Know any good legibility tests for typefaces?

[ebensorkin]

But please explain how measuring the “whatever-ability” of a group of items does NOT result in an index/table of their relative merit? This is pretty vague.

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? Serifs are particularly complicated. I won't say they can't be studied though. There are lots of aspects to break down if you were so inclined.

If all that’s different is one article of vocabulary, such as serif style, it’s the same typeface. Not the same. Very closely related.

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. Ah yes. But that isn't how you build up a strong case for something scientifically. Science would be better off not making a new design - that's your job. Instead science must take little baby steps. And at last with many many tests perhaps enough data is collected & interpreted so as to be useful to you. Maybe a whole matrix of these co-variant themes could be built up...

Sorry Eben, that is the point. I think it isn't so much the point so much as it's your favorite bugbear. Maybe after I read more about scientistic typefaces & the marketing of them I will get more worried. But even if I do it won't make me think you can always save people or committees from seeking false security in scientism. You can try. But it isn't going to work every time. I will also doubt that can you always save them from marketing hype. And even if you could; neither goal is advanced by demonizing science's potential participation in type. Instead it is the thinness of the claims that is to be pointed out, and counter examples offered and criticisms made - ideally both humanities and science based.

I am interested not in listing the hot type for a all purposes but in understanding how to design better fonts for a particular purpose, in a particular environment for a given group of folks etc. But I think that willfully leaving the potential benefit/information that science might offer out my bag of tricks is simply nutty. There are aspects of type design that I would just assume not even ask a scientist about because my sense is they are not interesting questions. On the other hand I bet there are some aspects of type that have eluded the best efforts of the humanities to grip that might be better plumbed using scientific techniques. To reiterate; I am not proposing that we worship science in our design process', and nobody here is that I can tell. But I do think there is room for them to help or collaborate with us.

[ebensorkin]

Thanks John!

[Steve Tiano]

What an interesting discussion to have walked into. It seems to me that—as is almost always where matters of aesthetics are involved—is finding some kind of ideal balance between the objective and the subjective. And I just don’t know that there’ll ever be a solution that works for all. Does make me want to re-read Colin Wheildon’s Type and Layout: Are you communicating or just making pretty shapes, however.

This discussion also makes me realize how easy it is to become lost in the trees, when there’s a forest all around that bears dealing with somehow. I’m glad I’m mostly busy enough making books to have time to consider this to the degree I might if I had more time.

I envy you folks your self-discipline, being able to take part in this running conversation and then move on to your work. I’m going to have to force myself to clean up and get ready for work this morning, instead of re-reading the whole discussion very slowly.

[William Berkson]

>But please explain how measuring the “whatever-ability” of a group of items does NOT result in an index/table of their relative merit?

Since multiple factors are involved in readability, I would not expect a single index that applies to all situations. But I would expect an indication of what is more readable in specific conditions when other variables are controlled.

For example, I would not expect differences to show between headlines of a few words in serif vs sans serif, when there is similar x-height, stem weight, width and spacing. However, a common view is that sans work less well for extended reading at text sizes, and specifically they require more leading to work decently. So if you control for x height and weight, I expect that serif type will be more readable at normal text sizes with tight leading. When we have good measures of readability this kind of difference will be detectable. One simple test will be whether the eye loses the line more often with tight leading in a sans that is otherwise similar to a serif. Also I have already suggested that testing the decline of comprehension of extended text, with time, might be another way to get at it. Another is to test for people's ability to catch errors in proof reading.

Also legibility is one factor in readability. I think that the tests of Clearview did show that they were more legible and more readable at distances under night conditions, particularly for older drivers. It doesn't mean they will be more readable in extended text at small sizes.

>What’s your point, Bill?
My point is that scientists will control for some variables to test others. In your post, you seemed to overlook that this possibility, a standard thing to do in scientific testing. The ability to control for some and detect others requires creativity. You assume or assert that scientific method is helpless in such situations, and as I repeatedly have pointed out, it's not.

[enne_son]

What this discussion — to my mind — shows, is that we need a broader, more incisively denominated and diversified approach to functionality and value than the blunt instruments of ‘readability’ and ‘legibility’ provide. Considerations of ‘readability’ and ‘legibility’ can get us started, but a table of comparative measures can’t be where we end up.

