was hard enoughfor existing clerical personnel to adapt to theprocess of mechanization without having toface changes in the system as well.” Given theshambolic process by which many orders wereplaced, this attitude might have unavoidablyaffected companies, as they pondered the alter-native of a computer sitting idle while man-agement hired operations research specialistsand haggled over politically charged corporatereorganizations.32

This conservative application of computertechnology, as an extension of the punchedcard machine, triggered a rethinking of the roleof punched card specialists in the new order ofthings. Early computer studies tended to down-play the importance of punched card experi-ence when working with the new machines.The Harvard group advised those assemblingstudy teams that the punched card people werelikely to be inflexible and tied to outmodedthought patterns. As the group snidely put it,“company executives need have no concern ifpunched-card tabulating equipment men werenot available to work on an automatic dataprocessing project.” In practice, however, con-tinuity in the tasks to which the machines wereapplied was superficial—inside the new com-puter department were many of the people,attitudes, and occupational identities of the oldtabulating group. It was the revolutionaries,not the punched card staff, who were most like-ly to feel out of place there.33

Life in the tab roomTo understand the roles that eventually

emerged for the punched card machine staffwithin America’s new electronic data process-ing departments, we must explore the cultureand work practices of the “tab room.” By the1950s, the corporate punched card departmenthad been evolving for decades, but was still arelative novelty. Although the first punchedcard machines were used during the 1880s, itwas not until the 1930s that the machinesentered the mainstream of corporate adminis-



tration. Punched card technology advancedgradually over the intervening decades, pro-ducing machines that could print (as well asdisplay numbers on dials), manipulate lettersas well as numbers, and store more informationon each card. It spread slowly from its earlyniches in actuarial and cost accounting appli-cations into a host of others, such as billing andpayroll. By the 1930s, IBM had begun to referto them as accounting machines, signifying anattempt to reorient the machines toward thework performed by conventional bookkeepingmachines and away from purely statisticaltasks. Historians recognize the New Deal in theUS as a turning point in the fortunes of thepunched card machine industry, already dom-inated by IBM. Thanks in part to the massivenumber of machines supplied to the SocialSecurity Administration from 1936 on, IBMdoubled its revenues to finish the Depressionas America’s most profitable office machinecompany.34

Punched card machines were specialized toparticular tasks. The tabulator was the centraland most complex of these machines, but itcould do little on its own. When appropriatelywired, the tabulator could print reports basedon the cards fed through it. These reportsincluded totals, subtotals, headings, and textu-al information (such as names and addresses)from the cards. To include in these totals infor-mation from only some of the cards (say, thosefor a particular department or job code)required that the cards be run first through adifferent machine, known as a collator. Themachine operator would then pick up theappropriate cards and carry them over to thetabulator. The card order was important. To listsalaries, by department and then by job classi-fication, required collating the cards into sepa-rate piles for each department and sorting eachpile by job classification. Only then, when itsensed the department or job classificationcode changing, could a tabulator properlyinsert the appropriate totals and headings.

Even running a simple salary report requireda functional installation to have at least threemachines—one keypunch machine, a tabulat-ing machine, and a sorter. Most installationsalso had a collator. The machine operators hadto wire each one for the new task, then split theoverall job into a series of steps in which theyran the cards through each machine and man-ually moved card stacks between them.Transferring information onto the cards in thefirst place required a keypunch (like a type-writer, but it put holes into cards rather thanletters on paper). This operation was often

repeated using another machine, a verifier, toensure the initial input was correct. Keypunch,tabulator, sorter, and collator were just themost common kinds of machine—by the1950s, IBM offered gang punches, interpreters(for printing text onto cards), reproducingpunches, multipliers, calculating punches, anda number of hybrid machines. The most tech-nologically advanced of these already includedelectronic components and could be config-ured with a measure of programmability.35

Renting these machines was not cheap, so alarge and well-run department would try toschedule its jobs so that most machines were inuse at any given time. Their use was labor-intensive. First, each machine had to have itscontrol board specially wired for a given task(printing a report, copying certain parts of acard, or sorting by a certain field). Second, oper-ators had to feed batches of cards in and out ofeach machine, deal with jams and other prob-lems, and transfer cards between machines.Most punched card departments of the 1950swere small, probably employing a nationalmedian of fewer than 10 people. The largestcompanies, however, had enormous punchedcard departments. By 1951, PrudentialInsurance already spent more than $1.6 milliona year on salary alone for 600 tabulating staffspread among its 13 separate punched cardinstallations, plus another $700,000 on keypunching and verification.36

Punched card departments included twoalmost entirely separate classes of worker.Keypunching was a clerical job, and operatorswere invariably female. As in most other white-collar women’s jobs, their most realistic avenueof advancement from this position was to over-see other women, in this case as a supervisor ofkeypunch operations. On average, two key-punch machines kept up with the work han-dled by one tabulator, though this ratio variedaccording to the type of work. The only changethe computer made to keypunch work was togreatly increase its volume.

The other kind of role, typically known as amachine operator, was usually filled by men,particularly in the larger and more formallyorganized installations. The punched cardmachine operator was in essence a skilled craftworker. Although manufacturers offered shortcourses in punched card skills, most learnedtheir trade on the job. Trainee machine opera-tors carried cards between machines and ran themost routine jobs. As they progressed, they weremore likely to be trusted with the execution ofcomplex jobs and the wiring of machine pan-els. The most experienced operators were


charged with establishing wiring diagrams andprocedures for new jobs, and supervising morejunior employees. This progression remainedstandard practice well into the 1950s.37

Most punched card installation heads hadworked their way up from machine operators.Few punched card groups enjoyed high organi-zational status—the heads of most installationswere formally known as supervisors rather thanmanagers. They often gained experience inlarge, frequently governmental, installationsbefore accepting more senior posts in smaller,newly established departments. Few punchedcard staff members had college degrees,although almost all had graduated from highschool. The rapid proliferation of punched cardtechnology from the 1930s to the 1950s meantthat openings were plentiful for the ambitioustechnician. This boom continued well after thecomputer was introduced. A 1958 survey of 42installations in Oklahoma City found that mosthad been created since 1952, and that only onewas more than 15 years old. In 1961, the USboasted more than 35,000 punched card instal-lations. Not until 1962 did IBM’s revenue fromits computer activities finally overtake the con-tinuing flow of rental income from its punchedcard business.38

Working the computerThis growth might have been assumed to

offer a steady future to the punched card tech-nician. But in 1953, Richard W. Sprague con-fronted his audience of punched cardsupervisors with the disturbing question, “Arepunched card machines on the way out?”Sprague was in charge of applications and salesfor Computer Research Corporation, then asmall manufacturer of computers owned par-tially by the cash-register giant NCR. Spragueanswered his own question with a qualified“yes.” While he accorded the punched carditself a secure future as a medium for informa-tion exchange, he suggested that the tradition-al punched card machine was about to be cutdown in its prime. Sprague’s reasoning illumi-nates for us the power attributed to the com-puter at a singularly optimistic moment in itshistory. Electronic computers had been shownto work, and their commercial developmentwas proceeding apace. The computer’s poten-tial was clear, but as yet unsullied by the prac-tical frustrations and limitations that wouldemerge when the computer was applied toadministrative problems. While Sprague’s salesposition gave him little incentive to expressreservations, he had cofounded the company,and his address was marked by the unmistak-

able fervor of a true believer.39

According to Sprague, the computer’s powerlay not just in its speed or in its superiority to cer-tain existing punched card machines, but in itspower to replace the combination of humans,machines, and procedures that made up apunched card installation. He suggested that

the electronic machine is capable … of perform-ing automatically and with no human interven-tion not only all of the functions beingaccomplished by all of the card machines, butalso all of the functions performed by all of thepeople involved in a punched card system up toand including the head of the department.

Sprague saw the replacement of the punchedcard staff as a boon to efficiency: “People dropcards, forget to pick up cards, get called away forsomething and let cards pile up, forget whatthey are supposed to do with them …” Whenlisting the human foibles to which the com-puter would be immune, Sprague mentionedcigarettes, Cokes, pregnancy, resignation, psy-chological problems, and union membership.Furthermore, modifying a punched card opera-tion or introducing a new one required retrain-ing the operators of each kind of machineryand enduring a period of inefficiency and inac-curacy until the operators mastered the newjob. A computer, however, could take a newprogram and run it perfectly every time.40

The flip side of this flexibility was that ittook a great deal of work to program the com-puter to undertake a single, specific task. Toachieve such automatic operation, the programhad to perform the operations of the more spe-cialized punched card machinery and also toreplace the procedures, judgment, and excep-tion and error handling formerly supplied bythe machine operators. When Sprague wrotehis article in 1953, almost no administrativetasks had yet been programmed. His expecta-tions for the programmer were high:

He can program in all of the sets of rules being fol-lowed by all of the present card machine opera-tors, supervisors and department heads. He caninclude all of the exceptional cases that haveoccurred, and are likely to occur, and instruct themachine as to what to do about them. He can pro-gram the function of upper management in mak-ing decisions if he puts in all the possibilities.41

In reality, operators proved as essential to thesmooth operation of the computer as theyremained to the operation of the punched cardmachines running alongside it, though the



character of the job was rather different. Earlycomputers did not have operating systems. Likepunched card machines, their continued oper-ation from minute to minute demanded con-stant attention from human operators, just likepunched card machines. True, there was noneed for an operator to rewire a control boardin order to ready the computer to run a job.And, within a program, the computer couldadvance from one instruction to the next with-out waiting for the operator to pick up a pile ofcards or flick a switch. But readying the com-puter to run a program and then shepherdingthe job to completion was not the effortlessoperation that Sprague promised. The most sen-ior operators would schedule jobs and work thecomputer console, basically a large desk filledwith switches. Tasks here included resetting thecomputer and configuring it for each job, load-ing programs into memory, restarting the com-puter after hardware or software errors (ideally,without losing all the work in progress),responding to errors raised by the applicationprogram, terminating a program that had mal-functioned, and supplying the programmerwith clues needed to debug it. Tasks for themore junior operators included the mountingand de-mounting of tapes into drives, copyingcards onto tape, sorting cards, configuringprinters, and loading them with appropriateforms.

Sample 1956 staffing figures presented byWallace’s report for the Controllers Institute sug-gested that a large computer installation wouldrequire three operators per shift plus a chief oper-ator and a tape librarian—a total of 11 operatorsfor a three-shift, five-day-a-week computer oper-ation. In contrast, only four programmers werebudgeted for, which was likely a substantialunderestimate. A 1957 survey of installations ofthe large, administratively oriented IBM 702computer suggested that a typical installationmight have a dozen tape drives, while PrudentialInsurance had hooked up an impressive 19 to itscomputer. With two large printers and two cardreaders accompanying these machines, it’s easyto see why so many people were required towork them. Despite the number of operatorsrequired, business computing literature of thelate 1950s paid less attention to the problem ofhiring operators than to hiring programmersbecause a convenient source of operators layclose at hand: punched card machine operators.As Computing News advised its readers in 1957:“As a rule, your good tab operators will makegood EDPM [electronic data processing machin-ery] operators…. Your operators know their pres-ent jobs—a paycheck is still a paycheck, even

when processed by EDPM. Through experience,they know the pitfalls and exceptions.”42

Throughout the 1950s, computers weremore likely to supplement than to supplantpunched card systems. Even the medium-sizedIBM 650 still relied on punched cards for inputand output. The traditional assortment ofsorters, collators, and tabulators worked along-side this electronic marvel. The first 650 mod-els could not print without the intervention ofa regular tabulating machine, and even thecapability to handle alphabetical charactersinvolved an optional upgrade. While nobodywas in danger of mistaking a “large” computer,such as the Univac I or the IBM 702, for a meretabulating machine, even these room-fillingmonsters were almost invariably used alongsidepunched card machines.

