The Tabulating Machine Company had its origins in Herman Hollerith’s punched-card business in the 1890s and became the prime mover for the use of punched cards in statistics processing. As punched-card systems were developed to handle bookkeeping between 1907 and 1933, the Tabulating Machine Company and its successor company, the International Business Machines Corporation (IBM), retained the prime mover position and became one of the most influential companies in the United States. This was achieved through the company’s shaping of punched-card-based bookkeeping and in spite of the emergence of several challengers. The successful transformation of the company’s primary application field from statistics processing to bookkeeping operations was facilitated by fundamental changes in the company’s sales and machine developments.
The Tabulating Machine Company merged into the new Computing Tabulating Recording Company in 1911, which was renamed IBM in 1924. Thomas J. Watson Sr. joined the company in 1914, organized its sales to be the most efficient among office machine producers, and established a machine development department. The machine development staff first managed to bring the company abreast of the Powers company by 1921, and then they went on to shape a punched-card system for bookkeeping that became the industry’s de facto standard.
During this process, several conflicting paths of technological development were pursued in the search for not-yet-defined goals. The core of this process was identifying and interpreting a new prime application in bookkeeping.
In 1911, Herman Hollerith and his directors sold the Tabulating Machine Company to a conglomerate established by the entrepreneur Charles Ranlett Flint. Company growth through mergers and acquisitions had become increasingly widespread in the United States since the 1880s. Charles R. Flint was a prominent promoter of mergers, and among his more important promotions were the United States Rubber Company (1892) and the American Woolen Company (1899).1
There were two main reasons for the sale of the Tabulating Machine Company. First, the latent reason, that Hollerith’s one-person management style could not control his fast-growing company much longer. The company’s revenues had grown on average 50 percent per year from 1908 to 1911. Second, Charles R. Flint’s great offer was hard to turn down and served as a trigger. Flint’s offer of $2.3 million was about twenty times higher than an offer made three years previously that had been turned down. Hollerith claimed failing health as a reason to accept the offer.2 He might have had health problems, but he was only fifty-one years old and would live for another seventeen years.
The newly merged company was named the Computing Tabulating Recording Company (CTR). In addition to the Tabulating Machine Company, the components were the Computing Scale Company of Dayton, Ohio, the International Time Recording Company of Binghamton, New York, and the Bundy Manufacturing Company of Endicott, New York. The Computing Scale Company produced scales equipped with a chart enabling clerks to determine an item’s price as it was being weighed. The Bundy Manufacturing Company and the International Time Recording Company produced machines that recorded when workers entered or left their workplace.3
Charles R. Flint later described CTR as his most successful trust construction,4 but the shape appears curious. It was neither a horizontal consolidation of concerns of similar or competing products, nor was it a vertical combination. It was a conglomerate of four producers and manufactures of fine mechanics-based business equipment, which largely supplemented each other. The main advantage sought seems to have been the consolidation of their facilities for manufacturing and distribution. Of the four companies, only the Tabulating Machine Company did not produce its own machines. It is true that in 1901 Hollerith had bought a machine shop, the Taft-Peirce Company of Woonsockett, Rhode Island, but it had been sold in 1905. In 1911, the Tabulating Machine Company moved production to the amalgamated company’s various production facilities.5
Another advantage of the merger was the use of skilled engineers for inventions and for developing projects across the borders of the old companies.6 However, the new company formed a weak umbrella organization for the four merging companies, which remained largely independent. Therefore, the Tabulating Machine Company kept its former style regarding customer contacts and patents.
In 1911 the four parties to the merger seem to have been of comparable sizes, but as time went by, punched cards became the most prominent activity in the company. In 1918, the Tabulating Machine Company earned half the total revenue of the merged company, and adaptation of punched cards for bookkeeping purposes offered a far larger sales potential than did smart scales or time recording systems.7 In 1935, the company sold its scale interests, and the time recorders followed in 1958.8
From the outset, the Computing Tabulating Recording Company’s headquarters was in New York City, where the Tabulating Machine Company’s management moved from Washington, D.C.9 After the merger, Hollerith resigned as general manager and later left the board.10 This freed him from executive responsibilities and enabled him to concentrate on machine development, the work he always enjoyed most. He stayed with the company as a consulting engineer for a decade, and he was very influential until 1914 because all proposals for change in machine design were submitted to him for approval.11
Hollerith was far from burned out and made important improvements to his system and filed eight patent applications between 1911 and 1914. His main collaborator was still Eugene A. Ford, who continued to work up in Massachusetts.12 Machine development in this period was influenced by the new strength of the Powers Accounting Machine Company, and the management of the Tabulating Machine Company noticed the emergence of John Royden Peirce as a punched-card producer.13
At that time, Hollerith was not interested in extending the scope of applications to encompass bookkeeping. He worked instead on four machine improvements, of which two were influenced by competition from the Powers company. Electric brush reading of cards was susceptible to error due to tiny holes or specks of conducting material in the cards that could cause the machines to register incorrect figures. Hollerith’s early pin reading, which James Powers had adapted for his mechanical punched-card system, was less sensitive because the reading only took place as a pin passed through a perforation on a waiting card. To solve the problem of increased errors from brush reading, Hollerith invented a speck detector in 1914 that located flaws in card stock using an electric test current.14 The date indicates an influence from Powers’ system, as brush reading had been on the market for more than ten years.
