IPA Phonetic fonts encoding

peter bilak's picture

Hello, i am wondering how can i create a font that would function like standards IPA fonts - e.g. Stone IPA or Times. I have designed all necessary characters, correctly named them, created a custom FontLab encoding file (.enc) - based on Stone IPA, and tried genereting some PostScript Type 1 fonts. However the character mapping is different when typing on a Mac. When i try e.g. alt-w, i shoudl get reversed w, but i get something else instead. Does anyone has experience with phonetic fonts? How shoudl them be generated on a Mac? Any advice is greatly appreciated.

peter bilak's picture

Well, the above might be a difficult question, i haven't found answer for it myself, but here is another one:

Does anyone know when were the first phonetic fonts used? Was it at the time of photocomposition or punchcutting? I am just curious about some design decisions, and suspect that it was not always intentional.

John Hudson's picture

The first IPA fonts (not necessarily the first phonetic fonts per se) were metal. I'm not sure if any were cut by hand, or if even the first ones were made with the pantograph.

Victor Gaultney has written a paper that touches on this subject, which I believe will be published in the next issue of Typography Papers.

peter bilak's picture

As you can tell, I am working on a phonetic alphabet based on IPA standards.

Many of the characters in the IPA set are direct derivatives of characters of the Roman alphabet. For a designer it is tempting to reuse the already designed glyphs to save time. While rotating characters 180 degrees results in maintaining the correct contrast, flipped (mirrored) characters look inconsistent with the rest of the set.

application/pdfStone, Times and Fedra
correction_of_contrast.pdf (45.8 k)

The first IPA (Passy 1888) does not include any mirrored characters, but many rotated ones. There have been many subsequent revisions, and additions. Earlier phonetical alphabets use more accented characters (Lepsius 1863), later ones introduce many new glyphs (Pallum and Ladusaw 1986). 1993 revised IPA seems to be what we use today.

My question is whether the 'wrong-looking' contrast is a desirable effect of IPA. Should an 'ereversed' looks like a mirrored 'e', or should it work like a letter on its own. Is it necessary to relate 'Ksmallreversed' to a regular 'k'? If there were to be letters on their own, it would make sense to 'correct' their contrast. Please have a look at the attached document. The red ones look wrong to me, but maybe I am missing something. I would like to avoid replicating the already made mistakes and look at the intentions of the alphabet. Any advice is very welcome.

jfp's picture

I understand your concern about that, but the result you show is very disturbing. Don't take me wrong, your skills are good, just that reversed constrast of an oblique axis font such as Fedra looks odd. Terminal of reversed e suddenly heavy is very strange?

(point of view from a type designer, not phonetic specialist)

peter bilak's picture

I know it may look strange, I am working on the area where I have little experience, that's why I posted this. But at the moment I more disturbed by flipped characters. I don't know if this is a matter of habits, or a specific reasoning. I will also check with some linguists.

hrant's picture

This is an interesting issue. For one thing, I wonder why Stone (made by a chirographic designer) would have "incorrect" weight distribution - maybe somebody at ITC added those? Otherwise your hunch that it's supposed to look like that might be true.

One thing to consider is that the layman doesn't consciously care about "correct" strokes, so maybe it doesn't matter either way! In which case you should just save yourself the effort. The other thing to consider is that IPA is not meant for immersive reading, and in fact has horrible readability. This means that even if laymen do in fact mind "incorrect" weights in the subconscious, in the case of IPA that's moot.

So if you make the strokes "correct", you will be appealing to certain designers, and you won't be harming the reader. If you make the strokes "incorrect" you will be saving time, which you could say benefits the world more!

> Terminal of reversed e suddenly heavy is very strange?

This is a very good point, and causes an interesting observation: Even if perception of letterforms relies partly on the chirographic pattern of strokes, it certainly relies on symmetricity as well - especially in the consciousness (where IPA works).

And this brings on a VERY interesting question that I'd never thought of before: if symmetricity is valuable, does that mean that within the standard/conventional Latin alphabet itself there are places where "correct" stroke weight causes discomfort? Is this why fonts with vertical stress, or constructed fonts, have a certain appeal?

> I will also check with some linguists.

Just be careful: when gauging the layman, the questions you ask have to be very carefully considered -and presented- otherwise you will be guiding the answers. You have to carefully prod the consiousness to see what the subconscious is actually feeling.


I vote for "incorrect" stroke weight.


John Hudson's picture

The Stone IPA fonts were not made by Sumner Stone; they were made by John Renner at Adobe.

I disagree with Hrant about the stroke weight, but that's par for the course.

I recommend treating the rotated and flipped letters as individual letterforms, and design them not as flipped forms of other letters but in a manner to make them clear and readable forms in themselves. When I drew the Sylfaen IPA characters, I tried to balance the forms just as I would any other glyphs meant to work together. This is also the approach taken by Victor in his Gentium typeface, which I recommend taking a look at.

