One more piece of science that confirms the obvious.

So "sub-fluent" sizes are used for pocket bibles, footnotes and stock tables? Well, yeah.
And at the other end, 20% of headlines outside the range have very few words. Uh-huh.

…where typography and the science of reading meet…

They are still far apart. They would be a little closer if studies like this showed specimens (at actual size), and were set in a manner that typographers might find more appealing.

This study looks just like any other study written for scientists, and shows absolutely no outreach to typographers.

Nick, one of the things I found notable about this article -- I've only read in detail the section dealing with spatial frequency, since I've read earlier papers on that subject and believe it provides insight into the how and why of type design for different sizes -- is that it directly references the writings of typographers such as Harry Carter and Bigelow himself (Caslon, Dreyfus, Frutiger, Fry and others appear in the bibliography and are similarly referenced elsewhere in the article), and seeks to reconcile the findings of the study to the craft description provided by those writers.

So, for example, the authors cite Carter:

“Legibility is all that matters in 6- to 12-point types; so that their successful design is a technical, and not in the ordinary sense an artistic achievement. In the design of founts from 20- to 72-point the artist comes into his element.”

And then attempt to express an understanding of this statement in terms of their findings with regard to spatial frequency. In what way is that showing 'absolutely no outreach to typographers'?

I think it's quite clear I was referring to its appearance of boilerplate academese.

Typography is visual and practical, not academic—why must it meet its science on science's home turf?

Nick, I would have thought you might find the science interesting, also that you might find the documentation of historical trends informative and the discussion of metrics impressive.

I already know that x-height is the main effect on functional type size, and that newspaper types have a larger x-height than book types, and am not surprised that Times Roman falls bang in between.

The conclusions as to type size in the different genres are of course common knowledge to typographers, as we are the ones who determine those sizes!

I'm very impressed that type size is discussed in terms of visual angle, and that cultural categories of publications are analyzed en masse, statistically.

The documentation of historical trends is not interesting per se, as again, this is well-known to me as a typographer with some knowledge of type history. However, I do find the hypotheses presented to explain these trends interesting, although this is not really the realm of science, as I don't need science to tell me that, for instance, Incunabula type was quite large.

Statistical analysis of type history is fascinating, but it's about social culture, not human visual processing.

For instance, the paper notes that subfluent sizes proliferated in the 16th-18th centuries, and hypothesizes that economic reasons restricted type size. But I would argue that while that may have been the case, the sizes themselves were not subfluent, because people would have just held the books closer to get the angle up to 0.2°.

The paper's historical analysis of historical types is, IMO, flawed in the frequent manner of relying on extant type specimens, rather than type in use. Robert Thorne, for instance, issued no type specimens during his foundry's most prolific years; too busy. I have published type specimens, but the amount of usage of the types, based on sales or volume of readership cannot be construed from the specimens.

I theorize that hardcover books have larger type because they are heavier, so people are less likely to hold them close, which is tiring, so they rest them on their laps, further away.

I'm not impressed that fluency is assessed by speed in lab tests.
Of course, that can;t be done for history.

I find it particularly meaningless to typographers that the paper comes to the conclusion that almost all running text and most display text falls within the "fluent" range of 8 to 80 points.

Do we really need scientific confirmation of a theory that most type is set in a convenient size to read, which is 8 to 80 pts?

I don't therefore believe that the concept of "fluent reading range" is of any scientific use.

You might just as well say that people generally wear shoes that fit them.

I'm afraid I'm not really interested in this kind of material in the guise of science, because I'm not interested in pure science, and only read this paper reluctantly, as it was recommended by Peter. But I would be interested in it if it were in the form of a well-illustrated, well designed book A History of Type Size, viewing type history through the prism of size, with all the facets of the ecological approach of which this paper speaks, one of which might be sidebars with its charts and graphs, but not too much else in the way of science. I don't think such an "anecdotal" cultural study would suffer from a lack of significance.

Nick, I tend to see the paper’s greatest strength where you see it’s most outstanding weakness — in showing convincingly that legibility (the fluent range / the zone of robust affordance) is not a narrow ledge but a wide plateau. I suspect this can be generalized to include factors other than size.

This perspective might counteract the tendency prevalent among design clients to rely on sources claiming 12 point is the most legible size and Times New Roman the most legible font.

Yes designers and typographers tend to stay within the fluent range. But designers also know there are optimums within this range whose impacts are not effectively measured by reading speed. These optimums exist in the realms of boldness, spacing, configuration, and have to do with visibility and readability as explored by Matthew Luckiesh and Frank Moss. Optimums are niche-sensitive, and subject to countervailing pressures. These pressures include issues of gestural-atmospheric thrust (‘personality’). Type professionals are expert at managing the pressures in relation to the niche. Visibility and readability constraints appear to differ for signage and way finding (including way-finding on screen and in publications), as opposed to continuous reading for extended periods of time.

So for me the paper is excellent in what it tries to do, but doesn't go far enough.

For me the section summarizing scientific research on type size and reading speed is the most informative survey I've seen since Kevin Larson's paper on word recognition. It is, I'm afraid tough sledding, and I think I will have to study some of the linked papers to really understand it. Larson's essay was easier to get through. But on the other hand, I think Larson,'s conclusion—that we always identify letters before we read words—is probably wrong, and the conclusions in Legge and Bigelow are probably right!

I agree with Peter that we could use a lot more strong research like this, using other measures than reading speed, such as Luckiesh's visibility and readability (fatigue) measures. But personally I am grateful for such a summary. To me much of the "legibility" research is poor, and what is summarized here is some excellent and informative work.

