The Physical Infrastructure of Human Computing

If you cannot calculate, you cannot speculate on future pleasure and your life will not be that of a human, but that of an oyster or a jellyfish.
– Plato in Philebus

There is considerable evidence of the mainframe era, the rise of the PC, and the early internet, but the more mundane details of computing history are often lost to software updates and hardware trends. The era before digital computers, when “computer” meant a human performing calculation, has left considerable records of the “how” of this work: where it was done, for what purpose, and by whom. The work of scholars studying human computing, notably Grier and Campbell-Kelly, focuses on the social, political, and scientific aspects. We know the details of William Samuel Stratford’s Nautical Almanac in England and the long narrative of the Mathematical Tables Project in New York City, but little has been written on the physical infrastructures of human computing such as office layouts and furniture, or the ephemera like the worksheets used to complete calculations. Because most of the physical infrastructure exists only in snippets of written material or in the background of photographs, I will only sketch some of the physical objects used in human computing, and mostly focus on 20^th-century projects.

This account omits discussion of mechanical, electronic, and digital computing aides for human computers. Those devices are well researched, and many artifacts and detailed descriptions exist. Instead, this essay focuses on the lost fragments of physical infrastructure, namely worksheets, offices and furniture, and posters and other ephemera.

WORKSHEETS

A sample worksheet, reprinted in “Weather Prediction by Numerical Process”.

Computing has always been about mathematics. The first use of the word “computer” is found in an 1613 book by Richard Braithwait titled Yong Mans Gleanings: “I haue read the truest computer of Times, and the best Arithmetician that euer breathed, and he reduceth thy dayes into a short number.”1 While a single scientist might toil over a calculation for weeks or even months, a group of distributed computers, often with little formal mathematical training, could work on small parts of the problem separately and arrive at an answer much more quickly. This is essentially what a supercomputer does, but with distributed human labor instead of a room of servers.

In order to facilitate calculation, the problem needed first to be broken into discreet steps. Often this meant reducing more complex mathematical operations into simple addition or subtraction, accomplished with the use of logarithms to replace multiplication and division. Also termed “methods of difference,” this process was used in the 1760s in the creation of Tables du Cadastre, the first factory-like table-making operation.2 In small organizations, the task of preparing the work worksheets fell on the scientists, though later, larger organizations would employ “planners” to create them.3 In his book Weather Prediction by Numerical Process, Lewis Fry Richardson describes creating worksheets (an example can be seen above):

It took me the best part of six weeks to draw up the computing forms and to work out the new distribution in two vertical columns for the first time. My office was a heap of hay in a cold rest billet. With practice the work of an average computer might go perhaps ten times faster.4

The resulting worksheets, either hand-drafted or printed, were distributed to the computers. Worksheets were created for specific tasks, such as the “tally” and “consolidation” sheets used for the 1880 US census (a decade before Hollerith’s famous mechanical transformation of the census process in 1890).

In the same way we discard scratch paper, no computing worksheets exist today.5 After the answer to a particular worksheet was determined and verified6 and the project completed, the worksheets were thrown away. We do have a few written accounts of the worksheets and how they were filled out, the best of which comes from David A. Gibbs’ 1915 book A Course In Interpolation And Numerical Integration, For The Mathematical Library.7 In it, Gibbs suggests best practices for human computers, including:

• Worksheets should be on 26×16” paper that is “divided by a faint ruling into ¼” squares, each of which is capable of holding two digits”
• Numbers should be written in pairs – “10, 24” instead of “1, 0, 2, 4” as it is “found conducive toward accuracy and speed”
• Entries should be made in ink, not pencil, for a more lasting record and to prevent eyestrain

Later, at the Mathematical Tables Project, the last major human-computing effort, which began in the 1930s and continued until mid-1950s, worksheets “generally had 100 lines and were on graph paper.”8 Black pencil was used to indicate positive numbers, red for negative. It is unknown exactly what became of the worksheets when the calculations were finished – they were probably stored until the project was completed and then discarded.

