The Integrating Tabulator

1900 · Automatic Card Feed · Work 572 of 1,871

The Integrating Tabulator, 1900, Automatic Card Feed

A tabulator with an automatic feed hopper above and a card stacker below, cards travelling through the reading station on their own in a steady stream, the counters accumulating totals from the punched fields, a stack of cards in the hopper and a growing stack in the stacker close to the lens, the cards plain buff punched with round holes, the counter dials and wiring behind the feed in soft focus, the machine in black iron and brass, a tabulating office lit by daylight from a side window.

What it was used for
The adding tabulator was the machine for the agricultural statistics of the 1900 United States census. The counting machine of 1890 could only count unit by unit: one card, one person. Agriculture demanded sums - acre upon acre, bushel upon bushel. Truesdell puts the difference thus: what was needed was "an adding (or 'integrating') tabulator, rather than simply a counter". Inside the counter lay a cylinder with nine conducting strips of graduated length; a brush swept across them and gave one to nine impulses depending on its setting, and a carry mechanism took the amounts into the next digit position. The machine for the farm card had ten such adding devices plus 21 dials; the one for the crop card only three.
What was new
What was new was that the punched card now carried quantities, not just characteristics. The patent for it, No. 430,804 "Electrical Calculating System", Hollerith had filed as early as 4 January 1887; it was granted on 24 June 1890. The first documented need came not from the Census Office but from the Office of the Surgeon General, where machines were installed in 1889: there the days soldiers lost through illness were to be summed, monthly figures running from 1 to 31. Twelve machines ran in 1901 for the farm cards, in June 5,483 cards per machine per day, by late August 8,130. The reversal lies in the sequel: for decades the Census Bureau tried to build such a machine itself. Truesdell writes that it "never" developed it far enough to be worth producing in quantity.

What the sources leave open: The year in which the first adding tabulator was built was not found; what is documented is the 1887 application, the 1890 grant and the use in 1900-01; Martin already illustrates a "Hollerith Integrating Machine" in 1891. On the number of columns on the crop card Truesdell gives 16, Lars Heide 20 in one place and 26 in another. No figure in cards per minute exists for this machine in any source consulted; only daily output is documented.

Connections

Every connection names its kind. “Forerunner” on its own would be a claim.

  1. Follows fromHollerith's Tabulator1890Hollerith developed his own machine further. The 1890 census machine could only tally one and one over and over; the integrating tabulator housed separate adding machines and totalled several fields of a card at once. It debuted in 1896 and was used in the 1900 census.Source: Columbia University Computing History: The Hollerith Integrating Tabulator
  2. Works withThe Hollerith Card1890The integrating tabulator works on the fields of the punched card: its separate adding machines simultaneously add the totals recorded in separate areas, or fields, of one card.Source: Columbia University Computing History: The Hollerith Integrating Tabulator

Further reading