Nick, I can understand not wanting to plow through the scientific section, but your idea of "pure science" seems to me off. When science gets it right, it describes and understands reality correctly. So it's up to us then to make use of that understanding. And "us" includes practitioners as well as applied scientists.

@Peter: This perspective might counteract the tendency prevalent among design clients to rely on sources claiming 12 point is the most legible size and Times New Roman the most legible font.

Good point. But hence rather ironic to observe the typography of this paper!
Which leads me to the following—

Bill, my comments about pure science relate to Peter's "where the science of reading and typography meet", i.e. more about the visual presentation of this paper than its text content.

I'm no more interested in pure science than scientists are interested in pure typography.
However, I suspect I know more about science (a couple of A levels) than most scientists know about art, design, and typography. Part of this is because art is dropped as a compulsory subject in high school, while one is forced to continue with maths and science.

The science I keep up with is mediated: New Scientist magazine, and books like Guns, Germs, and Steel, The Invention of Clouds, This is Your Brain on Music, Awakenings, and the totally awesome The Physics Of Consciousness: The Quantum Mind And The Meaning Of Life (Evan Harris Walker).

As Peter observes, the "ecological" approach of the content is a step in the right direction, but the design of the document indicates that the paper is targeted at scientists, not typographers.

So it's up to us then to make use of that understanding.

No, I suggest it's the responsibility of scientists to make sure their research is properly presented and understood by the profession into which they're poking their noses, and by the general public of font users, bearing in mind that the press will mangle their findings horribly.

This is particularly germane as they are presenting their work in the medium which they are investigating.

>to make sure their research is properly presented and understood by the profession into which they're poking their noses

Poking their noses?

Psychologists can do research about sex, marriage, racial prejudice, and sports, but as soon as they "poke their nose" in typography, they'd better look out!

Damn right.

But what do you mean by "can do"?
Are you implying that there is no criticism of their work interfering with the work of others, in the areas you mention?
Even in their own field—mental disorders—a great many psychiatrists (and the informed public) are upset that their reach exceeds their grasp. Have you come across the controversy surrounding the changes to the DSM?

Here's what another guy with a bow tie has to say.

Psychologists can do research about sex, marriage, racial prejudice, and sports, but as soon as they "poke their nose" in typography, they'd better look out!

Since when is science beyond criticism? Almost by definition, it needs to be the lease sacrosanct of all pursuits.

I totally think science should not be above criticism, especially as criticism from observation and experiment is the heart of scientific method. And of course it should not be above ethical scrutiny either. My presentation at TypeCon had plenty of criticism of science and scientists, both about method and ethics—especially of Tinker, but also of Luckiesh.

But typographers and graphic designers also shouldn't be immune from criticism. "Poking their nose" into typography implies that it is somehow illegitimate or improper for science to do research in this field. Typographers are so superior that they can tell scientists to keep away. That's where I think Nick's complaint is pompous and wrong.

@Bill: …reactionary rubbish…

Bill, I have repeatedly made the distinction between reading research (against which I have no complaints), and readability measurements of named type products.

Once more: readability is a function of document design, which is the business of designers, and is not something that can be indexed to a typeface by measuring how fast a lab document is read.

Readability research which comes to conclusions such as "font A is more readable than font B" is the real rubbish.

So yes, you scientists, please explain how we read, but kindly refrain from opining on which typefaces are most readable.

Nick, how, in your practice, are your judgments of readability, legibility and aesthetics made? Does a model of reading play a role? Can a psychophyisical understanding of seeing play a role? Does an understanding of — or intuitions about — the perceptual-processing impacts of type-form manipulations enter into it?

To what extent are your judgments empirical knowledge-based? If they are empirical knowledge-based, how is this knowledge-base constructed or acquired? If your own individual experience is their basis, how does your judgement become conditioned, corrected or refined? Do you have personal touchstones in these areas?

You want an autobiography?

Having followed the science of reading from way back (Cyril Burt!) I presume it's had some slight effect on my work, but really, David Ogilvy's two how-to books were probably the biggest theoretical influence, and my mentor Raymond Lee, from whom I learned how to lay out copy.

I work like most designers — keep an eye on what's happening, follow my interests, engage in some professional development, please the client, target the work according to reader demographics, take account of the medium, try different approaches to a brief and see what works best. Occasionally there is feedback from focus groups.

I don't think there is any way that scientific knowledge of how reading works can add anything useful to what we, as typographers and type designers, already know through centuries of empirical practice and the culture, institutions and traditions we've developed.

[Nick] “I don't think there is any way that scientific knowledge of how reading works can add anything useful to what we, as typographers and type designers, already know […]”

Nick, is confirmation or refinement of accumulated craft knowledge of use? Does it help or add something to have explicit empirical confirmation (in the realm of establishing statistical significance), and a graphic visualization, for the tacit knowledge of designers that, from the point of view of general legibility, there is a fluent range?

Is it not of interest to you that — as the Legge / Bigelow paper explains — at text-letter sizes the spatial frequency channel used by the visual system filters information at the density of 3 cycles per letter? In a typical letter this corresponds to the stroke on the left hand side of the letter, the counter adjacent to it, and the stroke on the right. I take this to confirm the fact that in reading at normal text sizes, cues to the identity of the words are drawn from all three areas (as well as perhaps, the between-letter spaces), so designing the white (or what happens in the counter ‘domain’) in conjunction with the black has a real significance.

I know, you’re allergic to certain types of claims, but I think your negativity about this paper is warranted. It’s a step in the right direction, though I might have questions about some of the assumptions made in the explanations section.

…is confirmation or refinement of accumulated craft knowledge of use?…

It banalizes. The privileging of speed as the measure of readability says it all—the hammer is the only tool in the box.
Other measures, e.g. whether the reader remembered the message or responded to it, are so much softer, but these are what really matter to the practitioner.

