An Electric Tabulating System
Herman Hollerith (1860–1929)
April 1889

Cold Library · Open Stacks · S-007210 · https://coldlibrary.com/s/007210
Rights: Published in April 1889; Hollerith died in 1929. Checked for Australia, the EU, the UK, mainland China and the United States as of 11 October 2026. Guidance, not legal advice.
Source: Internet Archive, The School of Mines Quarterly, vol. 10 (1888–89), https://archive.org/details/schoolminesquar10chemgoog (retrieved 2026-10-11)
Generated 2026-10-14 from https://coldlibrary.com/en/open-stacks/hollerith-an-electric-tabulating-system

----

Few, who have not come directly in contact with a census office, can form any adequate idea of the labor involved in the compilation of a census of 50,000,000 persons, as was the case in the last census, or of over 62,000,000, as will be the case in the census to be taken in June, 1890. The fact, however, that Congress at its last session in "An Act to provide for the taking of the eleventh and subsequent censuses," fixes the maximum cost of the next or eleventh census, exclusive of printing and engraving, at $6,400,000, will perhaps impress one with some idea of the magnitude of such an undertaking.

Although our population is constantly increasing, and although at each census more complicated combinations and greater detail are required in the various compilations, still, up to the present time, substantially the original method of compilation has been employed; that of making tally-marks in small squares and then adding and counting such tally-marks.

While engaged in work upon the tenth census, the writer's attention was called to the methods employed in the tabulation of population statistics and the enormous expense involved. These methods were at the time described as "barbarous," and the suggestion was made that some machine ought to be devised for the purpose of facilitating such tabulations. This led the writer to a thorough study of the details of the methods used, which were no doubt the most approved ever employed in compiling a census. After a careful consideration of the many problems involved and considerable experimenting on quite a large scale, the method which forms the subject of this paper is confidently offered as a means for facilitating this work.

* * *

A census is often spoken of as a photograph of the social and economic conditions of a people. The analogy can be made, not only with reference to the results obtained, but also to the methods of obtaining these results. Thus the enumeration of a census corresponds with the exposure of the plate in photography, while the compilation of a census corresponds with the development of the photographic plate. Unless the photographic plate is properly exposed it is impossible to obtain a good picture, so likewise, in case of a census, a good result is impossible unless the enumeration is made properly and with sufficient detail. As the first flow of the developer brings out the prominent points of our photographic picture, so in the case of a census the first tabulations will show the main features of our population. As the development is continued, a multitude of detail appears in every part, while at the same time the prominent features are strengthened, and sharpened in definition, giving finally a picture full of life and vigor. Such would be the result of a properly compiled and digested census from a thorough enumeration. If this country is to expend $3,000,000 on the exposure of the plate, ought not the picture be properly developed?

* * *

Heretofore in census and similar compilations essentially one of two methods has been followed. Either the records have been preserved in their proper relations, and the information drawn off by tallying first one grouping of facts and then the next, or the records have been written upon cards or slips, which are first sorted and counted according to one grouping of facts and then according to the next.

To form some idea of the questions involved in the first plan, let us assume that the record relating to each person at the next census be written in a line across a strip of paper, and that such lines are exactly one-half inch apart, it would then take a strip of paper over 500 miles long to contain such records. These must be gone over, again and again, until all the desired combinations have been obtained. This is practically the method followed in compiling the tenth census. On the other hand, if written cards are to be used the prospect is hardly more encouraging. One hundred comparatively thin cards will form a stack over an inch high. In the next census, therefore, if such cards are to be used it will require a stack over ten miles high. Imagine for a moment the trouble and confusion which would be caused by a few such cards becoming misplaced. This method of individual cards was employed in the census of Massachusetts for 1885. The 2,000,000 cards there used weighed about 14 tons. Were the same cards to be used in the next United States census it would require about 450 tons of such cards.

In place of these methods it is suggested that the work be done so far as possible by mechanical means. In order to accomplish this the records must be put in such shape that a machine could read them. This is most readily done by punching holes in cards or strips of paper, which perforations can then be used to control circuits through electro-magnets operating counters, or sorting mechanism, or both combined.

