I've been using LaTeX for about two years; I'm only learning TeX now, to get a better understanding of the foundations.

In Chapter 6 : Running TeX of his TeXbook (more specifically, on page 23), Knuth writes about how to get started with TeX:

Go to the lab where the graphic output device is, since you will be wanting to see the output of what you get---it won't really be satisfactory to run TeX from a remote location, where you can't hold the generated documents in your own hands.

This passage left me a bit confused: back in the day, when TeX was released, did one have to print out the output document (i.e. produce a hard copy) in order to visualise it?

If not, what was on-screen visualisation like? Was the quality sufficient to get a good idea of what a hard copy would look like?

I'm just being curious, here... If on-screen visualisation was indeed possible, would anybody have screenshots they'd be willing to share or link to in an answer?

EDIT: Since all of the answers I've got so far are equally valid, I cannot reasonably accept a particular one.

enter image description here

  • 69
    Real men don't need previews.
    – topskip
    Commented Apr 20, 2013 at 13:53
  • 40
    @topskip Let's see what female users of TeX think of that :)
    – jub0bs
    Commented Apr 20, 2013 at 14:04
  • 18
    A very interesting question and answers indeed! People have it so easy these days and yet they use Comic Sans for reports and clip art in presentations… -_- Commented Apr 20, 2013 at 15:01
  • 9
    Real men don't need previews? Real men don't use TeX. The use ASCII for documentation. If they want something "evenly grey" they put a coat of primer over it. (And you can believe it is not latex).
    – Kaz
    Commented Apr 21, 2013 at 7:12
  • 16
    @Jubobs -- how about "real nerds don't need previews"? when i started inputting to-be-typeset math to a computer, it was on punched cards. didn't even have real upper/lowercase. and had to wait for output on photographic paper to see what it looked like. folks got it too easy nowadays. Commented Apr 30, 2013 at 14:01

11 Answers 11


It depended on the available hardware. My first tries with TeX were on a VT-100 terminal and the only "previewer" was dvi2tty.

Just for fun, here's what could be shown on the terminal with it

\sum_{k=0}^{n} k^2 = \frac{1}{6}n(n+1)(2n+1)

enter image description here

Actually the screen background was grey and the text was green, but this gives an idea. (I was just too lazy to prepare a Plain TeX source file.) Oh, and the printer connected to the mainframe (a Vax/VMS system) was in another building a couple of kilometers away.

  • Wasn't it called "runoff" back then? Commented Apr 20, 2013 at 14:59
  • @PieterGeerkens RUNOFF was a different typesetting system, from which nroff was originated.
    – egreg
    Commented Apr 20, 2013 at 15:08
  • 2
    And the character font was a lot less readable :) Commented Apr 25, 2013 at 11:43

I started using TeX in 1987 on SunBSD (later sunos later solaris) using the dviview previewer from the vortex project. That is still one of the best viewers around, you could select individual characters ask what font they were in, and the rendering was as good as the (monochrome) screen could handle.

This wasn't the setup I had but from a vortex project thesis from 1988, here's a picture of dvi preview synchronised with TeX source in an emacs buffer:

Page 151 of Pehong Chen's 1988 thesis

  • 22
    You lucky bas... ehm, guy. ;-)
    – egreg
    Commented Apr 20, 2013 at 13:55
  • 9
    @egreg compensation for abandoning the math department and switching to computer science:-) Commented Apr 20, 2013 at 13:59
  • 8
    I should downvote this answer because it has emacs on it. ♥ But I won't. :) Commented Apr 20, 2013 at 15:34
  • 5
    Sun workstations were awesome! Huge screens with Unix GUIs in the days when everyone else was peeking at BSD (or worse) on 24x80 character terminals. Such a pity Sun lost the battle for the workstation market...
    – alexis
    Commented Apr 20, 2013 at 16:33
  • 2
    In 1988 or 89 I had a fancy terminal which had a character-grid mode and a graphics mode (gasp!), and there was a DVI previewer application (on VMS) which could drive that, at about 5-10 seconds to raster one page. I was most envious when, a couple of years later, I saw folk with Suns and fancy previewers. 5-10 seconds per page: hmm (young 'uns), there's a reason why \include exists. Commented Apr 20, 2013 at 16:51

OK. I was officially the first TeX user in Imperial College, London, using it to write my PhD thesis. Malcolm Clark was the one person in Computer Centre who supported TeX.