These machines were particularly importantfor input because there was no means of punch-ing input data directly onto magnetic tape—thefirst such machine was produced only in 1965.No business computer of the 1950s alloweddata entry directly into its internal memoryfrom a keyboard; instead, all input was punchedonto cards. Whenever a new job was convertedfor the computer, it required an army of key-punch operators to enter every account code ortransaction record onto cards. Keypunch work-ers made up the majority of all data processingstaff at most computer installations—outnum-bering programmers, operators, analysts, andmanagers combined. Keypunching was excep-tionally time-consuming—in the first textbookof administrative programming, Daniel D.McCracken noted that conversion could takemore man-hours than everything else com-bined, while Business Week reported that anunnamed insurance company spent $12 mil-lion readying its files for the computer. Eachrecord was often entered twice, to verify itsaccuracy. Consolidating existing files couldrequire a great deal of computer time and spe-cial programming. Paper records themselvesoften contained many errors and inconsisten-cies, so although the “cleanup” mandated by acomputer conversion could be rationalized as along-term benefit, it was also difficult enoughto cause many early efforts to overrun severelyon both time and cost.43

Even once a job had been converted to thecomputer, punched card machines remainedmuch better at some tasks, notably the sortingof records into a particular order. In a punchedcard system, each record had its own card. (Forthis reason, unit record equipment gained wide-spread use as a formal term for punched cardequipment during the 1960s.) Sorting was


highly important in punched card or earlycomputer routines, because neither machinecould do much more than work on one recordat a time. For example, to update account bal-ances or print statements, it was necessary tosort all new transactions records in order oftheir account codes and then merge informa-tion on the latest transactions with informa-tion from the master account records. Sortingrecords on punched cards was easily accom-plished by shuffling them into order. Becausetape could not be cut into little pieces andstrung back together in the right order, a sortrequired, minimally, two tape drives for read-ing and one for writing, a complex program,and many repetitions before the file was finallysorted. Coupled with the unreliability of earlytape drives, the repetition was often enough tonegate their raw advantage in terms of speed.Consequently, sorting the punched cardsmechanically before transcribing them ontotape could be beneficial. As it struggled toimprove its efficiency, GE’s pioneering payrolleffort was eventually forced to go one step fur-ther: It used three clerks to sort the paychecksafter they were printed, rather than wastinghours of valuable machine time on the task.44

Sorting was perhaps the most common taskwhere computers fell short. The sequentialnature of tape storage meant that it was pro-hibitively slow for cases when a single recordneeded to be looked up—the computer mighthave to read through the entire file to find it. Ifthe file was big enough to justify a computer,then it was too big for this search procedure tobe cost-effective. Clever programming allowedsuch requests to be batched and combined withroutine updates, but, even in the best case thismeant that the information could not beretrieved until the next day. The only solutionwas for the computer to print out enormousreport books, enabling the required detail orsummary information to be looked up manu-ally. Because these operating reports could eas-ily reach a size that demanded the use of ahandcart or small truck to deliver, this situationled to a great deal of interest in microfilm as apossible distribution mechanism for comput-erized reports.

Business computers of the 1950s were alsounable to exchange data with each other direct-ly or to drive remote terminals. To transmitinformation from a remote site for computerprocessing—for example, to send orders fromsales offices to a warehouse, or time sheets fromplants to the payroll office—a companyrequired another set of intermediary technolo-gies, known collectively as integrated data pro-

cessing. The same forward-looking firms thatflocked to the computer were also fitting theiroffices with communication systems such aspneumatic tube networks and centralized dic-tating systems. But the most versatile technol-ogy for data transmission was the five-trackpaper tape—a ticker tape punched with up tofive holes across its width—widely promoted asa “common language” for the interchange ofinformation between different kinds of officemachines. It could transfer information frombookkeeping machines into punched card sys-tems, or automatically operate specially adapt-ed typewriters called Flexowriters.45 Some firmstied punched tape readers to leased data linesand telegraph-style message forwarding tobuild their own national networks.

Throughout the 1950s, therefore, the com-puter remained just one technology in largeradministrative systems, supplementing ratherthan replacing punched cards, paper tape, andmultipart forms. Its symbiotic relationship withthese technologies created many opportunitiesfor people already familiar with them to shiftinto computer operations. Computerizationcreated new demand for their skills, and in gen-eral this was a move up.

The shock of the new furnitureFor the punch card staff who became opera-

tors of the new machines, the computer’s arrivalbrought an immediate upgrade in status andorganizational prominence, if not in formalauthority. They moved, quite literally, upwardinto the light. Unlike punched card workers,who progressed gradually from the slavish exe-cution of existing procedures to the design ofnew ones as their careers developed, computeroperators were neither expected nor allowed tomodify the programs that they ran. They mighttherefore be seen as (in the terminology of laborhistory) deskilled in relation to their predeces-sors. But, in practice, computer operators werebetter paid and more respected than punchedcard operators, if less well paid than program-mers. In addition, their work was closely alliedwith programming and could become anavenue for upward mobility within the dataprocessing department. Work as computer oper-ators provided many punched card personnelwith a bridge into the computer age.46

The useful punch card machine and its hum-drum technology was an unlikely attraction forvisiting dignitaries, although it stood at the apexof office machinery as one of the most complexmechanical devices ever to be mass-produced.Its thousands of parts, miles of wire, enormousspeed, and formidable accuracy earned it a place



in the heart of the mechanically inclined. In itsearlier years, it had itself been no stranger tohyperbole. But the spread of punched card tech-nology had been so gradual that the machines’capabilities—and their limitations—were a mat-ter of fact rather than fancy.

Nowhere was this difference more pro-nounced than in the physical environment ofthe two technologies. The tabulating depart-ment was a noisy, often uncomfortable place.The machines clanked and rat-tat-tatted as hun-dreds of cards per minute poured through them.Even those that did not punch new holes, suchas sorters, made a racket as they read the cardsand dropped them into different chutes.Duplicators and punches thrust rods in and outof the cards, and tabulators printed columns ofresults. The department was also likely to be hotmuch of the year—neither the imperviouspunched card machines nor their operators werelikely to command the comforts of air condi-tioning. Punched card machine installationswere also crowded. According to a contempo-rary survey, most installations were crammedinto the “left-over space” of existing buildings,so that “the space is often overcrowded by themachine and operators have almost no room inwhich to maneuver.” This was not an altogetherwhite-collar environment. As one punched cardveteran of the 1940s recalled, “When the weath-er got too hot (and after the women secretariesand control clerks left), we men would stripdown to our shorts.” On one occasion a suddenthunderstorm left many cards damp, afterwhich they had to be ironed carefully beforethey could be read.47

If there is one aspect of corporate adminis-tration in which early computer installationstruly launched a revolution, then it is an unex-pected one: interior design. IBM apparentlypioneered the design in 1948 with its SSEC, aone-off machine whose role served public rela-tions more than anything else. It boasted flash-ing lights, glass panels, and a publiclyaccessible location on the ground floor of itsworld headquarters in midtown Manhattan.

Nomadic computer pioneer Herb Groschtook this aesthetic with him when he left IBMto head the first scientific computer installationat GE. Although the computer was originallyslated for basement installation, Grosch soonhad it housed in what he claimed to be theworld’s first specially designed computer build-ing. Grosch chose futuristic Herman Milleroffice furniture, modernist design, large win-dows, and a carefully coordinated color scheme.The building’s unveiling was part of an eventattended by a host of senior GE and airline

executives, IBM’s leadership, and the top brassof the armed services. But while its woodenfloors and tropical fish tanks impressed interna-tional visitors, they also attracted the ire of GEmanagers working in more dowdy surroundings(or, as Grosch put it, “my peers and their jealousminions”). Meanwhile, the gradual transfer ofpower within IBM between Tom Watson seniorand junior led to a redoubled interest in indus-trial design in its regular product range. Theresult was IBM having its own prestigious show-piece 702 installation behind plate glass win-dows and on perpetual display to passersby.48

A distinctive architectural style developed, asfirms maximized their computer room’s visibil-ity in the most literal sense, placing the com-puter behind huge plate glass windows andapplying diffuse lighting to illuminate it againsta brilliant white background. Throughout the1950s, a large computer installation’s noveltyand symbolic modernity could be counted onto unleash a flood of upbeat local news storiesabout the newly installed giant brain. Whetheror not the computer was yet doing useful work,it became the site of a Potemkin village of clat-tering printers, spinning tape drives, and flash-ing lights. Visitors and reporters could not judgethe usefulness of what was being produced, stillless its cost savings, so it was more importantthat the computer be seen to operate than thatit improve managerial effectiveness.

After acquiring a Univac, the comptroller ofPacific Mutual Life went so far as to hire a full-time publicist to speak before local groups, con-duct tours, and entertain the policyholders,agents, and international visitors who passedthrough the specially constructed viewinggallery. He was “convinced that our companyhas benefited from the publicity of having thefirst commercial installation of this type westof the Mississippi ….” As late as 1960, a trademagazine reported with approval that, “A com-puter installation can have tremendous publicrelations value to a company. Attractive, longwindowed corridors permit an unobstructedview for the visitor without interfering with thesystem …. The dull and drab grays and blacks,once the official colors of the machine account-ing industry, have given way to the rainbow.”49

While the fish tanks, glass windows, anddesigner furniture were of symbolic importance,the computer—because of its large numbers oftemperamental electronic components—demanded a different environment than thepunched card machine. Despite the improvedsurroundings, the vacuum tubes were prone toburn out unexpectedly, especially when themachine was switched on. Many companies got


their first taste of this with electronic punchedcard machines, such as the IBM 603 and 604.When a 604 was first installed, it might beturned off each weekend. After discovering thatturning it back on usually caused a vacuumtube to fail, however, its operators realized thatthey should leave it powered up permanently.Such was their introduction to the electronicage. When computers arrived, things only gotworse.50

The only thing that early computers couldbe relied on to do was break down, exemplifiedby the first scientific computer installed at GE,an IBM 701. During 1955, GE kept its comput-er turned on for 6,600 hours—paying the addi-tional rent to IBM for a three-shift operation.About 1,400 hours of this was lost to mainte-nance (most of it scheduled). Unreliable elec-trostatic memory, used on the 701 and theadministrative 702, was to blame for most ofthis downtime—within a year or two, the supe-rior core memory of the 704 and 705 hadreduced it dramatically.

Magnetic tapes were even more tempera-mental. Early users of the large IBM computersfound that they needed to clean each tape twicedaily to ensure reliability. Even the tape storagearea had to be kept at carefully controlled lev-els of humidity and temperature. Tiny amountsof dust could wipe out data, meaning that acomputer installation might require its ownmeticulous janitor. Although IBM engineersdesigned much more reliable models than theircounterparts at Univac, even IBM drives func-tioned well only under a very narrow range ofenvironmental conditions. Early users reportederror rates of one per 2,000 records before theenvironmental conditions were adjusted and asuitable cleaning regimen introduced.51

A Univac I installation occupied around2,500 square feet, and the 5,000 vacuum tubesin its central processing unit burned enoughelectricity to require a two-and-a-half-tonpower supply. Cooling all this involved a sys-tem of internal water pipes and an air-conditioning system. Air conditioning was stillnovel for office use, and the equipmentremained bulky and temperamental. As a mag-azine article reported, “For the averagepunched card installation these things wereconsidered luxury, but for a computer systemthese are absolute necessities.”52 The mass ofcables, pipes, and power lines demanded by theinstallation prompted most firms to installraised floors and false ceilings. These not onlypreserved the clean, modern look of the com-puter installation but also spread the enormousweight of the equipment more evenly and

absorbed noise. Fitting this equipment into anexisting building might require the temporaryremoval of a large part of a wall, the installationof many temperature and humidity monitorsto ensure even airflow, the removal of existingsprinklers, the sealing of walls and floors toreduce dust, and the installation of expensivevinyl flooring.53

Systems analysis and flowchartingWhile operation of the computer fell to the

staff and supervisors of existing punched carddepartments, its application to administrationproblems was coordinated by a different figure:the systems analyst. The analyst was expectedto serve as intermediary between the technicalinternals of the machine and the needs of dif-ferent managerial and operational groups with-in the corporation. The primary tool of thesystems analyst was the flowchart—a symbolicdescription of administrative proceduresintended to provide an unambiguous descrip-tion to guide their successful computerization.As one analyst wrote in 1957, “It is not too faramiss to think of the systems profession asstemming in large part from the developmentand refinement of the flowchart.”54 Yet neitherthe role of analyst nor the techniques of flow-charting originated with the computer. As aresult, their relationship to the rest of the com-puter department was initially uncertain. Theirtools and approaches sometimes had more todo with managerial aspirations than with thespecific demands of early computer technology.