In 1942, the group did work on the LORAN Project, which aided in navigation of ships using synchronized radio towers. Because the project was top secret, there were extra protocols for dealing with worksheets. This stern warning accompanied the calculations:

You will not show this paper to any member of your group.
You will make no reference to this paper in any of your own work.
No copies of this paper will be made.
The paper will be returned as soon as it has served its purpose.9

These worksheets were stripped of “any reference to the physical setting, including units of all quantities involved in the work.”10 After calculations were complete, “computers packed their equipment, burned old computing sheets” to comply with security protocol.11 A later “hand-computing group” at Los Alamos, known as T-5, also did top secret computing work using human computers. Instead of worksheets, index cards were passed around as different calculations were made.12

OFFICES AND FURNITURE

Computers working at astronomer Edward Charles Pickering’s laboratory at Harvard University.

In its earliest days, especially in the United States, computing was a cottage industry. Worksheets would be dropped off at the homes of workers, completed, and picked up at a later date. Spaces created specifically for groups of computers were rare. Astronomical research offices, where calculations for complex equations like orbits could take a single researcher weeks to complete, might have full-time computers on staff. Since the focus of this work was on the science of astronomy, not computing, we have little record of the physical spaces computers occupied.

The Greenwich Observatory’s Octagon Room was literally that: an octagonal room. Other computing spaces were equally ad hoc and unique, based on what space was available – a trend that would continue into the early development of electronic computers. In a particularly poetic turn, H.T. Davis ran a computing program at Indiana University in a former artist’s studio. Keeping in the artistic spirit, Davis would computer “various things as they occurred to him,” often choosing projects with “epic irrationality.”13

But coinciding with the Industrial Revolution and changes in labor management, starting in Europe and later in the US, computing became more centralized. Operations like Banker’s Clearing House in London, described in detail by pre-electronic computer pioneer Charles Babbage, housed hundreds of clerks computing financial transactions in a large hall. In a nod to Adam Smith’s Wealth of Nations, one account of the Tables du Cadastre project describes the “manufacture [of] logarithms as one manufactures pins.”14

A group of clerks in the London Railway Clearing House, sorting tickets.15

Similarly, other “clearing houses” became centralized hubs of computing. The Railway Clearing House, described by Campbell-Kelly in his essay on Victorian-era data processing, had 1,325 clerks in 1874, completing 4.9-million transactions a year. By 1921 they employed 3,423 clerks.16 The clerks worked in pairs and could record 36,000 entries per month, or about one per minute. Working in large rooms with long tables, the clerks would tabulate the receipts on worksheets; once a month these worksheets were “sewn together under cover of brown paper.”17

Computers working in the “Nautical Almanac” office – note the hodgepodge assortment of desks, each customized to the worker’s preferences.

With centralized computing offices came more consideration for how the computers should work, including the furniture. Various accounts of 18^th and 19^th-century office layouts exist, but few specifically for computers. We know, for example, that United States Signal Corps computers, whose data was brought in by telegraph three times a day, worked at standing desks (harkening today’s startup-fueled fad for alternative office furniture) but little more.18

Gibbs’ A Course in Interpolation and Numerical Integration offers perhaps the most prescriptive description of computing tables. Desks “used in the mathematical laboratory of the University of Edinburgh are 3’ 0” wide, 1’ 9” from front to back, and 2’ 6-1/2” high. They contain a locker, in which computing paper can be kept without being folded, and a cupboard for books, and are fitted with a strong adjustable book-rest. Thus the computer can command a large space and utilize it for books, papers, drawing-board, arithmometer, or instruments.” Each desk was also supplied with useful books such as multiplication tables, as well as a slide rule.19 It is likely that such specific infrastructure was rare: photographic evidence suggests it is much more likely that office furniture was made up of whatever was common, available, and cheap, the same approach often taken to finding rooms for the computers to work.

Workers at the WPA-funded Mathematical Tables Project performing calculations. Because of the scale of the project, employing over 450 computers at its peak, furniture was standardized and mirrored corporate offices of the day.