The start of Powers’ production of a printing tabulator in 1914 also influenced Hollerith’s considerations. In 1908, the first printing tabulator in the Census Bureau had demonstrated the feasibility of such a design. This printing tabulator required manual release by the operator for each print and only printed totals. Another possibility was to “list” cards, that is, to print selected information from every card. In statistics processing total printing was an advantage as few results were produced compared to the number of entries. In contrast, in bookkeeping the number of results was high compared to the input, and listing data on the cards was also important. By 1911, the listing and total printing Burroughs adding machine had been a great success for two decades. Furthermore, a listing and total printing tabulator was a major feature of John Royden Peirce’s punched-card system, marketed in 1912.
Hollerith had worked on a total printing unit as early as 1899, and he filed patent applications for this capability in 1905 and 1913.15 In 1913, he also filed a patent application for a listing tabulator.16 However, Hollerith stuck to the view that only a total printing tabulator should be built; his vision of punched-card applications only encompassed statistics processing. He even resisted an attempt to build a listing tabulator as late as 1917.17 Since the 1880s, Hollerith had been the imaginative designer of punched cards for statistics processing but never appreciated the vast business and technical potentialities of punched card for bookkeeping. It was true that in the spring of 1912 the Tabulating Machine Company collected information about the emerging John Royden Peirce punched-card system for bookkeeping, but the initiative rested with the management succeeding Hollerith, and he was only informed subsequently.18
Hollerith’s first machine improvement of his closure for general statistics processing aimed at improving processing cards with information on many small groups, for example, for processing statistics on the age distribution of women in a county. For that task, the cards were sorted according to age, and then the cards for each age group were taken to be run separately on the tabulator. The operator had to stop the tabulator each time, copy the results from a Hollerith tabulator by hand—or release the printing mechanism on a Powers tabulator. Subsequently, the operator removed the small punched-card stack, put in a new stack, and reset the tabulator. Hollerith filed patents on batch processing of several individual stacks of punched cards separated by introducing “stop cards” in 1912, as did James Powers shortly afterward.
These special cards were introduced into the punched-card stacks manually or through sorting, in the above example, after the cards on each age group. The tabulator came to a halt when it reached a stop card, and the operator either copied from the counters or released the printing mechanism before the next group was processed. In the Hollerith system, the stop cards were blank cards provided with a particular cut out. When a stop card was hit by the seizing arm, the tabulator stopped and the sum of the group could be read. The adding unit was reset manually and the tabulator restarted.19 The Hollerith company introduced stop card control in about 1914.
The use of stop card control was unsatisfactory, as inserting and removing cards was time-consuming. Hollerith devised a way to control processing separate groups without stop cards using information punched on the cards, read twice in two separate reading stations. The simultaneous reading of two consecutive cards enabled the tabulator to compare without storing any information. When a change occurred in a selected field, the “group indicator,” the machine stopped. In the age statistics example, the age could be used as the group indicator. The technique was called “automatic group control” and was the earliest example of conditional programming of a punched-card machine, that is, control of processing of punched cards based on information on the cards. This technique significantly eased processing, and it became essential for developing efficient programming of electronic computers after the Second World War. Hollerith’s patent application ran into a lengthy litigation in the United States against James Powers and several other inventors. Hollerith’s important patent was only granted in 1931.20 Automatic group control was first implemented by the Tabulating Machine Company in 1921.
Hollerith’s second machine improvement of the closure in general statistics processing was the design of an easy touch key punch. The pantograph punch had been lever-based, which eased punching, and the Powers key punches used an electric motor for punching holes in the card. Hollerith’s idea was that key action would energize an electromagnet that would then perforate the card. A patent was filed in 1914, but the “electric key punch” was first marketed in 1917.21 This launch date might have been influenced by the improved Powers punch, the “automatic key punch,” which emerged in 1916.
Back in the years around 1911, the Tabulating Machine Company enjoyed fast-growing revenues, and the sales volume is apparent from the company’s having distributed 113 tabulators and 97 sorters to customers in 1914.22 However, the Tabulating Machine Company suffered from operational problems that needed to be solved. They had unsatisfactory sales activity, problems with technical development, and the newly amalgamated company experienced hitches in coordinating production.23
Of these, unsatisfactory sales activity was considered the major problem.24 To enhance sales, the company established branch offices. Prior to 1911, Chicago had been the only office outside Washington, D.C. Over the next few years sales activity was reorganized, based on a general sales office in New York with sales offices located in key cities throughout the United States.25 At the same time, the use of the various production locations within the amalgamated company required coordination, which did not always materialize.26 Furthermore, the new company was heavily in debt, having financed its own merger.27
The board addressed these problems in 1914 by engaging Thomas John Watson Sr. as general manager for the amalgamated company. The following year he became president of the company, and he retained this position for forty-one years until 1956.28 During his appointment, Watson proved a paternalistic entrepreneur who managed to turn the company into one of the most profitable in the United States. Thomas J. Watson was born in 1874. He attended one year of business school and became a salesman for the National Cash Register Company (NCR) in 1895. He learned the profession well and became sales manager in 1910.29 The extensive development of cash registers at NCR demonstrated to him the importance of a development department.