During the 1950s, a new occupational groupwas claiming specialist authority over theimprovement of clerical procedures and thereorganization of interdepartmental systems ofcoordination. While the new specialists weregiven different titles, such as methods analystor clerical procedures specialist, their preferreddescription of “systems man” was reflected inthe name they gave to their association, theSystems and Procedures Association (SPA).55

While many of the systems men used punchedcard machinery when instituting improvedoffice procedures, they were careful to avoidundue association with office machinery. Theiroverriding concern was to be credited as trulymanagerial in their concerns and methods,rather than narrowly technical and machineoriented. In practice, the SPA functioned as aloose assemblage of specialists in obscure sub-jects like form design, work simplification, andprocedure documentation. But its leaders werekeen to recast these disparate activities as meretools in the belt of the true generalist expert inadministrative techniques.



Most of the articles written about comput-ers for a managerial audience came from mem-bers of this broader systems community: fromcomputer vendors eager to promote their ownskills in systems work, from corporate systemsmen, or from consultant systems experts.Indeed, many prominent data processingexperts switched repeatedly between theseroles. These systems-oriented experts were keento point out that computerization benefitswould come more from the attention given tothe rationalization and improvement of proce-dures in preparation for automation than fromthe computer itself. As a result, the idea that“quality systems analysis is the key” to real sav-ings had become a cliché long before therewere any real savings to examine.56

As experts on business procedure formaliza-tion and improvement, the systems men wereassured an important place in the preparationfor a computer’s arrival. Among the most wide-ly practiced duties of the systems and proce-dures department were the design of forms, thewriting of procedures manuals, and the evalua-tion of office machinery. Systems men believedthat their existing methods would prove suffi-cient for the computer. Indeed, adoption of thecomputer would trigger a massive boom in thebusiness of documenting and redesigning cleri-cal procedures. A 1957 survey of systems workobserved that “the reorganization and organi-zation of so many systems departments withinthe past decade has been kicked off primarily byrumors of what electronic data processing cando.” Systems men were often involved in the adhoc teams convened to perform feasibility stud-ies to evaluate the computer’s potential andselect a suitable model. These teams became thenucleus of the new computer departments thatgrew up around the machines. But althoughmany individual systems men were involved inadministrative computing efforts from thebeginning, the assimilation of systems workinto the new computer department was a slowand uneven activity.57

Although different firms’ organizationalarrangements differed greatly, the systems andprocedures department of the early 1950s wastypically separate from the punched carddepartment. Both departments usually grewout of a firm’s accounting operations and, inmost cases, remained under the corporate con-troller’s authority. The systems and proceduresdepartment, however, held a mandate that wassupposed to extend throughout the corpora-tion’s administrative activities. In contrast, thepunched card department was machine orient-ed and inward looking, with its members iden-

tifying closely with their machinery. Whilethey would work out appropriate machine rou-tines to handle the jobs they were given, thestaff typically had little say in designing theadministrative systems their machines were tohelp process. Contemporary observers some-times complained that this led to an unevenapplication of the machines. Jobs better han-dled manually or with a desk calculator werefinding their way onto punched cards, whileother jobs were performed manually becausethe responsible managers had little interest inpunched cards. IBM salesmen usually had moreexperience in establishing machine proceduresthan punched card supervisors did, and so,guided by a cookbook of standardized meth-ods, they often played a leading role in settingup new systems.

Most of the specific techniques that systemsmen used, and much of the framework theyimposed on systems analysis, had been pio-neered well before World War II. The biggestdifference between systems men and the earli-er office managers was not specific techniquesbut organizational position in a corporate staffdepartment, separate from daily business activ-ities or direct supervision of clerical workers.Around the turn of the century, expertise insystematization techniques was a hallmark ofthe modern and professional manager. As a spe-cialist activity, its roots go back to the 1910sand the first attempts to constitute office man-agement as a professional activity. The bestknown of the early office management enthu-siasts, William Henry Leffingwell, was inspiredby the work of Frederick W. Taylor to apply themethods of scientific management to theoffice. This involved carefully analyzing officeroutines, such as opening envelopes or writingletters, to simplify and standardize them.Leffingwell also pioneered the clerical applica-tion of flowcharting, in which the physicalpath of a form or letter through the office wascharted together with all the operations per-formed on it. Although Leffingwell and his col-leagues did not enjoy enormous success inhaving their ideas implemented, their workremained well known.58

As these techniques developed, the flowchartmoved further away from its origins as a literaldepiction of the physical flow of paper throughan office. By the time Richard Neuschel wrotehis 1950 Streamlining Business Procedures, themanifesto of the systems and procedures move-ment, a wide variety of charting techniqueswere in use. The layout flowchart—where doc-ument flows were superimposed on a drawingof the office—most closely resembled the origi-


nal. The popular vertical flowchart was muchmore schematic, using rows to show each stepin a procedure and columns to show differentkinds of activity—production, transportation,inspection, and filing. The more complex hori-zontal chart mixed organizational structure andclerical operations. By the 1930s, punched cardequipment companies used flowcharts to illus-trate the physical transfer of cards between dif-ferent machines, the role of humans inproviding inputs to the system and as machineoperators, and the various files, forms, andprintouts involved in machine operations.59

Flowcharts had already been applied topunched card operations by some of the largerand more organized punched card depart-ments, which found it useful to produce chartsat different levels of detail. An application flow-chart showed a job’s overall outline—how theoriginal records and accounting transactionswere combined to produce work files, reportsheets, and so on. This was intended to com-municate the purpose of a procedure to man-agement in terms it could understand, bydescribing the “accounting job which a proce-dure accomplishes.” In contrast, the opera-tional flowcharts were more voluminous anddealt with “machine and clerical operations intheir proper sequence and the movements ofcards from operation to operation.” Each stepwas numbered, and cross-linked to a writtendescription for the machine operator. Suchelaborate formal procedures were far from uni-versal in the punched card world. Manypunched card operations never bothered todiagram their control board wiring patternsand were content to leave elaborate proceduresin the heads of the punched card staff.60 Theimportant thing, however, is that more formaltechniques existed and could be adapted foruse with computers.

Earlier flowcharts and written procedures,whether written for clerks, bookkeepers, orpunched card machine operators, were invari-ably read and followed by humans, not bymachines. Humans can tolerate ambiguity.Even the most detailed instructions written forhuman clerks do not begin to approach thecrushing literalness required in those given toa computer. When an exception occurs—a casefor which the standard procedure cannot beapplied—then a human will be quick to ask acolleague or call for a supervisor. A clerk willnotice when codes, amounts, or dates writtenon a form are missing or obviously incorrect.Should a clerk run out of paper or need tosharpen a pencil, the clerk resolves this hard-ware failure without undue problems. Clerks

are also not forced to transcribe all informationbefore reading it, or to squeeze both instruc-tions and data into a few thousand charactersof memory.

Systems analysis for the computer, howev-er, came with its own set of problems. Mostmanagers who ordered computers during the1950s were aware that some effort would beneeded to translate existing procedures into aform suitable for machine execution. But fewgrasped the enormous gulf separating a satis-factory clerical procedure from a computer pro-gram. It was often assumed that the systemsanalysis task would concern itself only withmanagerially oriented questions of policy andprocedures; the work of translating flowchartsinto computer programs was a lower status,and often entirely separate, activity. Thus, inmany firms, the systems and proceduresdepartment remained a separate entity fromthe new computer department, and only pro-grammers were given detailed instruction inthe computer’s workings.

Programming—The new taskAs a corollary, the programmer’s job was

heavily circumscribed. The idea was that ana-lysts could give high-level flowcharts showingoverall runs and processes to programmers,who would fill in progressively lower levelcharts to create explicit block diagrams of pro-gram logic. At one company, the job wasdefined as follows: “The programmer is expect-ed to take the broad flowcharts presented tohim by the systems analysts and to develop thedetailed flowcharts for the computer runs.” Theprogramming of administrative applicationswas constructed as an extension of the higherlevel and more managerially focused work ofsystems analysts, to bring their results closer tothe form demanded by the computer. Mostauthorities of the 1950s did not consider thefinal translation of this detailed chart intoinstructions the computer could run as part ofprogramming but as a third activity: coding.

Despite some early hopes that the program-mer’s labors would automate those of top man-agement, the work of programming insteadinvolved heroic efforts to make underpoweredcomputers perform conceptually trivial admin-istrative tasks with a modicum of efficiency.Even if the programmer did not physicallyinteract with the computer’s hardware, it wasnever far from his or her mind. Effective pro-gramming demanded clever sequencing ofoperations and juggling of resources similar tothat practiced by a generation of punched cardoperators. For example, GE found that its elab-



orately charted payroll program, once convert-ed to code, required 80,000 instructions and 36hours to run.

Results such as this demonstrated the prob-lems involved in treating analysis as a self-contained operation, performed without closeattention to the details of the computer itself.As we have seen, the tiny internal memories offirst-generation computers meant that a job likepayroll was split into dozens of separate runs.During each run, a small program updated oneor more of the master files or produced tempo-rary working files to be processed further by theruns to follow. Some runs would do nothingmore than sort or check data. The computer’sefficient use demanded that the number of runsbe kept to a minimum. It was the systems ana-lyst’s task to break a job into separate runs andsketch the requirements for each—yet without astrong background in programming’s arcanedetails, it was impossible to judge exactly howmuch the computer could accomplish on eachpass. In one case, attempts to handle all excep-tions manually created a payroll program thatrequired 90 separate runs to produce the week-ly paychecks. Subtle variations in code qualitycould dramatically increase or decrease the fea-sibility of a particular overall structure.61

Because of the limitations of early machinesand programming techniques, most of theadministrative programs in the 1950s werewritten in low-level, machine-specific lan-guages. The GE payroll program was codeddirectly into the form executed by the com-puter—all instructions were punched as a seriesof octal (base 8) numbers. Most computerinstallations soon adopted “automatic coding”techniques to assist in code preparation—including the use of symbolic assemblers totranslate mnemonic codes into instructions,assign convenient labels to specific memorylocations, and assemble various subroutinesinto a single executable program. But the nota-tion had simply become a little more conven-ient—programmers were still in the business ofwriting each instruction that the computerwould run. By the mid-1950s, the use of gener-alized routines to perform input and output,generate reports, and manage files was also wellestablished. The highly constrained resourcesavailable to the programmer forced many com-puter installations to adopt rigid standards onthings like memory organization, the means bywhich subroutines communicated with themain routines that called them, and even thetape drives to be used for specific purposes.62

Successful systems analysis work thusproved to require thorough consideration of

programming and operations issues duringdesign stages. But exposure to programmingwas no panacea. The very nature of program-ming work thrust one deep into the world ofthe machine and away from the administrativeperspective on which analysts had pridedthemselves. This was a hard gulf to bridge.Some companies of the 1950s explicitly com-bined programming and analysis roles, or atleast grouped both programmers and analystsinto project teams rather than separate depart-ments. However, through the 1950s and 1960s,programming and analysis remained notional-ly separate in most firms, with analysis alwaysof higher status and better paid. Within dataprocessing, the question of whether program-ming and analysis should be separate careers ordifferent stages of a single career was widelydebated. In practice, the business-orientedaspects of analysis were often neglected, as thetitle became a way of giving additional statusand higher pay to someone whose job was real-ly that of a programmer.

