f + umlauts

Florian Hardwig's picture

[continued/excerpted from this thread: The Most Useful OpenType Ligatures]

I really wish to see a text face with a solution for f+umlauts.
In German language, ‘fä’, ‘fö’ and ‘fü’ are anything but rare, just think of ‘für’ [‘for’].
In a lot of classic serif faces the drop of the ‘f’ reaches out quite far to the right. That’s why there are ligatures like fl, ff, fi – to avoid collision, right?

So, dear type designers: before doing fancy things like including ‘ffj’ and ‘ſſl’ ligatures, 5 different ampersands and a truckload of ivies, please consider addressing this umlaut issue!*

See this illustration:

Yellow is (to me) only just acceptable, red isn’t anymore.
And, of course: the bolder, the more clotted it gets.

By the way, I didn’t pick 4 fonts by Adobe because they were especially ‘bad’ – no, this problem is inherent to all faces that have an extensive ‘f’. I used these fonts as they are commonly regarded as being – in all aspects – very well equipped OpenType text fonts – and still suffer from this particular issue.

In my humble opinion, merging the ‘f’ drop with the left dot of the umlaut (as in ‘fi’) can’t be the way to go – that would look really awkward. I’d love to have an alternative ‘f’ with a narrower top for those cases, preferably with automatic substitution, via the lig feature.

And while doing so, don’t forget the triple combinations :°)

I don’t think this problem is limited to German language; the Swedish among others will benefit aswell.

*) Don’t get me wrong; I very much appreciate all those special equipment. All I want to say is: first things first.


I see this came up before in this thread, though not especially focussed on umlauts – starting with Charles’ 4th comment. There, Thomas Phinney called that less sweeping ‘f’ a “Linotype f” – I suppose this then was due to technical limitations.


cerulean said:
MUFI fonts have fä, fö and fü ligatures. They use the solution you reject as awkward, but considering the nature of the project I must assume there is historical precedent for the form.

Yes, Andreas Stötzner’s Andron Scriptor Web lets the ‘f’ drop merge with the left dot – even though there’s no urgent need.

Peter S. Baker’s Junicode takes a different stance. While its ‘fi’ and ‘fj’ do merge, its ‘f’+umlaut combinations offer a different solution: the ‘long f’ gets automatically replaced by an alternate! Exactly what I had in mind, great.
(It additionally has ‘true’ ligatures, but to me, these ligatures with just an elongated ‘f’ bar feel like an end in itself – and ‘fä’ being discretionary)

Andron (with ligatures off/on)

Junicode (with ligatures off, w/ alternate ‘f’, w/ ligs on)

They both introduce a single-storey ‘a’ for ‘fä’. Is this only to avoid collision? Then, I think that’s not a good idea; just imagine how odd words/lines would look like which feature ‘fä’ and a regular ‘a’.
Or do I miss the point, and these ligatures are for a specific historic application only? Who can shed light on this? Highly appreciated.

And: Are there more recent & decent digital fonts with an alternate ‘short f’, like Junicode?

Thanks, F

Jos Buivenga's picture

I've been working the last few weeks on this. See pdf page 4 (bottom). I've also used contextual alternates to avoid collitions (as I did on a smaller scale with Diavlo).

William Berkson's picture

Florian, in the other thread you say that merging the arm of the f with the first dot on the umlaut doesn't seem like a good idea.

What do you think of the following?

This is from DTL Fleishmann, by Erhard Kaiser, who is also German, so I assume that it doesn't look grossly wrong to the eyes of a German reader.

Florian Hardwig's picture

Jos, that’s wonderful news! You just don’t stop amazing me … ;°)
Contextual alternates, that’s definitely the way to go. 4 degrees of ‘f’ drop spread – wow!

William: nice! I have to admit that this isn’t as ugly as I had imagined – especially the italics are beautiful.
I think one’d have to see it in printed running text to really be able to render a judgment. Still, I’m afraid it would hinder the train of reading; just like those fancy ‘ct’ or ‘st’ ligs – nice to have for special applications, not so good for the normal, not historicizing case.

Jos Buivenga's picture

Thank you Florian! For topic's sake I'll post an image here :)

k.l.'s picture

Bill: This is from DTL Fleishmann, by Erhard Kaiser, who is also German, so I assume that it doesn’t look grossly wrong to the eyes of a German reader.

