The LaTeX Fetish. Or: Choose Your Tools Wisely

Show Sidebar

Today I stumbled over a tweet mentioning an article where the author discusses reasons not to use LaTeX as an author:

The LaTeX fetish (Or: Don’t write in LaTeX! It’s just for typesetting) https://t.co/7vo0fsKFmk (always good to ask: 'why do we use this?')

— Wilfred Hughes (@_wilfredh) August 26, 2017

I never came across a more elaborated LaTeX critique which came to totally different conclusions I came to.

As we shall see, arguments in favour of writing in LaTeX are unpersuasive on a rational level: LaTeX is in fact quite bad for writing in (although it could be worse, i.e. it could be TeX). This doesn’t mean that people shouldn’t use LaTeX at all, but it does mean that people probably ought to stop recommending it as a writing tool.

What a statement. Let's have a closer look.

LaTeX Source Versus Output

I guess the main difference is that I do think that LaTeX is one of the best markup/environment to work with a text as an author if chosen deliberately. The author of the blog article derived perfectly clear that LaTeX source is not optimized for reading. While this is absolutely true but misses the point totally in my opinion.

When I am writing a book, I really don't want to see the output format as in WYSIWYG or even as in LyX. What I want to see is all my notes which are not part of the result. Further more, I do want to see for example reference definitions I am comfortably with and not the reference format of the end result which is totally irrelevant in the writing phase.

For example, when I choose to name a book reference "ProfJohnsPhD" instead of the not very specific "Johns1994", please let me see the definition I am comfortable with: this is the PhD thesis of Professor John and not something I would like to mix up like a paper of him from the same year. This is a big advantage, where the articles author argues, that this is a disadvantage. I, therefore, totally disagree with the authors reasoning.

As of the time of writing, the Wikipedia page on LaTeX, for example, says this: ‘LaTeX follows the design philosophy of separating presentation from content, so that authors can focus on the content of what they are writing without attending simultaneously to its visual appearance.’ (Wikipedia 2016, para. 7)


What I eventually realised was that while the argument is (as noted above) widely repeated, it is also wrong.

I honestly can not follow the argument of the author since he was not able to explain this to me later on.

To cut to the chase, LaTeX documents are very hard to read until typeset, which is inefficient for both writing and editing. This is a point that programmers ought to understand: if the readability of code is important, then so is the readability of text.

As I wrote above, I can not follow the author here. On the contrary. I do find it useful to have notes, my own style of naming references, my own style of naming link targets or even the possibility to name a link target in the first place.

What usability do I get with "see page 7" in contrast to see page\pageref{sec:Introduction}? As an author, I prefer the explicit one in any case. And not only to prevent errors: page 7 could be the wrong link target in this case. I'd have to go on page 7 and check it myself every time. Same holds for references to figures. I prefer see figure \ref{fig:GoldenCut} over "see figure 3". What a horror scenario.

Never mind boilerplate code like \documentclass{article} or \begin{document} at the start of your document – running into something along the lines of \parencite[706]{lena_peterson_2008} in the middle of a paragraph and having to mentally parse it into ‘(Lena and Peterson 2008, p. 706)’ interrupts your train of thought and makes it harder to do what you really need to be doing: reading your punctuated words back to yourself to make sure that they ‘sound’ like what you really wanted to say and don’t have any mistakes in them.

When I really do want to check the end result as the author desires, I do it with an on-screen representation of the resulting PDF file. And no: I usually don't want to read ‘(Lena and Peterson 2008, p. 706)’ for the reasons stated above.

What the markup is doing is good – unambiguously identifying the section heading as a section heading will help us later (e.g. when the designer wants to apply some particular style to all the section headings) – but doing it through markup (as opposed to doing it in some other way) is disruptive of the text for human readers (including editors and the original author): looking at the screenshot, we see the text mixed up with a lot of symbols that are not part of the text, and it is up to us to figure out what words and punctuation marks the eventual reader will see.

I object. When authoring a text, I do not have to check what the reader is going to see. I am defining what I mean and later on, LaTeX is translating my wishes to a suitable form according to the template definitions.

Besides, any decent LaTeX environment offers syntax highlighting which helps the author to read the content despite the markup characters in between. As shown in the screenshot, the author did not manage to set up AucTeX accordingly. Which is a pity.

By the way, one of the words in the screenshot is not right. It’s a typing error that I deliberately inserted. I know where it is because I put it there, but looking for it is hurting my eyes.

Just like above, the author was also not able to enable one of the various packets that offer spell checking in Emacs. This can never be an argument against or for LaTeX. The author compares his inability of activating spell checking in his editor with the activated default spell checker of LibreOffice Writer.

The author now follows the same reasoning he is criticizing when mentioning the MIT studies I mention in the section below. What a big mistake.


