Know any good legibility tests for typefaces?

[dberlowgone]

Bill: "David, I would say a good ’stakes’ would be a good lunch for all the participants of this thread at a type conference five years from now. "

I'm much more interested in targeted cash, and not at all in a bunch of freeloaders getting in on my winnings.
You wanna put your money where my mouth is, that's fine, the rest can make their own bets.

Peter: "The fine art is in knowing where to make the adjustments."
I agree, with the even finer art knowing where not to make the adjustments. ;)

John: "I don’t yet see a complete description of how we read, and that makes me very wary about trying to draw practical conclusions in terms of ’doing things differently’ in type design, which is what we’re all concerned about isn’t it"

It's already been done. The renderings of the OS all went to doing things differently without letting the type or the designers 'at it' properly. A year ago you were in utter denial, now... crappy 'variations' will be brought to us by.... 'filtering expertz'.

Cheers!

[Nick Shinn]

I don’t believe that concepts are there to be balkanized; in other words they don’t necessarily belong exclusively to one area of information or study rather than another.

Phrenology, graphology, the ego and the id have been banished from the scientific sphere, as has the study of race and intelligence.

By the same token, realism lost its credibility in art long ago. Art students don't study life drawing any more, or even drawing. Representation is the preserve of photography.

scientizing typography isn’t going to happen. I am fairly certain that’s a straw dog.

OK, how do you implement a disability policy that provides access for the reading challenged to important documents? Do you pass a Legibility Act that stipulates only certified typographic practitioners may produce such reading material, or one that stipulates certain physical criteria for typography?

The latter is what's happening, and, as has already been pointed out, scientific claims have been made in the marketing pitch for faces such as Read Right and Tiresias, which do not stand close scrutiny.

With regard to the food analogy, this is something I've considered, but I wish you guys would stop making analogies, there's nothing like type!

Good experimentation separates the influence of different variables. The idea you seem to be assuming—that in principle the influence of different variables can’t be separated—is just wrong.

You're assuming I'm assuming. What I actually said was there are too many variables. As Peter put it "...a decontextualized descision based soley on simple or generic affordance micro-advantages [is] short-sighted."

And again, I would ask you to make the distinction between studying reading, and studying readability.
Certainly, the variables can be limited in tests that are designed to address particular aspects of the reading process.
However, the putative study of typeface readability is impossible, because there are too many extraneous factors which skew the results.

OK, here comes an analogy: manners.
Science can study manners--anthropology or sociology --but it's not its job to offer etiquette tips.
Readability is like manners, typographers use certain types and "style sheets" for different kinds of publication.
If the wrong typespec is used, readability will tank.
How can science assign a particular typeface a readability quotient when it is good manners to use it in one kind of periodical, but a faux pas in another? Sure, you isolate the variable and say, THIS quotient IF these circumstances. And so on, as you take account of all the variables--the reader's education and eyesight, where they're reading, what the writing is, and so on, not to mention the typographic variables of size, leading, line length, and paper stock. Do you really believe that the appropriateness of typefaces doesn't vary drastically against one another as such circumstances change?

[William Berkson]

>too many variables

Nick, that amounts to the same thing. The point is that with good experimentation you can get the influence of one variable. Blood is incredibly complex--probably as complex as reading--but medical testing regularly isolates different components to detect and diagnose illness.

>OK, here comes an analogy: manners.
Science can study manners—anthropology or sociology —but it’s not its job to offer etiquette tips.

It may not be its job, but it can help those who want to be polite or to offer etiquette tips. For example, the book Questions and Politeness points out that there is a conflict between being clear, as is a priority in debate for the growth of knowledge, and politeness. The goal in politeness is avoid embarrassment to anybody in the conversation. For that reason vague, open-ended questions, such as 'How do you do', are standard polite questions. Pointed, personal, close-ended questions, like "How much money did you make last month?" and "When did you have last have sex with your wife?" are rude.

I think the lessons for etiquette tips are pretty obvious. As is the challenge of trying to have a debate for the sake of better understanding--such as this one--while maintaining politeness.

As Kant said "There is nothing so practical as a good theory." A good theory is true, and therefore has practical applications. I have no doubt the same would be true for advances in the scientific study of readability.

[ebensorkin]

It’s already been done. Phooey. Not the same thing at all. Letting technical folk try to solve something outside of their base of knowledge is not the same thing as having something studied from a new angle or better put : using a different process; ideally with some input from Typographers. Being technical is not the same as being scientific, even if science is used for creating technology. But your point about it being a mistake to not let "designers ’at it’ properly" is a solid one.