Here is a PDF of the Sylfaen IPA set, which has not been released. It was not intended for release, only to show IPA characters in a database, so it is not as polished as it would be if destined for release.

application/pdfSylfaen IPA
Sylfaen IPA.pdf (446.0 k)

Peter, when you say you are working on an IPA font, what is your actual intent: to support phonetic notation by linguists or to support the kind of subset of IPA often used in dictionaries. Full IPA support for linguists cannot be realistically done without Unicode and OpenType: the IPA alphabet is a generative system involving combining marks that need to be dynamically positioned. John Renner's Stone IPA tries to do as much as it can with kerning, but it is not very satisfactory: you need GPOS mark positioning to avoid clashes with extenders and in mark stacks. Ligation is needed to form word contour symbols.

Finally, it is useful to bear in mind that numerous IPA characters are also used in the orthographies of natural languages, especially in Africa, so even if your intended market for new typefaces is linguists or lexicographers it is not unlikely that people may use the fonts for setting normal text.

hrant's picture

Ah, but how shocking that there's a hint of disagreement with JF!! That's the second time in like a year. I guess every Coalition of the Willing has its cracks. ;-)

To be fair though, I think the distinctions you point out are indeed highly relevant. If this is for setting the occasional non-basic character in an otherwise plain-ASCII script, then it makes more sense to "integrate" the IPA glyphs. But if it's for setting "pure" IPA strings, it strikes me that the effort to give the glyphs "correct" strokes is better used somewhere more rewarding (to the user if not the ideological play of the designer). One other thing: the chirographic "rules" of stroke weight distribution are broken within the basic alphabetics (think of the "N" and "Z" for example), so why fret so much about IPA?

Sylfaen IPA:
1) The "lz" thing (in Lateral fricative) is too cramped.
2) Why not give the Velar-Approximant a full double serif on its descender? Since your "q" has that.
3) Why the plain descender on the beta thing (Bilabial-Fricative)?

BTW, #2 and #3 make me think that maybe IPA fonts are best when sans.


hrant's picture

One other thing: you should look at the "heritage" of the design at hand. Gentium is strongly chirographic, while Fedra is "synthetic", right? So they might benefit from different solutions.


John Hudson's picture

My recommendation to look at Gentium is mainly to see how Victor has handled the details of the 'reversed' and 'rotated' forms distinctly from the 'normal' forms. Consider, for example, the open o vowel: the IPA letter that many designers treat as simply a rotated c. In Gentium, and also in some older and less chirographic IPA typefaces, this letter has its own form, its own identity, largely independent of the form of the c. I think this is good and useful and enhances both legibility and readability.

Regarding the specific adjustment of contrast in Fedra, as shown in Peter's PDF, I think some of it works well (the flipped epsilon and the open o), but it is too stronly applied in the flipped e, and looks very wrong in the rotated forms. I recommend that Peter take some time looking at Cyrillic typefaces and observing how mirrored forms like zhe and reversed forms (in the sense of contrary to expected contrast pattern) like ya are handled. Rather than a strong application of typical contrast pattern, these forms employ fairly subtle weight shifts. This is often especially noticeable in the bowl of the ya.

jfp's picture

I don't take IPA as a writing system as Cyrillic, Latin are. More a tool to help to understood how to pronounce some words. I don't know if complete books set in an IPA font!? Its generally more unique words or couple of them used conjointly. At least in dictionnaries, language teaching books and so on.

I take more IPA as a list of codes than a pure writing system. So, I don't see how a true design for flipped forms will help anything.

May I wrong, perhaps, mostly "bon sens" here, i'm not specialist despite my recent design of an IPA typeface last summer and design of Dictionnaries using IPA this last winter.

peter bilak's picture

Fedra IPA is made specifically for a set of dictionaries, it will probably exist only in PostScript, and it has to work with the main body text which is set in Fedra Serif.

Thanks for your comments -- very useful, as usual, the truth is probably somewhere in between. I started making these revised contrast shapes when i realized that all flipped characters are fairly recent ones, probably at the times of photocomposition and early digital type. They didn't exist in metal, so i tried imaginging how they would look if they did.

It is a good idea to look at Cyrillic and Gentium. Also, Victor's modification are subtle but ellegant.

hrant's picture

> I don't take IPA as a writing system

That's a good way of looking at it.

BTW, are there no metal IPA fonts?


John Hudson's picture

I have a number of linuistic grammars in which a large quantity of material on every page is typeset in IPA, in which context I think harmony of forms is more important than for occasional words. Remember, linguists can read this stuff: they don't decipher it symbol by symbol. This is why I think Victor's approach is sound.

hrant's picture

> they don't decipher it symbol by symbol.

But the boumas are so horrid, I can't see much "real" reading going on.

And probably more significant: do you think linguists would in fact prefer "correct" strokes, or simply what they're already used to (for better or worse)? Really, it's hard to see the extra effort of making "correct" flipped forms as being much more than either an "insular" application of ideology, or maybe a desire to please certain other type designers.

It would be interesting to do some reading performance tests here.


Hannes Famira's picture

I am wondering, would anyone be willing to share an IPA encoding file with me? Please contact me privately. Thanks guys!

Syndicate content Syndicate content