In 1912, John Henry Patterson, Watson, and twenty-eight other NCR officials were indicted and brought to trial for antitrust violations. Reportedly NCR had over 90 percent of the cash register business. A year later, the jury reached the verdict that all defendants were guilty except one. Watson, Patterson, and one more of the defendants were given the maximum sentence of one year in prison. An appeal was granted in 1914, and the case ended with NCR signing a consent decree to clean up its business practices. During the appeal, tensions rose between Watson and Patterson, and Watson resigned. Indications exist that these tensions were a result of the antitrust case, but Patterson’s biographer found it but a sign of Patterson’s eccentric way of running his company. Seen from this perspective, Patterson fired Watson, along with several others, because he had become too powerful.30
Watson brought extensive knowledge from NCR of how to run a business machine producer, especially sales. Over the next several years, he gradually reshaped his new company, building an efficient sales organization, establishing a development department, and creating a corporate spirit. When Watson entered the company, the foundations of a sales organization had been established, but he made fundamental improvements. Each salesman was assigned his own exclusive sales district where he had the responsibility to develop sales, knowing that no other salesman from his company would try to poach his customers. The salesmen were considered the cream of the company, well-paid through generous commissions and smartly dressed as the company expected. Systematic selling was taught within the company.
The company’s sales strategy was first to analyze the customer’s problem, to demonstrate how IBM punched-card equipment could solve the problem, and finally to sign a contract that had the same rental rates for all customers. To monitor and improve sales, a quota system was introduced. The initial sales quota was based on the population and the business structure of the district, and the quotas for the following were adjusted according the sale in the district. When a salesman reached his quota, he became member of the “One Hundred Percent Club” of that year. Every year all members of the club were invited to and applauded at sales conventions. Watson organized the first One Hundred Percent Club in 1916 in each of the constituent companies, and they became a united event after 1922, which indicated the ongoing integration of the company. Over the next few years, these events evolved into huge conventions, which were held at impressive hotels like the Waldorf-Astoria in New York.31
Sales conventions became a principal component of the corporate spirit that included all employees. They met at various big gatherings, listened to talks, and sang company community songs, for example, “Ever Onward.” The lyrics were written to a stirring tune from a Sigmund Romberg musical. Community songs were typical of the age, as reflected in the distinctive songs of various political movements of the interwar years. The corporate spirit was also promoted through slogans, foremost “Think.” Other examples were “There Is No Such Thing as Standing Still,” “Time Lost is Gone Forever,” and “Beat Your Best.”32
When Thomas J. Watson became general manager of the Computing Tabulating Recording Company in 1914, the company produced four punched-card machines: a key punch, a vertical sorter, a nonprinting tabulator that could operated by using stop cards, and a gang punch, which in one operation could punch one perforation in several cards. The key punch exhausted the operator, as it required substantial force to operate.33 The tabulator was reliable, but the absence of a printing capability was an increasing disadvantage as the number of printing Powers tabulators grew.34 Hollerith had worked to improve the punch and to build a printing tabulator for three years in Washington, D.C., in cooperation with Eugene A. Ford, who worked in Massachusetts. But no solution had been found.
Thomas J. Watson came from eighteen years at NCR, which had had a highly active development department since 1888. Within a few months of becoming general manager, Watson made Eugene A. Ford move from Massachusetts to New York to establish the first development department for punched-card machines. Soon a model shop was equipped and ten model makers were hired.35 Hollerith never traveled to New York to contribute directly to the development efforts, but he contributed through an extensive correspondence for several years.36
The first new senior technical person to be hired for the development department was Clair Dennison Lake in 1915. He was an automobile designer with no previous punched-card experience. Fred Merchant Carroll was hired the following year. He was an experienced inventor who previously had filed patent applications for printing adding machines and cash registers.37
In 1917, Watson hired James Wares Bryce as supervising engineer of the time recording division of the company. Bryce studied mechanical engineering at the College of New York, but he quit after three years to take up a position as draftsman and designer. Since 1905, the patent rights for his time recording developments had been assigned to the International Time Recording division, which became a part of the Computing Tabulating Recording Company in 1911. In 1922, Bryce was appointed chief engineer of the whole Computing Tabulating Recording Company, which put him in charge of the development of punched-card technology.38
The approach to punched-card development at the Computing Tabulating Recording Company—and later IBM—was decided upon by Watson and diverged from the development at the challenger companies. In several cases the company pursued competing project paths, and only at the end of the projects did Watson choose the design from one of the projects for production. In addition, the Tabulating Machine Company and IBM bought inventions from individual inventors and they, in fact, hired several of these inventors. This supplied the management with alternative solutions. But it was a costly approach that only this punched-card company could afford.
