The History of Computing, Part II: Reaching a (con)census
Reinventing the Punched Card
A pattern has emerged, as we’ve seen, when it comes to the invention of computer technology. A job is manual, labour intensive, and repetitive, somebody notices this, and promptly automates it. This is precisely waht Herman Hollerith set out to do in late latter half of the 1880s, after seeing the US 1880 census take 8 years to process.
Hollerith then effectively rediscovered the idea of a punched card to store data, after seeing similar techniques used on train tickets. Along with this, he built a machine, which he called a tabulator, which could count the amount of cards that met a certain criteria - for example, one could set the machine up to count the number of employed males over the age of 50, by having one hole assigned to gender, one hole assigned to employment status, and a group of holes assigned to age band. This was used, to great success, in the 1890 US census, dramatically shortening the time it took to process the results.
The next development Hollerith devised was an integrating tabulator - a machine that could be set up not only to count cards that met a certain criteria, but to add numbers represented in certain combinations of holes. He sold this under the Tabulating Machine Company, which soon thereafter became the Computing-Tabulating-Recording company, CTR for short, along with a series of other companies. This machine was used in the 1900 US census.
In the years that followed, several smaller developments were developed, such as printing the results onto paper rather than on a numerical readout, printing summary cards, more complicated calculations and criteria, and so on. In parallel, several other machines were developed to work with tabulators, such as card sorters, collators, and punches.
Shortly thereafter, CTR changed it’s name to International Business Machines, incorporated, initialised IBM. IBM tabulators would be used for decades following this, ending production in 1976. Computers slowly replaced tabulators from the late 1950s onwards, and following the end of the IBM 407, the tabulator era was over. During this time, the capabilities of tabulators were steadily advanced, allowing more and more advanced selection, sorting, and arithmetic calculations on the cards that were fed in, as well as faster operation and improved use of other devices such as printers.
Tabulators were among the first automated computation devices used in science. While mechanical calculators were in use for decades, or indeed centuries by some counts, previous to this, the use was driven manually - a calculation was input, and the calculator would calculate it and display the result.
The tabulating machine, however, was something altogether more revolutionary. Now science had the ability to be automated - one could encode a set of data on a set of cards, program the machine with the correct calculations, and leave the machine to run, potentially for hours, performing calculation after calculation while a researcher performed other duties.
This was achievable primarily due to the basic purpose of a tabulator. Invented to perform statistical calculations for census data, nothing was in place to limit the statistics solely to census data - a punched card could just as well represent data on a person as it could a molecule or astronomical entity, and the calculations were practically identical.
One of the first people to use tabulators in this manner was Leslie Comrie, an astronomer in New Zealand. He used a tabulator to compute lunar position using Fourier synthesis, achieving dramatically higher accuracy than was possible by hand calculation. This work, in 1928, was unheard of in the scientific community, and eventually lead to his suspension from the Nautical Almanac Office in 1936.
Comrie later on founded Scientific Computing Service, Limited, the worlds first private company offering scientific computing services. The company offered services to the war effort, calculating balistics tables for the Department of Defence.
Comrie’s work was later built upon by Wallace John Eckert, another pioneer in scientific computing. Eckert, in 1933, developed work based around the principle of chaining tabulators together, to achieve more complex equations. He used this method, with an IBM 601 donated by Thomas J. Watson, the president of IBM, to build upon the results of Comrie’s lunar position, improving the accuracy. In doing so, the laboratory Eckert worked in was renamed the Thomas J. Watson Astronomical Calculating Bureau.
Eckert then went on to work with the US Navy, automating the production of navigation tables through the use of tabulating equipment in order to reduce human demand on the process. This work would be continued by the team on the Manhatten Project, spearheaded by Nicholas Metropolis and Richard Feynman.
Later, at Columbia University, Eckert commissioned from IBM a large computing machine, known as the Selective Sequence Electronic Calculator, or SSEC, bringing scientific computing into the era of computers, rather than tabulators. This work was completed in 1947.
Meanwhile, Feynman and Metropolis worked on the automation of computing on the Manhatten Project, once again using IBM equipment. Details on the use of this computation equipment are few, due to the nature of the project itself. Both Feynman and Metropolis were influential in computing in later years as well, Feynman with work in supercomputing and quantum computing, and Metropolis working on teams to develop bespoke computers as was the default in the 1940s, as well as pioneering the Monte Carlo method of simulation, a technique still used today for scientific computation.
Also visible at EsseX-Files