Programming’s relationship to operationand coding was also unclear, although theseboundaries were resolved more generally andappear to have been enforced more successful-ly. According to one consultant, when organiz-ing a data processing department, “thefundamental number one rule is to keep theprogrammers out of the machine room.” Theseparation of programming and operationduties occurred early in the history of adminis-trative computing, although programmers werenot always complete strangers to the comput-er’s hardware. Many pioneers recall with pleas-ure the chance to operate the machines duringtheir brief testing sessions with IBM’s installa-tions, a few months before their own comput-ers arrived. In some companies, operators weregiven the same basic training as programmers.At least a few companies of the mid-1950sexperimented by combining coding and oper-ation. They viewed these two skills as comple-mentary because both demanded an intimateknowledge of the computer’s workings.Sometimes programmers gained access to themachine during the night, although mostcomputers of the 1950s were leased from IBM,and running the computer for an extra shiftadded 50 percent to the cost of its lease.Companies using the smaller 650 computerswere less likely to enforce a rigid separationbetween programming and operation.63

Despite early efforts to rigidly separate pro-gramming and coding, by the mid-1950s it wasincreasingly recognized that both activitiesshould be performed by the same group of peo-


ple. One of the first detailed reports on com-puter use stated simply, “Experience indicatedthat the best programmers were also the bestcoders.” Efficient coding proved vital to projectsuccess, and there was no way for the pro-grammer to unambiguously communicateexactly what code was required without effec-tively doing the coding as well. Strict separa-tion of coding was apparently rare in practice,despite its continuing presence as a job title insome firms and in the standard job descriptionsissued by the US federal government.64

How did companies hire programmers andother computer staff? Early in the 1950s, as thefirst corporations ordered their Univacs andIBM 700-series machines, there could havebeen no more than a few hundred program-mers working in America. Pioneering scientificand university computing installations sup-plied a trickle of intelligent and experiencedprogrammers to industry, including many ofthose who headed application developmentefforts for computer manufacturers. Most busi-ness application programmers, however, hadno previous programming experience or formaltraining other than a short course from thecomputer manufacturer. As Wallace reported inhis early guide to computer acquisition,

Usually, the computer team should be made upfor the most part of men who know the companyand its operations. Their backgrounds might be intabulating, procedures work, accounting systemsdevelopment, or industrial engineering …. 65

In 1957 alone, IBM trained more than 14,000people to prepare them for work as program-mers. According to Business Week, most had nomore than a high school diploma.65

Some experts with practical experience coun-seled that the benefits of having at least a depart-ment head or a chief programmer with computerexperience would sufficiently outweigh the costof hiring such people. But a consensus soondeveloped that it was easier to teach the funda-mentals of programming to somebody whoalready knew something about business than toteach the culture of business to someone withprogramming experience in an unrelated area.The Harvard team examining computer acquisi-tion policies reported that “Several executivessaid … it was easier to train an accountant to pro-gram than to train a programming expert inaccounting ….” These executives were likely par-roting what they had heard from computersalesmen. This idea was easy for IBM to sell tomanagers, because it fitted their own assump-tions that managerial knowledge of business is

more valuable and harder to replicate than tech-nical knowledge or skills. The idea also benefit-ed IBM, since it assured potential customers thatthe terrible lack of programmers was not a prob-lem. Instead, IBM supplied its ProgrammerAptitude Test (basically a standard verbal rea-soning and mathematical examination) and ashort training course. Some companies offeredall their white-collar employees a chance to takethis test, while others recruited more narrowlyfrom the accounting, systems, and punched carddepartments.66

Punched card machine + computer =Data processing

Programming, systems analysis, and opera-tions were well established by the late 1950s asthe three main tasks of the data processingdepartment. The data processing manager wasalso responsible for a number of supervisors,and in most cases a group of keypunch opera-tors, usually women. His department was typi-cally responsible for punched card operationsas well as computers. But how did the comput-er’s administrative use become electronic dataprocessing? This question can be addressedboth literally (as a question of linguistic usage)and sociologically. Acceptance of the comput-er as a data processing device greatly strength-ened the efforts of existing punched cardsupervisors to redefine their identities as dataprocessing managers and to lay claim to thecomputer as their rightful domain.

Linguistically speaking, early computingwas a mess. No term, from program to file, wasso self-evident that it was not widely quibbledwith, and usually for good reason. Computer,for example, was a confusing name. First, tra-ditionally it referred to a person. Second, andmore importantly, it referred to a person whoperformed scientific or technical calculations.As Robert Mauchly, one of the Univac’s inven-tors, observed in 1953, “To call these devicescomputers is nowadays a misnomer…. ‘auto-matic clerical equipment’ would better describewhat we are talking about.” It would havemade more sense to call the machine a clerk oran accountant than a computer. While calcula-tor, IBM’s original term for its “calculatingpunches” and the large 701 scientific computer,was well suited to these machines, it was equal-ly unsuited to describe a machine used prima-rily for business applications. Tabulatingmachines had been known more formally byIBM as accounting machines since the 1930s,so electronic accounting machine fit the new elec-tronic models, but was altogether an inade-quate term to describe the larger devices. On



the other hand, electronics, as in “electronics forthe office,” was hopelessly vague.67

As we know, computer was the term that even-tually triumphed. By the mid-1950s it wasalready widely used to describe administrativelyoriented machines, albeit with occasional reser-vations. What might not be apparent today isthat during the 1950s and 1960s its use wassomewhat colloquial. IBM’s adoption of the termelectronic data processing (EDP) to formallydescribe the new activity of business computingwas one of the cleverest marketing moves in itshistory. Its masterstroke was to rename its exist-ing punched card gear as data processing (DP)equipment. Just as the name shift from tabulat-ing equipment to accounting machines in the 1930sreflected a symbolic broadening of the role oftheir products, so DP and EDP announced thatthe new machines were good for more than justaccounting. More importantly, the nomencla-ture showed that computers and punched cardmachines were different flavors of the samething and belonged together.68

IBM symbolized the unity of data processingthrough the retention of a standard numberingsystem for all its products. This gave rise to the“number soup” that makes reading any discus-sion of early computing a frustrating mess of602As and 709s likely to baffle all but the mostcommitted of readers. The uninitiated are liableto feel that they have inadvertently strayedinto a corporate history of Levi Strauss’s jeans.But how better to emphasize the continuumfrom the humble keypunch to the biggest com-puter than to quantify it as the differencebetween 024 and 705?69

The power of this approach was not lost onIBM’s competitors. A 1957 study commissionedby Burroughs on its “corporate image planningand development” suggested that IBM had suc-ceeded in using the data processing concept togain a competitive edge on other office equip-ment firms. Burroughs was still the leader inadding, calculating, and bookkeepingmachines and had acquired Electrodata as thefoundation of its computer range. While anearlier 1953 study concluded that “IBM’s tech-nical monopoly was due to disappear rapidly”in switching to electronics, the consultantswere now forced to report that the enemy hadturned this new technology to its advantage:

IBM has tended to use the term ‘data processing’to designate the general area. This term, thoughunsatisfactory in many respects, has become gen-erally accepted making it difficult for would-becompetitors to define the field and their equip-ment in it in any other way.

The report urged Burroughs to follow IBM inoffering a “stepped-up line of machines, pro-ceeding from the least complex and inexpen-sive to the most complex and expensive.” Thisshould “be a ‘data processing line’ rather than a‘business machines line’.” While IBM onlybegan to unite its computer families techno-logically during the mid-1960s, with thefamous System 360 range, it had alreadyachieved a much broader semantic compatibil-ity via the concept of data processing.67

Toward data processing managementWhat of the final category of data process-

ing staff, the supervisors and departmentalmanagers? The installation of thousands ofcomputers meant the creation of thousands ofsupervisory jobs. For IBM, the definition of thecomputer as part of a larger, hybrid activity ofdata processing, rather than as a revolutionarydeparture from punched card methods, was afoundation on which it could marshal its exist-ing dominance of the punched card industry tocrush insurgents, such as RCA, GE, Philco, andSylvania, with superior credentials in electron-ics. This evolutionary approach had equallyprofound, although perhaps less widely appre-ciated consequences for the punched card staff,who seized on the identity of data processingas an entry to the world of the computer.

As Thomas J. Watson Jr., the head of IBM,told his audience of punched card supervisorsat their 1954 conference, “those with the bestbackground for stepping into the ‘electronicoffice’ of the not too distant future [are] themanagers of today’s punched card installa-tions.” He reiterated this in 1958, challenginghis audience to demonstrate the professionalcharacteristics that their future in data process-ing demanded. Now, however, the generality ofelectronics had been replaced by the more clear-ly defined data processing. “[T]he very namethat we have applied to our jobs—DataProcessing,” suggested Watson, implied a newfocus on the provision “of relevant facts on atimely basis, on a basis equal or better than ourbusiness competitors. You will gain prestigeand responsibility from the excellence of thefacts, counsel, and advice that you supply.” Ifthey could “become more professional thanever before” then soon “top management willhave begun to look for data processors to infil-trate into the very tops of their businesses.”70

The deliberate attempts of punched cardstaff to construct data processing as a new occu-pation are most clearly evidenced through thetransformation of the main association for sen-ior punched card staff, the National Machine


Accountant’s Association, into a “data process-ing” association and its attempts to certify dataprocessing management as a professional field.Founded in 1951, the NMAA grew rapidly dur-ing the 1950s, passing the 10,000 membermark in 1957. Throughout the 1950s and1960s, it was by far the biggest professionalorganization identified with the computingfield. Its concern with business data processingwas unchallenged by the scientifically orientedAssociation for Computing Machinery or thecomputer groups of the engineering societies(the Institute of Radio Engineers and AmericanInstitute of Electrical Engineers). Its explosivegrowth was partly a result of its federal struc-ture—members joined individual chapters andwere only indirectly members of the nationalassociation. But although policies differedsomewhat between chapters, membership wasgenerally available only to supervisors andmanagers of punched card installations ratherthan to the rank-and-file machine operators.

The association’s strong punched card rootsensured its strength in the hybrid field of dataprocessing. In 1962 the association adopted anew name: the Data Processing ManagementAssociation (DPMA). The name change ulti-mately expressed the evolutionary progressionthrough which the new institutions of dataprocessing formed around the older ones oftabulating. Here, as in individual firms, thetransition was slow and incomplete. The asso-ciation faced many obstacles in its quest tobuild a new profession, foremost being its ownmembership. Reformers within the DPMA sawthe computer as something that could assistthem in their continuing attempts to build aprofession regarded as managerial rather thanpurely technical, focused more on systems thanindividual machines. But attempts to breakwith its past were thwarted or blunted serious-ly by the continuing power of older, more con-servative, punched-card-oriented men. Changetook place slowly—the adoption of a logoshowing magnetic tape spools and punchedcards side by side, the renaming of the associa-tion’s journal from the machine-oriented TheHopper to the ambitiously broad Journal ofMachine Accounting, Data Processing, Systems,and Management. Even the name change frommachine accounting to data processing hadbeen under consideration for six years beforeits final adoption. This was not a group fittedby temperament, abilities, or organizationalmandate for leading a revolution.71

Work from the mid-1950s onward toimprove data processing education, and soforge a recognized profession, culminated in

1962 with the introduction of the Certificatein Data Processing (CDP). Nowhere can thepainfully slow evolution of data processing bemore clearly seen. Reformers wanted to usethe certificate to raise standards and attractmore able young men into the field. Theirgoals had been strongly shaped by exposure tothe corporate accounting profession, especial-ly its managerial orientation and certificationsystem. Yet enormous pressure was exertedwithin the association to make sure that itwould be possible for at least some of therank-and-file members of the association toreceive the certificate—men who had no col-lege education and had come to their supervi-sory posts from jobs as machine operators. Ifsuch men were not eligible for the certificate,and so were excluded from the new profes-sion, then the association would be using theirdues to underwrite the destruction of theirown careers.72

The certificate was therefore an uneasy com-promise. Caught between the need to includeas many current members as possible and theconflicting requirement to raise professionalstandards, the certificate was unable to accom-plish either. Most who eventually passed itwere not members of the DPMA. But pressurefrom within the association ensured that thetest remained easy, and plans to demand col-lege training as a requirement were eventuallydropped. This deterred the better educated out-siders that the test had been designed to attract.The CDP qualification never acquired criticalmass, and few employers ever required or rec-ommended that their staff attain it.

