Talk:Unit record equipment

This article needs information about Nazi Germany's use of these machines as crucial tool in the extermination of Jews
Edwin Blacks book IBM and the Halocaust details this and has extensive sources. A start would be to add the 1939 Nazi racial census to the timeline, the largest, fastest and most extensive census at that time.

About this census, Black writes:

"The additional Hollerith machinery assembled was massive: 400 electrical key punches, 10 gang punches, 20 summary punches, 300 key punchverifiers, 70 sorters, 50 tabulators, 25 duplicators, and 50 D-11 VZ tabulators. The Reich had imposed seemingly impossible target delivery dates for November 1939. So to increase speed, Dehomag's engineers converted their versatile D-11 calculating tabulator into a pure counting machine dubbed the D-11 VZ. The improvised device could process 12,000 60-column punched cards per hour in sixteen counters and then precision-punch its own summaries onto 80-column cards. Eighty million cards were actually used. 5" IBM and the Halocaust by Edwin Black page 175.

Black gives these sources for the above statement (which I am not in a position to verify):

5. "Aus dem Volkszahlungshaus in Berlin," Der Stromkreis (Werkzeitschrift DEHOMAG), Berlin, 66 (February 1940): 1-8, cited in Friedrich W Kistermann, "Locating the Victims: The Nonrole of Punched Card Technology and Census Work," IEEE Annals of the History of Computing, 19:2 (April-June 1997); Hollerith-Tabelliermaschine D-11 mit Zahleinrichtung (D 11 VZ), (Berlin: Dehomag, 1939) cited in Kistermann, "Locating the Victims: The Nonrole of Punched Card Technology and Census Work"; Letter, J.G. Johnston to J.E. Holt, June 14, 1938, Department of Justice, War Division, Economic Warfare Section, NA RG60; also see Biehler, "Lochkartenmaschinen im Dienste der Reichsstatistik," Allgemeines Statistisches Archiv (ASA) 28 (1938/39): 90ff, 93.

CouldThatBe (talk) 16:53, 18 May 2015 (UTC)


 * If you have content that is appropriate to the article, backed by reliable sources that you can cite, please feel free to add it to the article.—Finell 05:43, 17 February 2016 (UTC)

"Unit record equipment" category?
Some pages for various forms of unit record equipment are pointing to this page as the "main page", but, as User:R. S. Shaw noted in his removal of one such link from IBM 407, "main is used for sections, not whole article". It might be more appropriate to add a "Unit record equipment" category and have the pages currently using instead use. Guy Harris 22:07, 9 October 2006 (UTC)


 * Another possibility would be to use Template:SubArticle on the articles about pieces of unit record equipment if Unit record equipment gives a summary of the information in an individual article. Guy Harris 22:17, 9 October 2006 (UTC)


 * The Category:Unit record equipment idea clearly fits here. A category groups related articles very well.  I'm in favor of putting in  and taking out  .  I don't think Template:SubArticle would work well in this context; it's meant for talk pages, not articles, and is only used by a couple of talk pages presently. -R. S. Shaw 23:42, 9 October 2006 (UTC)
 * This seemed so sensible to me, I've been bold and gone ahead and created the category. I'll be editing pages into it tonight, I suppose.  I think punch card probably doesn't belong, as it is not equipment.  Control panel (computer) probably does, as it is part of the equipment. -R. S. Shaw 00:21, 10 October 2006 (UTC)


 * punch card does belong - they are the glue that lets the machines work together. Note that punch  card is restricted to Hollerith cards, etc, it is not all punch cards.  (this is not to say I agree with using a category, only that if a category is used, then ....  The problem with a category is that it is not make the set of articles explict to the reader)tooold 01:23, 10 October 2006 (UTC)

UNIVAC 1005
Described on UNIVAC page as having a stored program and a programming language. If this is a unit record machine, then wouldn't the IBM 1401 also be a unit record machine? 69.106.232.37 20:45, 10 October 2006 (UTC)


 * I don't think so. I believe the 1005 still followed the unit record model, i.e. cards were the primary storage medium. I never used one, but my recollection was that the 1005 had very limited memory and no mass storage. It was more in the 407 class than a 1401. Here is one article i found via Google that discusses 1005 use in Vietnem: https://calldbp.leavenworth.army.mil/eng_mr/txts/VOL48/00000009/art8.pdf --agr 21:44, 10 October 2006 (UTC)


 * Information seems hard to come by on this and seemingly somewhat contradictory. The 1004 has a fair  reference at, but the 1005 seems thinner.  This says "the UNIVAC 1005 was a plugboard computer that you had to wire for each program" but this says "SAAL was my first real love affair: Single Address Assembler Language on a UNIVAC 1005".  Supporting the latter is this: "developed a program written in IBM DOS/360 Assembly Language to automatically translate pro­grams from UNIVAC 1005 Assembly Language (SAAL) to IBM DOS/360 Assembly Language".  So it does look programmable despite the first quote.  Seems like a computer in a accounting-machine-like form, a close call. -R. S. Shaw 06:36, 11 October 2006 (UTC)


 * "Cards primary ..." would have been my definition, as well. But now I want to change my mind.


 * Found this site,, guy worked on a 1005 with about 4k memory, card read/punch, and printer. Reads just like a basic 1401.


 * From IBM Archives (note the "Unit Record Systems Programming"),, "On October 28, 1970, the company rolled out the IBM System/3 Model 6 (IBM 5406). Rochester's Advanced Unit Record Systems Programming group had developed the Report Programming Generator II programming language intended for commercial applications on the Model 6."


 * I think we should exclude digital computers, even if they only have card & printer capabilities, from unit record equipment. Many IBM 650 sites were card only, for example, and logically were unit record machines, but the 650 is not a machine I want to have included, no more that the 1005 or 1401. Anyone want to try writing a definition for the article?  69.106.232.37 07:05, 11 October 2006 (UTC)
 * The UNIVAC 1004 was plugboard-programmed but electronic, unlike the electromechanical IBM tabulators. There was an add-on for a 1004 which plugged into the plugboard area, replacing the plugboard, and made it software-programmable in a limited way. The 1005 had something like that as standard. It was close to the IBM CPC in concept, but with more electronics and fewer gears. --John Nagle (talk) 18:24, 17 September 2009 (UTC)

unclear sentence
This sentence from the 2nd paragraph needs clarification:

"The automatic operation of some unit record machines was directed by control panels, wired to directed the operation for a specific application."


 * 1st attempt at rewrite. Ok? 69.106.232.37 23:53, 10 October 2006 (UTC)

Looks good, thanks!

How it works
This was just removed from the article by ClueBot as possible vandalism by 66.83.148.18, but I think it is fair criticism that belongs here: "but of course none of the smart people tell us how it works!!!!!!"--agr (talk) 19:46, 17 September 2009 (UTC)


 * I noticed that. And I agree, the article doesn't explain what one does with all those machine. I've been trying to find a reference that has a good overview of data processing with that equipment.  IBM had a series titled "Punched Card Data Processing Principles", but those books just introduce the various machines. --John Nagle (talk) 03:35, 18 September 2009 (UTC)

How it works, again
OK, here's a first cut at how it works.

A key concept for modern readers is to understand that none of the unit record machines have any significant amount of memory. The tabulators have the most, with 10 to 40 multi-digit counters and some "selectors" (relays) which can hold one bit of state. The sorters only read one column at a time and lack any storage devices at all. The cards are the file storage.

An invoicing/billing system might work like this:


 * Transactions come in from sales reps or a sales floor, typically as paper documents to be keypunched. Each transaction is manually typed in, using a keypunch, creating a single card for each transaction, which must contain at minimum the item number, quantity, and customer number.
 * Payment transactions come in from Accounts Receivable, with the customer number and payment amount.
 * Transactions need to have the dollar amount for the item and the total cost computed. This requires the following steps:
 * Sort transaction cards by item number with a card sorter
 * Merge the transaction cards with a the pricing file (a deck of cards with one card for each item number, containing the price for that item) using a collator like an IBM 77. The transaction cards follow the price card for each item number.  Transactions that don't match come out in the fourth output stacker of the collator, for manual attention.  Note that each card must have a field with a "card type" punched, so we can separate the decks later.
 * Run the combined deck through a reproducing punch, like an IBM 514, set up for "interspersed gangpunching", so that the price per item is propagated to the following transaction cards with the same item number and punched into it. This phase also might propagate the item name (but it will have to be short; we only have 80 columns total.)
 * Run the combined deck through the collator again, or through a sorter, to separate the pricing deck from the transaction deck using the card type field. The pricing deck goes back into storage.
 * Notice that what we've just done is equivalent to an SQL "join".
 * We still don't have the total price (item price x quantity), so the transaction deck has to go through a calculating punch like an IBM 602. The multiply is done, and the product is punched back into the card that contained the operands.
 * Now all the transactions have been priced. At this point, we can run the deck (which is in order by item number at this point) through a tabulator and produce a sales report by item number, if desired.
 * Now it's time to do billing. We first sort the transaction deck by customer number, and the payment transactions by customer number.
 * The transaction deck is then merged with three other decks. First, the company name and address deck, which identifies the company being billed.  Then, the outstanding balance deck, representing the invoice balances from the previous month. Finally, the payment deck. A typical order might be
 * Name and address card 1
 * Name and address card 2
 * Name and address card 3
 * Transaction deck
 * Outstanding balance deck
 * Payment deck


 * These go into a printing tabulator such as a IBM 407. For this job, the IBM 514 reproducer is plugged into the tabulator for "summary punching".  The summary punch will produce a new outstanding balance deck for the next billing cycle.  The tabulator is loaded up with preprinted invoice forms, probably multiple-part forms with carbons, with feed holes along the edge.  A "carriage tape" loop, which indicates on what print lines each invoice item should be printed is loaded into the reader on the tabulator carriage.  The plugboard for the job is placed in the tabulator's plugboard rack, and the big locking handle pushed down.  Then someone pushes START.  The tabulator grinds away, printing the name and address from the name and address cards.  The transactions are printed and added up. The outstanding balance is added. The payments are subtracted.  The tabulator advances the invoice form to the line for the amount due and prints that.  Then, with a loud grinding noise, the 514 starts up and punches one card, the outstanding balance for next month.
 * The forms go through a decollator, which separates out the multiple copies, a burster, which separates the pages, and perhaps a folder/inserter, which puts the statements into envelopes. Then a pass through a postage meter, and they're ready for mailing.

That's automated data processing. --John Nagle (talk) 04:22, 18 September 2009 (UTC)

We need a collator article
The one piece of equipment that doesn't have a article is the collator. The major collators were the IBM 77 and IBM 85/87; the IBM 188, from 1961, was the last, as by then computers were coming in. . The first collators were developed for the U.S. Social Security Administration.. Collators do matching and merging of previously sorted decks. They can combine cards from two input sources, or keep them completely separate. These were the machines that added "join" capability to unit record processing, which made the whole process much more general and allowed for more error checking.

"Unit record" not defined?!
I vividly recall encountering the term "unit record" in my earlier decades (I'm closer to 75), and feeling very frustrated because I seemed to be the only one who didn't know what it meant. (What in the blazes is the "unit"?) I scanned this article with a Find command, and feel confident in saying that "unit record" is not defined anywhere! As far as I can tell, a tab. card is a unit record -- it's one self-contained collection (record) of data. I'm still not sure enough to actually edit the article, however.

Fwiw, I can recall continuation cards, for more data than could fit onto one card.

Regards, Nikevich (talk) 09:45, 10 October 2010 (UTC)


 * Defined! 69.106.237.145 (talk) 08:38, 30 April 2011 (UTC)

Card lore
We didn't think of them that way, but tab cards were very-low-cost precision mechanical entities, made to tight tolerances.

One nasty trick was to shave a few thousandths of an inch off the bottom edge of a card. I was told that such a card would mis-read sometimes, but wouldn't appear to have anything wrong with it. Careful comparison with a known-good new card would reveal the discrepancy.

I worked briefly in one large installation, where probably several 100,000 cards were handled per day. I can recall flat surfaces with a low vertical wall to their left; an operator would place a poorly-stacked deck onto the flat surface, and repeatedly lift the right edge and drop it while pushing the deck to the left. In a few seconds, the cards were all neatly aligned.

Now and then, a card would become creased or otherwise cease to be flat. Instead of keypunching a replacement, sometimes it was enough to apply a hot clothes iron (no steam, just about sure) to flatten it. In this installation, the irons were kept hot, and placed into fire-resistant topless and frontless benchtop safety "bins" apparently made of bonded asbestos.

In addition to keypunching, one could request that a punched deck be verified; the procedure was much like keypunching, except that the card was read, a column at a time, as the operator read from the handwritten (or possibly printed) form. Programmers (I was one, for a few months) requested keypunching and verifying.

Singer developed a prototype drum memory for the electromechanical machines this article described; it was called the SEMA, for Singer Electronic Machine Accounting, iirc. My job was to integrate it with the existing machines, which used lots of electromagnets and had sparking contacts. I tried to suppress the EMI; had some luck.

Regards, Nikevich (talk) 10:07, 10 October 2010 (UTC)

Cultural references
In Arthur C. Clarke's early short story Rescue Party, the alien explorers find a "... wonderful battery of almost human Hollerith analyzers and the five thousand million punched cards holding all that could be recorded on each man, woman and child on the planet". Writing in 1946, Clarke, like almost all sci-fi authors, had not then foreseen the development and eventual ubiquity of the computer. PhilUK (talk) 22:00, 16 February 2013 (UTC)

"Unit record" terminology and origin
I removed footnote 1, which contained the following text:

"In the late 1800s, early 1900s unit record was a reference to the recording of all information about a transaction or object on one document. At that time the library index card was pointed out as an early example of a unit record. Even unit record desks were manufactured, a desk that included what later, for punched cards, would be called a tub file.  This quote We had records of every car and locomotive on the railroad on a key-punched card or other unit record, either generated in the Car Accountant's Office or through other means, from  shows that in 1888 1) users were applying the term unit record to punched cards and 2) the term's use was more general than just punched cards.  Markus Krajewski in Paper Machines: About Cards & Catalogs, 1548-1929, 2011, MIT, credits Conrad Gessner with developing the unit record concept."

Overall, it appears to be WP:OR. Also, it is hard to determine from the text what specific sources contain or support what specific content. In some cases, the lack of full citations makes it unclear what work is cited. Also, Wikipedia does not use italic to indicate quotations; either use quotation marks or, for longer quotations, block-quotation markup:

If someone reworks this text with appropriate citations, it probably belongs in the body of the article, possibly in a section called Etymology or Etymology and early usage. In my opinion, it is too long for a footnote. Also, having this material in the text would enable complete source citations in footnotes, without interrupting the flow of the text.—Finell 06:16, 17 February 2016 (UTC)


 * I restored the text before finding your "see talk". It is necessary text, responding to "What is a unit record?"  (likely a prior "talk" entry).  More... later 73.71.159.231 (talk) 16:47, 18 February 2016 (UTC)

Unit record equipment is too broad a title
1) It allows for text from any field of study with "unit records".

2) Further, "Unit Record ..." is NOT the common name. As best I recall, Wikipedia wants common names and the common name is "punched card machines" or "... equipment". See for example "Early Punched Card Equipment, 1880 - 1951" "This article was initially written as part of the IEEE STARS program."

To checkout the "common name", just do a Google search for "unit record" and see what you find.

Cortada's "Historical Dictionary of Data Processing" has index entries for "punched card" but not "unit record". Williams "A History of ..." 1st ed has entries for "punched card" but not "unit record".

A better title would be "Punched Card Machines". That could be prefixed with "Data Processing", but the machines had more uses than just data processing. 73.71.159.231 (talk) 16:51, 16 March 2016 (UTC)


 * Disagree. In my first year or so at college we still had a roomful of that stuff, and I assure you, "unit record equipment" was the common name, used by IBM and many others. That the term "unit record" occurs in many other contexts does not change that. The process, or usage, of the equipment was commonly called "unit record accounting", though it was used for more than accounting (narrowly defined). An alternate name was "tab card" (short for "tabulating card"). "Punched card machines" is not sufficiently specific as many of these machines, such as the card sorter and duplicating punch, had uses in the DP shop even when the actual data processing (accounting) was being done by computers. "Unit record accounting" described an entire pattern of record-keeping and transaction processing that used these machines, a concept not conveyed by titles like "punched card equipment". Your references (like your own suggestions here) appear strongly to have been written by people who weren't there. In the case of the "Early Punched Card Equipment" article, it (perhaps ironically) ends too soon; the heydey of this stuff was well after 1951. Jeh (talk) 15:15, 20 March 2016 (UTC)


 * Disagree. I agree with Jeh: I was using UR equipment in the UK as late as 1961, after which I (personally) moved on to 1401s. The applications I worked on may well have survived for longer, as they worked well.Jpaulm (talk) 13:44, 23 March 2016 (UTC)


 * Disagree. Punched card gear was being used through the 1960s and into the 1970s. Big shops had computers, but little shops were stuck with cards. IBM tried to migrate card shops to the IBM System/3, with its little 96-column cards, but few went that route. John Nagle (talk) 17:46, 29 March 2017 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified 5 one external links on Unit record equipment. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive http://web.archive.org/web/20160304194200/http://www.feb-patrimoine.com/projet/frbull/heide_bull.pdf to http://www.feb-patrimoine.com/projet/frbull/heide_bull.pdf
 * Added archive http://web.archive.org/web/20120507132416/http://www.feb-patrimoine.com/projet/bull_t30/tabu_t30.htm to http://www.feb-patrimoine.com/projet/bull_t30/tabu_t30.htm
 * Added archive http://web.archive.org/web/20130821174522/http://www.feb-patrimoine.com:80/projet/serie_150/serie_150a.htm to http://www.feb-patrimoine.com/projet/serie_150/serie_150a.htm
 * Added archive http://web.archive.org/web/20130727083838/http://www.feb-patrimoine.com/projet/gamma3/gamma3.htm to http://www.feb-patrimoine.com/projet/gamma3/gamma3.htm
 * Added tag to http://bitsavers.org/pdf/ibm/productDescriptions/A-4060_IBM_Products_1940
 * Added tag to http://www.columbia.edu/cu/computinghistory.html
 * Added archive http://web.archive.org/web/20130530040351/http://www.feb-patrimoine.com/PROJET/index.htm to http://www.feb-patrimoine.com/PROJET/index.htm#SYSTEMS_CATALOG_

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at ).

Cheers.— InternetArchiveBot  (Report bug) 20:26, 20 July 2016 (UTC)

-- http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/productDescriptions contains A-4060_IBM_Products_1940.pdf -- I wasn't able to access bitsavers.org. 73.71.159.231 (talk) 06:42, 23 August 2016 (UTC)

-- www.columbia.edu/cu/computinghistory.html seems to have changed to http://www.columbia.edu/acis/history. But... The "acis" (it relates to Frank's new job, or at least the org. that gives him access - I've not talked to him in a long time) works, but what is displayed is the old cu/computinghistory. And the www.columbia.edu/cu/computinghistory.html works when the ".html" is deleted. So what's out there right now is

* Columbia University Computing History IBM Tabulators and Accounting Machines etc...

and I've no idea as to what is going on. 73.71.159.231 (talk) 07:50, 23 August 2016 (UTC)

Are Powers machines mechanical or electromechanical?
Text such as this


 * 1911: James Powers forms Powers Tabulating Machine Company, later renamed Powers Accounting Machine Company. Powers had been employed by the Census Bureau to work on tabulating machine development and was given the right to patent his inventions there. The machines he developed were mechanical, as opposed to Hollerith's electromechanical products.

is common throughout the various punched card articles. But Powers machines were electromechanical; they were powered by electricity. Yes, mechanical linkages were used instead of relays, but is that the sole criteria for mechanical vs electromechanical? 73.71.159.231 (talk) 14:43, 9 October 2016 (UTC)


 * Do you mean they were mechanically cranked by a motor to move the cards, or something else?


