6

I am writing a document with XeLaTeX in which I use IPA characters in italicized text. After trying a few fonts with extensive IPA support like Gentium Plus, Brill or Charis SIL, I encounter the same problem: the vowels a and ɑ are given the same character in italics, even though they are differentiated in plain text.

While not a tragedy, this is both quite shocking and a little troubling. I assume that a number of people will also have come across this issue.

Is there a straightforward way in TeX to get actually different italic shapes for the two characters?

-- Edit for clarification:

One would think that losing the distinction between two characters is grave enough in and of itself. Still, it does have actual implications: you become completely unable to capture the distinction between words like 'Sam' /sam/ and 'psalm' /sɑm/ (in some English varieties).

Let's see it in action in a French gloss to realize that this can indeed be a big issue:

Le chien a mis sa patte dans ma pâte.    
/lə ʃjɛ̃ a mi sa pat dã ma pɑt/    
the dog has put its paw in my dough    
"The dog put its paw in my dough."

In italicized text (for many reasons: an italicized line in glosses, a transcribed word quoted in italics, etc.), because of the problem reported in this post the /pat/ - /pɑt/ distinction is completely lost.

Bernat Bardagil
  • 108
  • 9
  • For future reference, please always add a minimal working example (MWE). – Sverre Mar 17 '15 at 11:47
  • 1
    I think this is an extremely shocking issue, since the two symbols represent two different vowels. I had the same problem but unfortunately I'm afraid there is no straightforward solution. I opted not to use italics in transcriptions, but boldface or underline... Not the optimal solution at all... – Stefano Mar 17 '15 at 12:27
  • @Stefano: Why isn't my solution below satisfactory to you? And yes, most fonts lose the distinction between IPA [a] and [ɑ] when italicized, but IPA transcriptions should never be given in italics in the first place. I'm sure I have transcribed thousands of forms in IPA, but I have never come across this problem before - simply because I have never italicized an IPA transcription (nor do I think I have ever seen anyone else do it). – Sverre Mar 17 '15 at 13:54
  • 1
    @Sverre Whilst I think I agree with you that IPA transcriptions shouldn't be italicised, there's the matter of representing certain forms in the running text, which is standardly done by using italics. – Pavel Rudnev Mar 17 '15 at 14:01
  • @PavelRudnev Do you know of any such 'certain forms' that are not IPA transcriptions where the distinction between a and ɑ is meaningful? – Sverre Mar 17 '15 at 14:02
  • @Sverre That's beside the point. What matters is that for languages without a writing system that happen to make the distinction between a and ɑ there should be a better way of typesetting the distinction in the running text than using the slanted counterparts of upright characters. – Pavel Rudnev Mar 17 '15 at 14:08
  • @PavelRudnev But Unicode and fonts using it aren't meant to provide characters for languages that don't have a writing system (how could they?). For that purpose one uses IPA. And IPA shouldn't be used in italics. You shouldn't use IPA transcriptions for such a language and pretend you're representing its non-existent writing system. – Sverre Mar 17 '15 at 14:12
  • 4
    @Sverre I agree that you should neither italicise IPA transcriptions nor try and pass non-existent forms off as existing written forms. Nevertheless there are languages whose orthographies heavily rely on characters from the IPA. One example of such a language would be Mbembe, which also happens to encode the a - ɑ distinction in writing. See p. 12 of this spelling guide. – Pavel Rudnev Mar 17 '15 at 14:34
  • @PavelRudnev Very good - that is a very interesting case you found there (and precisely of the kind I was asking for). BUT do note that the distinction made in writing (according to the spelling guide you link to) is not between a and ɑ, but between a and α (i.e. a Greek alpha). It wouldn't surprise me if they chose α (alpha) over ɑ precisely because they knew that the distinction between the graphemes a and ɑ would typically be lost in italics. – Sverre Mar 17 '15 at 14:52