In DTL Fleischmann these ligatures seem to work because the dot stands out as such. (Top parts of 'f' and long-s are rather thin toward the end.)
Your inference is dangerous though. That one (typographically well educated) German does something, does not mean that all Germans need to like it. Same for uppercase eszett.  ;-)

As to one-storey 'a' ligatures -- if they exist for typographic reasons only then I think they are nonsense: In Andron, 'fö' 'fä' 'fü' ligatures are not necessary since there are no collisions in the first place. And in Junicode, the alternate 'f' is much nicer solution, visually.

I think Florian is right in that a contextual 'f' is the way to go. I do this too, but the other way round: the 'f' with short top part is the default one (since it is "safe") and is replaced by one with longer top part only if context allows.
But then, it is rather the various i-accents that are tricky since they cause problems with letters like 'T' 'V' 'W' 'Y' or punctuation/delimiters like slash, braces, brackets etc too ...


ralf h.'s picture

What do you think of the following?

These Fleischmann ligatures are truly beautiful and would be great for any logotype, but I would agree with Florian: Don't merge the the f with the dots, it will never work in the body copy, because no one is used to this connection. So it would be worse than having no ligature at all.


blank's picture

…it will never work in the body copy, because no one is used to this connection.

Nobody was used to sans body copy, but it worked out just fine once people got used to it. Given how passive the act of reading can be, I think that a reader might stop once or twice and then not see it again.

William Berkson's picture

>does not mean that all Germans need to like it

I said "not grossly wrong", not "like". I was asking about "like".

>it worked out just fine once people got used to it

Not. Extended text, such as in books and newspapers, is still rarely set in all sans. And for good reason in my view. I still think there are inherent problems with sans for extended text. Though I admit that modern sans with narrower proportions move more toward the readability of traditional serifs.

dezcom's picture

Great thread, Florian!


ralf h.'s picture

Nobody was used to sans body copy, but it worked out just fine once people got used to it.

Right, but that's a very different thing. If some typefaces start to have those ligatures, they would only annoy the readers and hardly train them.

We really should think about the ligatures we build in our fonts and not just putting anything in there, just because we can (thanks to OpenType.) Sure, it looks great in the type specimen, but that's not the purpose of ligatures, is it?
Here is another bad example: http://www.typografie.info/typoforum/viewtopic.php?f=17&t=4629#p35497
It simply doesn't work. Using a shorter f is the only way to go.


filip blazek's picture

František Štorm included f+umlaut ligatures in his Walbaum Text family. Check it at http://www.stormtype.com, there is a PDF showing all the ligatures.

Sebastian Nagel's picture

f-umlaut-Ligatures will never work, they are too rare to get used to (given the fact that not every typeface has them, and most even don't need them), and can be too missleading.

It is better to just adjust the spacing of these characters, or, if you really are into micro-aestetics, there is a wonderful solution for collision problems via opentype: contextual alternates.

ebensorkin's picture

Unsurprisingly I am on the contextual alternates bandwagon here too. The DTL Fleishmann is beautiful but I would prefer to use it for a headline or a logo rather than in a running text block.

Lovely Thread!

Miss Tiffany's picture

I rarely use umlauts, so maybe my opinion isn't necessary. However, while the Fleischmann is pretty they don't strike me as useful, even for display.

Nick Shinn's picture

I don't think it's possible to isolate this issue to one accent and one language.

(Unless one cleverly uses OpenType language tags to provide different versions of the same accent for different languages, but that in itself seems over-elaborate and thus poor design.)

In general, accents work better when they are closer to the median (top of the x-height), and not floating too far away from the letter. However, the dot over the i is different--it works better when it IS significantly higher, with space necessary to disambiguate it from "l", and this causes difficulty in harmonizing the tittle (dot accent) on i, ż, ė, and ċ, with the other accents, especially the dieresis.

The type designer is faced with a dilemma: the need for a high tittle but low dieresis, a disparity which, to most of us, just doesn't look right. Usually, one opts for a dieresis that is almost as high as the tittle. This looks quite smooth in most languages, but the high dieresis does cause difficulties in old-style typefaces with their overarching "f" arms.

A question I would like to ask the Germans, Turks and Finns: does it bother you when the tittle and the dieresis are at obviously different heights? (See option 3 below.)