The MIT Research Science Institute argument isn’t much better. If you look at the three comparisons carefully, what’s actually being contrasted is not LaTeX and word processors, but the effective use of LaTeX and the naive misuse of word processors: all three things that the Research Science Institute staff tell their students can be done with the wonderful LaTeX can in fact be done perfectly well with a modern word processor.

I have to admit, that I did not read the papers yet.

If the author describes the content of the papers correctly, he is absolutely right. This seems to be a scientific mistake. You can not compare people using WYSIWYG tools without being educated on (like most people) with people using LaTeX after a certain training.

Comparing good use of LaTeX with poor use of word processors is unfair; the most that can really be said is that you are more likely to be introduced to LaTeX in a class taught by someone who really knows how to use it, and more likely to be introduced to a word processor by playing around with it or under the informal instruction of someone who doesn’t understand it very well, and that, for this reason, the number of people who use LaTeX but don’t use its document-structuring features is probably close to zero while the number of people who use word processors and don’t is enormous.

Well, this is not valid reasoning, in my opinion. It is no one's fault that people who write scientific papers and books are more likely to learn LaTeX and have a preference for it. This does not prove LaTeX to be more or less suitable for anybody. It simply proves that LaTeX seems to be preferred by professional authors. No more, no less.

Document Templates

For example, the last three universities I have taught at all have formal specifications for the formatting of student work – formal specifications that closely resemble the default settings on popular word processors. When it comes to stopping people from creating documents in purple 28pt Comic Sans, teaching them all to use LaTeX is a lot less efficient than stating that you will refuse to read anything that doesn’t match the style guide. (Teaching them to use word processors properly might also help.)

This emphasizes my point of view. Badly designed (or defined) template definitions are no argument for switching tools, independent if we are talking about WYSIWYG or LaTeX.

I didn’t use [LaTeX] to typeset the public report from the Valuing Electronic Music project, for example, because that was a public report and I didn’t want it to look like a conference paper.

Well, then you probably should have used a suitable template which did not look like a conference paper. Here, the author argues as if he has no clue how to use different LaTeX templates for different purposes.

He further compares a one-page document with default LaTeX output with the same content produced with LibreOffice Writer. He further seems to suggest that someone should use LaTeX or Writer according to the match of the default output. This is a very worrying argument. As if LaTeX nor Writer are not able to produce different styles of documents.


LaTeX is based on the idea that it is better to leave document design to document designers, and to let authors get on with writing documents.
Think about this for a moment. If ‘How can we let authors get on with writing documents?’ is the question, can LaTeX really be the answer? LaTeX does less to prevent authors from getting on with writing documents than TeX does. But if neither of the two existed, and you had to come up with something, right now, in 2016 – would it really be a markup language?

The author is right here: it is a valid discussion whether or not LaTeX is doing a good job of preventing the author of tempering in areas she has no knowledge of such as basic parts of the preamble.

Any "normal" LaTeX author should not create preambles. Never. This is a rather sophisticated task where you ought to have a deeper knowledge of typography. IMO, a LaTeX author should stick to a well crafted LaTeX template (sadly not that there are many of them) and concentrate on the text she is writing. Download a template and start with the content. Do not mess around with choosing non-matching font families or messing up page areas without knowing anything at all what one ninth or something golden has to do with it. This is not something an average author needs to know. For the same reason, the average author should not modify those basic formatting things.

I once tried to come up with a good LaTeX template offering good usability to the average author. Go and judge my result on GitHub if you like. Pull requests are very much encouraged.

Authoring Environment

Well suited working environments such as the mentioned AucTeX/Emacs support the author so that she is able to add markup stuff as easy as possible. With AucTeX/Emacs, I don't have to type heading commands, list item markup or other things. It is just a matter of learning your tool. Of course, this is nothing for the occasional author writing with LaTeX. This is a tool for somebody whose daily job description contains authoring text.

As stated above, the article author was not able to set up a proper LaTeX environment in his editor of choice.


The outcome of LaTeX is a very good legible PDF document, which is designed to be printed on paper. There are tons of research results that prove that good typography is nothing for the consciousness exclusively. It is part for your conscious eye but also part of the unconscious things in the head of a reader.

Most of the texts I produce (if not forced to write in MS Word) are produced using Emacs/Org-mode with its PDF export function which itself uses LaTeX in-between. This is for convenience, especially because I've got all of my data in Org-mode format.

However, when a decent PDF output is necessary, authors should stick to LaTeX.

Please: you do want to use typefaces other than Computer Modern. Computer Modern is not some sort of universal, all-purpose typeface. It’s just the digital version of the typeface that happened to have been used for the first edition of the book whose second edition Knuth created TeX in order to typeset, and it really isn’t suited to some of the uses to which I’ve seen it put, especially in slide presentations.