Is it not of interest to you that — as the Legge / Bigelow paper explains — at text-letter sizes the spatial frequency channel used by the visual system filters information at the density of 3 cycles per letter?

Yes, it's fascinating. But the importance of designing white space is not a scientific discovery. I was taught it in first year Foundation at art school, according to Itten, and subsequently by my calligraphy teacher who started us off doing pages of "o"s and "n"s. It appears to be self-evident to most designers and artists, appearing in patterns since the early days of human culture. I would assume that we were inspired by Mother Nature's patterns in animal and insect markings, so figure-ground relationships may be a quality of fractal-growth shapes.

PostScript language, as a means of defining letter shape, places emphasis on vector edge lines rather than solid shape, but at the same time, FontLab offers one the option to draw glyphs in either solid or outline mode, and view "clean" by hitting the grave key. Following this line of thought, drawing type glyphs rather than punching them would seem to be the paradigm shift, implemented by pantographic router (woodtype) and LB Benton's device.

[Nick] “Yes, it's fascinating.”

Good!

[Nick] “It banalizes.”

Empirical documentation of the fact that there is a fluent range and getting a strong graphic representation of the fluent-range curve banalizes and makes you insensitive to the fact that there are other levels of concern, including those Luckiesh tries to isolate and those you list above?

I know that even for some very basic things like ease of visual word-form resolution, readings speed is not a revealing measure.

Nick you don't seem to be aware that there is already research showing that different typefaces have different readability. Even by the reading speed standard, I believe that italic faces are shown to be slower in reading text than roman. I believe this is a very robust result, but I'd have to check on the studies.

And as I have reported, Luckiesh distinguished readability of text by reading fatigue as indicated by blink rate, and did get differences not only on different typefaces, but also on light, medium, bold, and extra bold.

And using a distance test on signs, the testers of Clearview type found the medium Clearview weight to be readable at a greater distance than the federal standard Highway Gothic Bold.

There are different aspects to readability, but on reading speed, blink rate, and distance, all different measures, differences in typefaces have show up.

Also you see to think—absurdly on Typophile!—that practitioners *agree* on craft, which is far from the case. For example, Theodore Low de Vinne in 1885 strongly attacked Scotch moderns for their high contrast, arguing that they were less readable than old styles. And he influenced the design of Century, which thickened hair lines. You have revived and defended the readability of Scotch modern.

There is lots of disagreement about craft—just look at the hinting threads!—and tests have indeed distinguished the readability of different type faces. Wake up and smell the coffee.

There are different aspects to readability, but on reading speed, blink rate, and distance, all different measures, differences in typefaces have show up.

Yes, there are different ways to measure readability, but mostly speed is used.

Differences in typefaces—product testing.

…you see to think—absurdly on Typophile!—that practitioners *agree* on craft, which is far from the case.

I referred to "what we know as typographers and graphic designers", there was no inference of unanimity.
What I know about the Scotch Modern may be different from what De Vinne knew, but what we both know is informed by our experience as practising typographers.

…tests have indeed distinguished the readability of different type faces…

Product testing of a handful of typefaces in a limited environment.
But one Caslon is as good as the next, isn't it?

Isn't there something seriously wrong with pitting one typeface against another in the same layout, and assuming the results hold for all layouts, all typographers, all media, all readers, all content, today and tomorrow?

Don't you think typography plays a dominant role in determining the readability of typefaces?
Or is one typographer as good as the next?

Might it not be the case that types respond differently to different typographers?

How does one test that?

[Nick] “… pitting one typeface against another in the same layout, and assuming the results hold for all layouts, all typographers, all media, all readers, all content, today and tomorrow …”

Nick, in the best science that's not what’s assumed. Luckiesh takes his multifactorial pittings to indicate that readability is an integral effect [his words]. I don’t think Legge or Bigelow would have a different view.

Nick, the point is that because there is difference among practitioners, research may shed light on which solution is best for a given usage. So contrary to your haughty dismissal of scientific research as inherently having nothing to teach experienced practitioners, it has in fact taught them.

I think the tests on Clearview Highway are a clear example of that. Not everyone would have agreed that the somewhat lighter weight would be better. Research succeeded in finding an empirically demonstrated superior result. And that is also one that can be experienced by viewers. And by the way that is always true when the science is right.

As far as "product testing in a limited environment", of course testing is always limited. You have general theories and specific tests. That's the way it always works. You find out how far the general theories hold by further testing. Unfortunately, the theory has been so poorly articulated in reading, that the research has not been very informative. This is not an inherent problem of scientific method, but a weakness of legibility and readability research in particular.

I'll post later on the role of typography, which is an important question.

@Nick Shinn:
Typography is visual and practical, not academic—why must it meet its science on science's home turf?

When a scientist submits a paper for publication, a number of things are taking place.

One of them is that he would normally submit his manuscript according to the standards of the journal which is receiving it, and they will typeset it in their standard form. Which will be dull and boring, since it's bad enough that academic papers tend to be full of such "penalty copy" as mathematical formulas.

I've seen, of course, many exceptions to that rule - in conference proceedings where each paper is usually "typeset" in double columns by the submitter. So most papers are done on a typewriter or a line printer with a lower-case print train, and one or two lucky submitters have some exotic equipment like a 200 dpi electrostatic printer.

Of course, that was back in the old days. Today everyone and his dog has a laser printer and desktop publishing software. But they don't know how to use it. At least few scientific papers make use of Comic Sans and Papyrus.

The other is that these papers are basically intended for reference by other scientists in the same field. Even review papers, although an entrance into the literature of a field, aren't intended for laypeople or undergraduates.