[Nick Shinn]

John, I have a similar understanding as your wave and particle theories.

I would also compare the past to non-latin scripts, although we can't go back and test how we used to read.

I think it isn’t so much the point so much as it’s your favorite bugbear.

I thought bashing bundling and Helvetica were supposed to be my "favorite bugbears".

[Nick]: If all that’s different is one article of vocabulary, such as serif style, it’s the same typeface. [Eben]: Not the same. Very closely related.

If it's not the same typeface, but as close as that, someone's plagiarising (or there's a "revival" going on).

But I think that willfully leaving the potential benefit/information that science might offer out my bag of tricks is simply nutty.

As I've repeatedly said, I'm not proposing ceasing readability research, and behavioral testing may be very useful in product development. But I don't believe that performance rankings of the factors involved in typography will ever be any use to graphic designers and typographers.

You assume or assert that scientific method is helpless in such situations, and as I repeatedly have pointed out, it’s not.

I've given many examples of such situations and asked how the variables would be separated. You've beaten about the bush, generalized and analogized, and stated that scientists would use creative test-building. That's wishful thinking.

Let's return to Eben's example of making a study serif-oriented, and my rebutal. I do not consider it possible to isolate the variable of "serifness" from that of "typeface".

It would be possible to vary the serif of a typeface (Penumbra comes to mind) and test that, but testing, for instance, the "readability" of slab serifs versus bracketed serifs by testing Clarendon against Times makes no sense, as the test may be measuring the performance of angle of stress, or degree of stroke contrast, or something else, or combination thereof.

I doubt that a definition of slab serif is even possible, as may be infered from looking at the FontBook.

Another variable that effects the quality of serifness is inking/gain/raster mode. So for instance, you decide to test serifness against ink gain, varying Didot serif thickness (or length?) at 9 pt, on a particular stock, printed on a particular press, varying gain. With a certain class of readers, and a certain kind of text content. Let's assume you do this by modifying Bauer Bodoni. Well, I doubt the result would even hold for Berthold Bodoni. And what use would such information be to a designer, when all the variables of a particular job are different from those that weren't varied in the test scenario? Isn't there a scientific principle related to the "isolatability" of multi-variant phenomena? What do meteorological theorists call it? The science is getting soft, so rather than hard outcomes, probabilities.

I suspect that if the massive task of testing all the variables of reading against one another, for individual typefaces, were done, the best that may be said would be that there is a 30% probability that ITC Garamond is more readable than Adobe Garamond.

[John Hudson]

Nick: I would also compare the past to non-latin scripts, although we can’t go back and test how we used to read.

Robert Bringhurst wonders if people read differently when all literate people were writers as well as readers, i.e. if the manual familiarity with the making of letters, in a pre-typographic environment, affected reading.

[Nick Shinn]

William St Clair, in his Reading Nation, recounts how restrictive trade practices in England in the 18th and early 19th century made books too expensive for most people, and how it was quite common for readers to circulate "pirated" versions of works (especially narrative verse poems) in handwritten form.

[ebensorkin]

Nick I am glad to see your argument is starting firm up. I am also struck at, and also unsurprised by your ability to imagine testing scenarios - which is great. There is something to the idea that you could as both a scientist and a typographer talk about types in a soft way or perhaps in an informed manner ( informed differently ) rather than a stiff, absolute, list oriented manner. If the reality is complex this is the only accurate way to talk about them. This doesn't preclude having meaningful things to say. It's just that these things will be rich/complex. This, as opposed to, and in a sense as a substitute for what hellbox was originally after.

It may be that you have acquired a 3rd bugbear. If 2, why not 3? I am with you for all 3 as far as I can see. I just don't stamp my foot as hard, and in the case of the most recent one I have a substitute in mind. It is starting to look as if you might too.

