Steven, in these ~10 months that you've been on Typophile,
have you been reading the discussions concerning readability?

hhp

Yes, I have read a bit about it here. Sorry, I hope I dont annoy you with this thread!? Its more like a simple summary for myself and want to share it with you!? Im from Germany and sometimes its a bit hard for me to follow your discussions which also sometimes requires a high know how and good knowledge of the technical terms. So if something is totally wrong: please say it. Or if something very important is missing: please say it.

Please be patient with me :(

Steven

Your rules are mostly appropriate (though not extended width) to, say, signage legibility, rather than to readability - or at least I think that would be the majority opinion round here - just so you know.

Sorry if I snapped (and certainly don't worry about annoying me). I guess I have a Pavlovian reaction whenever I hear the Emigre mantra, after all this public effort some of us put in to explain how near-sighted it is. On the other hand, something like "sans serif is better than serif" - where is that coming from? Are you perhaps distinguishing between legibility and readability?

hhp

Research about boybandability:

Boy Member #1
* Drop dead gorgeous, but in a pretty way.

Boy Member #2
* Drop dead gorgeous, but in a rugged way.

Boy Member #3
* Must be goofy looking. Added bonus points for dyed hair.

Boy Member #4
* Plain looks - could be your boy next door

Boy Member #5
* Hot looking, but in a rough and tough way. Added bonus points for scars and/or tattoos.

--Kristina Frazier-Henry

I am a bit cunfused.
So what exactly is the difference between legibility and readability!? In germany you translate both with the same word "lesbarkeit"
I allways thought legibilty is the way the typeface has to look for best reading.
Readability is how the text in itself has to look. For example: textsize, textlength, layout, choice of the typeface etc.

Steven

It depends who you ask (not surprisingly), with some people saying there's no difference* (and that's assuming we smartly ignore those who say none of it matters anyway). What I've found to be most useful is to see: legibility as pertaining to individual letters, usually seen for a long time (meaning seconds), sometimes at a long distance, very small (or very large), etc. - you get the picture; and readability to mean the comfort with which people can read a font for a long time. The two are really quite different - certainly not diametrically opposed, but I like to think of them as two vectors 90 degrees apart.

* These tend to be the same people who don't distinguish between text and display fonts (and as a result can't make good ones of the former kind).

hhp

Readability is about continued reading, legibility about deciphering individual letters and words. There is a tradeoff between the two. The latter is probably best served by widely-spaced, monoline sans-serifs, for instance, but the former by contrasting seriffed fonts.

Here are some rather random notes on what makes for readability: they're hardly the product of research, and may all be nonsense.

Think of pattern not rhythm - i.e. robust and taut white shapes, not stem distance regularity.

Construction of stems, bowls, counters, cross-bars, diagonals, and how they join matters more than detailing.

Limited but clear contrast assists differentiation and adds life, about 3 or 4:1 being ideal.

Differentiation within a small range in many aspects of the design will obliterate a consistent chirography; however chirography provides two (broad and pointed pen) admirable starting points for design, and since real calligraphy is never so geometric as, say, Minion one can learn from it especially once one has internalised the effects of miniaturisation.

Variety matters especially around the top of the x-height.

bpdq bowl-height should be apparently less than o's height.

Vary axes for differentiation, both exterior versus other exterior axes and interior versus exterior axes.

Vary textures for interior and exterior outlines: rigid and angular in one, rounded in the other.

Basic forms may be selectively differentiated by e.g. giving c 16th-century Roman capital serifs, or e a terminal beak, or h a rounder right leg than n.

Text lowercase baseline serifs should be only minimally cupped, if at all. The primary purpose of lowercase serifs is to limit lateral interference: they should be long and strong. Near-equilateral triangles are not good enough.

Perceived blackness flows into stems from intersections and serifs; smooth colour-gradations require the management of intersections. Thus the horizontals of n and m should meet the left stem immediately below the serif.

Distance between stems of n and m should be a touch smaller at the bottom than the top if the characters are not to seem inflated by the open counter at the bottom.

There should be no superfluous neural spiking in reading - the text should be transparent. Difficult letter-combinations, such as gg and ra, may be ligatured.

x-height for 12 point should be about 55% of h-height, and descender length about 55% to 75% of ascender length.

Examples of quasi-Aldine proportions:
----------------16 pt---11 pt---8 pt
x-height--------4.5-----5-------5.5
Italic Cap------5.5-----6-------6.5
Roman Cap-------6.5-----7-------7.5
h-height--------7.5-----8-------8.5
p-height--------7.5-----7.5-----7.5
Leading---------1.5-----1.5-----1.5


In small text, small irregularities, white and black, should be relatively larger.

Both nearest points and stem-distance matter in spacing. At 12 point, optical space between characters should be about three quarters of that within the characters' counters. At 8 point it should be about seven eighths.