  • @Sverre Wouldn't surprise me either. And yes, I have seen the Greek alpha thing, but only after I wrote the comment. Also, fonts like LinuxLibertine and Junicode do make the distinction. – Pavel Rudnev Mar 17 '15 at 14:57
  • As the OP clarified in the comments to my answer, this problem has nothing to do with xetex, nor is the OP looking for a xetex solution. The OP is looking for an italic font that distinguishes between certain glyphs. The question is therefore not about TeX or friends, so I'm voting to close. – Sverre Mar 17 '15 at 15:14
  • To clarify: I was hoping that there would be something like an option to activate the differentiation of shapes for these characters in italics for fonts that don't do so out of the box. I absolutely did not neglect your contribution, I was just waiting in the hope of a solution involving actual italics. – Bernat Bardagil Mar 17 '15 at 15:20
  • 2
    there are certain other possible instances where italic ipa is desirable: if a journal style specifies italic section headings, it would be appropriate to use italic ipa forms if a transcription makes sense in the heading text. i thought some of the sil fonts did provide these shapes, for just that reason. – barbara beeton Mar 17 '15 at 15:25
  • @barbarabeeton Actually, no. Even if the headings in an article should be in italics, IPA transcriptions would be exempted. To make that even clearer, consider the common case where headings are in small caps. It would be absurd to put IPA transcriptions in such headings in small caps. The most popular and commonly recommended font for IPA transcriptions, SIL Doulos, doesn't even come in italics. – Sverre Mar 17 '15 at 15:29
  • 1
    @Sverre Quick note: the glyph shaped like Greek alpha (used in African orthographies) and the glyph shaped like lowercase script A (used in IPA) do in fact share the same Unicode character. – hftf Mar 17 '15 at 16:01
  • @Sverre -- okay, thanks. (it's been a long time since i seriously read a linguistics journal. but what if an ipa letter is a full cap -- that looks ridiculous in a small caps environment. for math, it's the reason small caps aren't used for running heads; instead, "smaller" full caps are used, since authors aren't disciplined enough to avoid using math notation in titles.) – barbara beeton Mar 17 '15 at 16:03
  • @hftf That's very interesting! It's somewhat unclear to me now what the language Mbembe is actually using, because the spelling guide does use the Greek alpha (U+03B1), not the Latin alpha (U+0251). – Sverre Mar 17 '15 at 16:07
  • @barbarabeeton There are no full size capital letters in the IPA. – Sverre Mar 17 '15 at 16:09
  • @Sverre It's possible they simply did this because they had no way to achieve the desired shape with ɑ - they use Ɑ as the uppercase (with nothing done to change its shape) - apparently the proper shape in both cases is that of the greek alpha, but it should be encoded with the latin alpha character (uppercase greek alpha, of course, has an entirely inappropriate shape for this purpose) http://en.wikipedia.org/wiki/Latin_alpha – Random832 Mar 17 '15 at 16:39
  • 1
    @Sverre I believe the orthography uses Latin alpha, but the authors opted for Greek alpha because in the font used, its glyph looks more correct to readers. Indeed, for a while, African orthographies have notably gotten the short end of the stick in digital representations; have a look at 1, 2, etc. (Another interesting fact: IPA uses both Latin and Greek script characters.) – hftf Mar 17 '15 at 16:49
  • 1
    I can offer you a solution via older tipa package: \documentclass[a4paper]{article} \usepackage{tipa} \begin{document} \textipa{aA} \textit{\textipa{aA}} \end{document} – Malipivo Mar 17 '15 at 17:15

3 Answers3

10

I recommend the well-equipped-for-linguistics Brill font with Stylistic Set 20 (SS20) enabled:

\documentclass{article}
\usepackage{fontspec}
\setmainfont{Brill}

\begin{document}
\begin{tabular}{l l}
upright       &                                           a ɑ \\
italic        &                                  \itshape a ɑ \\
italic + SS20 & \addfontfeature{StylisticSet=20} \itshape a ɑ \\
\end{tabular}
\end{document}

Note that the SS20 option also changes the shapes of several other characters (in particular, β θ λ χ a f g) to suit linguistic documents.