Peter, Thanks. I agree that "a table of comparative measures can’t be where we end up" is right. For one thing that simply isn't how the science would be done. Each experiment will have multiple ways in which it is out of sync with others. But what you can say is that if a variety of scientists run a variety of tests geared to a given hypothesis what you would be looking for is for the series to begin to accumulate support for or begin to undermine that hypothesis. This is a simple scenario. More complicated ones exist! You would be hoping to be able to say this hypothesis seems to hold some water or it doesn't. Sometime the hypothesis itself or the methodology was ill conceived and you don't get to say either thing. And there is room to disagree about these things. So the reason you cannot have a list is that Science ( and most fields of inquiry) is more complicated than a list or matrix can describe. Peter, Kevin, is this sort of fair so far?

But if you have successful experiments for a given question and you go on to do another series and another after a long building on those experiments that seem to actually show something useful you can begin to have a picture emerge of what seems to be going on. Do enough of this and a theory can start to look pretty solid. A theory like evolution for instance.

It will be a very very long time before all the variables of reading have hypothesis' and have been tested out meaningfully. And before you can attempt that you have to have a theory about what those might be etc. So at our current rate of burn the mere possibility of doing a meaningful ITC Garamond vs. Adobe Garamond "readability" test must be 1000s of years away. It's an awful lot of work for a pretty banal question. That's why I think that targeted testing of what seems like an interesting or useful variable that is hard to crack by other humanistic means has to be where it's at.

That is why I would agree that isolating the variable of “serifness” is probably not that interesting or useful. On the other hand I do think it would be interesting to craft experiments towards a better understanding of what is going on with serifs and different serif designs and so on. There are many ways of looking at that.

John, I think Robert is probably correct.

[Kevin Larson]

Eben, I completely agree with your assessment:

But what you can say is that if a variety of scientists run a variety of tests geared to a given hypothesis what you would be looking for is for the series to begin to accumulate support for or begin to undermine that hypothesis. This is a simple scenario. More complicated ones exist! You would be hoping to be able to say this hypothesis seems to hold some water or it doesn’t. Sometime the hypothesis itself or the methodology was ill conceived and you don’t get to say either thing.

Science is not a complete collection of facts, but a fluid process for learning about a topic. It requires creativity and a deep understanding of the topic. A study or piece of data is not the final product; It is the starting point for learning and developing more accurate theories.

[William Berkson]

>You’ve beaten about the bush, generalized and analogized, and stated that scientists would use creative test-building. That’s wishful thinking.

I am not doing wishful thinking. My point, which you have not responded to, it that your 'too many variables' argument has been refuted so many times in history that it's not credible.

Over and over in history it turned out to be foolish to say 'science can't study this' about any phenomenon. Maybe I am misreading you, but when it comes to readability I am very surprised that you appear to be taking the position of the medieval church on Galileo.

Contrary to what you say, I have also been specific on testing readability. I have proposed two ways to test for the readability of serif vs sans serif text size types in extended text with tight leading--using a decline in comprehension with time and a decline in quality of proof-reading as measures. I don't know if either of these measures will turn out to be good, though I think they have a shot. Can you explain to me why in principle these and every other measure will fail?

[John Hudson]

Maybe I am misreading you, but when it comes to readability I am very surprised that you appear to be taking the position of the medieval church on Galileo.

To clarify, what was that position? I ask because a lot of people think they know what the issue was about, but most of them have not read the trial transcript. You might have, Bill, and you might be ascribing to Nick a position that is more subtle than people might assume by the mention of Galileo.

[enne_son]

Bill your first test tells us how the sustainability of immersion with comprehension is modulated (if at all) in tightly line-spaced settings by the presence or abscence of serifs of a certain dimension at the terminations of strokes.

I like information about how sustainability of immersion with comprehension is modulated by type-related parametric change. What this does is pin-point something quite specific. What I think it tells me is that type-related parametric change has a “perceptual processing in reading” effect.

My question than is: what, in perceptual processing terms, do serifs really do. Saying they make a thing more readable doesn’t tell me about what I think I want to know. Do they increase the distinctive cue-value of the features they are part of? Do they make counters and the spaces between letters more tractible and riveting for the eye by encapsulating them more fully so that they acquire a cue value of their own? Do they blur the letter-assignments of individual letter parts so that the bouma-integrity of the word as a whole asserts itself more strongly than each letter by itself? Do they mask variations in interletter spacing so that these variations don't become disruptive to the kind of perceptual processing that needs to happen?

I think this kind of progressive and highly instructive interrogation is pre-empted by moving quickly to the conclusion that one thing is more readable than the other, and that this can be known by applying your test, meanwhile leaving behind a more specific awareness of what we have come to know. Having discovered what I have come to see, I don't any longer have any great urge to generalize this into a statement about the broad experintial reality we've come to call readability.

If I were asked to comment about readability, my strategy might be to answer indirectly and state in specific terms a few of the things I've found. All I'd really want to say is: serifs, depending on their prominence and extent can have a very beneficially value, and without them, the pressure on interline and interletter spacing for acheiving satisfactory perceptual processing results is greater.

Stated this way, does my resistance to the heavy reliance on the rhetoric of legibility and readability in type-analytic contexts make any sense to you at all?

[Nick Shinn]

I am very surprised that you appear to be taking the position of the medieval church on Galileo.

Well, that's one way to discredit my argument.
I guess I should never have mentioned the Borg :-)

[William Berkson]

John, on Galileo, I wasn't thinking of the trial, and the philosophical position of the Cardinal Bellarmine, but of the priests who allegedly refused to look through Galileo's telescope and see the moons of Jupiter. They just didn't want to see scientific research happen. I am shocked that this seems to be Nick's position on research into readability.

Yes, there are a lot of bogus claims about readability, but I don't think that is a reason to stop research; I think it is rather a reason to try to get it done right.

>Stated this way, does my resistance to the heavy reliance on the rhetoric of legibility and readability in type-analytic contexts make any sense to you at all?

Well I agree with you--and Nick--that there won't be a general single measure of readability for all contexts. But I think that the concept of 'readability' is useful. The point of introducing the distinction, I am guessing was to say that there is more to readability than just how clearly the individual letters may be distinguished (legibility). Within the general concept, I'm sure a lot of further nuances can and will be distinguished as research progresses.

[Nick Shinn]

They just didn’t want to see scientific research happen. I am shocked that this seems to be Nick’s position on research into readability.

Bill, your rhetoric is fallacious.
Those priests were opposed to scientific research in principle. I'm not, I just think testing readability is poor science.

[enne_son]

Bill, as accessible and ready-to-hand labels for broad experiential categories of affordance the terms readability and legibility have a relatively transparent, legitimate, even important, and stable use. As sharp and focused denominators of or designations for discrete and incisively differentiated areas of functionality the terms fall short. They lack specificity and bite.

Science by plotting thresholds and ranges of affordance in actual conditions of use, or looking closely at cognitively isolated dimensions of perceptual processing in reading does a service. It broadens and deepens this experiential knowldge of affordance with diverse and specific knowledge of actual functioning.

I think we all want to avoid the insidious typographic equivalent of racial profiling, thinking perhaps it might be evidence based. Nick is right in insisting it isn't warranted, either on humanitarian or on scientific grounds.

[ebensorkin]

Saying I just think testing readability is poor science. could mean any number of things.

You might mean that science & type can't mix on a sort of all or nothing, oil & water / science vs. humanities model. Hopefully not.

Maybe you mean that saying that Science has not unlocked the question of readability. And that claims to the contrary are bogus. You might even say that Science alone isn't enough to do so*. All of that that I could agree with!

* Or even that it can't on purely Semantic grounds. Peter's point that Science isn't going to concern itself with the concept of Readability per se because as a concept is too imprecise, that is to say insufficiently atomized to match phenomena is also great point. Maybe that's what you are referring to.

But what I haven't heard you say yet is that Scientific study of reading; using the terms and conceptual frameworks appropriate, might over time contribute to our understanding reading. And that this work might contribute to font design. That it might inform the way we look at Readability in the humanities. These are not contradictory or clever dodges. The opposite is also true, insight from the humanities will necessarily be applied in Scientific theories.

Does anybody have a problem with this framework?

[William Berkson]

>insidious typographic equivalent of racial profiling

I agree with you generally, but this phrase is a little bizarre. There is bogus use of science and bad science. But if you're saying type that has feelings and rights, I'm not quite enough of a fanatic yet to agree :)

>Does anybody have a problem with this framework?

Sounds good to me.

[enne_son]

insidious: "operating or proceeding inconspicuously, but with grave effect"

It's not that type has rights, but that profiling has consequences.

