This article was initially written as part of the IEEE STARS program.
Citation
Punched card tabulating equipment, invented and developed by Herman Hollerith to process data from the United States Census of 1890, was the first mechanized means for compiling and analyzing statistical information. Through continual improvements, first by Hollerith and then by many others, punched card equipment created and expanded the worldwide information processing industry and continued to play a significant role in that industry for more than two decades after the first commercial electronic computers were installed in the early 1950s.
Introduction
Soon after graduating in 1879, at age nineteen from the Columbia School of Mines, Herman Hollerith was hired to work for the U.S. Census of 1880. He and several other technically trained persons had been hired to help the U.S. Census Bureau respond to the desire of the U.S. Congress to acquire information on the country’s economy to help in making policy decisions. This task had been added to the relatively simple mandate of the U.S. Constitution that Congress conduct an enumeration of the population every ten years in order to help ensure fair representation and taxation among the states of the United States.
Seeking Solutions
In his role as a special agent of the U. S. Census of 1880, Hollerith was encouraged by the superintendent of the Census, Francis A. Walker, and by a senior member of the technical team, Dr. John S. Billings, to spend some of his time studying the operation of the census and to seek ways to mechanize it. This he did, and Dr. Billings, who was a physician with interest in statistical information about public health issues, generously provided important suggestions and encouragement.
During the years before the next U.S. census, Hollerith considered many alternatives, including the use of paper tape and cards with holes punched in them to store information. He began conducting experiments during his spare time while teaching for one year at the Massachusetts Institute of Technology (MIT) and then while an assistant examiner in the U.S. Patent Office in Washington, D.C. In 1884 he resigned from the Patent Office and started his own business as an “Expert and Solicitor of Patents.”
Later that year he filed his first punched card patent application. The device described in the patent made use of tape rather than cards, but the patent claims were broad enough to cover either case. Later that year, he divided the patent into two patents. Hollerith could not claim the invention of punching holes in cards or tape to store information or the electrical reading of these holes because these concepts had already been used in the telegraph system. Furthermore, half a century earlier, the British mathematician Charles Babbage had proposed (but never completed an operational model of) an “analytical engine” driven by data and instructions stored in punched cards; and even a quarter of a century before that, the Frenchman Joseph M. Jacquard, had used a sequence of cards with holes punched in them to cause an automated loom to reproduce very intricate patterns. What Hollerith could and did claim was a novel use of concepts well known to others to create new and useful capabilities “not obvious to those skilled in the art.” (It is interesting to note that Hollerith’s decision to use punched cards in his tabulating equipment was influenced by several contemporary sources, including telegraph transmitters, library index cards, and probably the Jacquard looms, but Hollerith had no knowledge of the work of Charles Babbage.)
With the help of Dr. Billings in 1886, Hollerith arranged for a practical test of his recently designed equipment in recording and tabulating vital statistics for the Department of Health in Baltimore, Maryland. Information about each deceased person was recorded by holes punched in selected locations in cards 3 1/4 by 8 5/8 inches (8.26 by 21.9 centimeters) on a side.
Information was read from each card by placing it on a hard rubber mat that contained an array of tiny cups of mercury, so spaced that there was a cup directly under each of the possible hole locations. A hinged “pin box” that held an identically spaced array of spring-loaded pins was pulled down upon the card on the rubber surface. Where holes had been punched in the card, pins passed through them and made electrical contact with the mercury. Each cup of mercury was electrically connected to a dial with two hands that rotated in a manner analogous to the hands on a clock and were able to record numbers from 0 to 9999. After many cards had been processed, the operator manually wrote down the number incrementally recorded by each dial.
Hollerith had also devised an apparatus for sorting cards into categories; it consisted of a wooden box with many compartments. Each time a card was tabulated, the electrical signals generated when the pins passed through holes in the card were used to open the lid of one of the compartments. The operator manually removed the card from the rubber mat and placed it in the open compartment and closed the lid. Using this method, it was possible to sort cards based on many attributes and also on combinations of attributes.
Success
Following the successful results of the test in Baltimore, he applied for a third patent, which described the Baltimore system and was quite specific in its claims. All three patents were issued simultaneously in January 1889.
During the Baltimore test, a major failing of the equipment was revealed. The small hand-held punch of the type typically used by train conductors was not suitable for punching thousands of holes per day. Hollerith’s hand and arm became nearly paralyzed after spending a day continually punching holes. To solve this problem, he devised and patented a new punch in which large motions of a manually moved stylus were mimicked by smaller motions of a punch that cut round holes in the card.
