Greek characters on IPA

[Peroyomas]

This topic was imported from the Typophile platform

Many characters in IPA are based on Latin and Greek forms. Is considered an independent alphabet for them but Unicode unifies them with Latin letters, except for three characters that are unified with the actual Greek ones (the beta, the theta and the chi). The IPA is not supposed to be neither Latin nor Greek set, but that unification feels awkward since the Latin and Greek characters have different traits, and I've seen the beta drawn with a serif at the bottom in some IPA resources. Do you think that Unicode needs to make "Latin" characters for these? If not, what's the best way to create both an "unified" set for IPA and providing proper support for the greek characters at the same time? Or is fine as it is and all the somewhat Greek characters must maintain his Greek form?

[John Hudson]

IPA is essentially an extension of the Latin lowercase alphabet. A few letters of non-Latin origin are encoded in Unicode with latin codepoints because these are also used in the alphabets of e.g. African languages and have uppercase mapped characters that differ in form from the mappings in the original script. Case mapping is a functional character property, so needs to be distinguished at the character level. As you note, a few Greek letters in IPA are not separately encoded as Latin letters, because the distinction between their Latin and Greek use is not functional, only stylistic, and hence is considered a glyph-level distinction.

In the current software environment, my recommendation to clients is to have separate fonts for IPA, and not to try to support IPA in the context of fonts that also provides more typical Latin and Greek. This is not only an issue for Greek letters, by the way, but also for some Latin letters. IPA makes a semantic distinction between e.g. double- and single-storey lowercase a, which has implications for the design of italic IPA fonts. Similar issues exist for the f and g. The easiest way to manage these design issues is to make separate IPA fonts.

That said, the new OpenType 1.6 specification list of language system tags includes entries for IPA and Americanist phonetic transcription, and these can be used to trigger 'locl' variant glyph forms. But I'm not aware of any mainstream typographic layout software that yet implements these.

[guifa]

As John puts it, right now the best way is to just have a separate font. But if you want to put it all in one big font file for the time being, you could design the IPA characters from the Greek block with a Latin style instead of a Greek one and include them as a style sheet option.

[Andreas Stötzner]

… to just have a separate font.

This is but anachronistic in the age of Unicode and completely archaic when it comes to text composing including IPA. There’s normally no need to have a “greek beta” alongside with a “latin/ipa beta”.
A good multiscriptual font blends Latin and Greek smothly without unduely equalizing them. Hence the fitting of some greek letters into the latinish IPA set should work smoothly as well. If not, choose a better typeface.

[Jongseong]

There’s normally no need to have a “greek beta” alongside with a “latin/ipa beta”.

I'll have to disagree here. Not all Greek typefaces for text are designed to blend smoothly alongside the Latin. Some historically-informed designs of Greek follow a different ductus from the Latin. Why should these designs be cast aside as anachronistic just because the Greek letters don't look like Latin letters?

There are other cases where straightforward equalization based on Unicode points doesn't work. In most normative Korean text fonts, the period and the comma will be a bit smaller and sit somewhat lower than would be appropriate for the Latin. But there are no separate Unicode values for the 'hangul period' or the 'hangul comma' (those would be ridiculous). There are loads of such cases, due to different stylistic preferences in different languages or different traditions. Until the OpenType locl feature is widely supported, having separate fonts is probably the best solution.

[John Hudson]

Andreas, your approach makes assumptions about the design of the Greek. If, as you have done in Andron or Victor Gaultney has done in Gentium, the Greek shares the same stroke contrast pattern as the Latin, then the Greek letters may serve as IPA forms. But if the Greek is designed with a traditional steep contrast pattern, then these letters look really out-of-place in IPA.

There is nothing ‘anachronistic in the age of Unicode’ about providing separate fonts for specialised purposes, since there is absolutely nothing in the Unicode Standard that demands or even recommends that multiple scripts be stylistically harmonised in a single font.