I bought Knuth's "TeX and Metafont" in 1983, and used a buggy implementation on the College mainframe. There was no preview, and there were no laser printers. I wrote the files at home on a borrowed BBC Micro, and took the files to College on 5 1/4" floppies and uploaded to my mainframe account. I would submit a batch file and I would see the resultant log file. If there was an error I would resubmit. The dvi was then submitted to the (Autologic micro-5) phototypesetter. I would return the next morning and collect the roll of photographic paper from my pigeonhole, cut it into pages using a guillotine, and inspect. Invariably there were errors and I would correct the file and resubmit for next day...

Those were the days!

As an amusing anecdote, years later I bumped into Malcolm and thanked him, because had it not been for him I could not have completed my PhD in TeX. He said on the contrary, he had to thank me, because were it not for me using TeX, he could not have justified his job as TeX support. ;-)

(Please don't ask me how I wrote graphics routines in Fortran to draw all the figures in my thesis on microfilm!!)

  • 3
    Thanks for your answer. I'd be curious to have a look at your thesis. Is it publicly available at some address?
    – jub0bs
    Commented Apr 23, 2013 at 23:18
  • 5
    Hi. The source of the TeX files are on 5 1/4" floppies somewhere, but here is a scan of the final output: docs.google.com/file/d/0B1o0Y9sDxF-VT25ZXzBnbFpjVk0/… Please note that because of lack of screen preview I gave up trying to leave gaps for figures, and I manually pasted the text from galleys on a long roll of phototypesetter output. Commented Apr 25, 2013 at 23:15
  • 4
    @Kaveh How did you write graphics routines in Fortran to draw all the figures in your thesis on microfilm? Commented May 25, 2013 at 11:58
  • 1
    @ADP: It was a long time ago, but we had drivers for plotters and microfilm. The latter was same width as 35mm film but without the sprockets. An electron beam drew vector lines on the film which was processed. I then used a conventional photographic enlarger to print this onto photographic paper. Those were the days!! Commented May 26, 2013 at 15:08
  • 2
    @Kaveh OMG. I still remember having to process, only 10 years ago, film in a darkroom for certain applications. Sometimes it is easy to forget how quickly things change. Thanks for sharing. Commented May 26, 2013 at 19:57

Screen? Previews? They didn't exist at the beginning.


it has been asked how knuth viewed his "first page". it was definitely not on screen. it was most likely printed out on the xerox xgp which was installed at the stanford ai (artificial intelligence) lab (sail). in the book "companion to the papers of donald knuth", don says (pp.58-59) that he'd been "playing around" with this machine for a few years. it doesn't say definitively that the "first page" came out of the xgp, but no screen output is mentioned at that stage for display of more than one or two letters at a time, and that's a terrible way to try to decide that a whole page is "perfect".

you can learn a whole lot more about don and printing (and about everything else that he's done) from that book. it contains an extensive index to all his books and a complete bibliography including all his papers that were in print up to 2011. recommended.

end edit

At Stanford, there were a couple of electrostatic printers, 200dpi, made by Versatech and Benson-Varian; output was on rolls of slightly slimy paper since it had to go through a smelly liquid bath. Stanford also had a xerox xgp printer (approximately the size of a refrigerator, or a computer mainframe unit, if you remember those) with about the same resolution that used dry toner, but it was coarser then the toners now used, and periodically it would clump into little grains that were large enough to punch tiny holes in the (plain) paper. Output on these was often magnified to 130% and photographically reduced to smooth out the edges.

The first really hi-res output was to an alphatype crs, with a nominal resolution of 5333 dpi. (This was generated on a crt screen at much lower resolution and larger size, and stepped down through lenses.) The photographic paper onto which it was imaged was on large sheets, each of which could -- just barely -- accommodate four letter-size pages, the size of tugboat pages in fact.

Early issues of tugboat -- look for the entries for "outout devices" -- have lots of examples and technical info.