The development of a printing tabulator was the first instance in which the Tabulating Machine Company pursued several competing development paths, but the scope of what inventors were developing remained limited to the capability to print numbers and the implementation of automatic group control. Watson assigned both Clair Lake and Fred Carroll the task of independently developing printing tabulators.39
Clair Lake’s development group developed printing units for the existing tabulators, and their tabulator could both list and print totals. First, they worked on a tabulator printer with printing wheels, which did not succeed.40 Then, they developed a printing unit that used type bars and completed a prototype in 1917.41
Simultaneously, Fred Carroll’s development group pursued the design of a completely new tabulating machine, which incorporated a novel rotating drum printer, a very complex design that only printed totals.42 Watson selected Lake’s design to keep development and manufacturing costs low. However, the printing tabulator was not introduced until 1920 and was supplied to customers the following year. The delay was probably caused by reliability problems.43 Though the development of a numeric printing capability became protracted, the company desisted from buying alternative designs, which were offered by independent inventors.44
The Lake tabulator supplied from 1921 was the first commercial tabulator with automatic group control, which was based on a patent application by Herman Hollerith in 1914. The new tabulator brought the company into conflict with patent applications from James Powers, John Royden Peirce, and Charles A. Tripp. The Patent Office decided to give Hollerith’s patent application priority over Powers’ application in 1919. The conflicts with Peirce’s and Tripp’s patent applications developed less satisfactorily for the Tabulating Machine Company, as these applications were assessed as being superior by the Patent Office. This caused the Tabulating Machine Company to buy Peirce’s patent in 1922 and Tripp’s patent two years later.45
The key punch was the second reverse salient of the Hollerith system, to use Thomas P. Hughes’s concept. The problem was that the operator had to punch the card manually, which was extremely tiring. In addition, the key punch became an issue of competition when the Powers Automatic Key Punch appeared in 1916. Even the Tabulating Machine Company’s own engineers considered the Powers punch to be superior.46
Again several paths of development were pursued to improve the punch. First, an electrically driven key punch was built by Hollerith in 1913 but only supplied from 1917.47 Second, a key punch developed by independent inventors was purchased in 1914, but it never reached the market.48 As with the tabulator, the electrically activated punch brought the company in contact with another independent inventor, John Thomas Schaaff, who claimed patent infringement. The Tabulating Machine Company settled his claim by acquiring his patents and employing him as a development engineer.49 The Tabulating Machine Company marketed a new punch with electrically driven keys in 1923, but the operator still needed to enter and remove each card manually. Only in 1929, did IBM get a key punch with motor-driven feed and ejection facility, which completely equaled the Powers company’s punch from 1916.50
The Tabulating Machine Company introduced a “verifier” in about 1917 as a means of checking the work of key punch operators. An operator used it to repeat the original key punch operator’s work, and if a keying discrepancy occurred the verifier signaled this by locking its keyboard. If an inspection revealed the questionable card to be incorrect, the verifier operator prepared a correct card with an ordinary key punch.51
By introducing the number printing tabulator in 1921 and the electrically enforced key punch in 1923, the Tabulating Machine Company now produced a total line of machines that were ahead of the Powers company’s products. This number printing tabulator became IBM’s basic tabulator during the 1920s and remained in service into the 1940s.52
The Patent License to Powers
Shortly after joining the Computing Tabulating Recording Company, Thomas J. Watson was approached by representatives from the Powers Accounting Machine Company requesting a licensing agreement to use some of the still-valid Hollerith patents. “You are in a position to put us out of business,” Watson several decades later recalled one of them saying.53 The Computing Tabulating Recording Company decided to grant the Powers Company a license to use these Hollerith patents to produce mechanical punched-card equipment for 25 percent royalties on the gross receipts from the rentals of tabulators and sorters and 18 percent on gross receipts from card sales. This accounted for about 20 percent of their total revenues, as the Powers punches and verifiers did not depend on Hollerith patents.54 The license arrangement enabled Powers’ production but placed the company in an inferior competitive position.
Thomas J. Watson’s biographers saw this agreement as an example of Watson’s “clean, high-grade and square dealing[s].”55 However, the approach from Powers in 1914 gave the Computing Tabulating Recording Company two additional alternatives: They could reject the approach, forcing the Powers company out of business, or they could buy the Powers company (in the following decade they acquired patents from many individual inventors and bought John Royden Peirce’s punched-card company). Further, the royalty of about 20 percent in the United States should be contrasted with the royalty of 5 percent in the similar, but court-imposed, license in Germany.56
No record exists of the deliberations at the Computing Tabulating Recording Company, but Thomas J. Watson was still facing criminal antitrust charges for his activities at NCR.57 If James Powers’ request for a license had been rejected, or if the Tabulating Machine Company had purchased the Powers company, Watson would have reestablished Hollerith’s punched-card monopoly. This would certainly have harmed the reputation of Watson’s new company at a time when NCR’s antitrust case was not yet settled, and it could have attracted the attention of the Department of Justice to investigate possible antitrust violations.
