A few years ago there was a discussion here about how to facilitate IPA entry. A user kindly provided this stack app to provide a fly-out keyboard, and in addition there are lots of online character pickers, such as

as well as the SIL IPA keyboards that can be installed at the OS level. So to my mind the issue of how to get Unicode-encoded IPA symbols into questions and answers seems to be adequately addressed.

What is inadequate, to my mind, is the typography of IPA characters on the site. It was suggested here that IPA be rendered using the existing code formatting mechanism (surrounding all IPA text with backticks), and instinctively that is what I had been doing anyway, even before seeing that suggestion. It works well for isolated glyphs, but is pretty terrible even for the most common diacritics. For example: (nasalization), (creaky phonation), (devoicing), (breathy phonation), (dental). It's even worse when you try to use multiple diacritics, e.g.ː n̠̥ (retracted + devoiced).

I think this issue can be solved by changing the style sheet for Linguistics SE, so that either:

  1. text enclosed in backticks is rendered with the Doulos SIL webfont or a similarly capable typeface, or
  2. text not enclosed in backticks is rendered with the Charis SIL webfont (Charis because it has bold, italic, and bold+italic faces, so it can handle the markdown formatting).

If option 2 is chosen, then obviously the convention would be to not use backticks for IPA text, in which case it could optionally be delimited in the usual ways (i.e., [square brackets] for phonetic transcriptions, /slash brackets/ for phonemic transcriptions, <angle brackets> for explicit reference to a glyph and/or its orthographic use).


Here is a screenshot of what I see on Firefox 37.0.1 on Linux:

Screenshot of diacritic rendering

Here's a second screenshot of diacritics in one of the comments (with a little shading added to make more clear the relevant parts):

Another screenshot of diacritic rendering

Presumably that devoicing ring is supposed to be on the final a, not on the e in "etc".

  • Uhm interesting. I tagged your question featured so it gets more attention.
    – Alenanno
    Apr 25, 2015 at 8:25
  • Are your examples ã, a̤, a̰, ḁ etc instances of "pretty terrible". They look fine to me, ignoring the generally loathsome sans serif font (which seems to be popular). Diacritic stacking is a problem, but at least under magnification one can see what ã́ is, and that it is distinct from á̃.
    – user6726
    Apr 25, 2015 at 16:49
  • @user6726, added a screenshot to clarify why I think the rendering is bad. I suppose it could be Firefox's fault...?
    – drammock
    Apr 26, 2015 at 1:45
  • 1
    This should be tagged as bug, not feature-request.
    – curiousdannii Mod
    Apr 27, 2015 at 0:23
  • I've raised this as a bug on meta.stackexchange.
    – drammock
    Apr 27, 2015 at 2:05
  • With Chrome on my worn out old netbook running Windows 7 starter it all renders great. This is in fact one of the reasons I continue to run Windows despite being an open source kinda guy. *nix has been behind in font rendering for decades, despite continued work on the problem. As a language enthusiast the need for good typography outweighs the need for an open OS, for me anyway. May 24, 2015 at 13:14
  • 1
    @hippietrail at least for this particular case, the font rendering issue hinges on which typefaces are installed, and should be unaffected by OS. As I mentioned in my comment to prash's answer, hacking the site's CSS to change the preferred font order makes the problem go away.
    – drammock
    May 26, 2015 at 11:54
  • @drammock: Stack Exchange should be able to change the site CSS for us. We should open a new QA with 'bug' instead of feature request asking for just this change. May 28, 2015 at 14:51
  • @hippietrail I already did that here and mentioned it in a prior comment.
    – drammock
    May 29, 2015 at 8:02
  • @drammock: Cool. Sorry I had Internet and browser problems for days while I was trying to respond so I think it got stale by the time it went through. May 29, 2015 at 15:34

3 Answers 3


I too browse from Linux, with both, Firefox and Chrome. My own rendering is not as bad as it is for you. Only the last two, t̪ n̠̥, are rendered wrongly for me. enter image description here

To the right, you see a snapshot of the CSS for code.

When I modify the CSS options manually, to use my own computer's default monospace font, it gets better.

enter image description here

I am against the idea of using anything but monospace fonts for our code sections, because we sometimes need monospace text. Monospace text is not only useful for typing code and pseudocode, it is also useful for preformatted text.