If, someday, the psychovisual study of typography got to the stage of producing monographs... well, that's an exaggeration - it has reached that stage on occasion - as Legros and Grant (Typographical Printing-Surfaces, Longmans, Green 1916) attests - let alone textbooks, I expect they will receive decent typography.

Likely dull and workmanlike instead of flashy and spectacular, though. Better not to attempt too much in someone else's field than make a fool of oneself. Scientists, unlike cosmetics companies, don't have the budget to hire graphic designers.

Luckiesh takes his multifactorial pittings to indicate that readability is an integral effect [his words].

Sorry, I exaggerated.
Nonetheless, I don't believe that the integral effect is a straight sum, as it were, of the the components of a layout—which is holistic.

So contrary to your haughty dismissal of scientific research as inherently having nothing to teach experienced practitioners, it has in fact taught them.

Bill, you're conflating science and technics. These are not the same:

1. Experiments which address general principles
2. Product testing
3. Product development

Clearview is just about the only example of the scientific method being used in type product development, but is it science or technics? What did the tests contribute to general theory or practice, other than to determine that a lighter weight of Clearview was better for highway signs?

This is not an inherent problem of scientific method, but a weakness of legibility and readability research in particular.

The scientific method is more successful in some areas than others.
The weakness of which you speak suggests that readability is too soft a subject for useful science.
Scientists may think otherwise, beguiled by the fact that type appears to be quite simple and many of the parameters of typography are numerical.

However, I might change my mind if they start doing readability experiments with metafonts.
Why has no-one thought of this? After all, Don Knuth is a star.

John, up until 1960 The Penrose Annual separated its content into General Articles and Technical Articles, which always had exemplary design and typography of charts and graphs.

Edward Tufte has opened the eyes of a lot of designers to the possibilities of The Visual Display of Quantitative Information, as have exhibitions at the Design Museum in London.

[Nick] “Nonetheless, I don't believe that the integral effect is a straight sum, as it were, of the the components of a layout—which is holistic.”

What do you suppose “integral effect” means?

“… readability / too soft … [s]cientists may think otherwise …”

Ease of rapid automatic visual word-form resolution for a particular type of a particular weight in a specific setting is in principle quantifiable. Luckiesh’s work with blink rate suggests that this is so. It seems to be a matter of the efficiency — or signal-to-noise ratios — in the cortical integrational routines. These routines comprise the gathering of statistics about the prototypical identity, local combinations and global distribution of the role-units spread across the word. The gathering of statistics about prototypical identity, a process I call quantization, relies on what is known today as feature-binding. To arrive at the knowledge that component x is an ascender, the visual system must bind statistics about length to statistics about aspect; similarly, to arrive at the knowledge that component y is a b or p-type bowl, the visual system must bind statistics about closure to statistics about curvature. Feature binding, local combination detection and global distribution discrimination are arguably — and presumably measurably — less efficient under crowding or with noise, or under conditions of non-optimal phase-alignment.

Ease of moving along the line for a particular type at a given size, leading, line length and inter-letter spacing is tractable as well, by studying ocular mechanics.

Your idea of readability involves more than ease, but (if I've got your usage right) embraces it. That”s okay, but saying readability is to soft is to hard and fast.

I meant of course to write: “but saying ‘readability is too soft’ is too hard and fast.”

Bill: Even by the reading speed standard, I believe that italic faces are shown to be slower in reading text than roman. I believe this is a very robust result, but I'd have to check on the studies.

I recall the same, and I reckon that this a potentially very fruitful area for further research. Italics are distinguished from romans by a variety of factors (form, slant, compression) applied in a mix of varying degrees. This means that it should be possible to measure relative reading speed when those factors are mixed in different ways.

>Readability research which comes to conclusions such as "font A is more readable than font B" is the real rubbish.

Nick, this is your claim that I was contesting by pointing out both the case of italic vs roman, and the specific test of Highway Gothic vs Clearview Highway. As I said what you can test and give results about is a comparison of fonts under similar conditions, with the same measure. And here you are just flat wrong: there are sound conclusions from research saying "font A is more readable than font B," under a specified set of conditions, with a particular measure.

Instead admitting that the evidence is indeed there, you drag in a red herring of whether it is science or "technics", or in other words applied science or engineering. This is irrelevant. It is still research using empirical testing, and your claim under any reasonable reading is still contradicted by the facts.

You can go on denying that empirical testing can inform us about readability, but frankly I don't think you will have any credibility in light of the hard evidence that contradicts you. For starters, even the one case of the Clearview tests refute your above claim.

The reading speed differences between roman lower case and italic are discussed in Miles Tinker and Donal Paterson's very first study in their STUDIES OF TYPOGRAPHICAL FACTORS INFLUENCING SPEED OF READING [1928] using the Chapman Cook Speed of Reding test. The study is titled INFLUENCE OF TYPE FORM ON SPEED OF READING.

SUMMARY AND CONCLUSIONS
1. A speed of reading record was obtained for texts set up in roman lower case, italic, and all capital letters. Comparisons were made between all capital versus lower case text, and between italic versus lower case text. There were 640 subjects used in the experiment.
2. The text in lower case letters was read 13.4 per cent faster than that in all capitals. The difference in favor of the lower case text is highly significant. There were 4.74 words read per second of the all capitals and 5.38 words per second of the lower case text.
3. The text in lower case letters was read 2.8 per cent faster than the italics. The direction of the difference was consistent for all sub-groups. The subjects read 5.08 words per second of the italics and 5.22 words per second of the lower case text.
4. The following conclusions seem justified by these data: (a) Text printed in all capitals is less legible than material in lower case letters and slows down speed of reading to a very marked degree. (6) Text printed in italics is slightly less legible than material in lower case letters and decreases speed of reading somewhat.

In their discussion Tinker and Paterson state:

Lower case text was read only 2.8 per cent faster than italic text. Reasons for this difference being so small are found when we compare the two texts: (1) The material in italics covered exactly the same space as that in lower case letters. Therefore no extra space is present to produce a probable increase in total number of fixations. (2) Word form in italic printing is very similar to word form in lower case text. However, we, must bear in mind that the difference between means in each sub-group was always in favor of the lower case text. This indicates that italic text is actually harder to read than lower case text. It is possible that greater effort was expended in reading the italics. This increased effort may have reduced differences between italic text and lower case text and thus obscured true legibility differences. Greater eye strain may have been involved in the reading of the italic material.

Luckiesh and Moss didn't do a comparable study using visibility and blink rate.

It would be interesting to determine if all capitals, roman uper and lower case and italics have different critical sizes and fluent ranges.

And here you are just flat wrong: there are sound conclusions from research saying "font A is more readable than font B," under a specified set of conditions, with a particular measure.

Bill, that's what I said, but drew a different conclusion.
It doesn't prove that font A is more readable than font B, except in a very specific set of circumstances.
How many asterisks should one add for circumstances other than highway signage?
This is product testing. What scientific hypothesis does it address?

**

I am being quite specific about readability.
It's the document that is being measured, not the typeface.

In the first place, a typeface is not a physical object the performance of which can be measured like a machine or a biological organism, or chemicals in a test tube.
It is a set of abstract shapes which exist in the collective and individual conscious.
Fonts, which scientists purport to measure, are another matter.
But they are under the impression that fonts are a set of hard and fast shapes.
They are not.
Fonts are instruments which are used by typographers (or whomever) to create documents.

What is being measured is not the readability of typefaces or fonts, but the readability of the handiwork of people using these tools, as well as the reading ability of the test subjects.
The measurement tool is not a direct inanimate gauge, but the response of a select cultural group, filtering the instrumentality through the medium of language.

What is being measured, the readers or the document?
It is certainly not the typeface, at anything other than a vast remove.
This cultural complexity is what makes the scientific concept of "typeface readability" so soft.

It's pointless to compare the readability of anything so general as typefaces, when their functionality is so specific to media — just look at the way that hinting methods and browsers produce an array of readability from the same typeface.

Compare font as instrument with musical instrument.
What is the playability of a musical instrument?
Different musicians can play the same notes and produce quite different effects.
The listening environment has some effect.
And one can hear the same music and be moved to tears, later, to boredom.

Another set of playable instruments: golf clubs.
I don't believe science can tell you whether to use the wood or the iron, to chip or pitch with your wedge, and so on.

The paradox of type is that that part of the instrument appears to be part of the output.
But it never is—the process always transforms the font.
Just is there is no homunculus within one's brain, representing one's ego, there are no letter shapes in digital fonts, they are pure data.

Bill, I am no reactionary luddite.
I'm a debunker of bad science, and I would have thought that an advocate of science and typophile such as yourself would support rather than admonish me.

Ok here is about the issue of layout and readability, which you Nick, asked about.

For a start, I think it is important to realize that there are different measures of readability, reflecting different aspects, and they may not agree. For example there are threshold measures regarding contrast, Luckiesh's "visibility". And there is of course reading speed, a performance measure. There is also the performance measure of ability to read at a distance.

In addition there are fatigue measures, such as Luckiesh's measure of increase in blink rate over time. If we take this supra-threshold idea of "ease of reading," meaning low effort or low fatigue, then there are obviously many factors affecting readability.

For starters, there are huge factors that don't involve visual features, once you are above visual threshold levels. The research on fatigue, such as that discussed in the book Cognitive Fatigue, edited by Phillip L. Ackerman, shows that interest and skill of a person makes a huge difference to time-on-task fatigue. If the subject is boring or too difficult for the reader, he or she will find it extremely fatiguing to try to read. Similarly, how well written the text is, from the point of view of rhetoric, will also have an impact on reading fatigue, as whether it is enjoyable to read will affect reader fatigue.

Above visual threshold levels, these non-visual aspects I suspect are in general more important than the visual ones—either layout or typeface—to reading ease or fatigue.

However, both layout and typeface have an impact, as Luckiesh found. Luckiesh found that both leading and line length for a given typeface affects reading fatigue. Luckiesh did not test this very broadly but his results seem pretty good.

Luckiesh did get somewhat surprising results here, finding that shorter lines than normal typographic practice in books are less fatiguing. Here I think there needs to be more research on fatigue vs reading speed. It may be that there is some gain in reading speed with longer lines even if they are less fatiguing. Also Luckiesh's blink tests generally need to be reproduced using his guidelines, which nobody has attempted yet.

Tinker tested leading and line length, but on these issues I don't think his Chapman-Cook reading speed is reliable, because it involved a search task as well as a reading task. Also his results run too contrary to typographic wisdom, and also lack any smooth curve. Frankly on this issue I dismiss Tinker's results as unreliable.

I believe that other researchers have duplicated Tinker's finding on italic, and I wouldn't buy it from Tinker alone, though the search task may not have had such an impact.

Luckiesh did find a difference between typefaces, and most notably between different weights of the slab face Memphis, with the medium weight being more readable than the bold or the light.

And yes, Nick, even different versions of Caslon might be expected to have different readability. Luckiesh found Caslon Old Face less readable measured by blink rate, though not by reading speed. However, judging by his book, this was Caslon Old Face printed on coated paper, which is thin and spidery. It had "low visibility" as measured by Luckiesh's visability meter, because of its lightness. Caslon Old Face when printed by letter press on rough paper was a much more dark and robust face at text sizes—really a different face. For this reason both Monotype and Linotype showed their Caslons on both rough and smooth paper.