The latter is what’s happening, and, as has already been pointed out, scientific claims have been made in the marketing pitch for faces such as Read Right and Tiresias, which do not stand close scrutiny.And what close scrutiny will show is that it was faux science if it is*. And actually, that will be more useful to say than simply "this is the realm of the humanities"! So if you had more scientists involved the faux science would be easier to debunk.

....the study of race and intelligence All of these turned out to be mistaken theories. "Race" for example turns out to be construct that doesn't hold up under scrutiny. It was just an over-fixation on one variable:skin. Instead; now a more complex reality emerges. So looking at genetics, environment, nutrition, behavior and abilities etc. hasn't stopped. If Readability turns out to be a flawed theory maybe the same thing will happen with type - a more complex reality can emerge.

* I admit that that I haven't spent time looking at either fonts of the marketing associated with them. Perhaps I will this week.

[Nick Shinn]

Eben, here is the definition:

Scientism: The belief that the investigative methods of the physical sciences are applicable or justifiable in all fields of inquiry.

So phooey to you, because "technical folk out of their depth" are the very people most apt to be scientistic.

[William Berkson]

>I’m much more interested in targeted cash, and not at all in a bunch of freeloaders getting in on my winnings. You wanna put your money where my mouth is, that’s fine, the rest can make their own bets.

But you're going to lose :)

In any case lunch will be a pleasure, whether I pay or you do, so that's fine with me.

The problem is finding the referee who's going to decide whether there was significant progress on 'readability'.

>the distinction between studying reading, and studying readability.

If you can study reading, I don't see why you can't study readability. As I said, you aren't going to capture everything, particularly in the beginning, but I don't see why you can't identify factors that contribute to or hurt the ease of reading, which is what readability is. I expect that researchers will be able to identify thresholds where bad leading and spacing start to significantly hurt ease of reading, and these will be quite objective.

[enne_son]

Perhaps a better way to describe what we should oppose is the illusory belief or expectation that adequate or fitting decisions about the right or best course of action can or should be based principally or exclusively on naked evidential knowledge of cause and effect.

[enne_son]

…ease…

The problem with studying readability under your definition is, the researcher has to propose an understanding of ease in order to isolate and investigate. Is ease to be thought of as the sustainability of immersion with comprehension over large periods of time, or is it to be thought of in terms of the lightness of the computational load in neurological processing terms, or is it to be thought of in terms of the rapidness or automaticity of visual wordform resolution, or is it to be thought of in terms of perceptual attentional demands, or is it to be thought of in terms of the abscence of physiological stress.

All these are somewhat more tractable than ease, and probably related, but none of them really covers all we mean in ordinary, everyday terms, by ease, so a result derived from only one of it's more quantifiable dimensions can't bring us to where we want to be. And there is a naturally tendency for that to happen.

[added “with studying readability under your definition” at the begining later]

[William Berkson]

The definition I am using is the one from my late Uncle J. Ben Lieberman's book ‘Types of Typefaces,’ from 1967:

“'Legibility' is based on the ease with which one letter can be told from the other. 'Readability' is the ease with which the eye can absorb the message and move along the line."

Generally as there is progress in science, there is refinement in the meaning of terms, such as the distinction between 'mass' and 'weight' that came in with Newtonian physics. So I'm sure there will be further clarification and refinement in the meaning of 'readability' as more is understood of the reading process. What I am claiming is that there is more to readability than the ease of distinguishing one letter from another. These additional aspects of readability is what I hope and expect there will be progress on.

Any of the alternative description you put forward may win out, but they will all be about 'readability' in the sense that they involve more than ease of individual letter identification.

[John Hudson]

David: It’s already been done. The renderings of the OS all went to doing things differently without letting the type or the designers ’at it’ properly. A year ago you were in utter denial, now... crappy ’variations’ will be brought to us by.... ’filtering expertz’.

I was talking about me doing things differently, not rendering engines. What a rendering engine does might force me to do some things differently, but that's different from reading science directly influencing design decisions. A rendering engine can make my life difficult whether it is based on reading science, focus group response, unfounded optimism about gains in screen resolution, or voodoo.

Is what the rendering engines do, in fact, based on reading science? I don't think it is. Retroactively, some reading science is being applied to figure out whether particular rendering models have advantages over others, and that may affect the future development of those models, but I think the decisions to ship ClearType, CoolType or Quartz, were as anecdotally driven as they come: they made it, they liked the way it looked, they shipped it.

Denial? I'm just a bit slow: slow enough to spend 18 months designing a typeface to address a conflicting set of requirements without really understanding how they are conflicting.