Steven, in my post of May 2, in this thread I gave what I think is the standard distinction between 'legibility' and 'readability' in English. It is controversial whether the distinction is really valid.

If you do a search on 'readability' in the typophile search box, you will find more than you want to read on readability!

Here are two of the longer threads:

"Readability and Readerability" and "What makes and italic easier to read?"

George: I like.

> taut white shapes

Ooooh baby.
;-)

> the standard distinction between ‘legibility’ and ‘readability’ in English.

Except we don't use standard English. :-)

Everybody knows that when a type designer talks
only other type designers believe he's even human.

hhp

Thank you all for your explanations.
I will absorb it and save it in my vocabulary.

Thank you very much George for sharing your notes. They are quite helpfull for a beginner like me to differ a readable and a less-readable typeface from each other.

But I have one more silly question:
What makes a typeface more readable for a visually impaired reader!?
Do you have to design a textface more "legible" that every single letter could be recognised directly!? (almost the same rules like for signage legibility)

I should think so, yes. There was a bit of a scandal over Tiresias, a "research-based" type design project making a TV subtitling font for the visually impaired; I don't know whether there's any better research on this specifically.

What do discussions, definitinal and substantive, about legibility and readability tell me about whether I should set an edition of Moby Dick in Cheltenham, Gill Sans or Eaglefeather?

Is there evidence that the average literate person is having problems understanding the myriad texts in myriad type faces that he or she encounters every day?

Should I be telling my optometrist that my new prescription may be wrong because the typography on the test that he gave me last week may not have been optimal?

Anyone know what type was used for the first edition of Moby Dick? I'm thinking of writing a PhD thesis on whether the novel was panned because the type had the wrong bouma shape.

All in fun...

George, thanks so much for the results of your research in to the Aldine face.

Fascinating, and of course all debatable!

Steven, you should know that the state of research on readablity and legibility is not yet that strong. There are a lot of claims made for various type faces, but when you look into them, they do not actually have a solid research result behind them.

>we don’t use standard English.

True, the distinction between legibility and readablity is only used by people interested in type--and not all of them. I suspect that it came into existence in the 1930s, and I am guessing was introduced Stanley Morrison or Beatrice Warde. I would actually be interested to know the history of the distinction.

Bill, stress the quasi in quasi-Aldine ;). It's the best MM scaling I could think of that gives useful proportions, compromising between Aldus' and Griffo's (and indeed Jenson's, for the 16 point) practice.

"...the distiction between legibility and readability is used only by people interested in type - and not all of them."

One of the things that I like about Bringhurst's book is that he treats the issue of choosing type principally as a design issue, and he is quite clear about his design considerations (which one may or may not accept).

The words "legibility" and "readability" don't even appear in his index.

There may be an argument for this kind of research as pure research, but as applied research it strikes me (with apologies to Luigi Pirandello) as research in search of a problem, and a non-existent problem at that.

I guess that I'd like to think that after a few millenia of writing, people would have figured out how to write in a way that is legible and readable. Indeed, I was under the impression, when I was in grade school, that that is what my teachers were attempting to instill when we spent all that time learning to write.

The trickier question, and the one that Bringhurst focuses on, properly in my view, is what type to choose for a modern edition of Moby Dick.

Cheers

But as Bringhurst recognises, the best text type designers rethink their letters from the inside out, in structural rather than merely ornamental terms, at the levels of glyph-component, word and textblock; and the features which make for readability often follow from those same structural considerations that determine much of a good text type's character.

The relevance of readability comes not from asking a layman if he can read the word "dog" in a given font; it comes from the very real -if extremely hard to measure- fatigue a reader feels after hours immersed in a text, where things like setting a book in a sans will get you (even though you might have no idea why). This is what separates the mimics from the men so to speak (no sexism intended) in text face design.

> The words “legibility” and “readability” don’t even appear in his index.

Also, he is a poet.
Treat RB's book (or really any book) as gospel, and you will get nowhere.

Choosing a culturally valid font for Moby Dick is exactly
only half the problem. And the superficial, easy half at that.

hhp

"...the best text type designers rethink their letters from the inside out, in structural rather than merely ornamental terms..."

My personal view is that structure is nesessary, but that ornament and other attributes of style differentiate between what is necessary and what is aesthetically pleasing.

Not everyone would agree with one or both of the foregoing statements.

This is getting a long way from the original post. I just think that it is useful, if one wants to learn about and use type, to focus first on broader issues than legibility and readability. Why? Because I think that as a practical matter there are a lot of legible, readable type faces and there is more to be learned by paying attention to the aesthetic differences between them in relation to what one wants to communicate.

> focus first on broader issues than legibility and readability.

I prefer to focus on both facets, from day one until the end.
Even though I find aesthetic issues quite boring. This is in
fact the problem with designers who demote functionality:
they can't get themselves to pay attention to something they
find boring. Generally, they're too artsy to care about what
users really need (even if they never realize that). In contrast,
I find that designers who enjoy functionality often have a more
self-disciplined, humble philosophy.