Maybe my “the ” in "the insidious typographic equivalent " should have been an “an”

[Nick Shinn]

Science isn’t going to concern itself with the concept of Readability per se because as a concept is too imprecise, that is to say insufficiently atomized to match phenomena is also great point. Maybe that’s what you are referring to.

That's what I've been saying all along. "Too many (unrelated) variables".

this work might contribute to font design.

It's all grist for the mill.
Samuel Coleridge used to attend Humphrey Davey's lectures "to replenish [his] stock of metaphors".

Take a look at the effect of scientific knowledge of the reading process so far, on type design.
To begin with, the very short descenders of Cheltenham (because the top part of letters were discovered to be more significant in the reading process).

Century Schoolbook was supposedly developed using reading research data, but I've never come across anything further about this. It certainly looks like an easy to read face, and has been much used in childrens' literature (the name helps!); but so has Garamond; both were favorites of Theodore Giesel (Dr. Seuss).

Lintotype's "Legibility" news fonts of the 1920s and 30s were concerned with opening up the counters: certainly the results produced a cleaner image, and the mantra has since always been in favour of open counters--but there is no scientific evidence that this actually improves reading! As John has suggested, let's consider non-latin scripts, and in Chinese and Japanese there are closed counters galore.

I think that closed counters may improve readability. For instance, consider the sequence "i_n", which may be confused with "m".

Saccadic scanning has been common knowledge since the 1940s -- there were Life magazine articles with call-outs on a photograph, showing "where men and women looked subconsciously" -- all very Freudian. I'm not sure whether that knowledge has ever informed a typeface design.

Claude Hopkins' work in the 1920s (he wrote Scientific Advertising, and developed split-run direct marketing as a means of testing the relative merit of different copy and layouts, measured in sales) was a great influence on David Ogilvy, who disseminated some of the successful readability principles in several of his How-To books. But not really much hardcore science there, more market research using statistical techniques.

Cyril Burt's A Psychological Study of Typography (with foreword by Stanley Morison) was noted in its time, but all Sir Cyril's oeuvre has been tarnished by his reputation for falsifying.

It's not as if scientific knowledge about how we read hasn't been available in the past. But has this been any use to the designers of the most-read faces, other than giving them a lever to create something new and different--has it produced more "readable" typefaces and typography?

Why are you, Eben and Bill, assuming things will be different soon? Isn't this the modernist fallacy which believes that the present age is the only high-tech era, and all previous eras were pre-modern? The blind faith in science, technology and progress that believes we are always on the threshold of discovering the way things REALLY work.

[ebensorkin]

Thanks for all the examples. Very interesting! That i_n example is pretty cool.

One quibble: Saying “Too many (unrelated) variables” isn't at all what Peter is saying as far as I can see. Party because you were claiming that nothing that science could offer up could inform our understanding of Readability. As far I understand it Peter is saying thinking about Readability per se isn't what we are going to get up to in Science - but he isn't eliminating the possibility that our (humanist) understanding of Readability won't or couldn't be impacted by science. And probably the problem is that the variables are related when it comes to their impact and the tricky bit will be characterizing their relationships. Peter?

assuming things will be different soon? I don't think we are. And I don't think we have faith so much as curiosity. And perhaps some hopefulness. That's not so bad is it?

[Nick Shinn]

No, and if one is able to take criticism constructively and answer one's critics, so much the better.
Which is why I'm optimistic about the multi-disciplinary nature of Kevin's team at Microsoft, and respect his efforts to make reading research more typographically savvy. (But that won't dismount me from my anti-science-of-readability hobby-horse!)

[William Berkson]

>Too many (unrelated) variables

But the variables are related. For example, many books on typography state that if you have a long measure (over 65 lc letters), you can compensate by increasing the leading *to achieve good readability* .

So the concept of readability is still a useful one.

I shouldn't have agreed above totally with Eben, because I think it is a better way to put it that in different situations different designs are more readable, rather than to 'atomize' the concept of readability.

>won’t dismount me from my anti-science-of-readability hobby-horse!

Yeah, I'm sure you won't notice that you're sitting backwards on the horse, and it's running the wrong way :)

[Dan Gayle]

Didn't stop Crazy Horse from whipping some Custer *ss!