The NMAA’s professionalization efforts illus-trate both the domination of administrativecomputing departments by a relatively conser-vative group and the strong limitations thisdomination placed on the development of amore broad-based approach to corporate com-puting. The certificate enshrined two desiredforms of social mobility for its holders, bothgeared to evolutionary development. The firstallowed punched card supervisors to strength-en their professional credentials for the new ageof electronic data processing. The second formreinforced the conceptual ladder of advance-ment in data processing that integrated differ-ent jobs in data processing departments as partof a single career trajectory. Punched cardmachine operators and computer programmerswould aspire to achieve professional status bybecoming supervisors and department heads,not by strengthening their craft skills. Whilebusiness application programmers were dataprocessors (although not at the professional



level), so were punched card machine operators,systems and procedures analysts, keypunchsupervisors (if not keypunch operators), anddata processing managers of all ranks. The cer-tificate ensured that its holder knew somethingof all these areas—meaning that to qualify forit a programmer or punched card expert had,theoretically, to acquire a broader appreciationof data processing as a whole. Though the cer-tificate itself was a failure, the vision itenshrined of administrative computing as partof the hybrid field of data processing remaineddominant during the 1960s, and with it theinfluence of punched card culture.73

Data processing and the history ofcomputing

In professional identity, as in other respects,ad hoc arrangements made during the mid-1950s proved remarkably persistent. The name,organizational location, occupational culture,internal structure, and corporate mandate ofthe data processing department had essential-ly standardized by 1958. It changed only slow-ly through the 1960s, despite the considerableexpansion and proliferation of such depart-ments and the deployment of two further gen-erations of computer hardware. Yet, as we haveseen, the data processing department’s initialtemplate came from a convergence of factorsquite specific to the 1950s. These factorsincluded the need for punched card equipmentto work alongside computers, and the need tointegrate formerly separate tabulating depart-ments and analysis groups. Well into the 1970s,Electronic Data Processing remained theaccepted name of the computer department.EDP was the subject to which textbooks, con-ferences, and journals on administrative com-puting were devoted, and the term adopted byadministrative computing staff when definingtheir own identity.

Although it has sometimes been noted inpassing that the administrative use of comput-ers has been called electronic data processingfor most of its history, few historians have seri-ously considered the data processing conceptor its evolution. A history of data processingwould look markedly different from the pres-ent-day history of computing. Particularly in itsearly days, the historical study of computingwas oriented toward actual “computing.” Amore recent shift toward consideration of busi-ness matters has not yet triggered a rethinkingof some of these old assumptions. For histori-ans of technology in general, and of comput-ing in particular, a sticking point has been thesheer magnitude of such a project. The use of

computer technology in a particular socialspace (such as the laboratory, office, or factory)cannot be addressed without also studying theearlier history of this setting, the people in it,and the objectives to which the machine is put.So, while coherent one-volume histories of thecomputer hardware industry and its technolo-gies can be written, it seems unlikely that wecan produce a single coherent narrative aboutthe use of computers or of associated tasks suchas analysis, programming, or operation.74

Historians have also paid much less atten-tion to analysts, supervisors, or operators thanto programmers. The term programmer hassometimes been used to encompass the entiredata processing staff. This may in part reflectthe interests of early computer scientists, andmore recent historians of science, in comput-ing’s scientific or theoretical aspects. Analysts,operators, and supervisors were irrelevant, fromthese perspectives, and even applications pro-grammers did little work of note.75

The development of programming as a cor-porate occupation had much more to do withwhat the corporation was already like than itdid with the scientists and mathematicians whoprogrammed the first experimental computers.Programming was neither the most commondata processing job, nor the best paid, nor theone held by managers and supervisors in thehighest esteem. The data processing depart-ment was structured in accordance with themanagerial conception that business knowl-edge was higher and more valuable than tech-nical knowledge. (The boundaries between thetwo are arbitrary but powerful). The intimaterapport with machine demanded of a good pro-grammer during the 1950s drew many pro-grammers ever deeper into a specializedoccupational subculture. Yet the data process-ing prestige ladder progressed upward fromoperator, to coder, to programmer, to analyst,to data processing manager and finally to gen-eral manager. At each stage, one moved everfurther from the machine itself and gained evermore prestige and pay.

Of the four main data processing jobs(supervisor, analyst, programmer, operator), theprogrammer’s was undoubtedly the biggestdeparture from the earlier practices ofpunched-card work. But even here, program-ming was initially viewed more as a redistribu-tion of responsibilities previously split betweenoperators and analysts than as a revolutionarydeparture. Managers preferred to view pro-gramming as a new activity that their existingstaff could pick up than as a new profession.The corporate applications programmer has


historically been squeezed between themachine-oriented, craft knowledge of themachine operator (later the systems adminis-trator) and the ostensibly business-orienteddomain of the analyst. The new role was con-structed on rickety foundations in space hur-riedly cleared between these two much olderoccupations. The gulf between these domainshas remained the most important stumblingblock in the troubled history of administrativecomputing. Successive waves of facilities man-agement, charge-back schemes, software pack-ages, programming methodologies, end-usertools, outsourcing, and nontechnical CIOs(chief information officers) have failed toresolve it.

The male domination of corporate comput-er programming should not, in this context, besurprising. Jennifer S. Light has recently arguedthat “the job of programmer, perceived inrecent years as masculine work, originated asfeminized clerical labor.” Whatever the meritsof this argument with respect to ENIAC, thefocus of her paper, it is clearly not viable in thecontext of corporate applications program-ming—the dominant programming activityfrom the mid-1950s on. Applications program-ming evolved at the fuzzy interface betweenpunched card machine operation (a predomi-nantly masculine activity) and systems andprocedures analysis (an almost exclusively mas-culine one). The clerical job was that of key-punch operator—feminized in the punchedcard era, feminized after the computer arrived,and (as data entry clerk) feminized to this day.Given that few corporations relied on mathe-maticians as administrative programmers, theinfluence of human scientific “computers,”whether male or female, on the culture of rank-and-file administrative applications program-mers is marginal at best.76

No wonder, therefore, that nothingapproaching a single community of computerprofessionals or programmers has ever existed inthe US. Had the computer really been a revolu-tion, had businesses rushed to adopt operationsresearch and to use the computer primarily as atool of managerial analysis and scientific calcu-lation, then things might have been different.Perhaps computer staff might have rushed toembrace a fundamentally new identity,premised on computer science, or software engi-neering, or management science. This is notwhat happened. While the Systems andProcedures Association provided a separate iden-tity to analysts, not until the end of the 1960swas there a serious attempt to organize a sepa-rate association for computer programmers, to

promote programming as a profession in its ownright, or to offer certification in programming.It was not until the 1970s and the surge of inter-est in software engineering concepts that anyserious attempt was made to develop a moretechnically rigorous yet theoretically groundedalternative to data processing as a professionalidentity for the ambitious applications pro-grammer. Even today, relatively few practicingadministrative programmers would identifythemselves as software engineers or computerscientists, or hold degrees in these fields.77

Revolution revisitedWhen we consider how computers were

purchased and used, rather than how they werebuilt and sold, we see a gradual evolutionalongside continuing expectations of a loom-ing and revolutionary shift. The seemingly rev-olutionary technology of computing had,within a few short years of its introduction,been brought down to earth and containedwithin the fundamentally evolutionary corpo-rate institution of the data processing depart-ment. What began as a panacea became aplacebo, as technological upheaval substitutedfor managerial reorganization.

The gulf between revolutionary dreams andincremental realities was not lost on contem-porary observers. In 1967, a decade after thepublication of his first books, Canning couldstill report that computers were used primarilyto automate individual tasks. He also reportedthat most data processing staff had little moreprofessional background than a couple weeks’training from IBM, and that “data processing islooked upon as simply ‘the old tabulating oper-ation with chromium plating.’” Few program-mers were interested in professional educationor learning more about business. In practice,the “data processing department has usuallyremained at the same organizational level asthe tabulating operation—at either the fourthor fifth level” and somewhere under the con-troller. Canning reported that in some firms thetabulating department had absorbed the “sys-tems and procedures” group, while in othersthe reverse had taken place, but that resultinghybrid had so far failed to achieve what he stillconsidered its manifest destiny and “elbow itsway up the ladder.”78

Canning remained confident that this situ-ation was about to change in favor of a moreintegrated, management-science-oriented ap-proach and a higher level of top managementinvolvement, thus illustrating a strange butremarkably enduring feature of corporate com-puting. Following a pattern already well estab-



lished during the 1950s, throughout the subse-quent four decades most discussion of the cur-rent state of administrative computing practicehas been remarkably critical of its failure to ful-fill its promise. A 1958 Business Week report dis-cussed in detail the serious problems that firmswere having in making computers pay off eco-nomically and the highly conservative way inwhich they were being used. Not once duringthis saga of disasters and broken promises didthe reporter’s faith in the computer falter.Criticism, however scathing, of current usagebecame a means of protecting the dream. Ifcomputers could be used correctly, then prob-lems would evaporate. Consider its opening:

Just four years ago, at Louisville, Kentucky, anew industrial revolution started …. it hasbecome perhaps a most perplexing and disgrun-tled—but inevitable—revolution. It’s perplexingbecause industry, which has adopted the mar-velously complex electronic computers with analmost religious fervor, often seems unsure ofwhat to do with them after it has them. It’s dis-gruntled, because early results have fallen farshort of the rosy dreams in which they camewrapped. Yet it’s inevitable, because the com-puters still hold the key to new systems of organ-ization for the sprawling giants of industry,commerce, and government …79

These new systems of organization contin-ue to tantalize. During the recent “new econ-omy” years, business travelers were unable toturn on CNN or open a magazine withoutencountering commercials for a new Internetservice, communications tool, or personalcomputer product that will shatter corporateorthodoxy and put the viewer back in con-trol. Yet recent claims for the revolutionarynature of computing products stemmed large-ly from their promise to liberate us from cen-tralized computing facilities associated withthe IBM mainframe—the self-same technolo-gy that was originally destined to usher in therevolution.

It is now 40 years since these once-preciousfirst-generation machines were taken fromtheir glass-fronted, air-conditioned, meticu-lously clean showpiece installations andrehoused unceremoniously in dumpsters. Inhindsight, did the computer bring a secondindustrial revolution (or a third, depending onhow you count)? Nothing in economic figuresfrom 1955 to 1995 suggests that an industrialrevolution occurred in the office during thisperiod. Companies did not massively slashtheir clerical or administrative work forces.

Administrative productivity increases contin-ued to lag those in other sectors. Even beforefirst-generation computers were obsolete, it wasa truism that most administrative computinginstallations cost far more than they currentlysaved. Assuredly, these particular computersdid not usher in a revolution—as for their dis-tant descendents, the jury is still out.80

It seems unlikely that there ever was, or willbe, a generally applicable administrative tech-nology that is truly revolutionary in the senseused to sell first-generation computers, man-agement information systems, e-commerce sys-tems, and many other electronic tools—atechnology that changes business rules so rap-idly and fundamentally that to wait for provenresults before applying it is to flirt with disaster.We can readily document new products or pro-duction technologies that so quickly become soimportant in particular industries that failureto swiftly adopt them would prove disastrous.Think, for example, of the successive waves ofrapid technological advancement that affectthe manufacturers of computer parts such asdisk drives or processors. But the incrementaladvantages provided by even the most impor-tant administrative technologies have rarelybeen compelling or proprietary enough toreward those early-adopter companies thatrush in to pioneer new techniques at their ownexpense.