The success of the Baltimore test was broadly reported and led to opportunities for Hollerith to demonstrate his equipment in many places inside and outside the United States. In 1889 Hollerith’s equipment was chosen to tabulate the U.S. census of 1890. Hollerith contracted and paid for the construction of approximately one hundred of his systems. He rented these systems to the government to process the census, and he accepted responsibility for maintaining each system in good operating condition. Renting equipment remained the basic practice of the industry until the 1950s.
Because of the limited opportunities for processing census returns, Hollerith turned to commercial applications and focused on railroads. To obtain the business of the New York Central (the second largest railroad system in the U.S.) he made significant modifications to his equipment. Unlike cards used for census work, in which punched-hole locations represented personal characteristics such as age or marital status, the new cards were entirely numeric; a single digit could be recorded in each column by a hole in one of ten positions.
No longer did simple counting suffice. Hollerith made substantial modifications to his tabulating equipment to perform addition—that is, to add a number recorded in a field on one card to those in the same field on preceding cards. The dials were replaced with accumulators having sufficient digit positions to handle sums from the assigned fields. He also replaced the pinbox card reader with a mechanized card feeder and the manually operated sorter with a fully mechanized sorter. In September 1896, Hollerith signed a contract agreeing to supply the New York Central with sufficient punched card equipment to audit up to four million waybills during the next year.
In 1897, he incorporated his business as the Tabulating Machine Company, and he continued improving his general statistics processing equipment over the next many years. Several of his improvements became industry standards: the 45-column card and dynamic reading of the cards while in motion introduced in 1907 and automatic group control, a facility to control generation of subtotals and grand totals which significantly eased the processing of punched cards. It 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. Hollerith had filed a petition for this in 1914, but his patent on “Automatic Group Control for Tabulating Machines” was not granted until 1931 because of conflicts with other patent applications.
Competition
Despite the success of his ever improving equipment, Hollerith began to have serious problems after a permanent U.S. Bureau of the Census was created in 1902. The newly appointed director of the Bureau, Simon Newton Dexter North, determined that Hollerith was charging the government excessively high fees. To break the hold Hollerith had on the Census Bureau, the director obtained funds from the Congress for the Bureau to develop its own tabulating equipment. Hollerith fought back vigorously, even taking his case to the president of the United States. But ultimately, the census of 1910 was taken using equipment designed by employees of the U. S. Census Bureau combined with several rebuilt sorters originally built by Hollerith and sold to the Census. Herman Hollerith sued the government for infringement, but the court dismissed the suit, in 1912. It nullified the sorter patent because it had been disclosed before the patent application was filed.
In 1911, Hollerith had received an attractive offer to sell his company that he accepted, in part, because of his frustration with the Census Bureau and the problems of operating a large company. The sale resulted in the company being merged with three other companies, creating the Computing-Tabulating-Recording Company (CTR). In 1914 Thomas J. Watson, Sr., was hired to manage CTR, and in 1924 the company’s name was changed to International Business Machines Corporation (IBM).
Also in 1911, James Powers (a key person in developing tabulating equipment for the Bureau of the Census) left the Bureau to establish the Powers Accounting Machine Company to compete with Hollerith. In his new company, Powers was permitted to use all inventions he had created at the Bureau. The Powers Company and its successor companies were the primary competitors of IBM from 1911 until long after computers began to replace punched card equipment in the 1950s. The first successor company was Remington Rand, Inc., which acquired the Powers Accounting Machine Company in 1927.
Printing of entered numbers and results was the most important advantage of Powers’ tabulators. This relieved an operator from recording manually the numbers recorded in dials or accumulators. However, this advantage was not sufficient to overcome the difficulties it was having in producing reliable machines.
A basic difference between equipment offered by the two companies was that Hollerith detected holes in the cards electrically, whereas Powers detected them mechanically. Hollerith transmitted electrical signals by wires throughout the system to activate electromechanical devices, such as those that opened the doors to the original sorting boxes. By contrast, Powers transmitted information mechanically, i. e. by use of levers, chains and rods. He used mechanical techniques throughout because he was more familiar with such technologies, it would help circumvent Hollerith’s patents and, like many engineers and inventors of that time, he believed mechanical couplings were faster and more reliable than electrical ones.
To minimize the difficulty customers would have in changing mechanical connections, Powers localized all linkages between machine parts in an interchangeable mechanical “connection box.” This made it possible for customers to change from one application to another simply by replacing one connection box with another.
Fortunately for Hollerith, he had replaced his method for changing applications in 1909 from the tedious rewiring by hand of many electrical connections throughout the system to a method that centralized electrical terminals in a plugboard where they were more easily interconnected by plugwires. (Interestingly, the person responsible for this innovation was Otto Schäffler, of Vienna, who had been licensed by Hollerith to build and service tabulating equipment for processing the Austrian census of 1890.) Now to meet the challenge presented by the Powers connection box, the plugboard used by Hollerith was moved from the back of the machine to a more convenient location in the front; and this improvement was soon followed by introduction of the so-called “automatic plugboard” that could be quickly removed from, or inserted into, the machine with its plugwires in place.
In time customers would acquire numerous connection boxes or pre-wired plug boards to be inserted into their system to carry out desired tasks. An important difference was that each specially designed mechanical connection box had to be obtained from the Powers Company, whereas plugboards could be wired and modified by the customer. IBM capitalized on this advantage by encouraging customers to create novel plugboard interconnections and to share them with each other. Soon ingenious plugboard wiring arrangements created by customers and IBM employees were distributed to customers in an IBM news bulletin called “Pointers.”
Entering Foreign Markets
Herman Hollerith visited many European countries to promote his business beginning in the late 1880s. Important early successes were the adoption of his equipment for census purposes by Austria, Canada, France, Norway, and Russia. Significant penetration of the European market did not occur, however, until substantially improved systems became available during the first decade of the twentieth century.
In 1904, the Tabulator Limited was incorporated as an agency distributing Hollerith equipment in the British Empire. The agency was renamed British Tabulating Machine Company (BTM) in 1907. (It was succeeded by an IBM subsidiary in 1951.) In 1910, the Deutsche Hollerith Maschinen Gesellschaft (Dehomag) was founded as an agency to serve the German market. In 1922, Dehomag was acquired by CTR. Starting in 1920, CTR (then IBM) got agencies and subsequently established subsidiaries in most other European countries and later in Latin America and other continents. Powers established agencies in Germany in 1913 and Britain in 1915 and subsequently in other countries.
The British and German affiliates operated quite independently of CTR (and later IBM) and maintained their own development and production facilities for punched card equipment. Similarly, the British Powers company had separate development and production facilities and produced the first alphabet printing (not alphanumeric) tabulator in 1921. During the years 1933 through 1946, Dehomag designed and produced tabulators with advanced calculating facilities that were later produced by IBM after World War II.
Compagnie des machines Bull was the only significant producer of punched card equipment that was independent of IBM and of Remington Rand, Inc. (which had acquired the Powers Accounting Machine company in 1927). Machines Bull commenced operations in France in 1931, based upon designs by Fredrik Rosing Bull and Knut Andreas Knutsen in Norway in the 1920s. Like IBM, Bull read holes in cards electrically and applied electromechanical technology throughout the system. It produced innovative equipment that it sold across Europe. Bull’s offerings included reputedly the best punched card sorter ever produced. It invented wheel printing and produced the first wheel-printing tabulator in 1931; and it separately developed an alphanumerical system and produced the first alphanumeric printing tabulator in 1932.
The IBM Card
In 1927, IBM President, Thomas J. Watson, assigned two of his top inventors, Clair D. Lake and J. Royden Pierce, the task of creating a punched card with greater capacity than the forty-five-column card that was then standard for IBM and Remington Rand, Inc. Greater capacity was needed for many accounting applications.
Working in secrecy from each other, the Lake and Peirce groups took different approaches. Peirce proposed continuing to use the standard card, while introducing a multi-hole, combinational code for each character. This provided the ability to store twice as many characters per card, half of which could be alphanumeric. Lake, in contrast chose to use smaller holes, with rectangular rather than round shapes, so that each card could accommodate more columns. Rectangular holes were more easily read as cards passed over a metallic roller under metal brushes that made electrical contact when the hole passed between the brush and the roller. This design required the development of new punches and readers, but (unlike Peirce’s approach) the coding of the holes and the underlying numerical operations were not affected.
The decision between the two approaches would be made by Watson, who frequently sought the views of James W. Bryce, IBM’s top inventor and chief engineer. Bryce’s accomplishments were still little known outside IBM, but that would change in 1936 when he was honored as one of the ten “greatest living inventors” by the U.S. Patent Office. When he died in 1949, Bryce had over five hundred U.S. and foreign patents.
Bryce recommended Lake’s design because it could be implemented more rapidly and would retain the long-established machine timing methods for reading numbers on cards. He recommended against Peirce’s design because there appeared to be little market demand for storing alphabetic information and because of possible patent interference with Powers relating to the multi-hole card punch.
Watson selected Lake’s design for all the reasons cited by Bryce, but perhaps even more important to him was the distinctive appearance the rectangular holes would gave the card, and the fact that a patent being applied for would prevent others from copying it. Even without patent constraints, competitors using mechanical sensing of holes would find it difficult to make the change. Watson was eager to begin promoting it as the “IBM Card.” He did not want it to become just another punched card.
IBM introduced the “IBM Card” in 1928. The new card had 80 columns, almost twice as many as in the then-standard 45 column card. As with earlier cards, it had 10 rows for coding numbers from 0 to 9. But in 1930, two more rows were added to the top of the card to facilitate alphabetic coding. When a position in one of these two rows or in the row for zeros was punched, together with one in the numeric field, the combination was interpreted as alphabetic rather than numeric.
Remington Rand countered the IBM Card in 1930 with its own 90-column card, using one-to-two holes to code a number and, in 1938, two-to-three holes to code a letter. This provided for the storage of more data and a wider variety of characters than the IBM Card, but it was never as popular for many reasons – one being that the more complex coding scheme made it difficult for users to know what was stored on the card simply by looking at it, another being that the IBM card by then had become widely used throughout the industry.
An important coup for IBM was acquiring the huge the data-processing jobs of the old-age pension system introduced by the U.S. Social Security Act of 1935. Two years later, the administration of Social Security established a system recording the frequent payment of 33 million people based upon IBM punched card equipment. Subsequently, all recipients of Social Security payments received checks written on IBM Cards with numerous rectangular perforations for automatic data processing. For many recipients, it was their happy introduction to the punched card era, the workings of which were generally confined inside large organizations, out of sight of the general public. For IBM, it was the end of the economic depression that had begun in 1929.
During World War II, punched card equipment was used by the United States and its allies to keep personnel records and payroll information for large companies, government organizations, and the armed services; to facilitate the design and manufacture of munitions, airplanes, warships, and other equipment; to handle the logistics of supplying armed forces operating throughout the world; to decipher enemy codes; and to predict the weather. The Germans used punched card equipment to monitor and manage munitions and armament production so as to achieve increasing war production until the summer of 1944 despite relentless bombing of its factories and transportation systems. Thus it can be stated that, on both sides in the war, punched card equipment successfully performed many activities now associated with electronic computers.
Electronics
James Bryce had begun inventive work on electronic vacuum tube devices at IBM in 1936, and in 1940 an engineer in his department (A. Halsey Dickinson) made what is believed to be the first patent filing for an electronic computer circuit. From time to time during the war, IBM engineers considered using vacuum tube circuits in specialized war-related equipment, but reverted to using time-tested electromechanical devices to meet delivery-time and reliability requirements.
Nevertheless, this small research effort on vacuum tube devices was continued throughout the war; and in 1946 IBM introduced the industry’s first commercial product to employ electronic digital computation, the IBM 603 Electronic Multiplier. Production was limited to one hundred units, and in 1948 customer delivery began of the more sophisticated IBM 604 Electronic Calculating Punch. In its most basic version, the IBM 604 could perform addition, subtraction, multiplication, and division with electronic speed, and execute up to twenty plugboard-controlled program steps between reading data from a card and punching out the result.
When the IBM 604 was combined with other equipment to create the IBM Card-Programmed Electronic Calculator (CPC), announced in 1949, its performance was remarkably competitive with stored-programmed computers that had not yet reached the market. The CPC’s program sequences consisted of instructions in a deck of punched cards that could be changed or modified with relative ease, but this still did not provide as much versatility as stored-program computers that loaded instruction sequences into the computer’s main memory for processing data.
The replacement of punched card equipment by stored-program electronic computers in the commercial market began when Remington Rand’s first UNIVAC was accepted by the Bureau of Census in March 1951 and three more UNIVACs were delivered to other organizations within eighteen months. IBM responded with its 700 series of computers. The first installation outside IBM was in February 1953, and fifty more of the 700 series were installed or on order by mid-1954. All of these were high-cost, high-performance computer systems. It would be several years before electronic computers could compete with punched card equipment in the market place for low-cost systems.