My own Caslon follows the weight of Linotype Caslon Old Face printed on rough paper. Particularly at 10 point, I think it looks more robust than Adobe Caslon, and I would expect mine to be better, though the differences may be small.

Overall, different typefaces call for different settings to be at their best, and I expect that traditional text faces will not have great differences when set at their best. But I do expect differences.

Now because how inviting the overall layout is, and aesthetics influence readability in terms of fatigue, I expect for good text faces the differences will be a smaller part of the pie, so to speak, compared to good writing and good aesthetics.

But typeface being a small part of the pie on issues of fatigue doesn't mean that it can't be tested. By rigorously controlling other factors, including interest of reading material, illumination and distance to reading material, Luckiesh was able to get results that I think should be reproducible.

Finally, let me say that I expect that some of the most boring and most difficult materials are likely to benefit most from good typography. Thus timetables and phone book and other similar reference material most benefit in readability from good typography. Similarly text books, which are often demanding and sometimes not the most interesting material I think are helped greatly by good layout.

Overall, I think the fatigue measure is most likely validate a lot of typographic layout practice that reading speed alone doesn't do a good job of differentiating.

Nick, my last was cross posted with yours.

The test of Clearview vs Highway Gothic were NOT just a test of documents, they were a test of typefaces against each other for highway signage.

What is being tested is the hypothesis of what is most readable at a distance when taking up a comparable size of sign. This is a test of typefaces in a given CONDITION, not a given document, and thus they do refute your statement, and your false claim that science can't ever compare typefaces, only particular documents. They can test for typefaces under given conditions, that is, broad CLASSES of documents (unlimited in actual number), such as highway signs.

The hypothesis being tested is not of the generality as would be an hypothesis say about the ideal combination of stroke weight and width of character. But it still has generality because it applies to a broad class of events, and predicts that Clearview Highway will be consistently readable as greater distances. This has some universality and so is a scientific hypothesis.

Nick I don't know what exactly your views are about science, but your declarations on what science can't research seem to me not only anti-scientific, but demonstrably wrong. You don't seem to understand how controls work in science, and how these can be used to test particular variables in particular conditions, even when other factors that normally influence the results are present. You control those other variables—such as size and distance—and test the impact of one variable—typeface in the case of Clearview vs Highway Gothic.

[Nick] “It's the document [my contrastive emphasis throughout] that is being measured, not the typeface.”

I think this statement needs to be refined.

Strictly speaking, response patterns to specific settings under controlled environmental conditions using manufactured documents containing type are being measured. This is to make claims about affordances of a specific typeface at a certain size, line length and leading, or to make comparative claims about affordances at different sizes, leadings, etc for different types or settings. The focus of attention can also be on types of types.

Since affordances change at different sizes, etc., any claims about the legibility of the letters in the font or the readability of the type in use will have to be conditional, not absolute.

@Nick Shinn:
John, up until 1960 The Penrose Annual separated its content into General Articles and Technical Articles, which always had exemplary design and typography of charts and graphs.

Edward Tufte has opened the eyes of a lot of designers to the possibilities of The Visual Display of Quantitative Information, as have exhibitions at the Design Museum in London.

Good visual communication and layout is a very desirable thing, and I don't mean to suggest otherwise. But the Penrose Annual is a place where designers are published, and it isn't reasonable to expect psychologists to be designers even when they're attempting to study design.

Readability, as opposed to legibility, is a subject that at least seemed to be soft enough to be difficult to study. Recently, in these forums, it was noted that an early method used to test readability by measuring how often a reader blinked as a measurement of the effort made in reading had been rediscovered after a period of neglect.

I consider that to be very good news, since it seems to mean to me that there is finally some hope of there being some useful valid scientific input on readability.

But I will admit that we can't expect too much. If one looks at science news about, say, discoveries about human behavior and society through observation of primates or other scientific studies, what one will often find is scientific evidence of things everyone knew about human nature - on the level of platitudes.

So typographers aren't likely to learn a lot of things they didn't already know from readability research. Having a "scientific" basis for what they already did know isn't usually going to be directly useful; we're past the point where a designer needs "scientific proof" in order to convince a client that six-point type is not optimal for readability.

Eventually, though, working in their own way from a solid foundation, instead of taking what "everyone knows" for granted, I think that psychovisual researchers will find occasional unexpected nuggets of information that will enrich typography.

But it won't be much or soon - yes, they are in their own little world, as it were, with only limited relevance to the practising typographer. That's not something to worry about, it's just the way it is.

Though I have great respect for Nick Shinn's understanding of graphic design, including both history and type design, I am distressed to have spent so much time controverting his concerns, because I think they are groundless. Efforts to understand readability, including of particular typefaces, pose no threat to type design or graphic design, but will help it as the science gets more and more informative and robust results.

And this paper in fact has a lot to teach us as a very valuable review of the scientific research and type history. Let me start with just one fact. Legge defines "critical print size" as "the smallest character size for which reading is possible at maximum speed." Critical print size is something quite different from the smallest size that a person can distinguish letters, which is the "visual acuity" threshold, similar to what we all have been tested for by eye doctors. The striking finding here is that "critical print size is at least two times larger than acuity letters for normally sighted subjects, and the difference is often much larger for people with low vision." (p. 6-7)

This is a very significant finding, because it says that other factors than the ability to identify separate letters are crucial to fluent reading. What this says is that even by the standard of reading speed, there is a huge difference between legibility, defined as the ability to distinguish letters in isolation, and readability.