To see it in action, take a look at this TikZ/PGF linguistics vowel chart typeset in Brill.

I do still concur with the other users in discouraging IPA set in italics, even (as mentioned in a comment above) in headings.

Community
  • 1
hftf
  • 2,283
  • 1
    This would also be a great solution for a language like Mbembe, as mentioned by @PavelRudnev in a comment. But as you also say, highly unrecommended for IPA transcriptions. – Sverre Mar 17 '15 at 15:36
  • IRT using IPA in italics, I beg to differ. Sometimes there's no straightforward alternative: http://oi58.tinypic.com/iz3kg6.jpg

    (example from Salanova 2006. The sense of Mebengokre nominalizations. WSCLA 11 Proceedings.)

     – Bernat Bardagil Mar 17 '15 at 15:37
  • 1
    @BernatBardagil That case you link to should have had the constituent within brackets [] in an upright font. Just because someone once made a poor choice in a document doesn't mean you should too. – Sverre Mar 17 '15 at 15:41
  • 1
    Though I agree with @Sverre above me, Brill is still an ideal font for linguistic (and other specialist) documents, and looks great even when making questionable typographical choices. – hftf Mar 17 '15 at 15:46
  • 2
    @hftf's solution also works with the SIL fonts, only it's the ss05 set rather than Brill's ss20. – Pavel Rudnev Mar 17 '15 at 15:58
  • Does this g have a particular linguistic meaning? It's not in the IPA. – Random832 Mar 17 '15 at 16:16
  • @Random832 In many serif fonts, the italic variant of g (normal letter with two storeys) is written with one storey: g. Thus it would look too similar to the italic ɡ (IPA symbol with one storey): ɡ. The SS20 option makes both roman and italic g have two storeys; there is no change to roman or italic ɡ, which still both have one storey. – hftf Mar 17 '15 at 16:24
  • I don't know what context you would be using g in that could be confused with IPA, though. I also would think it unusual to see an italic two-story a in a non-IPA context. Are there any alphabets other than IPA that use both versions of either as distinct letters? EDIT: Apparently both a and ɑ are used in Cameroon, but ɑ should look more like Greek alpha in this context according to wikipedia. – Random832 Mar 17 '15 at 16:26
  • @Random832 I also don't know a context in which italic g and ɡ are distinguished, and was simply describing the ambiguity. Pavel Rudnev did mention above a non-IPA context distinguishing a and ɑ: the orthography of the African language Mbembe. Remember that these characters are not only used for IPA purposes: take a look at these uppercase forms. – hftf Mar 17 '15 at 16:33
  • @hftf as I noted in my edit Wikipedia says ɑ should look more like Greek alpha in this context (The real Greek alpha can't actually be used for this because the uppercase version needs to look like a bigger version whereas uppercase Greek alpha looks like A). – Random832 Mar 17 '15 at 16:35
  • @hftf You don't always get full agency in these choices, though. If you get something published at IJAL, for example, your glosses will be in italics. – Bernat Bardagil Mar 20 '15 at 10:48
  • 1
    Is this the style guide for the same IJAL? – hftf Mar 20 '15 at 12:19
  • 1
    Look at the bottom of that style guide: "Updated December 17, 2015" :) – Sverre Mar 20 '15 at 12:53
5

The straightforward solution would be to use a font that has italic shapes for both a and ɑ. LinuxLibertine and Junicode both seem to do the job.

