Punch Cards Revisited

In the beginning, there were punch cards.

As far back as Jacquard automating the power loom in the early 19th century, the idea of storing information as a series of perforations in various media was an established idea.   Shiplofters, blacksmiths, machinists, and horologists  used perforated templates to make complex parts.  The concept was known to Babbage and Wheatstone.  Samuel F. Morse used a form of binary code in his telegraph and came very close to modern punched tape with the pen-register for his telegraph.  Gamewell used a notched cam to automate signalling in telegraphic fire alarms. Notched or perforated disks, and later punched paper tape, were used to send “canned” messages via telegraph and early wireless systems.

Even with all of this prior art in mind. Herman Hollerith credited a railroad conductor’s practice of creating a “punch photograph” on train tickets to indicate the biometric characteristics of a passenger as his inspiration for using punched cards to record census data.  His development of tabulating equipment for the 1890 census was to launch the company today known as International Business Machines (IBM) and a method of automatic data processing that ultimately led to the stored-program digital computer.

Hollerith card as shown in the Railroad Gazette in 1895, with 12 rows and 24 columns.

Early forms of punch card storage used primarily the edges of the cards for recording information.  A standard ticket-punch could easily be used to create holes along the edge of the cards and cards could be sorted and selected by various criteria by simply threading a large needle through the deck of cards and “shaking” the deck to remove selected edge-punched cards.  I recall a card-file database using this method in use as late as 1984, used to schedule college classes and ensure compliance with various prerequisites.  Royal McBee was still marketing a commercial version of these cards into the 1970’s.


Royal McBee Card Notching Punch

When Hollerith developed his punched-card system for the 1890 census, he developed not just a single item, but rather a complete system of machines for punching his cards, tabulating census results, and even sorting the cards by various criteria.  The first machine we will consider is the “Pantograph Punch” that enabled the use of holes throughout the interior of the card to represent information, rather than just along the edges.

The Hollerith “Pantograph Punch” developed for the 1890 Census.

This particular punch allowed for the accurate placement of punches in the interior areas of the card and the longer lever made repetitive strain injuries less likely (at the time, what we now know as “carpal tunnel syndrome” was common among telegraphers and similar occupations and was known as having a “glass arm.”)

At this point in the development of punched-card data processing, there was very little in the way of standardized coding.  Individual punch locations were assigned to represent a single binary value such as sex or marital status; complex values like age or race used multiple columns to indicate a range of answers based on yes or no questions.  Census enumerators would punch a card based on original documents collected by field workers.  Batches of the punched cards were then forwarded to the tabulators who used a surprisingly sophisticated electrical machine to compute the totals.

Census Tabulating Machine developed by Herman Hollerith for the 1890 Census. The card reader is on the right side of the desk. A pantograph punch sits of the left side of the desktop.

These machines could tabulate up to forty different values during each pass through the deck.  A plug-board programmer was used to assign each of the clock-like tabulating dials to any of the 288 holes in a standard 12 row by 24 column cards.  The card reader on the right consisted of a “bed of nails” array of rods that were somewhat free to move in the vertical direction.  If a card was perforated in a particular location, the rod would pass through the hole and come in contact with a charged pool of mercury in the base of the reader.  The charge would then be conducted through one of the 288 wires connecting the reader to the plug-board and thence to the selected recording dial on the face of the machine. Totals were then manually recorded on summary forms and stored for later use.

In addition to the machines designed for data collection and tabulation, the original patent included a semi-automatic sorting machine that was typically placed next to the tabulator.  When a card had been read and its contents recorded by the tabulating machine, an electrical signal from the tabulator would trigger an electromagnet in the sorter that would open the appropriate slot into which the operator would insert the card.

Hollerith sorting machine, 1890.

Later developments of this basic system would eventually produce dozens of different machines that would make large-scale automatic data processing a reality during the first half of the 20th Century.

To Be Continued…