Those "other factors" are what I and others working on text type have long labored on in trying to make our typefaces readable, and it is encouraging that however much we succeeded, at least our quest has had a point to it.

And the specific research using gratings with fixed, regular spatial frequencies, and crowding in the parafovea, do seem to tell us something about these "other factors" in critical print size being twice the acuity size.

One of the aspects that is intriguing to me is that the critical print size is very close to the ideal print size, as measured by reading speed. In other words, the smaller type, just above the threshold, is the best for reading. This is not discussed in the paper, but my feeling is that at these small sizes, optical effects are quite strong, and need to be compensated for. This is one of the things I did in my Williams Caslon Text. My feeling was that some of the things done for very small text, typically for readability of type 8 point and below, for ideal readability should be done to some extent at 10 or 11 point sizes. In particular this involves not having too fine hairlines, which visually seem to disappear more quickly, in a non-linear way, as you make type smaller. There is a lot more to it than this, as I have discussed in my ilovetypography article and most recently in the current issue of Printing History.

Overall, I think the appropriate response to this article is not resentment at science "poking its nose" into typography, but gratitude for good work that can help us in the field of type.

Incidentally, Bigelow is the co-creator of the Lucida type family, winner of the MacArthur Prize and Goudy awards, and head of RIT's School of Print Media. The idea that he is in this paper with scientist Legge "poking his nose" into something he doesn't know about is laughable.

@nickshinn

"David Ogilvy's two how-to books were probably the biggest theoretical influence"
As relevant today as when they were written. IMHO.

@William Berkson:
I am distressed to have spent so much time controverting his concerns, because I think they are groundless.

Of course, the concerns are understandable. In our present age, science enjoys so much prestige and power, that there is at least a potential for a new "scientific" way of doing things to be taken up at the expense of the experience and wisdom of a craft that has been built up over a long period of time.

I agree that this is very unlikely in the field of typography, and that a contribution from science is to be welcomed, but I also admit that some degree of vigilance is still in order.

Knowledge is power, as Bacon said, so there is always a concern about how that power might be misused. But frankly I don't think the danger today is of science being paid too much respect, but rather too little. On climate change and economics there seems to be little attention paid to very hard won knowledge.

It is folly to pit science and art (and design) against one another. They both are high culture, and both need support and respect.

In climate change, I agree, but if the only choice available to prevent catastrophic global warming is a drastic cut in energy use, with all the economic consequences that entails, I'm not surprised at the preference for wishful thinking.

The thing is, I believe that we have an option that both solves global warming and gives us secure and plentiful energy supplies. Nuclear power.

In economics, while there is much about economics that is scientific, the discipline as it exists at present has, in my opinion, blind spots.

Marxism may be to scientific economics as astrology and alchemy are to astronomy and chemistry; but the rich and powerful of free-enterprise societies have managed to partially corrupt the discipline of economics as it is practiced here as well, I fear. Although the history of George III and the American Revolution may be part of the problem.

Here's a simple example.

Let's say it takes you three hours to mow your lawn. It would cost $30.00 to have a professional lawn care company come in and mow it fore you. At your regular job, you are paid $15.00 per hour.

Now then: your local city council passes a law that says it's illegal to mow your own lawn, and you must hire a lawn care company to do it.

Have they just saved you $15.00?

The answer is, not necessarily. Maybe things are slow where you work, and so you're on a three-day work week. So you don't have the option of working an extra two hours at your regular job to earn an extra $30.00 to have your lawn mowed.

Therefore, you would be better off if you could transform an additional three hours of your labor into additional wealth, even if less efficiently, by mowing your own lawn.

But if you go to a conventional economist, and suggest that the country would benefit if tariffs were raised, so that unemployed people could produce things that are currently bought with (scarce! - we keep having to fight with other countries to get them to lower their trade barriers) foreign exchange as cheap imports... he will tell you you're crazy, and Adam Smith proved that this is all a load of bunk, and so on.

Now, quantum mechanics and special relativity do tell us that when a scientific discipline contradicts "common sense", it's not necessarily wrong. But Adam Smith's arguments were presented in plain English (and they're applicable to some cases; India's ban on gold imports are an example) not differential equations. So we're not dealing with stratospheric heights of reason far beyond a layperson's understanding.

Of course, when other economists are right, they're ignored - I suppose you may have been referring to the repeal of Glass-Steagall and things like that. But there were economists who said it was safe to do that.

As "The Adventures of Unemployed Man" recounts, only one renegade Economalien warned of the danger...

@Bill: The hypothesis being tested is not of the generality as would be an hypothesis say about the ideal combination of stroke weight and width of character. But it still has generality because it applies to a broad class of events, and predicts that Clearview Highway will be consistently readable as greater distances. This has some universality and so is a scientific hypothesis.

That's not a broad class of events, when considering the huge variety of media and documents in which fonts can be used.

This is a font product test of Highway Gothic vs. Clearview, for a very specific kind of document (a certain kind of highway sign), and has very little universality or generality. Highway signs are a tiny, odd niche of typography.

Bill, do you seriously believe that the hypothesis "ClearviewHwy 5-W is a better US highway sign font than Highway Gothic E(m)" is hard science, and not merely font product comparison? The extreme conditionality (to use the term introduced by Peter) of such "research" (with a dose of product marketing) is what makes the science so soft if any universality is to be claimed.

Highway signs are designed, as are driving conditions (by highway engineers and legislators), and therefore it is wrong to even assume that the US results today would be applicable elsewhere or in the future, although they probably are, given the crudeness of Highway Gothic fonts, and the top-notch design of James Montalbano.

I hypothesize that 99% of graphic designers would choose Clearview.
How could anyone with a modicum of typographic acumen not pick Clearview, with its larger x-height and more open counters, to be the better face for this job?

However, it is wrong to see the Clearview vs. Highway Gothic font tests as a refutation of Highway Gothic as a typeface—Interstate, a version of Highway Gothic with comparable sophistication, weights, and styles to Clearview, would no doubt give it a run for the money.

You don't seem to understand how controls work in science

I have consistently stated, in the many Typophile posts in which I dismiss typeface readability as a scientific notion, that the multiplicity of variables involved in typography, and the elusive quality of many them (hard to control), makes this aspect of reading research so soft it's bunk.

Incidentally, Bigelow is the co-creator of the Lucida type family, winner of the MacArthur Prize and Goudy awards, and head of RIT's School of Print Media. The idea that he is in this paper with scientist Legge "poking his nose" into something he doesn't know about is laughable.

Yes it is Bill, and I would never suggest that. If you review what I said, I criticized the scientific profession in general for poking their noses, and Mr Bigelow is not a scientist AFAIK, so you are twisting my words.
Bully for him for getting involved in this inter-disciplinary project, and including design literature sources in the review.
Now let's see him finish the job and present this work to his own profession in an appropriate manner.

Nick you are equivocating about the meaning of "science", and how general something has to be to count as science. But in any case you originally were talking about sound "conclusions from research." This is empirical research, and scientifically valid because it is testable through observation and experiment, and is repeatable. And it is valid in spite of the many variables involved. They controlled the other variables and got a scientifically valid comparison. That is what is at issue, not how universal the theory is. Again, a red herring.

You are also simply repeating that there are too many variables, begging the question. And again, this just shows an ignorance of science. There are a myriad of variables involved in biological processes in the body, but this hasn't stopped science from testing the influence of particular processes, such as red blood cells in respiration or insulin in sugar use.

At any rate, I have concluded that it is futile to try to shake you from your certainty that you will never have anything to learn from science about the readability of typefaces, and I'll stop here.

I am very interested in hearing more about the content of this paper though.

Nick you are equivocating about the meaning of "science", and how general something has to be to count as science.

I am differentiating between hard science and science which is so soft it's bunk.

They controlled the other variables and got a scientifically valid comparison. That is what is at issue, not how universal the theory is. Again, a red herring.

If so, you are the one who introduced it: "This has some universality and so is a scientific hypothesis."

There are a myriad of variables involved in biological processes in the body, but this hasn't stopped science from testing the influence of particular processes, such as red blood cells in respiration or insulin in sugar use.

Science has its limits.
It's not just the quantity of variables involved in reading that's problematic, but their intractability.
Measuring the physical properties of biological matter is not quite the same as measuring cognitive response to cultural artefacts.

@Nick Shinn:
Clearview is superior to Highway Gothic in most respects. However, I can see one advantage Highway Gothic has: it is bolder.

So, under one set of poor viewing conditions, the Highway Gothic characters might become white blobs. Under another, the Clearview characters might become invisible. At least it is possible to see the shape of a blob.

This is not to say that I think Highway Gothic is better; it clearly isn't. Just that it seems as if it should be possible to do even better than Clearview - if it were made slightly bolder, its superiority over Highway Gothic (if that's its main competition) would be made more obvious and indisputable to even the most typographically unsophisticated individual.

Ah, but after a short search, I see that it was deliberately made significantly lighter because at night, when there legibility is usually the most serious problem, signs hit by headlights and seen by drivers with vision problems will likely have the white area appear to expand, and so the design was intended to deal with that specific problem. And the adoption of Clearview as the replacement for Highway Gothic is proceeding apace, and so there's no problem with acceptance by state road authorities.

…this just shows an ignorance of science…

Now you're being as hubristic as you accused me of being!

.

It's not hubris, it is just an observation based on long experience. I spent a lot of years studying history and philosophy of science, got a PhD on the subject in the process, and have taught it and published books and articles about it. It doesn't make me right about anything in particular, but it does mean I have some familiarity with the subject.

The objection of "too many variables" in my experience shows ignorance of the history of science, as that could have been said of any field before its breakthrough ideas or discoveries. And some aspects of visual perception are in fact already well understood by science, contrary to what you seem to think. For example, the fact that the visual nervous system has a process that is structured to pick out edges is well established in spite of "many variables." And we already have distinguished individual type faces under specified conditions—NOT JUST INDIVIDUAL DOCUMENTS—so as I have said the "too many variables" objection is already refuted in numerous ways.

—And the point of my saying that the Clearview tests have a level of generality is just that they go beyond individual documents, and apply to conditions with potentially unlimited instances—unlimited numbers of signs. --Not bound to specific documents.

More to the point this very article that we would be better discussing has a robust result showing that critical print size is twice the threshold for visibility under the same light. That's another example of an interesting finding, in spite of "many variables."

It is hubris, and you're patronizing me.
But you're right, the meta-discussion of whether this research is relevant to typographers has been much too general, and should focus on specifics in this paper, such as critical size/acuity threshhold size.

Highway Gothic has: it is bolder

It's too bold. It just looks like a black blob from a distance.

Russell is right about what the research seems to indicate: that the boldness is a problem. The concept is that the increased counter space (white space) helps readability at a distance. By the way, I think it would be interesting to see whether this holds for visual acuity, or it is mainly for whole words. In other words, testing letters on their own vs several words.

Incidentally this result aligns with Luckiesh's finding that it was less fatiguing, measured by blink rate, to read the medium weight of the slab serif Memphis, compared to the light and bold weights. Frutiger has also argued that there is a "right" normal weight, and the concept goes back to scribal practice, with an ideal ratio of nib width to height of strokes.