[Andreas Stötzner]

Well, of course you can always use three or more different typefaces for a multiscriptive text. But scholars who publish a 300-pages volume loaded with mixed text simply don’t want to do this, they seek a typeface which covers *all* they need. 8-bit-encoding was launched to enable precisely that: typeset more than one script at once with just one font. That remains the target. All I wanted to express is that in the case of IPA it is possible to find a typographically convincing solution for that. – By this I do not want to generally question the possible reasons for adjusted local fonts (I did some myself).
Alas, Opentype fancies: I still doubt that the average end user’s vast majority will ever come to make serious use of it. But that’s another issue ;-)

[Andreas Stötzner]

… if the Greek is designed with a traditional steep contrast pattern, then these letters look really out-of-place in IPA

I agree. However, the reason why I, victor Gaultney and others have harmonized Latin and Greek (a difficult but very interesting task, btw.) is because such fonts are needed for scientific publishing. To use two different ripped-off Times fonts because they have various beta’s does not make much sense in such cases.

[John Hudson]

...such fonts are needed for scientific publishing.

Right, so you are targeting a specialised publishing need with a specialised font. I'm advocating a different kind of specialised font for a more limited need, specifically IPA. Since I have customers who need to publish both linguistic texts and Greek texts, and who do not want a Latinised Greek for the latter, they are probably going to need two fonts, unless their page layout software adds support for the 'IPPH' language tag before I'm finished designing the types.

[Andreas Stötzner]

“… who do not want a Latinised Greek”

I was not talking about a ‘Latinized Greek’ but about a ‘harmonized Greek’, that is pretty different. I have customers who publish Greek text mixed with Latin and they appreciate having a typeface which offers both in a classical harmonious style.

[Jongseong]

For your amusement, John Hudson's SBL Greek letters masquerading as phonetic symbols:

Sorry!

[John Hudson]

Andreas, one person's ‘harmonised’ is another person's ‘Latinised’. What's ‘classical’ in this context? Obviously not classical Greek letterforms, since there was no minuscule alphabet in the classical period. When Greek minuscule letters developed, they did so in a Byzantine context and they share the characteristics of most other near-eastern writing, notably the steep nib angle. So the normative forms of Greek letters derive from a particular tool, the reed nib, held at a particular relationship to the page, and on the other hand there are Latin minuscule letters whose normative forms derive from a similar tool but held at a very different relationship to the page. Changing the stroke contrast model of the Greek letters so that it harmonises with that of the Latin letters is, in my reckoning, a form of Latinisation. Sure, people have been doing it for quite a long time now, and depending on a number of factors -- tellingly, on the nature of the Latin design to which the Greek is harmonised -- the results may be pleasing in themselves. But that doesn't change the nature of the exercise.

[guifa]

I would think in scientific publishing one would want the difference in ductus for Greek, given that that means one can distinguish easily o and ο.

While agree that two scripts can be harmonized to a point with a single style, once you have three or five scripts, you realize quickly that what might work between Greek and Latin to make them look the same, won't work with, say, Georgian. So, better (IMO) to find other ways to harmonize them.

[Andreas Stötzner]

What’s ‘classical’ in this context?

Admittedly, the term “classical” is somewhat misleading here when referring to Greek typography. Let me talk briefly about a personal experience.

When I traveled to Greece two years ago I noticed many many magazines and newspapers sold were using typefaces which may be termed ‘hardcore latinised’. Then I talked about that to M. Savidis, a Greek scholar from Athens with a remarkable alertness of ‘Greekness in typography’. He said: oh, that is the wrong direction; many do feel ‘modernised’ when using such a typeface, but many others already claimed that this is not the way to go for Greek typography, because it’s ‘not really Greek’.
Considering that situation, one may conclude that the traditional byzantine Greek faces are thought of being ‘not modern enough’ to the Greeks themselves whereas the ‘modernised’ faces are ‘not Greek enough’ – to the same Greek eyes. An issue which already has been widely debated. – Why is the situation as such?

My personal answer to that is, though rather political incorrect: the typical (byzantine) traditional ductus of the Greek minuscule typography is inferior to the typical (venetian) ductus of Latin minuscules (yes, bash me for that /////). Many do argue: why/how can Greek become “equal“ to Latin – regarding the close kinship of the two systems? The wrong interpretation of that question would be: “how Latin can Greek be”. The other possibility of dealing with that question is: “how can Greek get equally ‘modern’ to match (the rank of) Latin?”. This is something very different.
The attempt of finding a way through the middle is to carefully modernise Greek so as to make it ‘less byzantine, more modern, yet *not* less Greek’! I think that approach has a future and I am convinced that many Greeks themselves do feel like that.