The earliest mention in tugboat of a previewer seems to be in 1982, at Stanford, for the data-disc displays on a decsystem-10. a previewer running under vax/vms was mentioned in the first 1986 issue. The last 1986 issue announced previewing on an early macintosh, complete with a screenshot -- just look at those jaggy edges!

Times have certainly changed.

  • 15
    to the person who capitalized my entry, i object, but i'll leave it. Commented Apr 30, 2013 at 16:27
  • Now, this is pretty much what I had hoped for (also, I was expecting it to come from barbara). It's hard to imagine getting a more authoritative answer from someone who's been in the business longer, unless DEK himself drops in (and even then, I'm not sure).
    – Ryan Reich
    Commented Apr 30, 2013 at 19:58
  • 5
    @barbarabeeton I'm the one who capitalised your entry. Sorry, I didn't know of that policy of yours until Speravir pointed it out to me :) I won't do it again.
    – jub0bs
    Commented May 2, 2013 at 8:16
  • Another corroboration for some of this: The earlier version of The TeXbook (though not called that at the time), namely the 1979 book "TeX and METAFONT: New Directions in Typesetting" has the following in Part 2 (i.e. the TeX manual) on page 198 which is Appendix S ("Special notes about using TeX at Stanford"): "2. The standard TeX program produces output for the XGP. To produce output for the Alphatype (when it is available) we will use another program “texa”." Commented Jan 8, 2017 at 8:52
  • 1
    @ShreevatsaR -- thanks for this. there was actually an earlier tex manual, a small yellow and green book; i'll have to see if anything is mentioned there (don't remember, but it's definitely a demagnification from 120 or 130%). Commented Jan 8, 2017 at 9:13

At a beginning I was using TeX on PC XT without HDD and with the Hercules graphics. The difference between on-screen visualization (not more than 70 dpi, inversion of colours) and the print on the printer giving near 240 dpi was very visible.

Alas, the printing time was not in pages per minute but rather in minutes per page. An extreme case: about 30 min/page. It sounds like a joke, but it isn't.

  • printing time in minutes per page? indeed, but for quite a long time the processing time could be in minutes per page as well. go out for coffee, take a nap, ... that's one good reason why tex was run in interactive mode, so (if feasible) a trivial error could be corrected on-the-fly, rather than restarting the whole job. (harder in latex than in plain tex though, what with the need to match \begin ... \end.) Commented May 1, 2013 at 20:28
  • @barbarabeeton Yes, I remember: working on two or three pages, many \input's, and compilation of the whole document at the and of a day, giving a time for preparing and eating a supper. The printing time was worse, because printers were terribly noisy. And two hours of probably 85 dB during printing was horrible. Commented May 1, 2013 at 20:37
  • Was that in TeX distribution for DOS in the XT? What editor did you use?
    – alfC
    Commented May 2, 2013 at 0:29
  • 2
    @alfC Yes, it was a TeX distribution for DOS. It was prepared by me to fit a working part to 1.2MB disquette. And an editor was the Norton Editor (about 34 kB). Commented May 3, 2013 at 12:15

I bought an Atari ST in 1985 which had a 640x400 monochrome screen. Based on the purchase date of Lamport's LaTeX book I started using TeX on it that year. The onscreen preview was OK, but needed scrolling and zooming for checking things in detail because of the Atari ST screen resolution.

That however was way better than printing things out on my 9-pin Star SG-10 printer and reviewing that. The printer would go over each line 3 times before advancing and need more than one such pass for a line of text. Resulting in minutes/page out (around 10 IIRC) instead of pages/minute. For a friend's thesis I let that run overnight and went to sleep somewhere else.

  • 6
    Welcome to TeX.sx! Thanks for your answer. Feel free to elaborate and perhaps post a picture of the beast.
    – jub0bs
    Commented Apr 21, 2013 at 13:50
  • @Jubobs Sorry, I somehow missed that the end of my edit did not show up. I will search for pictures, so far nothing but general (non-Latex) shots.
    – Anthon
    Commented Apr 22, 2013 at 4:19

I used a GP 29 terminal that would preview TeX. That was in 1987.