The Powers company’s problems between 1918 and 1927 prevented them from paying their royalties to the Tabulating Machine Company. Their debt accumulated, and they got the royalty rates reduced by half in 1922.58 During the 1920s, the legal position of the conflicting patent applications on automatic group control remained unsettled until Hollerith’s patent was published in November 1931. Well before the license agreement from 1914 would expire in 1929, IBM filed a patent infringement suit against the Powers company’s successor, Remington Rand. To ward off this suit, Remington Rand entered a new five-year license agreement in March 1931, which also repeated the previous agreement’s market arrangement.59 The agreement between IBM and Remington Rand became the basis for the antitrust suit the following year against the two companies.
This history of the patent licenses to the American Powers company illuminates the different views on monopolies in Germany and in the United States. Alfred D. Chandler compared the cooperation among German companies with the competitive tradition in the United States, which was enforced through the antitrust legislation.60 However, the patent system in Germany comprised court-enforced licenses at a moderate rate, which facilitated competition between a first mover and challengers. In contrast, the United States had ultimate patent protection and antitrust enforcement. The patent protection forced a challenger to obtain a license from the first mover on the first mover’s conditions, which were hardly positively inclined toward the challenger’s business. This strengthened the patentee’s monopoly and weakened antitrust legislation.
Shaping Equipment for Bookkeeping
Before the introduction of the printing tabulator in 1920, the Tabulating Machine Company established a growing business with many manufacturers, railway companies, and insurance companies. Using their nonprinting tabulators, they were able to process business statistics with a limited number of outcomes, as the operators had to copy the results by hand. The printing tabulator eased this problem, but bookkeeping assignments were more complex to mechanize than processing statistics.
Access to a substantial extension of the punched-card business could be found either through the development of punched-card systems for core bookkeeping or by introducing less costly machines that could extend punched-card applications beyond the existing big company customers to encompass small and medium-size concerns. Both options were pursued by the Tabulating Machine Company. By 1920, core bookkeeping was carried out manually or by use of key office and address plate machines. Key office machines could subtract, multiply, and print letters. Address plate technology was a well-established, though cumbersome means to address letters and other correspondence.
The transition to bookkeeping was more complex than the development of the printing tabulator, produced from 1921. The development of the printing tabulator focused upon three tasks: list printing, total printing, and automatic group control. After the introduction of the printing tabulator, four more general elements were involved: subtraction, multiplication, letter printing, and an enlarged capacity card. How these were put in order of priority for implementation depended on the kind of bookkeeping application selected in the development process. Further, it was a very open-ended process, where even the basic punched card of twelve rows or punching positions was contested.
In 1922 the Tabulating Machine Company acquired all the patents held by the engineer John Royden Peirce, who was a small, independent punched-card producer, to get control of his important group control patent. It also hired Peirce. He brought with him his machines, which were different from the Tabulating Machine Company’s electromechanical punched-card machines. Peirce’s machines applied exclusively mechanical technology with pin reading, and he applied a double-deck 86-column card, quite different from the standard 45-column card.
Peirce had strong and attractive conceptual ideas about punched-card-based bookkeeping. He also had experience with printed lists of numbers and letters for policy administration in insurance companies. But for ten years he had been struggling with the problem of building reliable machines. For several years after he had joined the Tabulating Machine Company, his double-deck 86-column card remained the basis for policy administration at the Metropolitan Insurance Company in New York City and in the Prudential Insurance Company of Newark, New Jersey, and his new company built additional machines for this distinctive punched card.61 In spite of these shortcomings, Peirce was considered so essential for his new company that it paid an amount for his patents equaling 25 percent of its total net earnings in the year he was hired, plus his salary.62 The high price paid to hire Peirce and the conflicting paths of development pursued until 1930 indicated a company in jeopardy with respect to finding a path of development for their punched cards.
At the Tabulating Machine Company, the development process consisted of building several machines that contained various combinations of the four general elements of punched-card-based bookkeeping. There were up to four parallel machine-building projects at the same time. The two major lines of machine development were carried out by engineering teams headed by John Royden Peirce and Clair D. Lake.