By the way, the typical byzantine minuscule mode (in typography) was developed – entirely without Greek participation. In fact, it’s a mere product of lousy central European lettermaking. Dwelling on Byzantinism for Greek type today is to lock Greek into foreign ungreekness. – No wonder that nowaday’s Greek eyes do imagine something else.

[John Hudson]

Andreas, a couple of contra points:

I am not advocating a particular style of Greek typography, and not the Byzantine cursive style per se, although I have designed a couple of types in this style. I am pointing out that, for any script, one needs to be aware of the tools and techniques that contributed to the basic shapes, and to recognise that it is possible to do violence to those shapes and how they relate to each other if one applies the logic of different tools and techniques.

The most common typographic text style in Greece for serious publishing remains the Didot styled types that in Greek are commonly referred to as apla, i.e. ‘normal’, and typified by Monotype Greek 90. The corresponding italic is of Leipzig origin. These represent an authentic modernisation of the Byzantine lowercase, in the Romantic period corresponding to the Greek war of independence and hence strongly associated with the rebirth of the Greek nation. Typographically, they take into account both the change in common writing tool and developments in paper manufacture and printing refinement, as do the Latin types of the same period. But they do so in a way that preserves the distinctive character of the Greek letter shapes.

It simply is not true that the Byzantine minuscule style of Greek typography was developed ‘entirely without Greek participation’. I recommend to you Constantine Staïkos' Charta of Greek Printing, which catalogues the contribution of Greek émigré scholars, scribes, and printers across Europe.

[Peter Farago]

Andreas asked me by email to talk about the history of Greek glyphs in the IPA and discuss some options for the modern designer, and I figured I might as well make my reply here, where others can view and discuss the matter. To review Peroyomas's opening post, the IPA mostly consists of variants on latin minuscule characters, with a few small caps and handwritten forms thrown in for good measure. It also includes eight glyphs of greek origin:

0251 Latin small letter alpha (ɑ)
0252 Latin small letter turned alpha (ɒ)
028A Latin small letter upsilon (ʊ)
0263 Latin small letter gamma (ɣ)
0278 Latin small letter phi (ɸ)
03B2 Greek small letter beta (β)
03B8 Greek small letter theta (θ)
03C7 Greek small letter chi (χ)

The Handbook of the International Phonetic Association of America, which is as close as we can get to the holy word on the issue, has this to say in its Appendix 1:

"In the construction of the IPA attention has been paid not only to the appropriateness of each symbol from a phonetic point of view, but also to the suitability of symbols from the typographical point of view. The non-roman symbols of the IPA have, as far as possible, been made to harmonize with the roman letters. For instance the Greek letters included in the IPA are roman adaptations; as the ordinary shape of the greek letter [inclined beta] does not harmonize with roman type, in the IPA it has been given the form [upright beta with serifed descender]."

The extent to which the Greek characters of the IPA must be latinized is thus somewhat open to debate - on the one hand, merely setting them upright and unifying weight and ductus with the corresponding Latin characters might be adequate. On the other hand, the Handbook says that the characters should be harmonized "as far as possible", and most type designers have gone quite far indeed.

To begin with, the first four characters listed above are, in essentially all IPA faces, treated as symbols distinct from their Greek counterparts, and have received separate codepoints in the IPA. 0251 (ɑ) is normally drawn as a handwritten single-story lowercase "a", although in handwriting many phoneticians draw it as a greek-looking fish-alpha. 028A (ʊ) has absolutely nothing to do with greek upsilon (υ) and is instead always drawn as a rotated small-cap Omega. 0263 (ɣ) is always drawn upright, horizontally symmetrical, and with a visible counter in its descender.

Of the remaining characters, the IPA version of phi has its own codepoint, but the rest do not. In general, chi is the least harmonized of the Greek IPA characters, usually retaining a cursive aspect to set it apart from lowercase x, with which it must be frequently contrasted. Beta, theta and phi, however, are usually drawn to closely resemble their latin counterparts.