Pavel Rudnev
  • 605
  • 1
    They're slightly different in Linux Libertine, but I think one would have great difficulties telling one from the other unless they are put right next to each other. – Sverre Mar 17 '15 at 15:03
  • It's a reason for optimism that some fonts actually provide italic forms for these two characters. Sadly that limits one's options quite a bit, but at least we do have some. – Bernat Bardagil Mar 17 '15 at 15:05
2

I don't agree that this is a problem or an issue, and it's definitely not shocking. Almost all roman fonts will change the character a to an ɑ when it's given in italics, see http://en.wikipedia.org/wiki/Italic_type#Examples.

Since you put IPA in the title and tagged it with linguistics, could you please add an example from linguistics and IPA transcriptions where the difference between an italic a and an italic ɑ is meaningful?

The simplest solution would be to slant the upright font. I wouldn't recommend doing it, of course -- not because it fails to fix your problem, but because it looks weird to see an a in italicized text, and because you shouldn't ever italicize IPA transcriptions in the first place.

\documentclass{article}
\usepackage{fontspec}
\setmainfont[
    SlantedFont={Gentium Plus}, % use upright font as slanted font
    SlantedFeatures={FakeSlant=0.15} % slant the upright font when used as slanted font
    ]{Gentium Plus}
\begin{document}
\textit{aɑ} \textsl{aɑ} 
\end{document}

enter image description here

Sverre
  • 20,729
  • I have added a clarification. I do agree with you though that the slanted a option is not an ideal solution. – Bernat Bardagil Mar 17 '15 at 13:53
  • @BernatBardagil But why doesn't my proposed solution fix your problem? I didn't say that slanting the a is a sub-optimal solution for your problem, what I meant was that I wouldn't recommend using an upright a in italic text. But that's what you want to do, so I provided a solution for you. – Sverre Mar 17 '15 at 13:55
  • There are dozens of cases in which someone will present and manipulate written linguistic data without the interference of a spelling system (and run into situations in which using italics will be required). The problem is that fonts that are meant to support IPA characters fail to do so in something as ordinary as italicized text. I am just shocked that this should be blatantly overlooked by fonts that otherwise do provide italicized versions of IPA-specific characters. – Bernat Bardagil Mar 17 '15 at 14:06
  • @BernatBardagil I still don't understand why my proposed solution doesn't fix your problem. Please explain how it fails, and then revise your original question accordingly to make it clearer. Also: (1) a isn't an "IPA-specific character". Most fonts will follow the standard and the tradition of italicizing it as an ɑ. (2) If people "manipulate" IPA transcriptions of a language to pretend they're writing the language and not transcribing it, then they're abusing the purpose of the IPA. I don't find it shocking that most fonts ignore that approach. – Sverre Mar 17 '15 at 14:15
  • 3
    I suppose that you don't see it as something meaningful beyond IPA. The thing is, it's as simple as this: we can manage to keep the ɔ - o distinction in italics in fonts with a broad enough support. Why do we seem doomed to lose the a - ɑ distinction in the exact same context? – Bernat Bardagil Mar 17 '15 at 14:32
  • @BernatBardagil The main purpose of this site is to ask questions about some problem and to provide solutions for it, not to have lengthy discussions. I have provided a solution for you, yet you do not accept it or comment on it. You only say it's not "ideal", without explaining why. So, could you please explain why my proposed solution doesn't fix your problem? – Sverre Mar 17 '15 at 14:54
  • It is a way to deal with this problem, albeit as I said it's less than ideal: it means that it turns out you cannot have proper italics for these two characters. I specifically narrowed the question down to fonts that allegedly provide ample IPA character support, and the contradiction that two characters are merged in their italicized shape, in the hopes that an alternative existed. – Bernat Bardagil Mar 17 '15 at 15:02
  • @BernatBardagil I assumed you asked this question on a TeX site because you wanted a TeX solution to a TeX problem. I provided a TeX solution for the fonts you specifically mentioned in your question. If your actual question was "please tell me about some fonts that encode this distinction", you should have asked that question explicitly. But that would be a question about fonts, which have nothing to do with TeX and your question would be closed. – Sverre Mar 17 '15 at 15:07