Peirce’s group worked to develop machines capable of handling alphabetic as well as numeric information. Their work was based on Peirce’s experience from before 1922 and on his nonstandard double-deck punched card with 86 columns. Before joining the Tabulating Machine Company, Peirce had built a “notice writing” machine for the insurance division of the United States Veterans Bureau (Bureau of War Risk Insurance until 1924). This machine issued and addressed notices to be sent to policyholders, with a punched stub to be detached when paid to facilitate subsequent bookkeeping. After Peirce had joined the Tabulating Machine Company, a similar machine was built and supplied to the Metropolitan Life Insurance Company in New York.63 During the 1920s Peirce continued this line of development at IBM, and it was only terminated in 1930 after an additional tabulator had been supplied to Metropolitan. This tabulator could print numbers and letters in separate printing positions and it used Peirce’s double-deck 86-column card.64
Clair D. Lake had been at the Tabulating Machine Company’s development department since 1915, and his group based their machine development work on improvements to the existing machines. They focused on bringing the existing tabulator to perfection until 1925 when they turned to a phased extension of its capabilities to encompass bookkeeping: first an increased-capacity card, then a subtracting tabulator, and— finally—letter printing. For the increased capacity card, Lake and his group chose to keep the existing numeric punch code on the 45-column card, and they squeezed the columns together to achieve an enlarged capacity. This required building new punches and card readers, but unlike Peirce’s development approach, the basic numerical operations were not affected.65 The available material does not disclose if this group had a prime application field in the 1920s.
Peirce notice to insurance policyholders in 1922 with punched stub for subsequent bookkeeping. (J. R. Peirce, [U.S.] Patent 1,506,382, figure 9. Explanatory numbers removed.)
Watson selected Lake’s simpler design that used the 80-column punched card, and the machine was introduced in 1928. This design was chosen even though Lake’s prototype alphanumeric tabulator was not yet completed. Lake’s design was based on the representation of each character by up to two perforations, which was much simpler than Peirce’s representation. The ability to store alphabetic information in addition to numbers was judged, as yet, to be not sufficiently important to the customers. The 80-column card offered 78 percent more storage capacity than did the old standard 45-column card, and the rectangular perforations were its most explicit characteristics.
The new cards were rectangular instead of round, because this facilitated electric reading. Further, cards with rectangular perforations proved sturdier and made the new design patentable. The rectangular perforations provided the card with a distinct appearance that could not be copied by others.66 The age of the standard 45-column card was over and was about to be succeeded by proprietary cards. IBM protected their “IBM card” through litigation in France in the 1930s. In addition to handling the new card, Lake’s new tabulator had the capability to be wired to process the old 45-column card, which eased IBM’s service to its remaining 45-column-card customers.
The “IBM Card” of 80 columns introduced in 1928 with the standard English alphabet (four zone) representation from 1933. (IBM Corporate Archives, Somers, New York)
A “reproducer” was introduced to punch 80-column cards from existing 45-column cards to help customers to convert their records to the new 80-column card. The importance of being able to reproduce cards became evident through this experience, and copying from 45- or 80-column cards became a standard feature.
The second improvement from the Lake group was a tabulator with subtraction, which also was introduced in 1928. For this tabulator, negative numbers were identified by a perforation in row 11. Previously, positive numbers (credit) and negative ones (debits) were punched in separate fields on the card and separately totaled by the tabulator. Then balances were determined manually. Alternatively, the complements of negative numbers were manually determined and then keypunched in the same field as the positive numbers so as to achieve subtraction through addition of the complements.67
The Lake group also completed a prototype alphanumeric tabulator in 1928. This alphanumeric tabulator had a slightly reduced alphabet and enabled writing a full address, but it was not reliable and the company discontinued its further development.68 The dismissal of the Lake group’s work on an alphanumeric tabulator in 1928 and the termination of Peirce’s alphanumeric system in 1930 disclosed the IBM management’s low assessment of the importance of this feature for its customers. It reckoned that the customers primarily wanted increased capacity punched cards and subtraction.
