Designing Phonetic Characters

André G. Isaak

As someone whose background is in linguistics rather than type design and who has thus read lots of material containing phonetic characters, you’d think I’d have better intuitions on this topic but I don’t so I thought I’d ask...

The International Phonetic Alphabet employs a variety of letters which are rotated, reversed, or inverted versions of standard latin letters, and historically these letters (at least the rotated varieties) were actually created by simply rotating sorts, but I’m wondering if this is really the appropriate approach when designing type from scratch.

In some cases, rotating a character produces perfectly reasonable results. For example, open o (0x254) can be produced by simply rotating lowercase c:


However, in other cases, it seems that simple rotation doesn't harmonize as well with the latin alphabet as it could. Below are examples of rotated m and h (0x26F and 0x265), along with an alternate version which involves more than simple rotation (n.b. all examples are based on Adobe Text):




It seems to me that the rightmost instances in the above examples harmonize better with the latin alphabet than the center versions. However, a survey of various phonetic fonts indicates that the center version seems to be far more common (in fact, of the faces I looked at only John Hudson's Brill didn't use purely rotated versions).

What I'm uncertain about is whether departing from the more traditional forms (which as I said often resulted from simply rotating sorts rather than redesigning characters) may introduce a source of unnecessary confusion and whether that should override some personal notion of ‘harmony’

Thoughts?

André

Andreas Stötzner

I also prefer to treat the flipped glyphs individually.

Theunis de Jong

Here is an older Typophile thread discussing the same thing: http://www.typophile.com/node/58725 (in which John Hudson switched from 'bland' mirroring to actually redrawing "upside down" characters!).

The problem is that linguists are not type designers. They want to represent some sound with a unique character and somehow decide that it's best indicated with a regular character turned upside down. So that is what they do, never mind stress direction and what-if-it-needs-italicizing.

So what they expect to see is the plain character, mirrored and/or rotated. The only thing you can hope for is to educate them and explain it should be treated, design-wise, as a totally new kind of glyph.

I imagine the first designers of the /w glyph faced similar backlash. "VVhy not just dravv it as two vv's, that has alvvvvays vvvvorked bevvore!"

Vasil Stanev

I agree with Andreas and Theunis. The shape may change but the angle and width of the nib, from which everything derives, does not.

Another somewhat related topic is how much the IPA should actually harmonize with the body text? To my mind words set with IPA symbols should stand out, similar to the Paragraph sign. What I mean is don't linguists prefer IPA to be set in a very limited set of fonts, so they are instantly recognizable? And concequently only in roman weight, with no italics and so on. Like computer code. 

André G. Isaak

As a linguist, I much prefer IPA fonts which harmonize with the body text.

André

André G. Isaak

Thanks, Theunis, for the link. I haven't been on Typophile for a long time and I missed that thread.

André

André G. Isaak

So apparently I'm not alone here. I’d stated that only Brill followed a similar approach, but that’s clearly not the case. My brain has Andron hardwired as a ‘mediaevalist’ font rather than a phonetic font so I hadn't thought to look at that, and I somehow managed to overlook Gentium in my initial survey (I've never been a big fan of Gentium, but it's hardly a font to ignore when discussing phonetic fonts).

André

Ben Blom

(Off topic:) André, can you please refrain from repeating “André” at the end of each post? Your name is already at the beginning of your posts. It’s a waste of space. It is not customary practice here.

Andreas Stötzner

Vasil Stanev said:

… words set with IPA symbols should stand out…

No they should not.
The phonetic characters are not symbols but letters. These letters are a part of the (extended) Latin writing system. This needs to get materialized by the font design. Because the correspondence of the phonetic letterforms with the common a to z letterforms is of the essence for understanding what the phonetic letters represent, so it is not only desirable but neccessary to make the relation between the ordinary and the special letters visible and thus enable the recipient to draw the correct conclusions.

Vasil Stanev

Andreas, André, thank you for clearing things up. 

Jens Kutilek

For an overview of more phonetic fonts, see this series of blog posts by Christopher Bergmann: https://www.isoglosse.de/2016/06/fonts-for-phonetic-transcriptions/

John Hudson

For example, open o (0x254) can be produced by simply rotating lowercase c:

It's acceptable, but I find a rotated c often tends to be less stable on the baseline, and will tend to have a slight optical tilt to the left, so needs some adjustment to make an overall fuller, rounder shape than the c, more closely related to the o. For the same reason, it helps to make the lower terminal a bit larger, reaching more to the left. Here's the Brill Roman c, rotated c, open o, and o, for comparison.

Vasil Stanev

For a rotated /c, I would follow the same principles as е оборотное (uni044D), i.e. it is it's own beast.

John Hudson

Indeed, and also this of course.