He is right, that the font was created and used for Don's books. If you happen to mis-use a serif font like CM for slide presentations, you definitely have to learn a bit about very basic typography decisions. Or learn to choose better templates. Either way.

Installation and Storage Space

Free and open source software has a strong tendency towards being difficult to install and get up and running. TeX and LaTeX are no exception. Also, if you want to do anything really wild and crazy – like using a typeface other than Computer Modern – then plain vanilla TeX and LaTeX won’t do.

Oh come on. Really? This argument is pointless on so many levels I won't even try to mention them all.

Tex Live wraps up almost everything that a TeX or LaTeX user could ever possibly want into a single, handy download. A single, handy unbelievably large download that takes up over two gigabytes on disk: to be exact, 2.4 gigabytes for the Mac version, MacTeX. For comparison, LibreOffice takes up about one and a half gigabytes of disk space on a Windows or Linux machine, and less than one gigabyte on a Mac. And LibreOffice is a word processor, a spreadsheet, a slideshow presentation program, a drawing package, and a database, all in one. LaTeX is just a typesetting program.

This rather large package you installed on your Mac is not "just a typesetting program". On the contrary.

It contains TeX Live which itself contains the typesetting tools, a whole type management subsystem, METAFONT, tons of fonts in different resolutions and formats, installer for various operating systems such as MacTeX and enhanced proTeXt, and hundreds of extension packages of the CTAN archive. You get thousands of much different software packages additional to "just a typesetting program" that I do have to come to the conclusion that you did not even care to read the documentation you also installed. Not even the release notes. Not even a small Wikipedia page about it.

And even if you decide to weigh your hard drive down with TeX Live, you still have a lot of work to do in getting things to work properly, and almost nothing ever seems to be clearly explained. Sometimes it feels as if getting LaTeX to work has become a sort of hazing ritual through which the pledge must suffer alone.

Yes, did not even care to read the documentation before whining.

You, Sir, have no clue what you have installed on your computer. You did not care to read its documentation but complain on things not working which are explained in the documentation. This does not produce a nice picture of your ability or motivation to understand the tools you are using.

Compilation Errors

The author is quoting:

‘I get “There were undefined references” errors and [I] can’t fix it after two days of trying. I have tried switching editors from Sublime Text 3 to TeXStudio on a Mac, then trying both on a PC. I am willing to try anything at this point. …
I have read about doing a compilation trick but I’m not sure how to do this in either SublimeText or TexStudio. … I have run into many problems and taken many detours that led to other problems. I’m at a loss. Can someone please give me a few hints or keywords I can search for to fix these problems, or a complete solution? I can’t even get a minimum working example up. I will install anything.
(user2205916 2014, paras 1 & 9)

Well, this user is obviously new to LaTeX. Switching the editor does not effect the underlying error at all. This is just like your car breaks down and you try to fix it by opening the side-windows.

He then continues:

Every time somebody new tries to get started with LaTeX, that person is set up for hours or even days of this kind of thing – plus a lifetime of fiddling with TeX and LaTeX’s quirks [...]

In my opinion, this happens most of the time when the user is not using a good template, or the user tries to modify the template without knowledge of what she is doing, or when packages are used but their documentation is not read.

Another error, another helpful hint he found which actually solved his problem:

Well, how silly of the would-be LaTeX user not to have realised that the only way to get LaTeX running properly with Biber was to run it twice, once before and once after!

I am so sorry. But this is very basic knowledge for anybody using references and LaTeX. I'll explain: The first LaTeX run generates meta-data which is used by biber/BibTeX or even references to pages, headings or figures. The second follow-up LaTeX run generates the ready to read PDF file. If you have compiled at least one document which is using references, you ought to learn that in your first LaTeX course.

This is not intuitive, yes. But it is mentioned in every LaTeX introduction I know of within the first lines or pages.

When Not To Use LaTeX

The author suggested to use tools like Org-mode, Abiword, Pandoc or Markdown to author a text which is then processed using LaTeX.

While I do use Org-mode to LaTeX to PDF myself a lot, this is not a substitution of a hand-crafted LaTeX document when it comes to output quality.

In this blog article I argued about this notion.

There are good reasons NOT to use LaTeX as well. If you have to generate ebook in EPUB format or HTML files, you better should stick to a format that is a bit more abstract as LaTeX. Org-mode or even Markdown for example.

If you're an occasional author, please do stick to an easy to learn environment like LibreOffice Writer or MS Word. It gets the job done with a mediocre but easy to accomplish result. Research even showed that people are beginning to develop better legibility for bad output in contrast to sophisticated LaTeX PDF output. The reason is puzzling: people are exposed to so much bad typography that their inner eye somewhat gets irritated when they finally are reading a nice piece of work crafted by somebody who knows her job.

Comment via email or via Disqus comments below: