The self‌less butterfly. Or: Using and avoiding ligatures.


  • Craig EliasonCraig Eliason Posts: 1,022
    I think there are many type designs for which standard ligatures are going to be the least distracting solution, vs. both crashes and altering shapes to prevent crashes. I think the author overstates the degree to which connected ligs stand out (at least at text sizes), and perhaps underestimates the degree to which highly retracted f-hooks themselves might stand out in certain designs. 

    (Perhaps it's a measure of how long its been since Linotypes ruled the earth that the argument against the necessity of f-ligatures can again hold sway.)

    That said, there's a logic to the German rules about ligatures not crossing compound-word elements, and there may be merit to extending that more generally to keep morphemes intact. And it does seem responsible for designers to insure that pairings still work well if ligatures are unwanted. 
  • Ray LarabieRay Larabie Posts: 1,000
    I don't know about other people but ligatures really annoy me when I'm reading.
  • Paul MillerPaul Miller Posts: 240
    This is an interesting article and I find myself agreeing with a lot of what was said.  And if people want to ork cows in their own time who are we to judge :D
    I agree with @Ray Larabie, many ligatures can be highly incongruous, particularly the 'st' and 'ct' ligatures.  I have included these ligatures in some of my fonts but I made them discressionary so they could easily be turned off.
    But there are some instances where the use of ligatures can change the letters perceived by the reader.  In the Turkish language there are upper and lower case versions of both dotted and dotless 'i'.  These are four different letters, the upper case version of 'i' is'İ'.
    Where an 'fi' or 'ffi' ligature removes the dot of the i it is actually altering the sequence of letters to a Turkish reader.  I am told by a Turkish friend of mine that most Turks are inured to this and many don't even notice it anymore but where they do notice it they just roll their eyes and curse western typography for mangling their language.

  • Mark SimonsonMark Simonson Posts: 1,220
    edited April 19
    I find the idea of giving semantic meaning to ligatures problematic. Not all typefaces have (or should have) connecting ligatures (a lot of sans serifs don't), so a rule where ligatures are used for some words but not others could not be followed consistently. If so, how would readers ever come to understand the distinction?

    In any case, ligatures are not a problem in terms of affecting the meaning of a word, except for fonts that don't account for Turkish, and maybe Germans reading non-German text.

  • John HudsonJohn Hudson Posts: 1,931
    I think pretty much anyone making a font that supports Turkish knows they need to give special handling to the i and dotless i, not only with regard to ligatures but also in smallcap substitutions. This was some of the earliest demonstration of OpenType language system processing. So the issue for users is language that isn't correctly tagged as Turkish and software that 20+ years on still doesn't handle this the right way.
  • John HudsonJohn Hudson Posts: 1,931
    edited April 19
    I'm wondering whether the latter doesn't do more damage to readability. I suspect that neither has a measurable impact for almost all readers, though, so it's probably impossible to decide.

    I’ve yet to see any empirical evidence of either typical Latin ligatures or variant letter shapes having a significant impact on reading speed and comprehension. I can imagine some forms in some typefaces having an impact in letter recognition testing, and maybe some regressive saccades, but whether either of these would significantly affect overall reading performance more than numerous other factors seems doubtful.

    I work with a lot of complex scripts in which letters not only ligature but, in doing so, radically change their shape and arrangement, including across morpheme and, in some cases, even word boundaries. And many millions of people read these scripts with facility. Yes, there’s good evidence that script complexity increases cognitive processing load and slows reading — some scripts are measurably easier to read than others —, but the standard, non-decorative Latin f-ligatures are barely classifiable as complex.
  • Nick ShinnNick Shinn Posts: 1,538
    edited April 19
    In English, the f_l situation kind of resolves itself. You can make a normal ligature for fl, which is appropriate as this is usually a morpheme, and dynamic ligatures for the other ascender glyphs, which are, unlike in selfless, usually* trans-morpheme. 
    sub f' [b h k etc] by f.narrowtopknot