[Nick Shinn]

Nonetheless, it must have been quite a ride.
But what I really want to know is, what's the scoop on Jelle Bosma's Greek and Cyrillic?
Did they fail the readability test?

[ebensorkin]

Nick, under your definition of Scientism it appears that you might be as "guilty" as I of this "ism" because you appear to appreciate Peter's work -as do I. This is in part because as provided the definition itself is more charged than clear.

Certainly if by Scientism you mean that the investigative methods of the physical sciences are to be considered not merely potentially applicable; but further that they are priviliged; then I am am as ready as you are to condemn Scientism. Usually an "ism" supposes a privileged position for the thing being mentioned before the "ism". "Ism"s are just dogma, not theory; and they offer false certainty and stunt curiosity.

This is ( I think what peter is talking about when he says Perhaps a better way to describe what we should oppose is the illusory belief or expectation that adequate or fitting decisions about the right or best course of action can or should be based principally or exclusively on naked evidential knowledge of cause and effect.. Peter please disabuse me if I am wrong.

What worries me almost as much as Scientism though; is is reactionary Anti-scientism. It's no better.

Although I am no doubt flogging a dead horse by now I will say again: It isn't necessary that "the investigative methods of the physical sciences" be either subservient, equal or superior. What needed is mutual respect, curiosity and interest among the various "modes of inquiry" as they used to say back at my college during the 80's. All this bristling and fuming at "Science"* in the thread does nothing to help that along. Still from what I have read I wonder if we don't agree agree on far more than we disagree on. We both have significant respect for the caprices of the eye when making fonts for instance.

Perhaps what is left is mostly a question of how relatively worried or unworried we are about threats from faux-science being wielded by marketing.

* I am using quotes here because as often as not it wasn't Science per se but the claim that it was.

John, Bill & Peter thanks for your interesting posts!

[enne_son]

[reacting to Bill’s] "Any of the alternative description you put forward may win out…"

I don't see them as alternative descriptions but functional-specification proposals. And the issue for me isn’t which wins out but what gets looked at, and how what gets looked at is named. They all need to acknowledged for what they are, and explored for how they function. My prefered genral category to describe the domain I’m in is perceptual processing in reading.

The danger with designating a science of legibility or readability as the research goal is the danger, in the case of readability, of turning a value into a number, and in the case of legibility, af creating a reliance on threshold statistics (in the domain of perceptual discrimination affordance) for relevant information about functionality at a plateau (in the domain of visual wordform resolution).

That being said, I'm all for constructing legibility quotients and readability profiles. I'm just sensitive to how they are constructed and used. If concerns for readability and legibility are included in the brief, these shouldn't be a threat. And they shouldn't be all there is to consider.

[dberlowgone]

"Phooey."?
I was at an early publication of the latest mass discontinuity between test type and user: first the dreamer gave his ideas, then the scientist gave his scientifically studied presentation, then the type designer showed how simple it was going to be to make fonts, and then the typographer showed beautiful stuff a bit too far away to see. Kevin, before he was Kevlar, gave the scientific explaination for reading, which he still believes and which still inspires this CT effort, I think. It revolves around letters, not sylables, words, or lines — people read letters. I got close to him at Typecon, not just because I like him, but also to make sure he would say it three times while I was right there. "Chinese?" "No," "the word 'a'?", "NO," "youth vs expertise?" "Noooooo, people read letters one saccade at a time!

So, unless 'Ready Rendering Skills' lead science to bad reading science, and influenced type design decisions, an unthinkable thing, then Science did it. ;) By, ready rendering skills, in this case I mean the ability, for better or for worse, to cleave the resolution into components, not triple the resolution, as is so often claimed. Read the MS 'white paper' on anti-aliasing type, which 'scientifically' led this off. It measures, prods and pokes around the solution without ever mentioning type, or typography. This last study, if Sheedy only lets MS change the filter, (and not the oil or driver;), it is not going to change the fundamentally saccade-hiccupic nature in low resolution text across the Windows universe. This is not my fault, who ever let that French guy name this bluddy eye movement? it's their fault.

" I’m just a bit slow:"
Me too, has something to do in my case, with having to ceaselessly pry into the seemingly endless 'overlays' upon the concept of text for readability. I'd give up, but that each overlay leaves such fertile ground for custom work...I'm just a busy bunch of monkeys in the orchestra of type.

"...is what the rendering engines do, in fact, based on reading science? I don’t think it is."
I great question, No, in the strictest sense rendering engines do not do anything based on reading science, they are non-sentient e-world objects. But, who, or what, are these scientific studies of CT done for, if not for the engineering staff(s) who render rendering engines? Kevlar just said, 5 filterings were tested in their scientific study, and 2 filterings were released to the developing public. Maybe he knows then? Is what the rendering engines do, in fact, based on reading science? Or, are the renderings which are released, released based on reading science?

Cheers!

[Nick Shinn]

What worries me almost as much as Scientism though; is is reactionary Anti-scientism. It’s no better.

I am not a reactionary, I am progressive, and that has nothing to do with science. I just think we should move forward and leave a lot of past garbage behind (suitably recycled).

I am not anti-science, I am "anti-scientism", because I dislike all forms of totalitarian-ism.

which he still believes and which still inspires this CT effort, I think.

Well, if the folks in the Microsoft typography department want to stay busy and keep their jobs, they have to have something to develop. The boss will fund projects with measurable results; he is an engineer and runs a large-volume, small margin business, so that's understandable. Having a scientific readability component in a type-development project, one that is able to produce a result such as "5% faster", does the trick. This much I gleaned from the CT documentary video that MS published.

So although the people involved may be variously inspired, the CT project itself exists as a condition of the way the business works.

[eliason]

So I am happy to point out the obvious counter-argument of all the heavily processed food made with the help of “food scientists”. It’s a hell of an counter example.

There was an interesting show on public radio this morning about how food science (and marketing, and journalism) encourage people to eat much less healthfully. I know this seems off-topic, but I think the analogy might have some legs, and indeed Michael Pollan (the author who was the guest on the radio show) seems to be identifying similar dynamics in the food supply world that Nick is calling out in the type world.

[William Berkson]

Francis Bacon said: "Knowledge is power." Scientific knowledge is power that can be used for good or ill. It is up to us to use it wisely.

The romantic reactionary view of science always points to the misuse of science to oppose or limit science. Science is misused, but I don't think that's an argument against scientific research, unless humanity is hopeless at managing the added power that comes with science, and science-based technology. It is a serious question whether we can manage scientific knowledge for the good, but I don't see a lot of volunteers to become subsistence farmers.

I would like to see an effort to use science responsibly rather than abandon research.

Incidentally 'scientistic' according to the wikipedia article can mean pseudo-science or the misuse of science, and it can also mean the view that the methods of the physical science apply elsewhere, such as in the social sciences.

How far the methods of the physical sciences can be applied beyond their origins is an interesting question*, but in the case of cognition and perception I think the evidence is clear that it is possible to develop testable theories and test them.

*You can access my article on the subject here if you are interested.

[Nick Shinn]

The romantic reactionary view of science

It is a serious question whether we can manage scientific knowledge for the good, but I don’t see a lot of volunteers to become subsistence farmers.

Enough already with the false dilemmas!
I'm not a romantic reactionary--I'm making a distinction between reading research: good, and readability research: impossible.

And you don't have to be a subsitence farmer to question the agri-business complex.

[Kevin Larson]

David wrote: Kevlar just said, 5 filterings were tested in their scientific study, and 2 filterings were released to the developing public. Maybe he knows then? Is what the rendering engines do, in fact, based on reading science? Or, are the renderings which are released, released based on reading science?

The five filtering are all options that have been available in MS Reader for many years. Only the three most popular were made options in WPF, though I would have preferred all five been made available as options. The ClearType filtering option that was used in WinXP was selected by our typographers.

Our team now consists of two typographers, three engineers, and one scientist. We now have the ability to collect data from readers, which we use as an additional source of data for making decisions. As I wrote in the engineering magazine IEEE Spectrum, each of the three branches of our team plays an important role in improving the reading experience. There isn’t a fight between science and typography.
http://spectrum.ieee.org/print/5049

Cheers, Kevin

[William Berkson]

>readability research: impossible

As I have already explained, your argument against readability research doesn't hold water. 'Too many variables' is not a good argument against further research into a problem because it can always be said, including about problems where it has already been proven wrong by successful solutions.

When there are new insights in theory, and new experimental designs, then important variables are identified, and their influence tested for.

[Nick Shinn]

William, you can generalize till the cows come home, but it won't get you very far.

Do you really believe that it is possible to test enough of the variables involved in reading to create an index of readability for typefaces, which has any practical value?

The same is true for any of the variables.

Is 9 pt more readable than 10 pt?
--By people aged 16 or 60?
--With high or low IQ?
--For serif or sans faces?

Is 10 pt leading more readable than 12 pt leading?
--With old style or modern faces?
--With 8 picas line length or 28 picas?
--Ragged or justified?

Is Perpetua more readable than Times?
--On the bus to work, in bed, or in a reading lab?
--In bright daylight or poor artificial light?
--On newsprint or coated stock?