[from http://tiro.com/John/Hudson-Brill-DECK.pdf ]

André G. Isaak

Thanks, John, for that .pdf link.

Fascinating stuff!

Alan Bernard Hughes

Thanks from me too, John: a beautiful pdf

Denis Moyogo Jacquerye

Rotated or flipped letters do indeed look much better when their design is adjusted.

Especially since many of those letters have been used in standard orthographies for a while.

This is even more important for symbols like ɛ, ɜ and ᴈ (great example @Andreas Stötzner!), or ɑ, ɒ and ꭤ, where each needs to be distinguishable since they all can be used in the same transcriptions for different sounds.

However I’m slightly sad that ɟ has been redesigned to the point where it has become a dotless j with bar in many fonts instead of being closer to the turned f it originally was. In several fonts, ʄ has a completely different design and it’s not obvious anymore that ɟ and ʄ share the same origin or that in IPA the top hook of ʄ indicates it is the implosive form of ɟ. That said, the dotless j with bar shape has been common for a few decades so no bad feelings if you use it ;-).

About the lowercase rotated letters, the rule of thumb is that the center of rotation is the between the x-height and the baseline. This also applies to letters with ascender or descender. So:

• what is at the x-height ends up at the baseline
• what is at the baseline ends up at the x-height
  
• what is at the ascender-height ends close to the descender-height
• what is at the descender-height ends close to the ascender-height

The only exception to that rule, as far as can think of, is ʖ, which is a turned ʕ but it is rotated with the center of rotation between the ascender-height and the baseline.

That said, many phonetic symbols have been used in a variety of transcription systems, you may want to use a different rule depending on what you’re trying to reproduce.

With that rule of thumb, I would recommend adjusting ᵷ accordingly in several fonts which have it too low, considering the couple of known use cases:

See for example ᵷ in Александр Амарович Магометов [ალ. მაჰომეტოი], 1965, Табасаранский язык: исследование и тексты [ტაბასარანუმი ენა: გამოკვლევა და ტექსტები], Тбилиси, Мецниереба (taken from http://std.dkuug.dk/JTC1/sc2/wg2/docs/n2608r2.pdf)

Or also ᵷ in Association phonétique internationale, 1900, Exposé des principes de l’Association phonétique internationale, Leipzig, B. G. Teubner.

André G. Isaak

However I’m slightly sad that ɟ has been redesigned to the point where it has become a dotless j with bar in many fonts instead of being closer to the turned f it originally was.

This isn’t a particularly surprising development (nor do I share your view that this is sad). While [ɟ] was originally set using a rotated <f>, it’s pretty clear that it was the visual resemblance to <j> which inspired this character given that the sound which it represents (a palatal stop) doesn't even remotely resemble [f]. I suspect that most people view this as a variant of <j> even in cases where it is simply a mechanically rotated <f>.

John Hudson

However I’m slightly sad that ɟ has been redesigned to the point where it has become a dotless j with bar in many fonts instead of being closer to the turned f it originally was.

it’s pretty clear that it was the visual resemblance to <j> which inspired this character given that the sound which it represents (a palatal stop) doesn't even remotely resemble [f]

Pullum & Ladusaw call this 'barred dotless j' and note:

'Typographically, a turned lower-case f, but better thought of as a variant of ⟨j⟩, since in English words like job a ⟨j⟩ represents a palato-alveolar affricate [dʒ] that is at least somewhat similar to IPA [ɟ], though not identical.'

Not an overwhelmingly strong case, but reasonable.

My own reasoning in designing the character is that if I were writing the letter there would be no distinction of the main body between ɟ and j.

Denis Moyogo Jacquerye

Pullum & Ladusaw are not the only ones to call ɟ “barred dotless j”.

The IPA Handbook names the symbol “barred dotless j” but use a glyph that looks like a turned f but with a raised bar. Esling, 1988, “Computer coding of IPA symbols”, Journal of the International Phonetic Association also names the symbol “barred dotless j” but has a glyph that looks like a turned f without any modifications.

The issue is not that there is only a left top serif or two top serifs, but that the bar is not on the baseline or at least that it is not consistent with the bar of ʄ.

André G. Isaak

The issue is not that there is only a left top serif or two top serifs, but that the bar is not on the baseline or at least that it is not consistent with the bar of ʄ.

I think the main issue here isn't with <ɟ> but rather with <ʄ>. I agree that <ʄ> should follow the same design principles as <ɟ>, but it often is not. To me, the most sensible approach to the bar would be to treat <ɟ>, <ʄ>, and <ɨ> identically. Instead, you often see <ʄ> with a double-bar, as if it were constructed from <f> combined with a rotated <f>. I think the issue here is simply that [ʄ] is an uncommon sound with which many designers might not be familiar.

Kent Lew

To me, the most sensible approach to the bar would be to treat <ɟ>, <ʄ>, and <ɨ> identically.

If the bars of <ɟ> and<ʄ> are raised to mid-xht, more consistent with <ɨ>, wouldn’t that create a potential ambiguity between <ʄ> and <ƒ> (f with hook, if drawn properly upright for African use)?

Or are linguists unconcerned with that, given that ƒ is not used in IPA?

André G. Isaak

I hadn’t actually thought of that case. As you point out,, fhook isn’t actually part of the IPA, but it’s still worth thinking about.

Kent Lew

I guess the question would be if the voiced palatal implosive occurs in languages that also feature the fhook in their orthography (or closely enough related languages), such that a text that might be referencing words and pronunciations where the two different characters could occur in some proximity, and whether context would be sufficient to avoid any detrimental ambiguity.

Denis Moyogo Jacquerye

All you need is a phonetic description in Ewe which uses ⟨ƒ⟩ of a language like Fula or any other language with [ʄ].

Andreas Stötzner

André G. Isaak said:

... the most sensible approach to the bar would be to treat <ɟ>, <ʄ>, and <ɨ> identically.

This is what also appealed being sensible to me.

John Savard

Andreas Stötzner said:

The phonetic characters are not symbols but letters. These letters are a part of the (extended) Latin writing system. This needs to get materialized by the font design. Because the correspondence of the phonetic letterforms with the common a to z letterforms is of the essence for understanding what the phonetic letters represent, so it is not only desirable but neccessary to make the relation between the ordinary and the special letters visible and thus enable the recipient to draw the correct conclusions.

Somehow I don't think that having IPA text in Caslon, while the normal text is in Times or Univers, will lead people to think that the IPA glyph that looks like the Latin lowercase g might stand for some sound other than /g/.

Which is what you seem to be saying.

IPA characters are different from the letters of the alphabet. They are used, like letters, to write texts belonging to languages. But each character stands for a sound, whereas letters are conventional signs used to represent words according to the prevailing orthography for a language. Different languages may use a letter in different ways: thus, in English, j is a consonant related to /ch/; in others, it functions like the English letter y in creating dipthongs. And an orthography may only be partially phonetic, with the degre varying between languages.

The IPA notation writes the sounds in the same manner regardless of which language is being transcribed.

The phonetic characters are typographically letters rather than sorts, so they're not symbols in the sense that @ or & are symbols. But in other senses, their function does differ in a limited and hence subtle manner from that of ordinary letters.

André G. Isaak

Delimiters like […] and /…/ are sufficient to identify a string as being IPA. For a text such as (e.g.)

The English word ‘cat’ is pronounced as [kʰæt] or [kʰæʔ], depending on dialect.

I really don’t want to see the IPA characters in a separate typeface. The above is a single sentence, and should be set in a single font.

If you ever want typographic nightmares, flip through any conference proceedings or working papers volume in linguistics from the eighties or nineties. These usually wanted camera ready text from their authors, and since major operating systems of the time didn't include much in way of multi-lingual or phonetic fonts, the results were horrendous, with text in one typeface, non-English Latin examples and phonetic transcriptions in another, often mixing more than one typeface in the same word.

In fact, it was this situation that first got me interested in type design in the first place. Armed with only Fontographer and Minion (at the time, one of the few fonts with a relatively large character set which I considered suitable for text), I managed to produce what might have been the first typographically (though not necessarily theoretically) coherent MA Theses of the early nineties.

[EDIT: I don’t think I still even have the various Minion Supplement fonts I created. I suspect they were probably awful, but they were decent enough to fool a crappy 300 dpi LaserPrinter.]

Theunis de Jong

André, your July '93 thesis? Is it digitally available anywhere?

André G. Isaak

July '93 sounds about right (though I'm not sure where you got this from -- MA theses usually aren't catalogued). I doubt it's available (in fact I'm not even sure if *I* still have it anywhere).

[EDIT -- OK, you got me curious so I did some digging. It would appear that the University of Calgary started archiving MA theses in electronic form in 2005, but since there was no requirement for electronic submissions prior to that, they’ve retroactively added a catalogue entry, but since there was no .pdf submitted it is ‘embargoed’ until 2999 (which I assume just means forever). So that would suggest that no, there is no electronic copy available anywhere, just the paper copy which I submitted at the time. This is probably for the best since I suspect the contents of said thesis were embarrassingly bad.]

John Hudson

@John Savard Please remember that a significant portion of IPA letters are also used in natural language orthographies. In the case of North American first nations languages, this involves not only full size letters but also a variety of IPA superscript modifier letters. All these letters need to balance and blend into a typographic whole.