    top:  The default /f is designed to merge into the following ascender, but nonetheless, awkward fit and concentrated heaviness are hard to avoid, therefore the single-glyph ligature below smooths and thins that down.
    halfbacks: the first shows the “narrow topknot” <liga> substitution, the bottom one is the default glyphs, with added kerning to avoid morphemically-inappropriate touching—but which makes too much of a gap in the middle of the word.
    So the issue for me is not so much one of optimizing grammatical interpretation, but more of achieving good text colour.
    (Pratt Nova Text Regular)
    *I haven’t checked this. Am I correct?
  • Alex VisiAlex Visi Posts: 58
    The article seems to be a very modernist thought, and it’s 2020. People don’t care whether to use feet marks or acutes instead of apostrophes, dashes and hyphens, guillemets and less/greater — and the author is saying connected letters, which has always been this way, bother somebody. Come on, there are more interesting things to think about in typography.
  • Thomas PhinneyThomas Phinney Posts: 1,890
    The article is full of sweeping, unsupported assertions of supposed “fact.” As an argument for possibilities that would be worthy of study it would be fine. But I call bullshit on many of the assertions and assumptions in the article (also the call-to-arms tone, and pejorative wording).

    Ligatures supposedly affect “random” letters! (“Random” and variations appear seven times in the article.) This is nonsense.

    “Admittedly, we do not know if ligature glyphs whose shapes differ from the shapes of the isolated letters they represent are confusing indeed and, if so, to what degree. We can be certain, though, that there is no functional advantage to using ligatures. Readers may be mostly unfazed by them, but they can never benefit from them. And if there is no benefit in using ligatures, it may be time to get rid of them.”

    But ligatures are never “distributed randomly”—and I would be happy to bet any amount up to $1000 USD that third-party research with adequate sample size would find no statistically-significant negative effect of ligatures on reading, segmentation or comprehension for adult readers. Anybody want to take me up on this? Place the bet now in public, and when the research eventually comes in a year, or a decade, or two, I will be happy to collect.

    Contrariwise, the famous Microsoft studies (Larson & Picard, 2007; Larson, Hazlett, Chaparro & Picard, 2007) of the effect of good typography on the mood and future task performance of readers found positive effects from a combination of typographic factors that included ligatures. This is not especially convincing as ligatures were just one of numerous variables conflated together, but it certainly suggests both that (1) researchers in the area expected that ligatures would have a positive impact, and (2) if ligatures are actually “bad,” their impact was easily overwhelmed by other factors.
  • John SavardJohn Savard Posts: 622
    edited April 20
    In short: I find the call to avoid ligatures across morpheme boundaries groundless and purely academic, both within German and outside.

    I absolutely agree. While I can't speak to German, in English the idea that "butterfly" can have a ligature, but "selfless" shouldn't, is indeed silly or even bizarre.
    However, I also feel that the issue is largely moot. If, for whatever reason, it is necessary to provide contextual alternatives to ligatures - which I find plausible - then, since many people use simple word processors that don't support advanced Open Type features, it becomes necessary to avoid ligatures, even though the expedients, such as the infamous "pot-hook f", may not be aesthetically ideal.

    Where an 'fi' or 'ffi' ligature removes the dot of the i it is actually altering the sequence of letters to a Turkish reader.
    Back when typography was done by professionals with metal type, were ligatures for fi and ffi ever used in Turkish typography, and, if so, how was the issue dealt with then?
    I can't fault a font designed for use by people in printing English-language texts for not working properly when used for Turkish. Today, with OpenType, one can tell a font to do different things for different languages. But WordPad in Windows, for example, doesn't support that feature.
    It lets you create documents with proportionally-spaced fonts. It lets you choose the font, and use more than one of them in a document. So a typical naïve computer user is very likely to say: What do you mean, I need something better?
    Of course, Open Office is free, but it takes longer to load than WordPad. I do not know of a lightweight formatted text editor that is easy to use, doesn't take up too much space on disk or in memory, and yet supports the fancy Open Type features. However, I haven't spent much time looking for one, but something tells me I would not find one. Still, since there are excellent programming text editors for Windows around, like Notepad++, maybe I'm being unduly pessimistic.
    Looking around, I see that there is at least one fine WordPad alternative that is both small and free out there, AbleWord, but no, it does not seem to offer advanced OpenType support.
  • Adam JagoszAdam Jagosz Posts: 621
    edited April 22
    @John Savard Both macOS's TextEdit and Windows' Notepad support OpenType features, though only the former offers any sort of control over them. [Not to say that Notepad is an alternative to WordPad though, not really.]
    Unluckily, the LOCL feature seems to be disabled, at least in Notepad — which I think is argument in favor of avoiding ligatures as a default solution for the "fi" collision.
  • Adam JagoszAdam Jagosz Posts: 621
    edited April 22
    The article, however opinionated, was still an interesting read.
    When reading as a child, I remember often stopping at fl and fi and thinking "Wow, how cool that an ef and an ell look together like A, and an ef and an i look like h...". I had no idea they were made to look this way, and I thought it was a mere coincidence. If a reader thinks this way, I'd say the ligature is executed successfully. Though I must admit it slowed down my reading oftentimes, as I was stopping just to investigate these lettershapes (but long emdashes which are rare in Polish typesetting, also had this effect on me — I guess it's just my freaky self).
    On the other hand, the first thing that comes to mind when I think of Alright Sans is its fl ligature, with its arch ideally symmetrical — it looked nothing like an ef and an ell. Imho, silly and unnecessarily, naively "clever". (I am probably referring to v1, as I think v2 has that effect mitigated a bit.)
  • Nick ShinnNick Shinn Posts: 1,538
    edited April 22
    Ligatures have remarkable staying power.
    Deference to tradition.
    They are “on” by default in the OpenType spec, and the classic /fi and /fl are part of the basic Mac encoding I’ve been working in since the early days of Fontographer.

    91 years ago, when this ad appeared in Harper’s Bazaar, those ligatures were de rigueur, even for type designs which had absolutely no need for them, such as the new Futura concept. Same for old-style figures. Such things added class and value to fonts and typography, and are statements of quality—a signal from type professionals to their peers, as much as, if not more, for the client or end reader.

    I wouldn’t be surprised if the designer of this ad hadn’t specifically chosen to set the subheadings in all lower case, just to be able to start off a sentence with this wacky /fi ligature!
  • John HudsonJohn Hudson Posts: 1,931
    Both macOS's TextEdit and Windows' Notepad support OpenType features, though only the former offers any sort of control over them. [Not to say that Notepad is an alternative to WordPad though, not really.]

    NotePad is a plain text editor; Wordpad is a rich text editor. TextEdit is both: you can toggle plain and rich text modes.
  • John SavardJohn Savard Posts: 622
    TextEdit, then, is one nice feature of the Macintosh, no doubt among many.
  • Adam JagoszAdam Jagosz Posts: 621
    I wouldn't say TextEdit is nice, as it messes with your file extensions even in plain mode (though probably there's a preference buried down somewhere for that).
  • I remember in an interview that the late Evert Bloemsma had a problem with ligatures. He sort of wanted to keep the purpose of desktop publishing during the time of interview.

    In an interview dated 25 September 2002, he stated this:

    "Desktop publishing (DTP) has lifted type to the meta-level of digital media. Type is now cut off from its physical origins, the roots that determined its shapes: handwriting and letterpress. The return of features like ligatures and ‘old-style’ figures, the revival of monospaced fonts, and the use of ‘rough’ types like Interstate and Bell Gothic for text demonstrate our emotional desire for tradition, rooted in limitations and a certain characteristic imperfection. Paradoxically, DTP intended to liberate us from all this. These contradictions present a dilemma in which contemporary type design has to find its way."

    It was due to him breaking traditions and wanting to free the constraints of the physical medium that I went into typeface design.

    As other typeface systems got computerised, this reality bumped into Indic typefaces, some which have ligatures.
    When I saw how Thai somewhat Latinised in its design paradigm (using Greek as an example), I knew that Lao and Khmer will be next; it's a domino effect.  However, in Khmer, there is a ligature of ប  ា (ba and aa) to បា (baa).  If the modular system is to be followed, a ហ (ha) is formed, and it will not serve Khmer readers well.

    With that, I still see that unnecessary ligatures must still be dropped; typographic letters are not only written anymore, they are typed too with a keyboard and interact with the software.
  • With that, I still see that unnecessary ligatures must still be dropped
    Surely this is a stylistic decision, and the whole point of the flexibility of digital typography is that such decisions can be stylistic — matters of typeface idiom and aesthetic —, rather than dictated by limited technology or ideological commitment. South and Southeast Asian scripts — not to mention Arabic — have often been simplified for technical reasons, and in various typesetting systems traditional features of the scripts have been sacrificed for mechanical facility. Sometimes, those technical limitations have been dressed up as modernism or as script reform, but they're inevitably traceable to the inadequateness of imported technologies to faithful reproduction of the script. My take on this is that if there are cultural moves towards modernism, if there are genuine proposals for script reform to e.g. improve literacy rates, then the technical issues have to be removed from the equation. This is where the flexibility of digital type becomes important, because it enables type designers and font foundries to both give users options with regard to how text can be displayed as well as to contribute materially to those cultural discussions around modernity and tradition.

    While Evert saw the return of features such as ligatures and oldstyle numerals as nostalgia for things 'rooted in limitations', I see them as things to be thoughtfully engaged with in technologies that overcome previous limitations. If you're including ligatures in a font — for any script — just because other fonts have them and you've heard they are, somehow in themselves, 'good typography', then that isn't thoughtful. If you are considering the question of whether these things are suitable for this particular typeface design and how text set in this typeface should look, then I think perhaps the dilemma to which Evert referred goes away, in that contemporary type design doesn't have to find only one way,

    Talking about the technical side, many of the western technologies were positioned using the Cartesian grid.  We still see that legacy in our programming languages and the graphic design programs we have.  This is why there is relative ease in putting European typefaces in one superfamily.  The farm layouting, architecture and city planning of Greece and Rome were based on the gridiron, and I see that the page layout is also based on the same paradigm.
    Most Indic typefaces have a complex vowel positioning due to the radial layout concept derived from the Mandala.  This can be seen with the radial layout of the glyphs.  Some vowels are placed before the abugida itself.  Example is in Thai แล้ว (laew), in which แ is ae, ล้ is lo ling (default sound: lo) and ว is wo waen (default sound: wo).  All the sounds and tones revolve around lo ling, except for ว wo waen.  Some have the feature of subscript consonants (Javanese, Khmer, South Indian).  These features have created interesting scenarios in modernisation of the Indic typaces.
  • John HudsonJohn Hudson Posts: 1,931
    These features have created interesting scenarios in modernisation of the Indic typ[ef]aces.
    And some wacky script reform proposals to better fit Indian writing systems to the imported technologies. Just a few examples from the chapter ‘Notes on the work of the script reformers’ in Naik's Typography of Devanagari (first edition, 1965):

  • Rafael CasesRafael Cases Posts: 28
    These features have created interesting scenarios in modernisation of the Indic typ[ef]aces.
    And some wacky script reform proposals to better fit Indian writing systems to the imported technologies. Just a few examples from the chapter ‘Notes on the work of the script reformers’ in Naik's Typography of Devanagari (first edition, 1965):

    I only know that Adrian Frutiger had to study how to write Devanagari to be able to design it.  I don't know if these guys did their due diligence.  I haven't really read up on Devanagari and other South Asian scripts, but I do read up on the ones derived from the Old Mon, Old Khmer and Old Javanese script, but I look most especially at Thai, Khmer, Lao, Lanna, Mon-Burmese-Shan, Javanese, Lontara and Baybayin.  By looking at numerous scripts, this makes me able to classify by common origin and design elements.  Given the history of Latin, Greek and Cyrillic, as they evolved for more than a thousand years in Europe, as they shared design elements (by convergence or by similarity of tools or even the change of tools), it's unsurprising that Cyrillic and Greek are treated almost like Latin as a design strategy.  If one goes to Southeast Asia, the relationship is that with Khmer, Thai and Lao, Khmer is the odd one out in the features (subscript consonants and ligatures) but the form, space and line breaking treatment is nearly identical to Thai and Lao.  By designing typefaces in script superfamilies, that way, it harmonises and regularises the scripts further, and in addition, it saves more time.
Sign In or Register to comment.