1. The default height for dieresis, with a lot of extra kerning added after f.
2. The dieresis lowered, and a little extra kerning after f.
3. The dieresis lowered a lot, and no extra kerning after f.
4. A contextual alternate f to precede a-dieresis etc: however, the difference in "f"s may be a problem.
5. A narrow "f" throughout.

Options 1 and 2--with extra kerning--are quite good solutions, although type designers today, aided by class kerning, are not inclined to individually kern each "f" combination.

I favour option 5. After all, the old-style "f" with long arm was never designed with multi-lingual setting as a criterion. If Bill Caslon had been required to satisfy Finnish, Turkish, and German typesetting in the Antiqua (old-style), I doubt he would have given his "f"s such expansive arms. The reason for the large overhang was, I believe, to make the plain f more like the f in the f_i and f_l ligatures, and to thereby not draw attention to the ligatures. However, when one also considers the problem it creates for the accented characters, then the large overhang is NOT the best solution. It solves f_i and f_l (f_b etc), but at the expense of making f_adieresis etc. much worse. Therefore a narrow f throughout is the best design.

Also, the wide arm of an "f" at the end of a word can occasionally jump right over a space and interfere with a following T or W, and surely this should be avoided (e.g. A Tale of Two Cities).

It seems to me that the use of contextual alternates is a last resort, and that there is something dubious in adopting slightly different alternate letter forms to solve this kind of design problem--it dilutes the personality of the typeface. Nonetheless, what Jos has done looks fine, and I've been experimenting in a similar manner.

The "f" substituted into Caslon here was taken from Palatino. Hermann Zapf's narrow f works well with a high dieresis--and the separateness of f and i was even maintained in its foundry type ligatures.

rabattski's picture

Totally unproductive but this is an awesome thread, seriously interesting.

ralf h.'s picture

A question I would like to ask the Germans, Turks and Finns: does it bother you when the tittle and the dieresis are at obviously different heights?

Yes, it bothers me a lot. As a German reader I want them to be at the same height and also with the same size.

k.l.'s picture

Another German reader's choice:

The f -- version 5.
Dieresis positioning -- version 2.

More on dieresis positioning:
Version 5 looks strange to me, the dieresis is taking off.
It would be interesting to know whether readers whose languages include 'zdotaccent' or 'cdotaccent' feel disturbed if their dots sit as high as that of the 'i'.

[Edit: Maybe I am just a bit oldfashioned and dot and dieresis on the same height work if neither the (i-)dot sits too low nor the dieresis too high.]

Off-topic, but since you started discussing vertical positioning of accents:
What bothers me again and again is placing accents on uppercase. Given that accents vertically align on a common 'baseline' (rather than being center-aligned), accents on letters with overhang (like 'O') look closer to the letter than on letter without overhang (like 'I'). And it is particularly annoying with open letters (like 'U' or 'N'). In text setting this may not be visible, but in display setting it is ...

Jos Buivenga's picture

I nearly almost follow Adam Twardoch's view on the dotaccent and also place dieresis on the same height.

There was also a remark on my blog (about the 1st Diavlo) from someone who said " ... Some languages have very popular words when letters with dot accent are next to “i”, and your beautiful font harmony suddenly disappears ... "

Contextual alternates work very well imho ... especially in combination with the ligature and kern feature (see animated gif).

William Berkson's picture

On the i dot verses dieresis height:

a possible solution is to have a language specific alternative i with a smaller, lower dot for languages where the i and umlauted letters are in frequent proximity.

If you did this, for what languages would it be appropriate to substitute the i with the lower dot? Finnish seems to be one. Any CE languages?

Incidentally, one typographically aware German speaker told me there was not problem with a higher i dot. As I understand it, in German the i is very rarely followed or preceded immediately by an umlauted letter. What do others say on this?

I am completing a Caslon revival, and the high, heavy i dot is in fact important to the traditional 'look' of Caslon, so I am reluctant to change that.

ebensorkin's picture

Jos, that looks bang-on to me. Very very nice example.

Nick & jos what fonts are these?

Jos Buivenga's picture

Thanks Eben. My little Christmas workout is called Calluna.

guifa's picture

Ack, thread resurrection (sorry). In German the i might not follow often an umlaut, but in Spanish or Portuguese, about 50% of the time an i will follow an umlauted u.

nina's picture

OK, I'll help resurrect. In case somebody stops by and wonders about umlaut placement, let me just say here's a native German speaker who firmly believes there is generally no problem at all with tittles being higher up than umlauts. When people try very hard to make them the same height it can actually look quite off, especially when the dots of the umlaut are noticeably smaller than the tittle (which I also believe can work, it depends on the font).
The thing to keep in mind is that umlauted letters are different letters than their non-umlauted counterparts – meaning the diæresis must read as an integral part of the glyph. If say an "ä" looks like an "a" with something small floating way up (as can easily happen when the diæresis is small and at tittle height), it doesn't read optimally.

FWIW, this is how my eyes see things. Feel free to disagree:

Bendy's picture

Nina, I get a 404 following your link :(

nina's picture

Argh! Thanks for pointing that out, Ben. It should work now.

Bendy's picture

Yep, working. Interesting. I think you've seen Adam Twardoch's pdf on diacritic design since you mentioned the Polish creska in the other thread? It looks like the nicest diereses are substantially lower and smaller than the tittles. Very useful to know, thanks for your pdf :)

nina's picture

Thanks for that link Ben! I have indeed been looking at Adam Twardoch's writings with regards to Polish diacritics, but only on his web site.

Like I said, I'm not sure how authoritative the opinions expressed in my PDF should be taken to be. They come pretty much from a "user" perspective (but a slightly obsessed one ;) ).

Bendy's picture

>a slightly obsessed one

I think that's why we're here on Typophile ;)

FWIW the lower and smaller dieresis looks more 'correct' to me too, though I have had minimal exposure to this particular diacritic. I'll be lowering it in Mint.

Miguel Sousa's picture

> in Spanish or Portuguese, about 50% of the time an i will follow an umlauted u.

Can you give examples? Portuguese from Portugal does not use umlauts at all. Umlauts are occasionally used in Portuguese from Brazil, AFAIK to indicate when the 'u' in 'qu' needs to be pronounced (e.g. freqüente)

William Berkson's picture

Finnish I think its been said works better with i dot and j dot on the same lower level as diacritics. Are there other languages? Evidently it's not a problem in German.

guifa's picture

Miguel, I learned Brazilian Portuguese, but it's been a while (and I know they're still fighting over the orthography so maybe something's changed). The dieresis is only used in the sequence [gq]u[ei]. I don't think the rare qu[oa] sequence would have it (quasar, quorum). In Spanish, the dieresis is, outside of poetry, only used in the sequences gu[ei] to show the u is pronounced. The Portuguese is replaced with cu

Hence my estimate of roughly 50%: if the dieresis/umlaut shows itself, it will necesarily be followed by an e or an i.

Michel Boyer's picture

Miguel and Matthew

I just downloaded to my mac the Brazilian spell checker for Thunderbird, installed it following the indications, found it as a text file called pt_BR.dic in a subsub... folder of ~/library/thunderbird/Profiles/ and had a look at it.

If I grep ü in that file, I get 765 entries (that may be different words or lemmatizations of the same word). If I grep üi, I get 388 entries, starting (after sorting) with agüista ambigüidade ambigüifloro angüicida angüiforme angüiliforme antigüidade aqüicaldense aqüicalor aqüicultor and ending with ubiqüismo ubiqüitarístico ubiqüitário ungüi ungüiculado ungüiforme ungüinal ungüinoso unilingüismo úngüis but also containing (but not so many, at first sight) variations of the same word, like this nice list

sagüi, sagüi-amarelo, sagüi-branco, sagüi-caratinga, sagüi-comum, sagüi-de-bigode, sagüi-de-duas-cores, sagüi-de-mão-ruiva, sagüi-de-orelhas-compridas, sagüi-do-brasil, sagüi-do-rio-de-janeiro, sagüi-imperador, sagüi-ordinário, sagüi-pequenino-do-maranhão, sagüi-preto, sagüi-preto-de-bigode-branco, sagüi-veludo, sagüi-vermelho

and similar lists for sagüim, sagüins and sagüis.

To draw conclusions on actual frequencies, a native speaker or complementary statistics are required. Indeed, even if in those entries more than half of the time ü is followed by an i, that that does not imply that in actual texts the same proportion is to be found. That depends on the frequency of use of those words.