In 1962 IBM’s revenues from electronic stored-program computer systems exceeded, for the first time, those from punched card systems. Because IBM’s worldwide revenue from both of these systems was more than two times larger than that of all other companies combined, 1962 is a good estimate for the crossover date for the entire industry. Despite this rapid rate of installation of computer systems, punched card equipment continued to be used for many more years. It was needed to satisfy the requirements of small organizations and for special purposes, and it also served as information input-output equipment for many computer systems.
Acknowledgements
The authors thank members of the STARS Editorial Board and others for review and constructive criticism of this article, with special thanks to Editorial Board members Janet Abbate, Ross Bassett, Brian Bowers, Paul Ceruzzi, James W. Cortada, Bernard Finn, and Michael Williams for helpful comments and suggestions. Also, the authors gratefully acknowledge permission to use text, figures, and figure captions previously published in Building IBM: Shaping an Industry and Its Technology, by Emerson W. Pugh (MIT Press, 1995).
Timeline
- 1880, Hollerith works for U.S. Census and seeks ways to mechanize the census process
- 1884, Hollerith files first punched card processing patent
- 1886, First practical test of Hollerith’s equipment at Baltimore Dept. of Health
- 1890, Tabulation of the U.S. Census of 1890 is begun using Hollerith’s equipment
- 1896, Hollerith signs contract to supply equipment for NY Central Railroad
- 1904, Tabulator Limited is established to market Hollerith products in Britain
- 1910, Deutsche Hollerith Maschinen Gesellschaft is established for German market
- 1910, Punched card equipment developed by Census Bureau processes U.S. Census
- 1911, First number-printing tabulator is built by the U.S. Census Bureau
- 1911, John Powers leaves Census Bureau to establish punched card equipment company
- 1911, Hollerith’s company is merged with three other companies to form CTR
- 1914, Thomas J. Watson, Sr., is hired to manage CTR
- 1924, CTR is renamed IBM
- 1927, Remington Rand, Inc., acquires the Powers Accounting Machine Company
- 1928, The “IBM Card” is introduced with larger storage capacity and rectangular holes
- 1931, Compagnie des Machines Bull begins competitive punched card business
- 1932, Bull produces first alphanumeric tabulator
- 1936, IBM punched card equipment is adopted for new U.S. Social Security System
- 1946, Vacuum-tube electronics are first used in a punched card product
- 1951, Electronic computers begin competing with punched card equipment
Bibliography
References of Historical Significance
Herman Hollerith. 1884. Art of Compiling Statistics. U.S. Patent 395,782, filed 23 September 1884 and granted 8 January 1889.
Herman Hollerith. 1914. Automatic Control for Tabulating Machines. U.S. Patent 1,830,699, filed 11 March 1914 and granted 3 November 1931. Automatic group control described in this patent was a facility to control the generation of sub and grand totals. It is the earliest example of conditional programming of a punched card machine, a facility subsequently essential for stored-program computers.
References for Further Reading
Lars Heide. 2009. Punched-Card Systems and the Early Information Explosion. Baltimore: Johns Hopkins University Press. Provides the most complete coverage available of the development and usage of punched card equipment.
Emerson W. Pugh. 1995. Building IBM: Shaping an Industry and Its Technology. Cambridge, Massachusetts: MIT Press. Early chapters of this book describe Hollerith’s inventive activities and his role in the creation of IBM.
Geoffrey Austrian. 1982. Herman Hollerith: Forgotten Giant of Information Processing. New York: Columbia University Press. This is a good history of the life and inventions of Herman Hollerith.
About the Authors
Emerson W. Pugh is the author or coauthor of four books on the history of IBM and the computer industry. One of these, Building IBM, Shaping an Industry and Its Technology (The MIT Press, 1995) provides a primary basis for this article. Dr. Pugh has a PhD in physics and worked for IBM for 36 years in many capacities, including research scientist, product development manager, and corporate executive. He served as president of the IEEE in 1989, chair of the IEEE History Committee in 1996 through 1998, and president of the IEEE Foundation in 2000 through 2004.
Lars Heide is the author of three books and a large number of international articles on history of technology, including Punched-Card Systems and the Early Information Explosion, 1880-1945, (Johns Hopkins, 2009). Dr. Heide has a PhD in History of Technology and a Dr.merc. in Business History. He has been an associate professor at Copenhagen Business School since 2000. Before joining the business school he held managerial and research positions at universities and a museum. Currently he is member of the editorial board of the IEEE Annals of the History of Computing and a member of the advisory board of an international research network.