Is sans more readable than serif?
--for a novel or a technical manual?
--for paper or lcd monitor (and at what setting)?
--for short or long paragraphs?

And so on.

[William Berkson]

>index of readability

Nick, I never mentioned anything about an index of readability. You are arguing against a straw man.

I argued that there would be scientific progress in understanding readability.

Your options are so primitive that you must think scientists are required to be stupid, which is not the case. You might just as well argue that we can't prescribe corrective lenses because there are so many different ways we can see badly, and so many causes of bad vision. But they do measure these many variables, and prescribe successfully.

As any one who studies typography at all knows, the different factors of type face, leading, spacing etc. interact to produce a product which is more or less readable.

Scientists can test for a combinations of these variables, or for other variables which they might be able to isolate.

For example Bill Hill of Microsoft mentioned at TypeCon that the reason for the 65 character line width of a single book column being ideal, and more than 75 a problem is that at normal reading distances we have to turn our heads, and not just our eyes for longer measures. I believe the argument was that we are less efficient at moving along the line or to the next line, This is a kind of thing you can actually check by seeing how the readers' eyes move, and their change in reading speed.

I don't know whether this has been tested, but according to you it would be pointless and worthless, I gather, because it is 'readability research'--it tests how measure affects readability.

[ebensorkin]

To shift the debate into what I would guess might be more useful territory let me ask: What specifically would you be interested in having studied? Serif shapes? Contrast? Rendering schemes? What ( if any) studies if any have been most influential on you?

And to move away from dubious gladiatorial battles between fonts that we all seem to agree are not to the point - let me see if we agree instead to something else: That if something about type is to studied via a scientific process that the best idea is to take a single typeface as a control and then make a specific micro typographic adjustments to a version of it. A study could be serif oriented, or contrast or a whole variety of things. This would be a way of reducing variables massively but obviously not completely. But most important it would be a way of doing science useful to typographers. And not coincidentally it paralelels the way a type designer works. Of course it would be prudent/useful not note the variables you think are still in play as well, demographics for instance.

In a way I was a little hesitant to even mention the idea -it is no doubt the obvious thing. In fact I can well imagine Peter or Kevin rolling their eyes. Still, I am interested to hear if you do or don't agree.

Personally I would be interested in seeing individual examples contextual alternatives tested in this way because I am guessing that holds special promise to increase comfort. What do you think?

[Nick Shinn]

Your options are so primitive that you must think scientists are required to be stupid,

What's your point, Bill? ;-)

And to move away from dubious gladiatorial battles between fonts that we all seem to agree are not to the point

Sorry Eben, that is the point.
Bill says we're not talking about the scientific production of an index of readability for typefaces.
But please explain how measuring the "whatever-ability" of a group of items does NOT result in an index/table of their relative merit?

Eben, the test you are suggesting is, as you say, appropriate for font development, where the variable is a feature of a particular typeface. If you varied the typeface, how would you know what realm of difference you were measuring? If typeface A has bigger serifs than typeface B, how do you know it is serif size and not some other difference you are measuring? If all that's different is one article of vocabulary, such as serif style, it's the same typeface.

I'd say you need at least two or three co-variant themes of a typeface to be different from one type to another, for it to be considered an original design.

[John Hudson]

To shift the debate into what I would guess might be more useful territory let me ask: What specifically would you be interested in having studied?

Reading is a human activity carried out every day by millions of people who speak and think in many hundreds of languages and who read and write in dozens of different writing systems. This suggests to me that any understanding of the mechanics of reading that does not involve empirical data from multiple scripts and languages is not really an understanding of reading per se, it is only an understanding of reading a particular script (and most likely a particular language written in that script, given the limitations of most studies). So what I would like to see is studies that are designed to produce comparable data for reading in multiple writing systems.

I have an hypothesis, which is that word or phrase recognition is built up from recognition of the base distinct 'atoms' of any given writing system, with the latter recognition involving sub-atomic role architecture and/or molecular (multi-atom) compounds as needed by the individual reader encountering a given text.

The interesting aspects of this hypothesis, I think, is that for some writing systems I don't think we know yet what constitutes an atom. For the typical typographic Latin script I'm pretty sure the atom is the letter, and studies seem to confirm this. But what is the atom in Chinese? The ideograph or the radical? What is the atom in Hindi? The letter or the syllable? What is the atom in Arabic? The letter or the segment? And, for that matter, does the atom of the Latin script change depending on whether one is reading disconnected typographic letters or connected script?

I have another, contradictory hypothesis, which is that there are no atoms, there are only waves, and what we make use of in reading varies all the time depending on a large variety of in-textual and extra-textual factors.