Corporations have usually had cause toregret their periodic excursions to the “bleed-ing edge” of administrative technology.However great the apparent potential, it hasbeen wiser to learn from the mistakes of oth-ers—thus far at least, every important advancehas been debugged and packaged long beforefailure to adopt it would have spelled commer-cial disaster. The genius of the computer revo-lutionaries has been in their ability to blur thisfairly obvious distinction between universallyapplicable administrative technologies andindustry-specific technologies of productionand distribution. Again and again, they havesold the claim that here is a technology simul-taneously “disruptive” in a fundamental way(that is, its adoption compels an organization-al paradigm shift), offers a sustainable and pro-prietary competitive advantage to its earlyadopters, and applies to the core managerial oradministrative concerns of every business.When faced with such a technology, to wait forwidespread and sustainable results—as opposedto a few vague but widely circulated claims ofspectacular success—would be irresponsible. Ifsuch a technology exists, we can now state withconfidence, it was not the computer.81


AcknowledgmentsThe author would like to thank William Aspray,Burton Grad, Rosemary Stevens, Jeffrey Tang,Jeremy Vetter, and Atushi Akera for their com-ments on earlier versions of this article. Thisresearch was supported by the IEEE Life MemberFellowship in Electrical History and the TomashFellowship from the Charles Babbage Instituteof the University of Minnesota.

References and notes1. The need to examine the usage of computers is a

major theme of J.W. Cortada, Information Technol-ogy as Business History: Issues in the History andManagement of Computers, Greenwood Press,Westport, Conn., 1996. The quotation is from p.160. See also P.N. Edwards, “Making History:New Directions in Computer Historiography,”IEEE Annals of the History of Computing, vol. 23,no. 1, Jan.–Mar. 2001, pp. 85-87. One exceptionto the general lack of attention to use has beenthe work of JoAnne Yates; see particularly J. Yates,“Co-evolution of Information-Processing Technol-ogy and Use: Interaction between the Life Insur-ance and Tabulating Industries,” Business HistoryRev., vol. 67, no. 1, Spring 1993, pp. 1-51 and J.Yates, “Early Interactions between the Life Insur-ance and Computer Industries: The Prudential’sEdmund C. Berkeley,” IEEE Annals of the History ofComputing, vol. 19, no. 3, July–Sept. 1997, pp.60-73. Yates argues that the demands of userorganizations played an important part in shapingthe hardware of punched card machines andearly computers. For a first-hand account of thechallenges involved in constructing the firstadministrative computing installation (one facingthe additional challenge of building its own com-puter), see J. Aris, “Inventing Systems Engineer-ing,” IEEE Annals of the History of Computing, vol.22, no. 3, July–Sept., 2000, pp. 4-15. For anexcellent and historically informed overview ofchanges in the development of computer applica-tion systems, see A.L. Friedman and D.S.Cornford, Computer Systems Development: History,Organization and Implementation, John Wiley &Sons, New York, 1989. The corporate use of com-puters from 1960 onward is discussed in R.L.Nolan, “Information Technology Managementsince 1960,” in A Nation Transformed by Informa-tion: How Information has Shaped the United Statesfrom Colonial Times to the Present, A.D. Chandlerand J.W. Cortada, eds., Oxford Univ. Press, NewYork, 2000, pp. 217-256, while a set of historicalcase studies are presented in J.L. McKenney, D.C.Copeland, and R.O. Mason, Waves of Change:Business Evolution through Information Technology,Harvard Business School Press, Boston, 1995, and,for the Dutch case, D. de Wit, The Shaping of

Automation: A Historical Analysis of the Interactionbetween Technology and Organization,1950–1985, Verloren, Hilversum, Netherlands,1994.

2. The evolutionary progression from punched cardtechnologies to electronic computers is a majortheme of M. Campbell-Kelly and W. Aspray, Com-puter: A History of the Information Machine, BasicBooks, New York, 1996; J. Cortada, Before theComputer: IBM, Burroughs, and Remington Randand the Industry They Created, 1865–1956,Princeton Univ. Press, Princeton, N.J., 1993; andJ. Cortada, The Computer in the United States:From Laboratory to Market, 1930–1960, M.E.Sharpe, Armonk, N.Y., 1993.

3. R.S. Cowan, “The Consumption Junction: A Pro-posal for Research Strategies in the Sociology ofTechnology,” in The Social Construction of Techno-logical Systems, W.E. Bijker, T. Pinch, and T.P.Hughes, eds., MIT Press, Cambridge, Mass.,1987, pp. 261-280. The classic examination ofthe concept of computer revolution remains L.Winner’s “Mythinformation in the High-TechEra,” in Computers in the Human Context, T.Forester, ed., MIT Press, Cambridge, Mass., 1991,pp. 82-132.

4. What I call administrative computing is moregenerally known as business computing. Duringthe 1950s and 1960s, it was frequentlydistinguished from scientific computing—the useof computers to automate scientific andengineering calculations, and for purposes ofmodeling and simulation. The latter was some-times also known as technical computing. Thereason I use administrative rather than business orcommercial is one of precision—much scientificcomputing took place within aerospace andother engineering companies, while a great dealof administrative computing took place withingovernment and military institutions.

5. The concept of a social institution is important,implying that when senior managers establish newdepartments, evaluate their success, anddetermine their organizational position they do soprimarily with reference to the prevailing consen-sus on what is appropriate. Thus the process bywhich a particular organizational form (like a newdepartment or function) comes to enjoywidespread acceptance is only indirectlydetermined by objective experimentation in indi-vidual firms, and different firms come to look moreand more alike. See W.W. Powell and P.J. DiMag-gio, The New Institutionalism in OrganizationalAnalysis, Univ. of Chicago Press, Chicago, 1991.

6. J.M. Thesis, “Practical Application of ElectronicEquipment,” J. Machine Accounting, vol. 6, no. 3,Mar. 1955, pp. 5, 7-8, 16-17, reprinted from theAugust 1954 issue of the Nat’l Assoc. Cost Accoun-



tants Bull. The term automation was coined with-in Ford during the late 1940s, but only becamewidely known following its tireless promotion byconsulting prodigy John Diebold during the early1950s. See J. Diebold, “Automation—The NewTechnology,” Harvard Business Rev., vol. 31, no.6, Nov.–Dec. 1953, pp. 63-71.

7. E.C. Berkeley, Giant Brains or Machines That Think,John Wiley & Sons, New York, p. vii.

8. J.W. Haslett, “The Coming Revolution inPaperwork,” Systems and Procedures Quarterly, vol.1, no. 1, Mar. 1950, p. 1, and J.W. Haslett, “Is theModern Office Vanishing?,” Systems and ProceduresQuarterly, vol. 2, no. 2, June 1951, pp. 11-13.

9. W.B. Worthington, “Application of Electronics toAdministrative Systems,” Systems and ProceduresQuarterly, vol. 4, no. 1, Feb. 1953, pp. 8-14.

10. Modern consultants call this the fear factor,pointing out to executives that their companieshave lots of money but little time, and so itwould be wise to part with a lot of the former togain a little of the latter. As a best-selling businessbook published in 2000 claimed, things movedso fast in the Internet age that to adopt the “fastfollower” tactic of adopting only provenapproaches was enormously irresponsible. In thisargument, the lack of demonstrated savings ordebugged systems is actually a selling point. SeeP. Evans and T.S. Wurster, Blown to Bits: How theNew Economics of Information Transforms Strategy,Harvard Business School Press, Boston, 2000. Fordiscussion of the mass hysteria whereby investorsthrew money at startups without clear projectedprofits or even revenue streams, and establishedcompanies threw money into creating unviablespin-offs to stave off their challenge, see A. Har-mon, “What Have E-Consultants Wrought?,” inthe New York Times, 13 May 2001, p. 1.

11. R.F. Osborn, “GE and UNIVAC: Harnessing theHigh-Speed Computer,” Harvard Business Rev.,vol. 32, no. 4, July–Aug. 1954, pp. 99-107; M.L.Hurni, “Decision Making in the Age of Automa-tion,” Harvard Business Rev., vol. 33, no. 5,Sept./Oct. 1955, pp. 49-58.

12. R.L. Ackoff, “Operations Research—Its Relationshipto Data Processing,” Proc. Automatic Data Process-ing Conf., Graduate School of Business Administra-tion, Harvard Univ., Boston, 1955; R.R. Ross, “CanYou Afford the ‘Practical’ Approach to Electronics?”Management Methods, Nov. 1956, pp. 36-37, 56.This thinking reached its zenith in the celebratedH.J. Leavitt and T. L. Whisler article, “Managementin the 1980s,” Harvard Business Rev., vol. 36, no. 6,Nov.–Dec. 1958, pp. 41-48.

13. F. Wallace, Appraising the Economics of ElectronicComputers: An Approach for a Company to Deter-mine the Feasibility of Acquiring a Computer, Con-trollership Foundation, New York, 1956, p. 50.

14. V.F. Blank, “Electronics—Possibilities and Limita-tions,” Systems and Procedures Quarterly, vol. 6,no. 3, Aug. 1955, pp. 8-15.

15. P.B. Laubach and L.E. Thompson, “ElectronicComputers: A Progress Report,” Harvard BusinessRev., vol. 33, no. 2, Mar.–Apr. 1955, pp. 120-128.

16. J.W. LaForte, Market Analysis—Electronic Data Pro-cessing Machines Types 702-703-705, Feb. 1955,Cuthbert C. Hurd Papers (CBI 95), Charles Bab-bage Inst. (hereafter CBI), Univ. of Minnesota,Minneapolis.

17. T. Watson Jr. and P. Petre, Father, Son & Co: MyLife at IBM and Beyond, Bantam, New York, 1990,p. 244.

18. For the technical and business history of first-gen-eration computing, see C.J. Bashe et al., IBM’s EarlyComputers, MIT Press, Cambridge, Mass., 1986;M. Campbell-Kelly and W. Aspray, Computer: A His-tory of the Information Machine, Basic Books, NewYork, 1996, pp. 105-130; P.E. Ceruzzi, A History ofModern Computing, MIT Press, Cambridge, Mass.,1998, pp. 13-78. For contemporary descriptionsof two of the era’s most important administrativecomputers, see C.J. Bashe, W. Bucholz, and N.Rochester, “The IBM Type 702: An Electronic DataProcessing Machine for Business,” J. Assoc. Com-puting Machinery, vol. 1, no. 1, Oct. 1954, pp.149-169; F.E. Hamilton and E.C. Kubie, “The IBMMagnetic Drum Calculator Type 650,” J. Assoc.Computing Machinery, vol. 1, no. 1 pp. 13-20.

19. Figures on shipment value are taken from J.W.Cortada, Information Technology as Business His-tory: Issues in the History and Management ofComputers, Greenwood Press, Westport, Conn.,1996, in which they are attributed to a 1974International Data Corporation report. For orderbacklog figures and average rental and purchasecosts, see the early computer census included inR.H. Brown, Office Automation: A Handbook onAutomatic Data Processing, 1959, Market andProduct Reports Collection (CBI 55), CBI. For adiscussion of the experience of a preinstallationpractice session, see R.E. Porter, “First Encounterwith the 701,” Annals of the History ofComputing, vol. 5, no. 2, Apr.–June 1983, pp.202-204. On the difficulty of completing a thor-ough study without ordering a computer, seeP.B. Laubach, Company Investigations of Automat-ic Data Processing, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1957.

20. “The Office Equipment Industry: They SellAnswers to Problems that Executives Don’t KnowExist,” Dun’s Rev. and Modern Industry, vol. 64,no. 3, Sept. 1954, pp. 101-119.

21. The classic examination of organizational mimic-ry is P.J. DiMaggio and W.W. Powell, “The IronCage Revisited: Institutional Isomorphism andCollective Rationality in Organizational Fields,”


American Sociological Rev., vol. 48, no. 2, Apr.1983, pp. 147-160. For an examination of thedynamics of managerial fads, structured aroundthe concept of the “management-knowledgeentrepreneur,” see E. Abrahmson and G.Fairchild, “Management Fashion: Lifecycles, Trig-gers, and Collective Learning Processes,” Admin-istrative Science Quarterly, vol. 44, no. 4, Dec.1999, pp. 708-740.