The business opportunities based on these capabilities were increasing, as was indicated by the emergence of the new large application of public utility billing, that is, calculating the price for electricity, gas, and water based on meter readings and issuing invoices to be mailed to the customers. For this application, IBM modified a standard tabulator that came to perform automatic extra checks on its computation. Public utility companies were extremely sensitive to invoicing errors, as the invoices were produced on large numbers and went directly to the public. In this application, the result was either printed on forms or on postcards to be mailed to the customers, and a paper feed was built to enable printing on forms. However, public utilities only required their invoicing process to encompass numbers, while addressing was accomplished by the use of address plates.69
A new drive for alphabetic tabulator capability started in 1930. While Peirce had built punched-card machines with letters for insurance companies between 1916 and 1930, IBM now focused on chain stores that were undergoing a roaring expansion. The number of parent companies of chain stores had grown from 2,030 in 1914 to 7,046 in 1929, simultaneous with a rise in the number of store outlets from 23,893 to 159,826.70
In the mid-1920s, IBM had cultivated the chain stores by designing punched-card-based systems to control goods stored in their central stores. Every box in the store had a punched card stating its particulars, including product number and price. These cards provided a representation of the whole stock, which eased control. When an item was taken from the store to be dispatched to an outlet, the matching punched card was removed and used to produce the invoice.71
To extend this chain store application, in 1930 IBM started to develop a reduced alphabet tabulator to print invoices on shipments of goods from a central store of a chain store to the outlets. Reduced alphabet specifications were easier to verify by senders and recipients than product numbers. This development was based on the invention in 1927 at the Danat Bank in Germany of a simple reduced alphabet printing unit, which substituted the ten numbers on a numeric type bar with ten selected letters. This German invention had been acquired by IBM in New York by 1930.72 First, IBM implemented a reduced alphabet by substituting the ten digits on several type bars in a tabulator with ten selected alphabet characters. A few copies of this design were supplied to chain stores in 1931 and 1932. Then this design was developed to an extended type bar of twelve characters corresponding to all twelve rows on the punched card. The most used letters were spread over two adjacent type bars, and the machine was used to produce in-company invoices. The first tabulators of this design were shipped in 1931, and a total of 255 were produced. This tabulator remained on the market until the 1950s (though the company stopped soliciting orders for it in about 1937) as its features sufficed for in-company invoicing for chain stores and it was cheaper than the later full alphabet tabulators.73
Although a reduced alphabet worked for in-company invoice specifications, insurance companies needed a better representation to write lists of policyholders or to address letters. Around 1930, insurance companies typed addresses or printed them using address plates; they do not appear to have cared much for addressing with punched cards. When IBM terminated the production of equipment for Peirce’s double-deck, 86-column card in 1930, his engineering group switched their efforts to developing an alphabetical capability on the 80-column card that had been introduced in 1928.
From 1931, IBM produced a machine from this engineering group that provided for a reduced alphabet of twenty-one letters, with the remaining five letters were substituted by digits. The product was acceptable for lists for internal use in an insurance company or a bank, but not for letters to customers. The Twenshe Bank in Amsterdam, Holland, received the first copy of this tabulator in 1931, and the next year the Prudential Insurance Company in Newark, New Jersey acquired the second copy. A historian explained the launching of this reduced alphabet tabulator as a countermeasure against Remington Rand’s announcement in 1929 of their 90-column punched card.74 However, Remington Rand had chosen not to implement the American Powers company’s reduced alphabet from 1924 on its new punched card due to its lack of success. Consequently, the reason for developing and producing this reduced alphabet tabulator was more likely a demand conceived by the IBM sales organization, and it was perhaps also a measure to curb discontent in Peirce’s engineering group, who had had its proposal for a new card standard of a double-deck card turned down in 1928 and the production of its machines for the double-deck card terminated in 1930. The IBM tabulator with a reduced alphabet was a limited success, as only about ninety copies were produced, but it remained in use for several years just as the chain store tabulator was.75
The first of a series of full alphabet tabulators followed two years later, in 1933. A perforation in one of the top three rows combined with a perforation in one of the bottom nine rows were used to designate letters, which supplemented the old standard for representing digits on punched cards. This alphanumeric code provided for ten numeric and twenty-six nonnumeric characters, and it survived to the end of the punched-card era.76 This was sufficient for the English alphabet, but the German alphabet, for example, has several additional characters, which caused problems. Later types of tabulators enabled twenty-eight alphabetic characters in addition to two special characters, for example, & and * and were sufficient for most national alphabets.77
Addressing letters later became an important punched-card task, but IBM did not perceive sufficient demand for this application in the United States in the 1930s to prioritize production of this feature. The first alphanumeric tabulators could only print one line per card. This was not adequate, as three lines were needed for an address and one card of eighty columns could hold all the information needed for an address. In 1937, IBM began production of a tabulator for the French market, which could print two lines from a single card, possibly to attain a contract on a conscription and mobilization register. However, a tabulator that printed three lines from a single card only appeared on the American market in 1941.78
Extending the application of punched cards to encompass bookkeeping called for an improved multiplication capability. For invoicing, it was crucial to be able to compute the price of a commodity given the item price and the number of items. Further, attracting bank accounting applications would require the ability to compute interest rates. When using the existing IBM tabulators, multiplication was achieved through successive additions, through mental calculation, or by use of a separate key calculating machine.