 * Fwiw, both traditional electric typewriters (up to and including the Selectric 1, 2, and 3) and traditional electric adding machines are considered electromechanical, even though the only electrical components are an electric motor and a power switch. There are no contacts under the keys. The rest is all levers, cams, clutches, pulleys, and (in the case of the Selectric) digital-to-analog converters built with levers (really!). Jeh (talk) 00:09, 10 October 2016 (UTC)

About the 80-column punched card
I made this edit to call attention to a sentence fragment.

It is ok for item lists to be e.g. simple lists of terms, names of things, etc. It is also ok for an item list to contain descriptions of events (e.g. company A acquired company B). The item list here contains some of each and is this inconsistent. In this particular case, it is not at all clear why the 80-column punched card is mentioned in the entry for 1928. Was it introduced in that year? In that case the word "introduced" or similar should appear. Or was it solely associated with The Tabulating Machine Company? Or what?

Since many entries in the list are complete sentences, with verbs, they of course all should be. Jeh (talk) 18:56, 8 November 2016 (UTC)

What exactly did Hollerith invent?
The article states that "In the 1880s Herman Hollerith invented the recording of data on a medium that could then be read by a machine." However, Babbage's Analytical_Engine design used punched cards for both instructions and data, in the 1830s. In the Wikipedia article on Punched_card it lists five prior designs, at least one of which was built, but still assigns the invention to Hollerith. — Preceding unsigned comment added by 81.187.188.94 (talk) 14:41, 2 January 2022 (UTC)
 * Babbage only proposed the idea; he didn't actually implement it. I rephrased to avoid the loaded word "invented".--Srleffler (talk) 20:33, 7 January 2024 (UTC)

Etymology needed
As mentioned by someone before, this article is in need of a section about the origins of the term unit record equipment and who used it first. --217.149.171.189 (talk) 00:25, 3 January 2022 (UTC)
 * The article also needs to mention more examples of unit record equipment. All of these have been described as unit record devices:
 * Card punches, whether 80 column of 90 column
 * Card readers, whether 80 column of 90 column
 * Collators
 * Duplicators
 * Interpreters
 * Key punches
 * Printers
 * Sorters
 * Tabulators, whether 80 column of 90 column
 * I would assume that keypunches and such for 96-column cards were also referred to as unit record equipment, but have no firsthand knowledge.

FWIW, I've never heard the 1401, UNIVAC 1005 or any other stored-program computer referred to as unit-record equipment. --Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:24, 3 January 2022 (UTC)

Labeled Section Transclusion?
Currently the Keypunch section has only a short paragraph with a template, along with a comment that the individual units are accounted for in the main article. Is this a poster child for the use of Labeled Section Transclusion (LST)? Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:33, 29 November 2022 (UTC)

Mention common but politically incorrect nomenclature?
In the 1960s the term keypunch girl was more common than keypunch operator. Should the Keypunch section mention the term even though it is potentially offensive? Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:46, 29 November 2022 (UTC)
 * I was thinking of keypuncher also, but including these might be TMI. The article does mention that operators were usually women. When I first started working most keypunchers were older than I was, so girl didn’t fit.Peter Flass (talk) 14:31, 29 November 2022 (UTC)
 * I recall the terms keypunch girl and Kelley Girl being used regardless of age. --Shmuel (Seymour J.) Metz Username:Chatul (talk) 07:44, 30 November 2022 (UTC)

Minicomputer anachronism
Unit record equipment claims however DEC was using the term minicomputer well before then. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 15:38, 8 January 2024 (UTC)

The paragraph doesn’t claim first use of the term.Peter Flass (talk) 19:07, 8 January 2024 (UTC)