22. The books referred to are, respectively, F. Wallace,Appraising the Economics of Electronic Computers:An Approach for a Company to Determine the Fea-sibility of Acquiring a Computer, ControllershipFoundation, New York, 1956; B. Conway, J. Gib-bons, and D.E. Watts, Business Experience withElectronic Computers: A Synthesis of What Has BeenLearned from Electronic Data ProcessingInstallations, Controllers Inst. Research Founda-tion, New York, 1959; R.G. Canning, ElectronicData Processing for Business and Industry, JohnWiley & Sons, New York, 1956; R.G. Canning,Installing Electronic Data Processing Systems, JohnWiley & Sons, New York, 1957; and P.B. Laubach,Company Investigations of Automatic Data Process-ing, Graduate School of Business Administration,Harvard Univ., Boston, 1957. An initial version ofthe latter appeared in article form as P.B. Laubachand L.E. Thompson, “Electronic Computers: AProgress Report,” Harvard Business Rev., vol. 33,no. 2, Mar.–Apr. 1955, pp. 120-128.

23. An enormous amount of work is conducted withinlarge organizations to hide the true causes of theadoption of new technology (often located inorganizational politics, emotion, and occupationalsubculture) and to present it as the result of arational process of cost-benefit analysis originatingat the top of the managerial hierarchy. For a theo-ry of this process grounded in ethnographic casestudies, see R.J. Thomas, What Machines Can’t Do:Politics and Technology in the Industrial Enterprise,Univ. California Press, Berkeley, 1994.

24. F. Wallace, Appraising the Economics, 1956, p. 47and 65.

25. H.S. Levin, Office Work and Automation, JohnWiley & Sons, New York, 1956, p. 95. His opin-ions echo those of J.M. Thesis and R.E. Slater,“What Management Has ‘Discovered’ aboutComputers,” J. Machine Accounting, vol. 6, no. 7,July–Aug. 1955, p. 24. On the depreciation ofpunched card equipment, see D.C. Niles, “Pur-chase versus Rental of Data ProcessingEquipment,” Systems and Procedures, vol. 8, no.1, Feb. 1957, p. 28. The topic became muchmore important around this date, as IBM wasforced on antitrust grounds to offer purchase aswell as rental options. Experts continued to dis-count the importance of obsolescence even aftertransistorized models had been announced and

disk drives were becoming common. See N.J.Dean, “Computer Installation—Will It Pay ToWait?,” The Management Rev., vol. 49, no. 3, Mar.1960, p. 25, and J.H. Dillon, Data Processing inNavy Management Information Systems,SecNavInst. P 10462.7, Dept. of the Navy, Wash-ington, D.C., 1959, p. V-9.

26. F. Wallace, Appraising the Economics, 1956, p. 21.The difficulty in achieving projected savings wasalready apparent—see P.B. Laubach and L.E.Thompson, “Electronic Computers: A ProgressReport,” Harvard Business Rev., vol. 33, no. 2,Mar.–Apr. 1955, p. 125.

27. R.G. Canning, “Planning for the Arrival ofElectronic Data Processing,” J. Machine Account-ing, vol. 7, no. 1, Jan. 1956, pp. 22-23, 30. Forcase studies of actual companies, see P.B.Laubach, Company Investigations of AutomaticData Processing, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1957, pp.29-121.

28. J. Dearden and F.W. McFarlan, Management Infor-mation Systems: Text and Cases, Richard D. IrwinInc., Homewood, Ill., 1966, p. 23.

29. P.B. Laubach, Company Investigations of AutomaticData Processing, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1957, p. 3.

30. A full decade later, a leading trade publicationcould still report that “experience has establishedthat the installation of a computer does not obvi-ate the need for punched card systems. In fact,the opposite may often be true,” in “TheIndispensable Card Machines,” Business Automa-tion, vol. 14, no. 1, January 1967, pp. 37-39.

31. Automation in the Office, Nat’l OfficeManagement Assoc., Willow Grove, Pa., 1957.These findings are broadly in agreement with theresults Cortada recently achieved by examiningthe frequency of discussion of particular applica-tions in the computing and specialist businesspress of the era. See J.W. Cortada, “Using TextualDemographics to Understand Computer Use:1950–1990,” IEEE Annals of the History of Comput-ing, vol. 23, no. 1, Jan.–Mar. 2001, pp. 34-56.

32. The precise quotation is from a case study of anunnamed shoe company in P.B. Laubach, Compa-ny Investigations of Automatic Data Processing,Graduate School of Business Administration, Har-vard Univ., Boston, 1957, p. 126. The Harvardteam gave the same firm’s computer planningexercise a book-length treatment in E.L. Wallace,Management Influence on the Design of Data Pro-cessing Systems, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1961.

33. P.B. Laubach, Company Investigations of AutomaticData Processing, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1957.

34. For figures on IBM revenue see J. Cortada, Before



the Computer: IBM, Burroughs, and RemingtonRand and the Industry they Created, 1865–1956;Princeton Univ. Press, Princeton, N.J., 1993, p.153. For the history of the tabulating machineand the importance of the Social Security Admin-istration to IBM, see also M. Campbell-Kelly andW. Aspray, Computer: A History of the InformationMachine, Basic Books, New York, 1996. Early useof punched card machinery is discussed in A.L.Norberg, “High-Technology Calculation in theEarly 20th Century: Punched Card Machinery inBusiness and Government,” Technology and Cul-ture, vol. 31, no. 4, Oct. 1990, pp. 753-779. Verylittle has been written on the clash between revo-lutionarily inclined computer specialists and old-school punched card experts. For an insightfulFinnish analysis, see M. Vehvilainen, “Gender andComputing in Retrospect: The Case of Finland,”IEEE Annals of the History of Computing, vol. 21,no. 2, Apr.–June 1999, pp. 44-51.

35. M.L. Edwards, The Effect of Automation onAccounting Jobs, unpublished doctoraldissertation, Univ. of Oklahoma, Norman, Okla.,1959, p. 156, lists the machines in use.

36. Ibid., and F.M. Johnson, “Control of MachineAccounting Equipment,” Systems and ProceduresQuarterly, vol. 4, no. 2, May 1953, pp. 18-22, 26.

37. The best source on punched card careers remainsM.L. Edwards, Effect of Automation, 1959. For afascinating discussion of punched card work onthe cusp of the computer age, see E.P. Daro,“Workshop Seminar on Selection and Training ofPersonnel,” in Data Processing: 1958 Proceedings,C.H. Johnson, ed., Nat’l Machine AccountantsAssoc., Chicago, 1958, pp. 324-338. For an earlyattempt to apply scientific testing to operatorselection, see E.J. McCormick and R.H. Finn,“Tests for Use in Selecting IBM Operators,” J.Machine Accounting, vol. 6, no. 2, Feb. 1955, pp.12-13, 17.

38. See M.L. Edwards, Effect of Automation, 1959, p.114, on department age. For IBM’s revenue, seeC.J. Bashe et al., IBM’s Early Computers, MIT Press,Cambridge, Mass., 1986. The estimate of thenumber of installations is from A.E. Keller, “Crisis inMachine Accounting,” Management and BusinessAutomation, vol. 5, no. 6, June 1961, pp. 30-31.

39. R.E. Sprague, “Are Punched Card Machines onthe Way Out?,” The Hopper, vol. 4, no. 7, July1953, pp. 2-7.

40. Ibid.41. Ibid., p. 4.42. For the survey, see “Survey of the 702,” Comput-

ing News, vol. 5, no. 94, 1 Feb. 1957, pp. 3-5.The quotation is from “Staff Organization andTheir Training,” Computing News, vol. 5, no. 95,15 Feb. 1957, pp. 8-11. Computer operatorsremained a substantial proportion of the data

processing workforce for decades, but continuedto be largely ignored by commentators.

43. “Business Week Reports To Readers on: Comput-ers,” Business Week, no. 1503, 21 June 1958, pp.68-92. This company might be Prudential Insur-ance—for premium billing, its first application,Prudential had to convert 13 million cards, hold-ing records of 320 characters each for three mil-lion policies. “Survey of the 702,” ComputingNews, vol. 5, no. 94, 1 Feb. 1957, pp. 3-5. For theeventual arrival of equipment to transcribe direct-ly onto tape, see A.L.C. Chu, “Key-to-Tape: K.O.For the Keypunch?,” Business Automation, vol. 17,no 4, Apr. 1970, pp. 52-59.

44. On the importance of sorting, see D.D.McCracken, H. Weiss, and T.-h. Lee,Programming Business Computers, John Wiley &Sons, New York, 1959, pp. 300-330. On GE’sreintroduction of manual sorting, see “BusinessWeek Reports To Readers On: Computers,” Busi-ness Week, no. 1503, 21 June 1958, pp. 68-92.

45. For the origins of integrated data processing, seeA New Approach to Office Mechanization: Integrat-ed Data Processing through Common LanguageMachines, American Management Assoc., NewYork, 1954. On the use of pneumatic tubesalongside computers, see A. Newgarden, “Men,Machines and Methods in the Modern Office,”AMA Management Reports, vol. 5, 1958, p. 7.

46. For early data on relative pay in computing instal-lations, see R.F. Clippinger, B. Dimsdale, and J.H.Levin, “Automatic Digital Computers in IndustrialResearch,” J. Machine Accounting, vol. 6, no. 2,Feb. 1955, pp. 7-11, 14-16; F. Wallace, Appraisingthe Economics, 1956; and R.G. Canning, InstallingEDP Systems, 1957.

47. The first quotation is from M.L. Edwards, TheEffect of Automation on Accounting Jobs,unpublished doctoral dissertation, Univ. Okla-homa, Norman, Okla., 1959. The second is fromJ.J. McCaffrey, From Punched Cards to PersonalComputers, 10 June 1989, John J. McCaffreyMemoirs (CBI 47), CBI.

48. On the SSEC (or Symbolic Sequence ElectronicCalculator) and the GE installation, see H.R.J.Grosch, Computer: Bit Slices from a Life, Third Mil-lennium Books, Novato, Calif., 1991. On IBMand design, see T. Watson Jr. and P. Petre, Father,Son & Co: My Life at IBM and Beyond, Bantam,New York, 1990.

49. A good overview of the new physicalenvironment of the computer is given in “BasicElements of Computer Environment,” Manage-ment and Business Automation, vol. 4, no. 5, Nov.1960, pp. 28-30, 32, 48; the quotation is frompages 29 and 48. These issues are also discussedin D.R. Daniel, “Getting the Most Out of YourComputer,” EDP: The First Ten Years: Highlights of


Management Experience and a Look Ahead, McK-insey & Company, ed., Am. Soc. for PublicAdministration, Chicago, 1961, pp. 15-21, andretrospectively in T. Kishi, “The 701 at theLawrence Livermore Laboratory,” Annals of theHistory of Computing, vol. 5, no. 2, Apr.–June1983, pp. 206-210.

50. The unreliability of the 604 is discussed in J.J.McCaffrey, From Punched Cards to Personal Com-puters, 10 June 1989, John J. McCaffrey Memoirs(CBI 47), CBI. On the need for air conditioningwith a 603 or 604 see B.J. Lappeus, “MovingDay ... In the Machine Accounting Department,”The Hopper, vol. 3, no. 3, May 1952, pp. 20-26.

51. J.B. Hughes and D.D. McCracken, “IBM 700Series,” J. Machine Accounting, vol. 7, no. 10, Oct.1956, pp. 26-29, 48.

52. “Basic Elements of Computer Environment,”Management and Business Automation, vol. 4, no.5, 1960, p. 32.

53. Figures are taken from an internal IBM intelligencereport on the Univac: W.R. Elmendorf and W.W.Peterson, A Study of the UNIVAC, IBM, 1954. Cuth-bert C. Hurd papers (CBI 95), CBI. For a detailedaccount of physical installation, see R.G. Canning,Installing EDP Systems, 1957, pp. 103-114.

54. R.W. Pomeroy, “Basic Flow Charting Techniques,”Systems and Procedures, vol. 8, no. 3, Aug. 1957,pp. 2-8. Pomeroy was a management consultant,and his article was, revealingly, a plea to his fel-lows not to abandon their unglamorous yetproven flowcharting techniques in the face of thenew fervor over electronics.

55. Inasmuch as historians have documented theexistence of the systems men at all, they havebeen conflated with computer experts, officemanagers, industrial engineers, or operationsresearchers. In fact, they considered themselvesquite separate from all these groups. See T.Haigh, “Inventing Information Systems: The Sys-tems Men and the Computer, 1950–1968,” Busi-ness History Rev., vol. 75, no. 1, Spring 2001, pp.15-61 for their story.