Michel Boyer's picture

Florian mentions Swedish above; here is what I get from the Swedish OpenOffice spelling dictionary : it contains only 66920 entries, but there are as many as 2920 occurrences of "fö" and 698 of "fä" (with 81 occurrences of "ffä", only 6 of "ffö", and no "fü"). Rem: those are occurrences of digrams in the file, not word counts.

[edit] I get 2901 words containing fö and 698 containing fä.


henrypijames's picture

In Chinese Pinyin, there are four "toned" version of u, each of which can follow f: U+016B (u with macron), U+00FA (u with acute), U+01D4 (u with caron) and U+00F9 (u with grave).

Jongseong's picture

In Chinese Pinyin, f can also be followed by toned versions of a, e, and o in addition to u. Thankfully, in an f+i combination the i is always plain.

henrypijames's picture

Actually, "fi" combination doesn't exist at all in Pinyin (at least not for Han Chinese).

Jongseong's picture

My Chinese dictionary lists "fiào" as a dialectal variant pronunciation of 覅, which is why I added the above comment. But checking again, I see that it's the only such case given in the dictionary, and seeing as it's one variant reading that has the f+i combination, it would be just as well to say that the combination doesn't exist for pinyin for standard Chinese.

henrypijames's picture

You're right! This single character with this unique pronunciation does indeed exist according to the Xiàndài Hànyǔ Cídǐan, even though the common standard Xīnhuá Zìdiǎn doesn't have it.

Now I'm unsure whether there are other "fi" combinations among the even more obscure characters -- some of which might even have tone marks on their "i". Unfortunately, I don't have the Hànyǔ Dà Zìdiǎn around.

ebensorkin's picture

this thread is getting cooler and cooler

Michel Boyer's picture

A relevant post: http://www.typophile.com/node/61027#comment-362977 (Thread on letter pair / diacritic pair frequency)

Florian Hardwig's picture

Daily collision:

Maxim Zhukov's picture
The ії, їі, and ïï combinations are common in Ukrainian, e.g.
„В она повела його до вітальні, де чекала родина — хлопчик і дівчинка ïï віку, дружинина сестра та ïï чоловік. Основні принципи і методи аудиту готової продукції в його організації. Він зустрів їі рік тому у Дебрецені, коли найнявся до їі батька доглядяти кони.“

Of course, Cyrillic і (U+0456) and ї (U+0457) are different characters from Latin i and ï, and yet.

Maxim Zhukov's picture

I’ve just spotted a typo in the above Ukrainian text. In the first sentence—from F.S. Fitzgerald’s Babylon RevisitedВона (‘she’) should be a single word, not two. Sorry.

flooce's picture

I never understood why especially Adobe does not pay more attention to this issue. They take so much care to support Greek and Cyrillic systems, but did not realize that some of their most prominent fonts are of diminished use in German and other languages using the Latin system.

Minion and Adobe Caslon are some of the few commercial fonts I own, but I avoid using them in German. For a person who does not need or own a lot of different font licenses, Minion is in many regards close to a perfect serif font, because of it subtle beauty and versatility. (Of course it might be generic, the more fonts one has to choose from.) It is just sad, that this seems to be such a non-issue for many.

William Berkson's picture

Flooce, probably because of this thread and related ones, in Williams Caslon Text I put in contextual alternative f's to avoid f + umlaut clashes.

I also put in a stylistic set that lowers the traditional high dots on the i's and j's for German or Finnish, if you prefer that.

The contextual alternatives for f + umlauts are default, and the lowered and smaller dots are stylistic set 2. Here are some of the alternatives, and here is a comprehensive list, with examples.

Nick Shinn's picture

Why not just publish a German language-specific version?
You could also include lowered cap umlauts and ck, ch, and tz ligatures as default.

William Berkson's picture

Nick, I'm not sure whether German speakers actually want those options. My cap diacritics are already pretty tight, and a German-speaker here complained about the umlaut lowered around the O as looking like Mickey Mouse ears. If someone wants some additional glyphs, of course I could do them.

charles ellertson's picture

I'm not sure why this thread got revived -- to complain about Adobe? -- but in any case, we've been reworking their fonts for years with a contextual f, sometimes more than one. And not just for umlauts. There are other accents that cause problems too. Try a f+grave, sometime. Or even the sequence "f+space+J" -- "of Jewish . . . " is one of the stock things I look at when kerning & preparing contextual alternates.

Syndicate content Syndicate content