Record-cards of suitable size are used, the surfaces of which are divided into quarter-inch squares, each square being assigned a particular value or designation. If, for example, a record of sex is to be made, two squares, designated respectively M and F, are used, and, according as the record relates to a male or a female, the corresponding square is punched. These holes may be punched with any ordinary ticket-punch, cutting a round hole, about three-sixteenths of an inch in diameter. In similar manner other data, such as relate to conjugal condition, to illiteracy, etc., is recorded. It is often found, however, that the data must be recorded with such detail of specification that it would be impracticable to use a separate space for each specification. In such cases recourse is had to combinations of two or more holes to designate each specification. For example, if it is desired to record each single year of age, twenty spaces are used, divided into two sets of ten each, designated, respectively, from 0 to 9. One set of ten spaces is used to record the tens of years of age, while the other set is used to record the units of years of age. Thus, twelve years would be recorded by punching 1 in the first set, and 2 in the second; while 21 years would be recorded by punching 2 in the first set, and 1 in the second set. Occupations may be arranged into arbitrary groups, each such group being designated, for example, by a capital letter, and each specific occupation of that group by a small letter. Thus, Aa would designate one occupation, Ab another, etc. If desired, combinations of two or more letters of the same set may be used. Thus, AB can be used to designate one occupation, AC another, BC another, etc. With such an arrangement, the initial letter may be used to designate groups of occupation as before. In this way it is apparent that a very small card will suffice for an elaborate record. For the work of a census, a card 3″ x 5½″ would be sufficient to answer all ordinary purposes. The cards are preferably made of as thin manilla stock as will be convenient to handle.

If printed cards are used, the punching may be done with ordinary ticket-punches; more satisfactory results, however, can be obtained with punches designed especially for this work, as will be presently described.

In a census the enumerator's district forms the statistical unit of area, and a suitable combination is arranged to designate each such district. A card is punched with the corresponding combination for each person in such enumeration districts, and the cards of each district are then numbered consecutively, in a suitable numbering machine, to correspond with numbers assigned to the individual records on the enumerator's returns. This combination of holes, and this number, will serve to identify any card. Should any card become misplaced, it is readily detected among a number of cards by the fact that one or more of these holes will not correspond with the holes in the balance of the cards. By means of a suitable wire or needle a stack of a thousand or more cards can be tested in a few seconds, and any misplaced cards detected. When it is remembered that in a census millions of cards must constantly be handled, the importance of this consideration is appreciated. With ordinary written cards it would be practically impossible to detect misplaced cards, and a few such misplaced cards would cause almost endless confusion.

As the combination of holes used for designating the enumerator's district are the same for all the cards of that district, a special machine is arranged for punching these holes. This machine is provided with a number of interchangeable punches, which are placed according to the combination it is desired to punch. Five or six cards are then placed in the punch against suitable stops, and by means of a lever the corresponding holes are punched through these cards at one operation.

The individual records are now transcribed to the corresponding cards by punching according to a pre-arranged scheme as described above. For this purpose what may be known as a keyboard-punch is arranged, in which the card is held fixed in a frame, while the punch is moved over the card in any direction by means of a projecting lever provided with a suitable knob or handle. Below the knob is a keyboard provided with holes lettered and numbered according to the diagram of the card, and so arranged that when a pin projecting below the knob is over any hole, the punch is over the corresponding space of the card. If the pin is depressed into any hole of the keyboard, the punch is operated and the corresponding space of the card is punched. With such a keyboard-punch it is, of course, apparent that a perfectly blank card may be used, one corner, however, being cut off to properly locate the card in subsequent operations.

To read such a punched record card, it is only necessary to place it over a printed form, preferably of a different color, when the complete record shows directly through the perforations.

* * *

With a little practice great expertness is secured in making such transcriptions, and a record can thus be transcribed much more readily than by writing, even if considerable provision is made for facilitating the writing by the use of abbreviations.

From the punched record cards it next becomes necessary to compile the desired statistics. For this purpose the apparatus shown in Figs. 2 to 8 is used. The press or circuit-closing device, shown in Figs. 2, 3, and 4, consists of a hard rubber bed plate, as shown in section in Fig. 4, provided with suitable stops or gauges against which the record-cards can be placed. This hard rubber plate is provided with a number of holes or cups corresponding in number and relative position with the centres of the spaces of the card. An iron wire nail is securely driven through a hole in the bottom of each cup, and a wire, connecting at its other end with a binding-post on the back of the press frame, is securely held under the head of each nail. Each cup is partly filled with mercury, which, through the nail and wire, is thus in electrical connection with the corresponding binding-post. Above the hard rubber plate is a reciprocating box provided with a number of projecting spring-actuated points, corresponding in number and arrangement with the centres of the mercury cups. The construction and arrangement of these pins is shown in Fig. 4. If a card is placed on the rubber plate against the stops it is of course apparent that, when the box is brought down by the handle, the pins will all be pressed back, excepting such as correspond with the punched spaces of the card which project into the mercury, and are thus in electrical connection with the corresponding binding-posts on back of the press frame.

A number of mechanical counters are arranged in a suitable frame, as shown in Fig. 5. The face of each counter is three inches square, and is provided with a dial divided into 100 parts and two hands, one counting units the other hundreds. The counter consists essentially of an electro-magnet, the armature of which is so arranged that each time it is attracted by closing the circuit it registers one. A suitable carrying device is arranged so that at each complete revolution of the unit hand the hundred hand registers one, each counter thus registering or counting to one hundred hundred, or 10,000, which will be found sufficient for all ordinary statistical purposes. The counters are so arranged that they can readily be reset at 0, and all are removable and interchangeable, the mere placing of the counter in position in the frame making the necessary electrical connections through the magnet.