56. The quotation, an early example of what wasalready a mantra, is from N. Chapin, “Justifying theUse of an Automatic Computer,” The Hopper, vol.5, no. 8, Sept. 1954, pp. 9-10, 14. While on theone hand, the stress on systems analysis was also acall for business reorganization (and so for revolu-tionary rather than incremental change), it wasalso a claim that office systems experts had madeabout dictating machines, accounting machines,and many other previous technologies. Thisemphasis on the analysis of business systems overtechnical knowledge was very familiar in theory—ifnot widely followed in practice.

57. I. Place, Administrative Systems Analysis—Michigan Business Reports, Number 57, Univ.

Michigan, Ann Arbor, 1957.58. Leffingwell included a detailed office flowchart in

his W.H. Leffingwell, Scientific Office Management,A.W. Shaw, Chicago, 1917.

59. R.F. Neuschel, Streamlining Business Procedures,McGraw-Hill, New York, 1950.

60. A. Whitney, “How to Prepare MachineProcedures,” The Hopper, vol. 4, no. 2, 1953,pp. 1-5.

61. P.B. Laubach, Company Investigations of AutomaticData Processing, Graduate School of BusinessAdministration, Harvard Univ., Boston, 1957, p. 52.

62. The best single source on early administrativeprogramming is the first textbook devoted to thetopic, D.D. McCracken, H. Weiss, and T.-H. Lee,Programming Business Computers, John Wiley &Sons, New York, 1959.

63. T. Stein, “Managing the Data Processing Depart-ment,” presented at Int’l Data Processing Conf.of the Data Processing Management Assoc., NewYork, 1962. For case studies in which operatorsand programmers are either selected as part ofthe same pool, or share training or tasks, see R.H.Brown, Office Automation—Selecting, Training,and Organizing Computer Personnel, 1959, Mar-ket and Product Reports Collection (CBI 55), CBI,and R.G. Canning, EDP for Business and Industry,1956, p. 45.

64. The quotes are from R.G. Canning, Installing EDPSystems, 1957, p. 41 and 85. Canning’s stress onthe separation of analysis from programmingstruck at least some of his contemporaries asunrealistic, suggesting that a variety of approach-es could be found between different companies.See McGee, W.C., “Book Revie –InstallingElectronic Data Processing Systems,” ComputingNews, vol. 15, no. 115, 15 Dec. 1957. These jobsmay have been more frequently separated inadministrative computing installations than intechnical computing installations. On the failureof efforts to establish coder as a separate job, seeB. Conway et al., Business Experience with Electron-ic Computers, 1959, pp. 88-90. Early on, detaileddiagrams of computer program logic were oftencalled block diagrams, to distinguish them fromthe true flowcharts that showed the flow of docu-ments through office systems.

65. F. Wallace, Appraising the Economics, 1956, p. 24.66. The quote is from P.B. Laubach, Company Investi-

gations of Automatic Data Processing, GraduateSchool of Business Administration, Harvard Univ.,Boston, 1957, p. 146. Similar sentiments areexpressed by the manager responsible for Gener-al Electric’s seminal Univac installation in G.M.Sheehan, “An Application to Payroll,” Proc. Auto-matic Data Processing Conf., R.N. Anthony, ed.,Harvard Univ., Graduate School of Business



Administration, Division of Research, Boston,1955, p. 155. The figure comes from “BusinessWeek Reports to Readers on: Computers,”Business Week, no. 1503, 21 June 1958, p. 87.See also R.H. Brown, Office Automation—Select-ing, Training, and Organizing Computer Personnel,1959, Market and Product Reports Collection(CBI 55), CBI, for a discussion of programmertraining, and B. Conway et al., Business Experiencewith Electronic Computers, 1959, pp. 83-93. Foran earlier statement of the theme, see P.B.Laubach and L.E. Thompson, “ElectronicComputers: A Progress Report,” Harvard BusinessRev., vol. 33, no. 2, Mar.–Apr. 1955, p. 127. Onprogrammer testing, see T.C. Rowan, “TheRecruiting and Training of Programmers,” Data-mation, vol. 4, no. 3, May-June,1958, pp. 16-18.Those with more intimate experience of comput-ing, such as R.G. Canning and D.D. McCrackenet al., tended to put more emphasis on the needto include at least some experiencedprogrammers in the new department.

67. J.W. Mauchly, “Electronic Accounting,” TheHopper, vol. 4, no. 12, Dec. 1953, p. 2. Prominentauthor, consultant, and publisher RichardCanning, for example, initially felt that theelectronic data processing machines should notproperly be called computers or vice versa. SeeR.G. Canning, EDP for Business and Industry, 1956,pp. 82-83.

68. Nowland & Co., Management Report: Burroughs’Corporate Image Planning and Devlopment, Nov.1957, Burroughs Corporation Records (CBI 90),CBI. Very early use of data processing does notfollow this neat divide between administrativeand technical work—the scientific 701 wasrenamed from the defense calculator to an elec-tronic data processing machine. But by the late1950s the distinction was generally observed, tothe extent that computing and data processingwere sometimes spoken of as distinct activities.See, for example, the reference to the “operationof a data processing or computing installation” inH.R.J. Grosch, “Software in Sickness and Health,”Datamation, vol. 7,no. 7 July 1961, pp. 32-33.

69. While Levis’ classic 501 denim jeans never sharedits designation with an IBM model, the 501 wasan RCA mainframe. In Japan, more obscure pantscarrying designations such as the 701 and 702are highly collectable today and share theirnames with IBM’s first large computers.

70. “Glowing Future for Machine AccountantsDescribed by Convention Speakers,” J. MachineAccounting, Systems and Management, vol. 5, no.7, July–Aug. 1954, pp. 3, 8, and T.J. Watson Jr.,“Address by Thomas J. Watson Jr., President, Inter-national Business Machines Corp.,” in Charles H.Johnson (ed.) Data Processing 1: 1958 Proceedings,

National Machine Accounting Association, Chicago,1958. For examples of the early discussion of thechallenges and potentials of computer technolo-gy within the community of punched card super-visors, see R.G. Wright, “Electronics Challenge toMachine Accountants,” J. Machine Accounting,vol. 7, no. 4, Apr. 1956, pp. 4-7,27, L.E. Hill, “TheMachine Accountant and his ‘Electronic’ Oppor-tunity,” J. Machine Accounting, vol. 8, no. 1, Jan.1957, pp. 12-14, 23-25.

71. The name change was defeated on separateoccasions by the Executive Committee and thechapter delegates who made up the InternationalBoard of Directors.

72. Although these tensions dogged the certificatethroughout its history, they received their cleareststatement during the initial planning period. SeeNational Machine Accountants Assoc., “ExecutiveCommittee Meeting Minutes—Verbatim,” 5-6Aug. 1960, Data Processing Management Assoc.Records, (CBI 88), CBI.

73. To receive it, a candidate needed to demonstratethree years’ experience with data processing(punched card or computer) and provide threereferences to show the “moral character”demanded of a professional. Candidates also hadto pass a short test dealing with three main sub-jects: punched card techniques, computer tech-nology, and “systems.” (This last was a grab-bagcategory that included analysis and managementas well as “systems machines” such as opticalscanners and data transmission.) The three sub-jects together counted for about 80 percent ofthe marks. The final 20 percent were splitbetween knowledge of paper tape (used to trans-fer data between telex machines, mechanicaloffice machines, and electronic systems) andsome elementary questions on statistics. Execu-tive Committee Minutes, 31 Aug.-1 Sept. 1962,Data Processing Management AssociationRecords (CBI 88), CBI, p. 123 and ExecutiveCommittee Minutes, 30 Nov.-1 Dec. 1962, p. 279.

74. Inasmuch as the concept of data processing hasappeared in the historical literature, it has beenused primarily as a neutral description of a particu-lar kind of work. M. Campbell-Kelly, “The RailwayClearing House and Victorian Data Processing,” inInformation Acumen: The Understanding and Use ofKnowledge in Modern Business, L. Bud-Frierman,ed., Routledge, London 1994, even extends dataprocessing back to the 19th century—a deliberateanachronism that succeeds in highlighting surpris-ing historical continuities, but threatens to obscurethe historical specificity and culturally chargednature of the concept. The data processing con-cept is inextricably tied to particular technologiesand to a particular concept of how they should be




used and who should use them. It does not mean-ingfully predate the computer.

75. Even some of the most insightful recent work,such as N. Ensmenger and W. Aspray, “Softwareas Labor Process”, in Mapping the History of Com-puting: Software Issues, U. Hashagen, R. Keil-Slawik, and A. Norberg, eds., Springer-Verlag,New York, (to be published) does not address thedifferent uses to which the computer was put orto differentiate systematically between the differ-ent social institutions within which the activity ofprogramming was practiced. Without such differ-entiation we cannot explore what is trulycommon or divergent between different kinds ofcomputer user, and risk the reification of ideaslike “the community of software workers” or“professional programmers,” which imply strongidentities where none existed.

76. J.S. Light, “When Computers Were Women,”Technology and Culture, vol. 40, no. 3, July 1999,pp. 455-483. Light’s formulation mirrors that ofP. Kraft, Programmers and Managers: TheRoutinization of Computer Programming in theUnited States, Springer-Verlag, New York, 1977. A1960 survey by Management and BusinessAutomation magazine examined more than7,000 data processing staff at 489 companiesacross the US. It found that less than 15 percentof the programmers surveyed were female.

77. While private trade schools offered educationand tests in programming, no generally agreedstandards were in place. The CDP was not a cer-tificate for programmers—no experience in pro-gramming was required, and the DPMA nevermade a serious effort to interest scientificprogrammers in the certificate or to adapt it totheir needs. The first serious attempt to provide aprogramming credential came with the DPMA’sRegistered Business Programmer (RBP) examina-tion in 1968, when the DPMA recognized thatcertification of programmers might provide avaluable alternative to its supposedly broader andhigher level certificate in data processing. TheRBP examination (a name deliberately chosen tosound less impressive than a certificate) wasgreeted by widespread apathy and proved afinancial disaster for the association.

78. R.G. Canning and R.L. Sisson, The Management ofData Processing, John Wiley & Sons, New York,1967, p. 9. The idea that the data processingdepartment remained a simple extension of theold tabulating department is also expressedstrongly in E.H. Brock, “Coundown to ADP Man-agement Crisis,” in Data Processing XII: Proc.1967 Int’l Data Processing Conf. and BusinessExposition, Data Processing Management Assoc.,ed., Chicago, 1967, p. 213.

79. “Business Week Reports To Readers on: Comput-ers,” Business Week, no. 1503, 21 June 1958, p. 59.

80. For a good summary of the “productivityparadox” discussion, see D.E. Sichel, “Computersand Aggregate Economic Growth: An Update,”Business Economics, vol. 34, no. 2, Apr. 1999, pp.18-25, and J. Madrick, “Computers: Waiting forthe Revolution,” in New York Rev. of Books, vol.45, 1998, pp. 29-33.

81. On disruptive technologies, see C.M.Christensen, Innovator’s Dilemma: When NewTechnologies Cause Great Firms to Fail, HarvardBusiness School Press, Boston, 1997.

Thomas Haigh is completing aPhD in the history and sociolo-gy of science at the Universityof Pennsylvania, writing thefirst full-length synthetic histo-ry of American corporate useand management of informa-tion technology from the 1910s

to the 1970s. He holds BSc and MEng degrees in sys-tems integration from the University of Manchester.Recent fellowships include a Fulbright Award forgraduate study in the US, the IEEE Life MemberFellowship in Electrical History, and the TomashFellowship from the Charles Babbage Institute of theUniversity of Minnesota. He is currently teaching atColby College ([email protected]).

For further information on this or any other com-puting topic, please visit our Digital Library athttp://computer.org/publications/dlib.

The Chromium-Plated Tabulator: Institutionalizing an Electronic

Documents

Transcript of The Chromium-Plated Tabulator: Institutionalizing an Electronic