Alternatively, I propose:

  1. The admins should change the preferred fonts for code. I tested with Ubuntu Mono, Roboto Slab, and DejaVu Sans Mono, and it looks fine with each.
  2. If the above change is not feasible for some reason, users should stop using code for IPA. So, just write it as [ãa̰ḁa̤t̪n̠̥], [ãa̰ḁa̤t̪n̠̥], /ãa̰ḁa̤t̪n̠̥/, or /ãa̰ḁa̤t̪n̠̥/ and let the existing selections for body handle its rendering.
  • brilliant! I had forgotten about the developer tools in Firefox. By deleting the fonts in the font-family definition for code one by one, I discovered that the terrible rendering is due to the use of Liberation Mono (I don't have installed any of the typefaces earlier in the list). Elevating DejaVu Sans Mono higher in the list (or demoting Liberation Mono) would count as a solution in my book. This is much better, as I agree that it's important to have monospace as an option.
    – drammock
    Apr 26, 2015 at 10:06
  • 3
    Users just shouldn't be using backticks for IPA.
    – curiousdannii Mod
    Apr 26, 2015 at 13:31

You shouldn't be using the backticks for IPA transcriptions. They're not code. They shouldn't be written in monospaced type. Linguistics papers don't present IPA transcriptions in a different font. They present them in slashes (//) or brackets ([]). Now there could still be a font problem with those, but I haven't seen one before.

  • 1
    Users of this site seem to be a mix of professional and armchair linguists, language teachers and learners, and the curious. I don't think it's realistic to expect them all to know or understand the difference between phonetic and phonemic transcription. Code formatting is a convenient and effective way of separating object language from metalanguage in cases where the -etic/-emic distinction is irrelevant or tangential, and a convention new users can easily emulate (it's documented). Also see my second screenshot; the problem doesn't go away if you stop using backticks.
    – drammock
    Apr 27, 2015 at 0:19
  • 2
    No professional linguist would naturally think of putting them in code blocks, and no armchair linguist would either unless it was suggested to them. IPA transcripts don't need to be separated! If they still break without the back ticks, then you should report it as a bug on Meta Stack Exchange.
    – curiousdannii Mod
    Apr 27, 2015 at 0:22
  • 1
    I am a professional linguist, and as I said in my question, I quite naturally gravitated to the backticks as a good way of separating object- from meta-language in cases where the phonetic/phonemic distinction was irrelevant.
    – drammock
    Apr 27, 2015 at 0:26
  • @drammock What do you meant object- and meta-language?
    – curiousdannii Mod
    Apr 27, 2015 at 0:36
  • 1
    The object language is the language being described; the metalanguage is the language used to do the describing. On this site the metalanguage is almost always English, but the object language varies. Using formatting to separate the two is, I think, uncontroversial: [ ] and / / serve this purpose, but so does ILG, or setting object language orthography in italics. You seem to strongly believe this site should emulate linguistics journals; can you tell us why? Why is monospacing inappropriate for IPA?
    – drammock
    Apr 27, 2015 at 1:45
  • I've raised a question about proper use of backticks on meta.stackexchange. Please weigh in on that question if you want to explain why you think monospacing is inappropriate for IPA transcriptions of speech.
    – drammock
    Apr 27, 2015 at 2:49
  • @curiousdannii: I like using a typewriter font for IPA because some letters/numbers/symbols can end up looking alike in a proportional font. Your arguments about not using code blocks should really be in the form of filing a bug report on Stack Exchange Meta to remove that formatting option from every Stack Exchange site except the few dealing with computer code. Otherwise it seems apt for each site to find some kind of analogous use for it. May 24, 2015 at 12:50
  • @drammock: But are you a true professional linguist? And are you a Scotsman? (-; May 24, 2015 at 12:51

Aha. I hate to point fingers... I sort of recognize the problem. Here is my screen capture of your original examples:


It's an interaction between font and software, on user's system. Some fonts provide spacing metrics for combining characters "correctly", and some do not (i.e. Open Type vs. True Type). But the software also needs to be up to speed. So for example Word 2003 handles OTF fonts almost correctly, and Powerpoint 2003 does not (I guess Powerpoint is generally one generation behind Word). I've had similar problems where Word 2010 documents appear to me (using W 2003) as "Vietnamese" and combining characters -- in some fonts -- were totally wrong (combining grave was ì and combining acute was Í or vice versa), but the same file in Open Office same era came out correct. The general rule is: apparently nobody understands how to deal with combining characters.

I don't have a "practical" solution for you, other than experiment with changing everything you can imagine, like browsers and fonts. Or maybe there's a system-level compu tweek that's way above my pay-grade. I actually think this would be worth reproducing on Stackexchange meta, because this place is full of compu wizards.

  • I've investigated text rendering in the past and there are many variables. The font. The font format. The featureset the font format has. The ability of the font creator to make best use of that featureset. Whether the font editor they use can make use of the full featureset. The Operating system and its "rendering engine". Rendering engines are things such as Uniscribe on Windows, Pango on *nix, Graphite, HarfBuzz, ATSUI, WorldScript, etc. Some browsers may even provide some extra layer. I know Firefox on some OSes used to have something like its own rendering engine. Very complicated ... May 24, 2015 at 12:59

You must log in to answer this question.

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