For the purpose of sorting the cards according to any group of statistical items or combinations of two or more of such items, the sorting-box, shown in Fig. 6, is used. This consists of a box suitably divided into compartments, each one of which is closed by a lid. Each lid, L, as shown in Fig. 7, is held closed against the tension of the spring, S, by the catch, *a*, in the armature, A. If a circuit is closed through the magnet, E, the armature, A, is attracted, thus releasing the lid, L, which is opened by the springs, and remains open until again closed by hand.

As the cards are punched they are arranged by enumerators' districts, which form our unit of area. The first compilation that would be desired would be to obtain the statistics for each enumeration district according to some few condensed groupings of facts. Thus it might be desired to know the number of males and of females, of native born and of foreign born, of whites and of colored, of single, married, and widowed, the number at each of centre groups of ages, etc., in each enumeration district. In order to obtain such statistics the corresponding binding-posts on the back of the press frame are connected, by means of suitable pieces of covered wire, with the binding-posts of the counters upon which it is desired to register the corresponding facts. A proper battery being arranged in circuit, it is apparent that if a card is placed on the hard rubber bed plate, and the box of the press brought down upon the card, the pins corresponding with the punched spaces will close the circuit through the magnets of the corresponding counters which thus register one each. If the counters are first set at 0, and the cards of the given enumeration district then passed through the press one by one, the number of males and of females, of whites and of colored, etc., will be indicated on the corresponding counters.

If it is desired to count on the counters directly, combinations of two or more items, small relays are used to control secondary circuits through the counters. If, for example, it is desired to know the number of native white males, of native white females, of foreign white males, of foreign white females, of colored males, and of colored females; these being combinations of sex, race, and nativity, ordinary relays are arranged as shown in the diagram, Fig. 8, the magnets of which are connected with the press as indicated. If a card punched for native white, and male is placed in the press, the corresponding relays are actuated, which close a secondary circuit through the counter magnet, native white male, thus registering one on the corresponding counter.

By a suitable arrangement of relays any possible combination of the data recorded on the cards may be counted. When it is desired to count more complicated combinations, however, special relays with multiple contact points are employed.

If it is desired to assort or distribute the cards according to any desired item or combination of items recorded on the card, it is only necessary to connect the magnets of the sorting-box in exactly the same manner as has been described for the counters. When a card is then placed in the press, one of the lids of the sorting-box, according to the data recorded on the card, will open. The card is deposited in the open compartment of the sorting-box and the lid closed with the right hand, while at the same time the next card is placed in position in the press with the left hand.

It is, of course, apparent that any number of items or combinations of items can be counted. The number of such items or combinations, which can be counted at any one time, being limited only by the number of counters, while at the same time the cards are sorted according to any desired set of statistical facts.

In a census the cards as they come from the punching machines would, of course, be arranged according to enumeration districts. Each district could then be run through the press, and such facts as it is desired to know in relation to this unit of area could be counted on the counters, while the cards are at the same time assorted according to some other set of facts, arranging them in convenient form for further tabulations. In this manner, by the arrangement of a judicious "scheme," it will be found that a most elaborate compilation may be effected with but a few handlings of the cards.

Two of the most important elements, in almost all statistical compilations, are "time" and "accuracy." The rapidity with which results could be obtained with the present method, in a census, for example, would be dependent upon: 1st, the rate at which a clerk could punch the record-cards, and, 2d, the number of clerks employed upon this part of the work. The first can readily be determined by experiment, when the second becomes merely a simple arithmetical computation. The work of counting or tabulating on the machines can be so arranged that, within a few hours after the last card is punched, the first set of tables, including condensed grouping of all the leading statistical facts, would be complete. The rapidity with which subsequent tables could be published would depend merely upon the number of machines employed.

In regard to accuracy, it is apparent that the processes of counting and sorting, being purely mechanical, can be arranged, with such checks, that an error is practically impossible. The one possible source of error is in the punching of the cards. If proper precautions are here taken, a census practically free from errors of compilation could be obtained. Even in this respect the present method would have manifest advantages. A card wrongly punched could involve an error of only a single unit, while by all previous methods single errors involving an error in the result of tens, of hundreds, of thousands, or even more, are possible.

It is firmly believed that in regard to cost, time and accuracy, this method would possess very great advantages in doing the work that has heretofore been done, but this is believed to be insignificant in comparison with the fact that a thorough compilation would be possible, within reasonable limits of cost, while such compilation is practically impossible, by the ordinary methods, on account of the enormous expense involved.