I understand that only either nonelectromechanical computing devices
or electromechanical computing devices were ever developed by humans.

I also understand that the first electronic general purpose computing device (GPCD) which perhaps worked on the same electromechanic pattern of today's common computer devices was ENIAC and was completed around February 1946.
When I saw pictures of it, I didn't saw a screen device connected to it.

If an electromechanical computing device is not screened (hence no CLUI/TUI/GUI) and/or is not sensored for general spatial activity (hence not a robot)
What way or ways left to interact with it?

  • 23
    – Ghedipunk
    Dec 12, 2019 at 20:50
  • 6
    ENIAC wasn't a general purpose computer as we'd think of it. It didn't have any stored program ability for most of its life, and was mostly programmed by rewiring it with jumpers.
    – RETRAC
    Dec 12, 2019 at 21:15
  • 13
    Switches, knobs, punch cards, punched tape. Also teletypes and terminals. Dec 12, 2019 at 21:15
  • 9
    Note that there's no dichotomy between mechanical + analog and electronic + digital. You can have digital mechanical computers (e.g. Digi-Comp) and analog electrical computers (e.g. Typhoon).
    – Luaan
    Dec 13, 2019 at 8:34
  • 6
    "I understand that only either nonelectromechanical computing devices or electromechanical computing devices were ever developed by humans." Wouldn't that encompass all possible computing devices?
    – Glen Yates
    Dec 13, 2019 at 16:21

11 Answers 11


I understand that either nonelectromechanical computing devices (analog) or electromechanical computing devices (digital) were ever developed by humans.

That distinction is a bit off, mixing unrelated categorisation. Analogue computers can be as well mechanical, as electromechanical as well as electronic (or pneumatic or hydoponic) - similar digital computers do not have to be electromechanical. For example Zuses Z1 is a fully mechanical-binary-floating-point computer that needs to be cranked to work :)

I also understand that the first electronic general purpose computing device (GPCD) which perhaps worked on the same electromechanic pattern of today's common computer devices was ENIAC which was completed around February 1946.

ENIAC wasn't the first, nor was it general purpose (as Retrac already pointed out). Equally, electromechanical doesn't imply digital.

If an electromechanical computing device is not screened (hence no CLUI/TUI/GUI)

A UI doesn't have to be 'screened'. A UI is whatever lets the User interact/interface.

and/or is not sensored for general spatial activity (hence not a robot) What way or ways left to interact with it?

After excluding everything, nothing should be left. So, lets rather see what has been used.

On the input side:

  • pulling levers
  • turning dials
  • plugging markers
  • setting pointers on gauges
  • pressing numbers on panels
  • flipping switches
  • typing numbers
  • punch tapes
  • punch cards

For output:

  • lever positions
  • wheel / knob positions
  • pointers/gauges selecting number(s) on a scale
  • flip-doors - much like flip-dot-displays today
  • lamps
  • nixies
  • bells
  • punch cards
  • punch tapes
  • printout

Imagine some 1940s Frankenstein movie, or alike, and it still falls short of what man has used to communicate with machines. Anything that can be thought of has been used on one machine or another.

  • 5
    Since we were talking about Jacquard looms, how about "patterns in woven cloth" as output. Admittedly, a Jacquard loom can't really be said to be doing any computing.
    – JeremyP
    Dec 13, 2019 at 11:50
  • 3
    Moniac used water levels in tanks/gauges for I/O. I'm sure there are even stranger indicators than that
    – Chris H
    Dec 13, 2019 at 12:10
  • 1
    @user253751 A tty still features a TUI/CLUI, despite having no screen, doesn't it?
    – Raffzahn
    Dec 13, 2019 at 14:45
  • 1
    Yes it does.... Dec 13, 2019 at 14:52
  • 2
    @JeremyP it would not be a stretch to find some linear mapping function that could be used to interpret the output of a Jacquard loom as the result of computer functions. After all, the Church-Turing thesis says basically anything enumerable is effectively calculable, so in fact that's all a Jacquard loom is doing is computing the result (the cloth) from the input (the cards). Dec 13, 2019 at 19:58

Switches and lights.