The semi-official typeface of the IPA during the 90s was LaserIPA, drawn by Linguist's Software. It's a very Times-like font; graceless and ordinary, but it gets the job done.

Note the serifed descender of the beta, which perfectly matches the p. The phi features a similar but narrower stem, with a serifed descender and an ascender that's not quite like anything else in the latin lowercase. The chi remains thoroughly non-latin, with a stroke weight that completely contradicts the x, no serifs of any kind, and a cursive crossbar.

The earliest professionally cut IPA fonts were variants of ITC Stone and Linotype Times, which Adobe still sells. Here is Stone Serif Phonetic, with some reference characters from Stone Serif Medium:

Again, the beta features a serifed descender which matches the p. The the stem of the phi is much closer to that the of the beta and p, although still noticeably thinner, and again featuring the unusual top-serif which seems to be drawn from the uppercase I. The theta is simply the numeral zero with a crossbar drawn from uppercase H. Chi is again exhibits the opposite stroke weight of the x, although an awkward attempt at furnishing the character with weird, stubby serifs has been made.

According to Adobe, Linotype's Times Phonetic is designed for eight points, and I don't have a matching optical size for comparison, so please forgive me for using Apple's twelve point edition of Linotype Times.

I think this face is particularly disastrous. As you can see, the beta is not a beta at all, but an ess-tzet (!) with a descender stolen from the lowercase p. (I hope we can all agree that this precedent should not be followed by future designers). Theta, again, is a zero with a crossbar, and chi remains almost entirely Greek. Compare the phonetic symbols with the Greek glyph's from Apple's Times on the second line.

The consensus among the early designers of IPA typefaces is therefore pretty clear: Letter forms are upright; chi maintains the Greek ductus while beta, theta and phi adopt the latin ductus. Phi has an ascender and descender with uppercase-I like serifs. The descender of beta matches the descender of p and q.

The SIL fonts Charis (above) and Doulos (below) follow this precedent precisely:

Neither one has an associated Greek type, so there's nothing to worry about with respect to the shared codepoints. Gentium, on the other hand, has full coverage of Latin, Greek, and IPA:

Gentium's calligraphic ductus allows the Greek and Latin to share the same stroke without doing undue violence to either (although I cannot emphasize this enough: despite being a linguist's dream in terms of glyph coverage, its texture makes it weird to the point of unusability for body text). Unfortunately, I think that the Greek beta just does not work with the other IPA glyphs. The use of the d's ascender on IPA phi seems logical enough, but I can't help but find the glyph a little disconcerting - maybe the stem robs the bowl of its negative space, and it needs to be wider to stand up to the p? Theta works fine in either a latin or Greek context, here (again, due mostly to Gentium's harmonizing, vaguely hand-drawn quality).

Just to round things off, here are some of the other (non-SIL) freely available IPA fonts. From top to bottom, Cardo, Junicode, LeedsUni, and Titus Cyerbit.

I think you will agree with me that these are all unmitigated failures.

Let me conclude, then, with a few thoughts.

1. Of the Greek characters without IPA-specific codepoints, beta is by far the most problematic. Really think hard about offering a latinized glyph variant.
2. If you make sure that your theta stands upright and that your Greek and Latin have generally similar weight and ductus, you might be able to get away with one glyph for both systems.
3. I consider chi to be an unsolved problem for the IPA. It seems to me that there must be a way of unifying its stroke weight and serifs with the Latin without making it and the letter x hopelessly ambiguous. Could a chi-like glyph based on the letter y (with serifed upper terminals and a ball terminal on the descender) be successful? Maybe a Q-like terminal on the downward diagonal? Previous efforts suggest not satisfaction but surrender; this calls for further experimentation.

[Ross Mills]

This is an issue I have been tackling, and so while I don't know that my solutions are the best or even recommendable yet, I can illustrate my experiences. In the example below I decided to not have a Greek with a Latin-like ductus, nor to hybridize it too much (either being options when you know that your font will support both Greek and phonetic transcription) so am left with the problem being discussed here.

I can't say these are final forms, or that I am completely satisfied with how they work yet, but they suggest a possible approach, at least for this particular typeface. Shown at the top are some glyphs from the normal Greek, followed by their ‘Latinised’ counterparts, some of which are encoded as IPA, some of which need to be mapped by some other mechanism. Without support for locl-variants, one option would be via stylistic sets, the other (more complicated) option would be to trigger the alternates display via [calt]. Either of these hacks presuppose you don't take the recommendation to have a separate IPA font for whatever reason.
In the above sample, there are a couple of implementations shown for different contexts: Americanist and IPA. Both have similar problems in harmonizing forms, but IPA will tend to discrete amounts of text rather then continuous text (whether or not that has an influence on how the forms develop is debatable). Another consideration for languages which use the Americanist system is that it is possible readers will be unfamiliar—or otherwise find odd—the overtly Latinised forms of some of the (IPA) Greek characters, as materials published may well have only used more ‘traditional’ Greek letterforms (e.g. the gamma, vs. the IPA gamma) and so there is a chance a third form needs to be made.

[charles_e]

Thank you Peter and Ross.

As I see it, the question about a separate phonetics font then, is simply how to draw the glyphs that must reside in the Greek Unicode spots. If that is the extent if it, I'd reverse myself on an earlier objection to separate fonts for IPA and Greek, as long as the full Latin character complement is included in the IPA font.

I've used both Stone & Times in the PostScript Type 1 days. It was a mess, having the (2) separate IPA fonts for Stone and (4) separate Times fonts in the Linotron 202 days. But of course that was in part due to the the limitation of 8-bit fonts.

For my own work, where I only have to coordinate with two other comps, I'll probably use a stylistic set to switch to the phonetic Greek characters. But these are end-user only fonts, and the work environment allows "voice" communication (as in "Hey, Charles, whatdidyoudohere?"). You can argue the separate/same font either way when both Greek and "phonetic Greek" are included.

Again, Peter, thanks for the good discussion on what these glyphs should look like.

[Chris Dean]

For the thread, IPA = International Phonetic Alphabet.

[Ross Mills]

‘...the question about a separate phonetics font then, is simply how to draw the glyphs that must reside in the Greek Unicode spots. If that is the extent if it, I’d reverse myself on an earlier objection to separate fonts for IPA and Greek...’

The answer is that there are options. The approach is dependent on the circumstance. The user may have some valid rationale for wanting a single font, or separate fonts. This decision may be informed by whether the typeface is new, and decisions can be made from the ground up, or whether you have to work with an existing design, whose IPA and Greek glyphs may or may not be usable, elegant, or functional. If this is the case, are you willing to re-draw the Greek to suit an environment in which they are used for phonetics? Does this compromise their use for Greek language text (if that is a priority)? Or perhaps you simply have to make do with what you have for fiscal reasons. Options.
Fonts are tools, and tools are mutable dependent on their application. The option I have shown for one particular typeface is one of maximum function and at the same time respecting the scenarios it may be used for; whether for typesetting Greek classics or Salishan—I can't discount either. Such an approach may well be outside the purview of a single project, but in the larger context of many diverse projects may be viable—and indeed preferable. So, it may be that re-working some glyphs is sufficient; or if you are willing to take on the baggage, a more comprehensive approach which will suit wider applications, if such development is not too onerous. Or, one could just license a typeface to suit, should it exist, and its design be suitable for its given tasks.

[Jongseong]

Thank you, Peter and Ross, for your fine posts. It's great to have an overview of the current options for setting IPA as well as a preview of what's to come.

The selection of freely-available non-SIL IPA fonts made me cringe. I hope at least that those who use IPA symbols regularly would know better than to use these.

Huronia (the design in Ross's sample) is looking really great! In addition to the IPA symbols needed for American languages, I like the organic design of the Canadian Syllabics, especially the cursive. My only complaint is that the small Latin letter upsilon, ʊ, looks too literally like a small upside-down capital omega at the moment; I think the serifs should be simple. But that's very much a minor complaint about a typeface that everyone dealing with American languages will be extremely grateful for. Any idea when it will be ready?

[John Hudson]

It should also be noted that IPA-specific forms are needed not only for some Greek characters, but also for Latin characters in italic fonts. Since IPA makes a sematic distinction between different shapes of the letters a, f, and g, this distinction needs to be maintained across all styles of a typeface family. Here is an illustration of this issue that I made for a client recently, using the 'IPPH' localised variants that we recently added to the MS Cambria fonts:

[Theunis de Jong]

Indeed, John, for me it's common to get the correction "please use italic 'a'" -- where, of course, I had used an italic font along, not knowing the author meant (semantically) 'the upright a' in an emphasize font.

The semi-official typeface of the IPA during the 90s was LaserIPA ...

... and that still haunts us at night, as elderly(?) linguists complain loudly if I have the nerve to use an actual beta instead of the ugly one (with the serif foot).

[Peter Farago]

Ross, I love your small-cap ƛ. Who did you design this for? Are there any languages that use the symbol in their practical orthographies?

John, you are correct to point out the need to distinguish the single- and double-story as in Italic IPA, and that g should always appear in a single-story variant. However, I'm not sure I know any system where your two f glyphs are contrastive - are you thinking, perhaps, of ɟ or ʃ?

Theunis, the serifed beta is ugly on its own terms, but it has to be considered in the context of the latin characters here. I realize dressing a beta up like a Latin character seems like a cruel sort of thing to do, but I think it's better than sticking a Greek beta in the middle of a bunch of latin characters.

For the thread, I recently stumbled upon IPAPhon, a very old (c. 1994) set of phonetic characters drawn up in the style of Palatino, created by Henry Rogers, a linguist at the University of Toronto. They are not totally unproblematic as typefaces, but I wanted to share a few of their attributes. For one thing, he gets the various a characters absolutely right:

I was also totally charmed by his latinized chi:

This is clearly just an x with a stretched-out downward diagonal and a normal-length upward diagonal, and it has no right to work as well as it does, but it strikes me as quite tolerable: noticeably distinct from x, latin, kind of clever and with a little bit of off-kilter personality.

One other thing to consider when talking about italic IPA faces that I don't think anyone here has brought up is the contrast between v and ʋ (028B). ʋ is itself derived from an italic v, much as ɑ is derived from an italic a, but because the letter shapes themselves do not drastically differ, keeping them distinct is really hard. This is another problem that I don't think anybody has really solved adequately - here are a few of the faces I have that include italic IPA, and how they treat the issue:

Gentium does it best, but I think it's just an irritating glyph in general. I never know how to handwrite it, either.

The selection of freely-available non-SIL IPA fonts made me cringe. I hope at least that those who use IPA symbols regularly would know better than to use these.

My desperation for a straightforward old-style phonetic typeface with italics is so deep that I actually seriously considered using Junicode for a few runs of IPA in a paper last year (I didn't happen to need any of the greek glyphs at the time). I'm beginning to think I may have to draw such a face myself. Unfortunately, my skills and experience are as close to zero as you could imagine. Can anyone here tell me about the methods and tools I would need to draw up the relevant glyphs in the style of, say, Garamond Premier Pro?

[Theunis de Jong]

>I’m beginning to come to the conclusion that the only way for such a font to be created is to do it myself.

It's a bloody ungrateful job :-) The linguists just cannot stop adding new "variants" into the mix. And more likely than not, you will need regular, italics, bold, and bold italics -- at least for the most frequent characters.

For tools, you are required to have a program that can output either only TrueType, or full-fledged OpenType. That is, unless you want to restrict yourself to a niche market with some propriety format! There are lots of editors, from nigh-on usable to excellent -- FontLab is considered the very best. FontForge is a free product, but you'll have to wrestle it into submission first.

You will need a fairly advanced editor if you want to add OTF goodness. Microsoft VOLT (free) might be useful if you want to add self-aligning diacritics, although I didn't try that myself so far.

As for methods: there seems no way around 're-using' parts of the target font. You need at least the same x-height and stem width for your new characters. Since most IPA characters are based on existing ones, you can use parts as base for the new ones. Some artistic appreciation comes in handy if you need to rotate an existing glyph; some regular fonts, and most italics, do not survive that treatment without extensive re-aligning of stems and diagonals. It's hell to create a pleasing italic schwa (rotated 'e') ...