GP-29 terminal (from New Products, IEEE Computer Graphics and Applications, November 1984)

  • I have no idea what a GP 29 terminal is, could you elaborate on that a bit? Perhaps some details on the part of the question that says “Was the quality good enough to get a good idea of what a hard copy would look like?”?
    – doncherry
    Commented Apr 21, 2013 at 1:01
  • It was a graphical terminal, unlike the "dumb termials" that were basically vt100s. It could display a TeX file with a reasonable degree of rez to the screen. Commented Apr 21, 2013 at 1:03
  • I was still using an Informer terminal like this in the mid 90's. Hooked to a modem, connected to the OSU dialup service. In fact, I wrote my dissertation on it. Does anybody know when dvi2vdu first appered? The dvgt README file written by Geoffrey Tobin in 1993 only says "It was invented long ago by Andrew Trevorrow ..." Commented Apr 30, 2013 at 20:05

Awesome contents that worth reading:

  1. Thoughts on TUG 2010 by Barbara Beeton, regarding the environments in 80's for TeX;
  2. The Design of TeX and METAFONT: A Retrospective by Nelson H.F. Beebe, and
  3. TeX82, Version 1.0 for Berkeley Unix, 4.1 and 4.2 bsd, an email from Richard Furuta to announce the availability of TeX82, Version 1.0, for Berkeley Unix, version 4.1 bsd and 4.2 bsd.:

enter image description here

  • 1
    The third one comes closest to directly answering the question. It seems (because of lack of mention of alternatives) that they really did print everything in 1982; no graphical option is mentioned. I'd be curious for direct confirmation.
    – Ryan Reich
    Commented Apr 29, 2013 at 17:02
  • @RyanReich I think you are right. From TeX 1L 12/4/83, available at math.utah.edu/cgi-bin/man2html.cgi?/usr/local/man/man1/… we can see: The output DVI file is written on name.dvi where name is the jobname. A log of error messages goes into name.log. Note that there have been incompatible changes in the DVI format between TeX78 and TeX82, so programs used to print TeX78 output will not work for TeX82. A number of output drivers are available. Ask your local TeX guru for information on what one you should use.
    – user13907
    Commented Apr 29, 2013 at 18:30
  • The correct link above is math.utah.edu/cgi-bin/man2html.cgi?/usr/local/man/man1/…
    – user13907
    Commented Apr 29, 2013 at 22:46

Note that the very first version of TeX - TeX78 - was written to allow for better typesetting of Knuths books - especially the math heavy bits - not as such for general consumption. The rewrite into TeX82 (along with MetaFont) was as literate programs (where you write an essay on your program containing snippets of actual code which was then assembled into the final program - this was due to the strictness of the pascal compiler).

The original programs used by Knuth is unknown to me but most likely something like rendering DVI files to something which could be manually inspected - and also the typesetter needed to generate proofs for proofreading. At the time I learned about TeX (late 80'es) we used xdvi to inspect output on screen (without figures and a lot of zoom to overcome the low display resolution, it was a bliss when we found out about anti-aliasing on 256-color devices) and dvips to generate final PostScript. dvi2tty could be used if you only had access to a character based terminal.

As a side note I would like to mention that a long while back I read an article where Knuth said something along the lines of that if Postscript had been available to him at the time he probably would have used that instead of DVI (or perhaps MetaFont). Unfortunately I cannot locate said article again, so my memory may be wrong. Note that this did not need to be a complete typeset document for printing, but perhaps just font descriptions (for rendering into DVI) or similar.


Guy Steele ported TeX from Stanford's WAITS operating system to ITS in 1978. Just like Stanford, the MIT port would output on an XGP printer. Daniel Weinreb wrote a program called XD to preview XGP files. It displayed on the bitmapped 576x454 monochrome Knight TV terminals.

Source: PDP-10 backup tapes from MIT.


In 1986, I used the Atari ST with its TOS system for LaTeX. Don't know much more than that anymore, but I used the ST for my PhD, and there was a clear and beautiful previewer, as the monitor was all B&W, and very sharp. The OS was a rip-off from the Mac of course. Some of the editors, Tempus, were bloody fast

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .