Legible type design literature/scolars

Mikhail Vasilev

I've started to write an essay on the topic of legible type design.
Main focus is on glyph structure and determining the optimal shapes for legibility/readability in general.  
It is practical work, based fully on experience. The essay should document the results and attempts at forming a more or less structured theory based on the results. At least there are some indications that it is feasable at this point, and that is my major motivation to start with documenting.

The thing is, I haven't found much, if any, directly related literature. It would be good to mention some of other scolar's works but nothing notable yet is found in the Web.
Maybe some of you know such books or articles? I mean directly related to what I do.

Since I haven't found much, it feels like almost no progress happens in the field (after some successfull movements of 16th century whithin Latin typefaces). Was not to expect in our high-tech world, but it seems it is so.

Craig Eliason

Matthew Lukiesh at General Electric and, more recently, Kevin Larson at Microsoft might be good leads.

James Puckett

Reading Letters by Sofie Beier. 
“Readability: discovery and disputation by Berkson & Enneson in Typography Papers 9.

Johannes Neumeier

My MA was touching on that topic, PDF available for reading from the link. Check the bibliography

Daniel Calders

Hi Mikhail,

A ton of research has been done (and continues to be done) by Dr. Sofie Beier. You can find lots of info at her website (http://sofiebeier.dk/). Most of her research articles and papers can be found and downloaded from Researchgate (https://www.researchgate.net/profile/Sofie-Beier).

Good luck with your essay!

Mikhail Vasilev

Сolleagues!
Good to see your interest in the discussion!

Sofie Beier is already in the list of interest.

However let me elaborate on the topic. There might appear some misunderstanding.
My topic is typeface _in general_. Not Latin or some other specific existing type.

See the crucial difference:

1. "Legible typeface design"
2. "Legible Latin typeface design"

The task definition of my work (1) is:
  "to determine the glyphs for optimal (maximal) readability in general"
 
The task definition of most works found by web search (2):
  "to tweak the strokes of Latin (or other historical) glyphs for better readability"

Yes, some of the studies try to "dig into the surface", still initially the _task_ definition is completely different.
I do realize there is a lot published on e.g. "Times vs Arial" and such, but that is not the topic.

Mikhail Vasilev

Craig Eliason said:

Matthew Lukiesh at General Electric

Which of his books is closer to the topic? I've tried to find "Light, vision and seeing..." 1944 but could not find any open access copy. Some sites mention it is possible to request by institution. I am not an institution. I am individual.

Craig Eliason

Mikhail Vasilev said:

Craig Eliason said:

Matthew Lukiesh at General Electric

Which of his books is closer to the topic? I've tried to find "Light, vision and seeing..." 1944 but could not find any open access copy. Some sites mention it is possible to request by institution. I am not an institution. I am individual.

If I recall correctly the Berkson and Enneson article mentioned upthread summarizes Lukiesh's work on this topic. 

James Puckett

Mikhail Vasilev said:

Craig Eliason said:

Matthew Lukiesh at General Electric

Which of his books is closer to the topic? I've tried to find "Light, vision and seeing..." 1944 but could not find any open access copy. Some sites mention it is possible to request by institution. I am not an institution. I am individual.

If you have a local librarian ask them if they can arrange an inter-library loan. Sometimes you can get books from academic libraries that way.

Helmut Wollmersdorfer

Johannes Neumeier said:

My MA was touching on that topic, PDF available for reading from the link. Check the bibliography

Excellent work, easy to read.

Helmut Wollmersdorfer

Mikhail Vasilev said:

[...]
My topic is typeface _in general_. Not Latin or some other specific existing type.

See the crucial difference:

1. "Legible typeface design"
2. "Legible Latin typeface design"

The task definition of my work (1) is:
  "to determine the glyphs for optimal (maximal) readability in general"

- Then you need to find similar articles for each writing system OR
- try to find some for general aspects in the field of design or cognitive psychology

The work of Sophie Beier focuses on design with empirical methods of cognitive psychology.

The principles for alphabetic scripts (writing systems) like Cyrillic, Greek, Hebrew etc. should be the same. Maybe Asian scripts like Han, Hangul or Indic (e.g. Hindi, baseline on top) are special in some way. Arabic is special in many ways.

You can begin at basics of visual (=human) recognition works. This falls in the science of physics, medicine (neurology, anatomy). In short: a complicated process recognising color, contrast, edges (contours), skeletons, angles, corners, weight (area), geometric components and does something with it. It can adapt very fast (see in Beier "familiarity") to new letters and symbols or fonts. There is also influence of a language model which allows fast reading (and not recognising mistypings or reading similar words). We can also read words if some characters are missing, reordered, damaged, degraded (overinking, fading), warped. Adapt to mirrored or rotated text.

Johannes Neumeier

Mikhail Vasilev said:Some sites mention it is possible to request by institution. I am not an institution. I am individual.

If the publication in question is by an author that is active and you can find their contact information, many researchers are also more than happy to share pdf manuscripts with your directly or even send you a paper copy. Had both happen myself more than once, just need to ask nicely. Many of the journals are gatekeeping information, but the authors are free to share their research outside of those journals — after all, it propagates their work and influence in the field.

Also use google scholar search, many papers are available from there, "by mistake", as it were, and the reference count gives you some idea about how relevant a paper is.

Thank you, Helmut

jeremy tribby

"Size-specific adjustments to type designs" by tim ahrens and shoko mugikura is an excellent book on the subject

Thomas Phinney

Mikhail Vasilev said:

However let me elaborate on the topic. There might appear some misunderstanding.
My topic is typeface _in general_. Not Latin or some other specific existing type.

See the crucial difference:

1. "Legible typeface design"
2. "Legible Latin typeface design"

The task definition of my work (1) is:
  "to determine the glyphs for optimal (maximal) readability in general"
 
The task definition of most works found by web search (2):
  "to tweak the strokes of Latin (or other historical) glyphs for better readability"

Yes, some of the studies try to "dig into the surface", still initially the _task_ definition is completely different.
I do realize there is a lot published on e.g. "Times vs Arial" and such, but that is not the topic.

One thing that is not often discussed directly, because we are embedded in it…. Legibility is about how easy it is to decipher which character is represented by a given glyph. This is largely dependent on what all the possible characters are, and the particular ways a given character differs from the other possibilities.

That is to say, in most real world cases, legibility is substantially dependent on what the range of possible characters is.

Let’s simplify it an awful lot for discussion purposes. Let us say a character set is only two characters. What makes those characters legible depends immensely on which two characters they are. Being boring and using Latin for a moment, what makes a “P” legible will depend immensely on what the other possible character is. What if it is R? A? O?

Now, what is interesting to me is that if you abstract too much above this, perhaps you ALSO won’t discuss this issue, because you don’t want to talk about character sets? I hope that is not the case.

Mikhail Vasilev

Thomas Phinney said:

This is largely dependent on what all the possible characters are, and the particular ways a given character differs from the other possibilities.

Yes and no. I have seen a lot of scolars speak about pairwise discernability (if there such a word) and that is actually normal to think about, when you approach the task but in the end and what follows from my results, that is not the most important factor in the design process. (Again, I am oversimplifying the matter, since it is all quite complicated in general)  

Actually my main thesis and result is that there are specific glyphs, kind of geometrical constants, that just work better and my work deals with describing those. The amount of them will be simply the amount I get. It is around 20 I can tell.
I call them "primal" or "reference" glyphs. Those that come out with pairwise similarities will be treated further somehow, but that is another story.

Thomas Phinney said:

Now, what is interesting to me is that if you abstract too much above this, perhaps you ALSO won’t discuss this issue, because you don’t want to talk about character sets? I hope that is not the case.

Certain application of glyphs is currently out of scope. In the end, e.g. for a phonetical coding it would be more like a matter of choice. Namely taking these "primal" glyphs and see what can be done to compromise the shortcomings. Either by extending the set by "not-so-good" glyphs or something, I haven't come to this point anyway. If much more than that amount is needed, then, well, you are out of luck .  And speaking of pure knowledge, these exercises are not so important at this point.

Thomas Phinney said:

That is to say, in most real world cases, legibility is substantially dependent on what the range of possible characters is.

Let’s simplify it an awful lot for discussion purposes. Let us say a character set is only two characters. What makes those characters legible depends immensely on which two characters they are. Being boring and using Latin for a moment, what makes a “P” legible will depend immensely on what the other possible character is. What if it is R? A? O?

That is connected with application. Yes, if we speak about special cases, like e.g. binary numbers representation, special tricks can be applied.  
For binary numbers I personally use the "dash and knot" presentation which is for example a string like "---s  -s-s" for the "0001  0101". The fact that there are only 2 digits, can be used to exploit the weight difference for legibility. But for greater amount of glyphs, one cannot use this "loophole". So in other cases, have to deal with, basically, optical properties of individual glyphs and determine the best working ones  regardless of others*, and this approach is what gave good results in the end.
(* that is an oversimplification of course, dependency arises, especially when adapting the glyph for close placement etc, but not major _structure_ dependecy)

For binary numbers, see above dash-and-knot example, e.g. I can pick other glyphs as a "knot" without losing much, but even in this case it is better to base off of those "primal" glyphs.

Helmut Wollmersdorfer

Thomas Phinney said:

One thing that is not often discussed directly, because we are embedded in it…. Legibility is about how easy it is to decipher which character is represented by a given glyph. This is largely dependent on what all the possible characters are, and the particular ways a given character differs from the other possibilities.

That is to say, in most real world cases, legibility is substantially dependent on what the range of possible characters is.

Sure, but we have some reasonable sizes of character sets, e.g. 70-80 in more trivial European texts, English more like 65. Newspapers are above 170 if foreign names are written with accents.

Mostly readability is more important. But if there are unknown words in the text legibility is important to decipher the spelling. In more formal texts (science, math, juridical) spelling and punctuation can be very important.

We have the approaches of Sophie Beier measuring "Typeface Legibility: Towards defining familiarity". That's not complete but good enough.

Also Johannes Neumeier tried to design "Legibility in typeface design for screen interfaces". That's an example of his work:



And that's not very different to what I did a few years ago to choose a fixed-width font for editing program code and proof reading of historical transcriptions, because low legibility costs time and strengthens the eyes:



I decided to use Menlo, but would like to have the enclosings ([{}]) of Monaco. Or Monaco with the \a of Menlo.

Maybe i will compile a font with "best of" fixed width shapes.

Yes, legibility can hurt elegance. 

Nick Shinn

My scholarly 10¢:

There are no ‘optimal’ letter shapes, any more than there are optimal notes or optimal instruments in music. It’s how the elements are put together that’s critical, and this is a matter of taste, not any theoretically objective measure of ‘readability’ or ‘legibility’.

In other words, type designers rely on their individual good taste to create functional fonts, and such things as formal aesthetics (e.g. texture and tonality en masse), and allusive qualities (e.g. an impression of elegance or a particular cultural ambience) are indeed functional, not just disambiguation between I, l and 1). And typographers rely on their good taste to format selections of fonts into layouts that will appeal to the taste of prospective readers, to the extent that documents will be sufficiently palatable for the nutrients to be digested.

Illegibility is like toxicity: it’s the dose that makes a poison. For reading, the dose is measured scientifically in time—the faster text may be deciphered, the less toxic it is; science mostly prioritizing speed as efficiency. But why not slow down and smell the roses? After all, there is another scientific theory, disfluency, which posits that illegibility is not without virtue and thus no doubt may also, in certain circumstances, be optimized!

To Thomas’ point about the inter-relation of characters, there are in fact optimal suites of letter shapes that have emerged in the field—through marketplace empiricism, one might say— and these are the accepted genres of typeface, identified by categories such as didone, geometric sans, and grotesque. These genres constitute a taxonomic zone between typefaces and alphabets.

For instance, the ‘Continental’ version of Gill Sans may be considered a set of alphanumeric shapes for geometric optimization, while the original was a humanist optimization. Which of these is more legible, more readable? If that can be decided at all, it depends entirely on context—the typography of a particular document, the nature of the media, the reading proficiency of the target audience, the text content, the cultural milieu.

Clearly, of course, the geometric has a legibility deficit when compared with the humanist, as the round \a is much more similar to \o than the two-storey version. But so what? If this were an issue, the genre would not be as pervasive as it has proven. 

Optimization, then, is nothing but good, appropriate style.

 



 

Mikhail Vasilev

Nick Shinn said:

My scholarly 10¢:

There are no ‘optimal’ letter shapes, any more than there are optimal notes or optimal instruments in music.

What do you mean here by 'optimal'? In case you address legibility alone, then my thesis is exactly the opposite to your statement. See my previous answer to Thomas' comments. 
With legibility by the way, I personally understand the intrinsic (physical) ability of the glyphs to contribute to readability. And as I see from special literature there is tendency to use this term with that meaning. Main purpose of the term, when used in studies, is to separate from other higher level aspects of readability and try to abstract from application. Although with some interpretations from author to author, but in general there is no big disagreements here as far as I understand. Yes, some of authors define legibility as pairwise glyph discernability, but in my view the latter is not synonym for legibility, but rather a [resulting] property contributing to legibility. One of major but not the core.

The task is quite specific and technical and initially has very little to do with context, taste, stylistics or historical establishments. In my case, focusing mainly on glyph structure. It is mostly related to physical (optical) properties of the image.

Yes, the tools I use to determine the glyphs is my eyes and experience, but nevertheless.

Nick Shinn said: But why not slow down and smell the roses?

It's like saying: why to work on this at all? Well, I have interest in it and it is all useful, not only for reading efficciency exploration, but also as artistic inspiration that can be applied to existing fonts.

Nick Shinn

Please pardon my distaste for reading science.
I like to have a rant, once in a while!

Without peripheral context: I vaguely recall that there is a method of reading sentences in which individual letters are rapidly flashed, in the same position. Apparently one can learn to read with some fluency using this method. Perhaps that might be useful to you.

Helmut Wollmersdorfer

Mikhail Vasilev said:

The task is quite specific and technical and initially has very little to do with context, taste, stylistics or historical establishments. In my case, focusing mainly on glyph structure. It is mostly related to physical (optical) properties of the image.

IMHO 

Charles Bigelow, Typeface features and legibility research, Vision Research 165 (2019) 162–172 https://www.sciencedirect.com/science/article/pii/S0042698919301087?via%3Dihub

is a good overview on 11 pages what happen the last 100 years in this field. It's a sort of meta study with many references.