As early as 1926, IBM had decided not to base punched-card multiplications on an improved tabulator but to build a separate nonprinting machine that could read figures from a punched card, perform the required arithmetic operations, and punch the outcome on the same or a successive card. In contrast, Deutsche Hollerith-Maschinen Gesellschaft (Dehomag), IBM’s German subsidiary, built a multiplying tabulator between 1931 and 1935. In the United States, James Bryce filed patents on a separate punched-card multiplier in 1928, and a prototype was built. At that time, a study identified applications for such a machine but did not find enough demand to start production. Two years later a big telephone company approached IBM. The telephone company wanted a punched-card multiplier to process invoices to their customers. This made IBM start production of the Bryce machine from 1928, which was marketed in 1931 (IBM Type 600). It could read two factors from a single punched card, multiply them, and punch the result onto a blank field on the same card. Two years later it was superseded by an improved model (IBM Type 601), which could calculate combinations of multiplications and additions.79
In 1928, during Bryce’s work on his inventions for the first punched-card multiplier, Dehomag brought attention to the Austrian inventor Gustav Tauschek’s patent application for a punched-card multiplier, filed in Germany in 1926 and granted in 1928.80 Tauschek’s German patent contained very broad claims, and the IBM patent department was troubled by its consequences; an application in the United States would have priority from the date on which the German application had been filed. In order not to reveal to Tauschek the assumed value of one of his patents, IBM bought all of his patents in 1930 and employed him in the United States for five years starting in 1931. This solution was similar to the purchase in 1922 of patents held by John Thomas Schaaff and John Royden Peirce. However, the United States Patent Office turned down Tauschek’s patent application in 1935, as it found his multiplying invention to be inoperative, and Tauschek’s contract with IBM was not extended beyond the original five years.
A historian regarded IBM’s purchase of Tauschek’s patents and his employment as merely an attempt to terminate a competing punched-card producer.81 This might have played a role, but IBM’s action was a repetition of the actions that ensured the granting of Hollerith’s crucial group control patent. It worked to prevent Tauschek from gaining control of punched-card multiplication in the United States in the event of his application’s being granted. In fact, IBM applied six out of Tauschek’s twenty-three patents in the United States, including a patent for the improved punched-card multiplier launched in 1946 that also could divide.82 As an IBM employee, Tauschek worked on design in several fields, as recorded by his patents, including multiplication and division.83 He also developed ideas beyond applications that IBM found relevant and filed applications for his own nonassigned patents.84
Reshaping Punched Cards
Reshaping punched cards was a prolonged process. The basic idea of using punched-card systems for bookkeeping tasks was conceived in 1906 by John Royden Peirce and published in an article in Scientific American in 1913.85 However, actually realizing this concept only took place in 1933. This process provides insight into the opening up of an established closure and demonstrates how a new closure is established.
In 1906, Herman Hollerith was still working to complete his closure of punched cards for processing general statistics. As his Tabulating Machine Company had a complete monopoly, he controlled this process, which included his customers and outsourced machine production. The reopening of this closure and the subsequent new closure were established through a more complex process as the users’ requirements grew extensively and challengers emerged.
Based on observations from several technologies, Thomas P. Hughes introduced a distinction between radical and conservative innovations. While radical innovations are system originating, conservative innovations are system improving. Hughes also found that independent inventors invented a disproportionate share of radical inventions, while organizations established around established closures, preferred conservative or incremental inventions.86 These findings assist the analysis of the reshaping of punched cards.
John Royden Peirce was an outsider to the producers of punched card for statistics processing and the related network of users, which confirms Hughes’ observation. In contrast, the two other challengers originated in the statistics-processing closure, which curtailed their imagination. The Census Bureau machine shop remained within the established closure and worked only on incremental changes. In spite of its name, the Powers Accounting Machine Company had the same origin, but by 1915, it was trying to attract bookkeeping tasks that would use statistics-processing equipment in which the only improvement was a number printing tabulator. It also had pin-box programming, but that was only a mechanical implementation of flexible programming, first implemented by Otto Schäffler in Vienna in 1890. Compared with the machine-building activities in the Census Bureau, the Powers company’s decision to market beyond statistics processing in the federal government provided the dynamics, which gave the company opportunity to contribute to the shaping of punched cards for bookkeeping and explore this market.
Hollerith’s negative reaction to printing capabilities confirms Hughes’ observation. Hollerith had shaped the previous closure. Further, his successor, Thomas J. Watson’s, performance between 1914 and 1921 is notable. Only first in 1921 did his company manage to produce a number printing tabulator. This observation supports Thomas P. Hughes’ observation of the preference for incremental innovations in an organization, in this case specifically an organization embedded in an established technological closure. Simultaneously, Watson improved the sales organization substantially, which increased the company’s dependence on the statistics processing closure. For the Tabulating Machine Company this dependence was based on the expertise and machinery in its production facilities and, especially, on its business strategy of leasing equipment that remained the company’s property.
The Tabulating Machine Company, and its successor IBM, was most influential in the shaping of punched cards for bookkeeping, as it was the biggest company involved in this field. It enhanced its momentum in 1922 by hiring visionary John Royden Peirce, but he brought prototype technology with different characteristics than IBM and for several years this threatened to displace the company’s basic standards, the electric reading of cards and the 45-column card layout. This shows the extent of the negotiation of the characteristics of punched cards for bookkeeping. The winning characteristics were developed for actual or imagined users, and IBM chose them because of their demand among the users. The number printing tabulator had given the Powers company a competitive advantage until 1921. It was, however, not able to exploit this because it lacked the ability to design and produce reliable machines—and because demand for their machines was poor. Further, the Powers company and Remington Rand were only able to adapt to the subsequent shaping of punched cards for bookkeeping to adapt to the new closure, as it was established by IBM.