The Mathematical Tables Project, a WPA-funded initiative based in New York City, housed the largest number of human computers in the United States. At it’s formation in 1938, the offices employed 450 out-of-work Americans, many with no mathematical skill. The computing office was headed by Gertrude Blanch, a mathematician who brought a rigor to the project’s office layout. Computers worked on the “computing floor,” working in groups. From photographs, we can see the various layouts of the Mathematical Tables Project as it grew in size through the early 1940s and declined to 25 members in 1948, when the offices moved to Washington, DC. Wood and metal desks fill massive downtown New York office buildings. Workers sit in rows like a classroom – a highly formalized arrangement very different than collaborative, open-office environments of today’s tech companies when they have different methods for efficiency and options for clocking employees.

The sound of the Mathematical Tables Project offices must have changed considerably over the years as well. What would initially have been the sounds pencils scratching on paper gave way to the mechanical clanking of adding machines and other mechanical and electrical devices.

1967 marks the end of human computing, and with it spaces for hand-calculation. Blanch retires, the National Institute of Standards and Technology closes its computing laboratory (the last remaining human computing office in the country), and the American Nautical Almanac moves to a punched card system.20

POSTERS AND EPHEMERA

Finally, alongside any large-scale project there exist ephemera that are not part of the work itself, but support it. This kind of material is often the most likely to be discarded. In the case of human computing, we have two interesting descriptions of the detritus of this work.

At the Mathematical Tables Project, most of Gertrude Blanch’s computers had little mathematical training. Based on their skills, they were divided into four groups: group one performed addition, group two subtraction, group three multiplied numbers by a single digit, and group four, which was described as the “elite of the computing floor” and consisted of only four members, performed long division.21 In order to assist the computers, Blanch prepared posters reminding the workers how to carry out their tasks. While no copies of the posters exist today, Grier cites a poster for the Addition Group describing how to handle color-coded positive and negative numbers:

Black plus black is black.
Red plus red is red.
Black plus red or red plus black, hand the sheets to group 2.22

Similar details, however rare, give a glimpse into the objects that might populate a computer’s desk. Gibbs, whose detailed description of worksheet layout is mentioned above, also suggests a very specific set of texts for each computer’s workstation. This includes Barlow’s tables (including square, square root, cube, cube root, and reciprocal of all numbers to ten thousand), Crelle’s multiplication tables up to one thousand, tables for trigonometric functions and logarithms, and “any books related to the subject at hand”.23 Gibbs also suggests referencing E.H. Horsburgh’s book Modern Instruments and Methods of Calculation for useful mechanical devices to aid calculation, such as slide rules.

NOTES

[1] Oxford English Dictionary entry for “computer”.
[2] Cambell-Kelly, Martin and Aspray, William. Computer History. New York; Basic Books, 1996. 12.
[3] The term “planner,” likely starting around the time of De Prony in the early 19^th-century, would carry all the way through to the end of computing in the mid-20^th-century.
[4] Richardson, Lewis Fry. Weather Prediction by Numerical Process. Cambridge University Press, 1922. 219.
[5] Or rather, none that I am aware of.
[6] Answers were calculated twice by two different computers, then verified by a third worker, sometimes called a “checker”.
[7] Gibbs, David A. A Course In Interpolation And Numerical Integration, for the Mathematical Library. G. Bell & Sons, 1915. 1-2.
[8] Grier, 214.
[9] Grier, 285.
[10] Grier, 275.
[11] Grier, 255.
[12] Grier, 275.
[13] Grier, quoting Thornton Fry. 183-84.
[14] Campbell-Kelly and Aspray. 12.
[15] For more images of Railway Clearing House offices, see: http://www.railwaywondersoftheworld.com/clearing-house.html
[16] Campbell-Kelly, Martin. The Railway Clearing House and Victorian Data Processing from Information Acumen, edited by Lisa Bud-Frierman.
[17] Campbell-Kelly (quoting Dowden, 1877, page 99). 67.
[18] Grier, 77.
[19] Gibbs, 2.
[20] Grier, 318.
[21] Grier, 213.
[22] Grier, quoting Ralph Slutz’s Memories of the Bureau of Standards SEAC (in Metropolis, et al, pages 471-77). 213.
[23] Gibbs, 1-2.