Here's the panel for an IBM 650: IBM 650 front panel

The top row of dials are used to configure in decimal values (the IBM 650 being a decimal machine). The repeated columns of 5 lights with 2 on top indicate 1 of 10 values for each digit of the 10 digit word the IBM 650 used. The lights below that row indicate operation and address. The other dials configure various settings, and the buttons trigger various actions based on the dials, like store to memory.

You can deposit instructions and data, examine any memory address, and begin execution from the console. That's all you need.

One of the first programs you enter in such a way are routines to run a punch card or paper tape reader, to read in the rest of your routines in a more humane fashion. Output was usually on punched card, paper tape, or printed to paper.

Interactive teletypes acting as terminals which could provide a command line interface would start to become common in the 1960s.


From the point-of-view of someone at the time, you interacted with early computers in the usual way.

From the 1930's onwards, business and some science used IBM and Remmington-Rand mechanical punched-card processing machines. You set up the specialized gear-based machines for your job (payroll, overdue accounts, super-hard math equation), punched data onto cards, fed them in, and the output was either printed (all caps), or punched onto more cards (for example, payroll checks). Some systems used a long paper tape (which had been used for telegraphs for 50+ years) or very early (1900) systems had numeric dials -- you hand-copied the answer after it finished. But however you did it, it was 1 run with 1 set of results.

A 1950's computer (UNIVAC, IBM702, Honeywell D-1000) essentially replaced the middle part of that system. If you worked at Prudential, you typed the same cards, fed into the same readers, and generated the same printed reports. That was on purpose. Businesses could buy a computer without having to retrain the entire data-processing staff.

It wasn't that simple. Quickly we started using new magnetic tape for I/O. And computers weren't seen only as "like the old machines, but faster" -- early on we knew they could also do more. There was some early work interacting using light pens (SAGE). But in general, we thought it was normal to have Batch Jobs, which were "1 run, 1 result, no interaction" for quite a while.

An early 1960's manual gives a nice feeling for how computers fit right into the old system: Chapter 1 explains how ADP (automatic data processing) consists of EAM and EDP. The old mechanical systems are Electric Accounting Machines, while computers are Electronic Data Processing machines. Using either of them is the same job. Ch2, page 11+, gives a nice history, then page 31 starts on EDP (again, computers). It explains how the major difference is only programming (which someone else does) and knowing how to mount tape reels sometimes.

  • 6
    I think this answer is perhaps nearest the mark in a general sense. Once computers became commercially available, by far the common method of use was: you submitted a program to be run, with its data, on some medium such as punch cards or paper tape. The program produced its "results" on punch cards, paper tape, or lineprinter copy. You often didn't get to examine the result until after your program had finished. No interaction and thus no need for "screens". (Of course, whether such batch systems count as "early" systems depends on how the OP defines "early")
    – dave
    Dec 14, 2019 at 18:18
  • The US Navy had an introductory training manual that covered the period from pure Unit Record to early computer systems. This might be worth including as reference material. scribd.com/document/93125207/…
    – PDP11
    Dec 15, 2019 at 3:18
  • As late as the 1970s, I remember submitting a punch-card deck and getting the printout the next day as the normal way of running a job. One spent a lot more time proofreading and examining the code before submitting it, as a simple one-character typo would mean a one day delay in working on a project. Dec 16, 2019 at 0:47
  • In the 1980s some first year university students in Australia had to use mark-sense or punch cards to submit programs (MONECS system) and received output on a line printer. There was no access to a console and the system only ran in batch mode.
    – PDP11
    Dec 16, 2019 at 1:18

There have been many different kinds of I/O devices since the early years:

There is some more history available at the Computer terminal Wikipedia page.


When your memory device is literally "a screen", you can just look and see what is held in memory.

(I think, but can't find a reference right now, that actually the display tube was a mirror of the actual storage tube - since the storage tube would have had a metal plate in front of it).

The video in the second link is worth watching to get a glimpse of the very first fully electronic stored-program computer. This is a bit of a fudge with respect to the question, since the Baby wasn't really intended to be used for 'real computing', but as a means to test the use of Williams tube storage.

  • While a fine mesh could be used instead of a plate, it reduced sensitivity, so they did just hook up a second tube to the same x/y/z output as a display. No need to rank the Baby lower than it deserves. To demonstrate the storage, it had to work, thus being a full machine, despite it's quite limited capacity.
    – Raffzahn
    Dec 13, 2019 at 1:53

The Harwell Dekatron is an excellent example of this type of machine, not least because it was designed for simplicity and reliability rather than speed, and this in turn makes it easy to understand. This is a machine that operated entirely on decimal numbers, not on text.

It has a few buttons on its control panel, which are essentially used to start, stop and reset it. It takes all its real input from a bank of perforated-tape readers, of the sort already in widespread use at telegraph offices at the time. These contain both instructions and data for the machine to work on.

Output can be to another paper tape (via a perforator), or to a printer. Another option however is to directly read the contents of the ninety memory words, by observing the position of the discharge glow in each Dekatron tube. Many early computers had an option similar to this, but it eventually grew impractical as the amount of memory increased.

  • Thumbs up for someone recognizing that "early computers" doesn't mean the 1970s :-)
    – dave
    Dec 14, 2019 at 0:11

The first computer I built was a COSMAC ELF. It didn't have a keyboard or a monitor, no CLIUI/TUI/GUI. It was a full-fledged digital electronic computer.

It had 8 toggle switches and a pushbutton for input and a dual 7-segment LED display for output.

  • Have you ever heard of the 1802 Membership Card?
    – supercat
    Dec 15, 2019 at 21:38

Before terminals, the most common method of interacting with large electronic tabulators, calculators, and computers (IBM mainframe, et.al.) was via punched cards for input, and line printer output for output. Some minicomputers used punched paper tape instead of punched cards.

The first terminals were mechanical teletypes (ASR-33, et.al.), with a keyboard plus a line-by-line printer above the keyboard. No glass display.

The first personal computers (Altair, IMSAI, KIM-I, Cosmac Elf) had switches on the front panel for input and LEDs for output. A serial port that could connect to a glass terminal was optional.

  • Sure, teletypes were the first interactive way, somewhat like a line-by-line scrolling screen, but all on paper -- you could play a game where you typed "go east" and the computer printed back "you see a vast cave...". But weren't they not until the 70's, on timesharing minicomputers? Dec 15, 2019 at 2:04
  • There were typewriter consoles even on vacuum tube computer of the Johnniac IAC era.
    – hotpaw2
    Dec 15, 2019 at 3:32
  • Refer to Page 304 from this training manual: scribd.com/document/93125207/…
    – PDP11
    Dec 15, 2019 at 14:15
  • That's lovely. Page 304 of the document is pg 315 of the digital scan. The heading "consoles" is 1 long paragraph where "provides external monitoring and supervisory control" is the most specific part. 1966. Written for the navy, but covers the while industry. I feel like supervisory control means the sysOp uses it -- not the people running normal jobs. Dec 16, 2019 at 0:22
  • I suspect the Flexowriter came before the Teletype :-)
    – dave
    Dec 19, 2019 at 23:37

Besides dedicated LEDs/lights, some of the earlier computers printed out there. In my first programming class, our college had an old PDP-11. Running a program would result in an answer on the printer. For batch programs running on other mini-computers, the output all went to a printer. Fortunately, we were able to use terminals with screens for the actual development of the computer programs, as programming with punch cards (the way of some of the computers of the 50s/60s) would be unnecessarily tedious and error-prone. But it is technically possible to be screenless and just type your answers and get your responses real-time on a printer.


When I started programing in 1974, I had a keyboard, printer, and punch card reader/writer.

Using a command prompt in Windows the text is printed to the window rather than a real printer. Just the way it was.

Early hobbyist computer kits had switches, one set for address, one set to set the value of that address, and two sets of lights showing address/value. The only useful program I saw was tennis.


I used to maintain a sonar system with an embedded computer. While the sonar system was certainly a sensor, and the computer could drive displays, it was also possible (but awkward) to use the computer itself with all the other cabinets shut down. Programs could be entered, single-stepped, or ran by pushing buttons. And results could be read from lights. Each bit of each register was connected to a light which was also a button to change the state of that bit.

You must log in to answer this question.