But it's OK, we need you guys too. :-)
Even if you fail to see that you in turn need us.

> there is more to be learned by paying attention to the aesthetic differences

For a layman reading 9 pt text immersively, this is quite implausible.
Your stance is confined to display typography.

hhp

I think terms like readability and legibility have an uncertain or unstable existence in the typographical domain because, while they seem to evoke relevant considerations, they have never been fully specified in perceptual processing terms. A more native use of the terms exist outside type: a sample of writing is legible or it is illegible; texts are readable or unreadable (in discourse-analytic terms).

However, understanding the perceptual processing impact of manipulating typeform microvariables might be applicable if the notational devices that make up our writing systems are "constrained by the propensities and limitations of the human hand and eye." (Bringurst)

And perhaps more important than aesthetic differences--which is less of a concern to Bringhurtst than cultural differences--are gestural-atmospheric differences.

enne_son,

When I use the phrase "aesthetic differences", I mean it as a shorthand phrase that includes artistic, cultural, historic and contextual considerations of the kind that Bringhurst discusses, especially in chapter six of his book.

Although the feeling of fatigue may be subjective and individually variable, eye fatigue is associated with specific muscle activity around the eye, which can be quite precisely measured with the correct tools. So fatigue, or rather the physiological activity that results in fatigue, is actually one of the easier things to measure.

But are you retreating to a position of fatigue reduction, Hrant? In the past you have posited possible improvements in speed and accuracy.

John,

Are you suggesting that this business about legibility and readability is ultimately about Evelyn Wood Reading Dynamics :)?

John, I feel those all boil down to the same thing (and this seems to be one thing Kevin and I agree on). Helping saccades get longer translates to less fatigue in the long run. Because in the end, it very rarely matters how fast you read something, what matters is, is the reader discouraged enough (via fatigue) to not finish the piece; is a usability threshold crossed. That said, the reason to concentrate on something like saccade length is that it's so much more direct and measurable.

hhp

For the typographer, optimizing readability doesn't just depend on the choice of typeface, but on other variables such as size, leading, line length, paper stock, page size, layout hierarchy, margin size, amount of text, and who's doing the reading.
Changing any one of these variables affects the others.

In the real world, one is faced with deciding between typeface A which looks better at 9.5/13, and typeface B which looks better at 9/14. Unless laboratory testing develops an index of readability for all the typographic variables, it's of no practical use in any but the most banal typographic projects.

For standard formats on standard equipment, such as templates in Microsoft Word, running on Vista with ClearType fonts, reading research has no doubt been useful in font development. But if people generally want some research-proven rules to follow about what typefaces are best for original layouts and design, that's not possible.

> For the typographer, optimizing readability
> doesn’t just depend on the choice of typeface

1) That goes without saying.
2) For the text face designer, it's extremely significant.

> it’s of no practical use

On the contrary. It can help Redge decide against considering Gill Sans for a book, it can reduce the chances of people using negative tracking to save space, and all kinds of other nice things.

> if people generally want

Well, people want all kinds of dumb things. But a lot of smart people feel they need some degree (not absolute) help from empiricism, because they're mature enough to realize they're not omniscient artistes.

hhp

>eye fatigue ...can be quite precisely measured with the correct tools. So fatigue ...is actually one of the easier things to measure.

I think there are other important dimensions of reading comfort. There is brain fatigue as well as eye fatigue. I think you might be able to test this by seeing after what time readers' comprehension falls off. I suspect that this would not just be the measurable eye strain.

As I remember, Kevin Larson was reported here has having tested people's ability to use concepts they read, and found lay-out mattered. I think the same thing would prove true with type face, though perhaps longer reader periods would be necessary to show the difference.

Nick, that lay-out factors are equally or more important than type face does not contradict the claim that there are important differences in readability of typefaces, especially at text sizes.

I think if even redge had to read a long book in 10 pt Helvetica or old-style italic, and had the same book set in 10 pt Sabon or Minion, he would be mighty tempted to switch to the book set in these real text faces.

As I remember, Kevin Larson was reported here has having tested people’s ability to use concepts they read, and found lay-out mattered.

Not quite. This relates to Kevin's presentation in Helsinki. It does not have to do with ability to make use of concepts in what has been read, but in the ability of the mind to solve puzzle-like tasks after enlivening mental activity. The basic observation had been established in experiments in which subjects were given tasks to perform after a drily delivered lecture and a humourous video. Consistently, subjects performed the tasks better after the humourous video. Kevin's tests found subjects performed tasks better after reading a well typeset document than after reading a poorly typeset document. The tasks are not related to the content of the reading.

>The tasks are not related to the content of the reading.

Thanks for explaining it. That's interesting also. Is it because the well type-set text and humorous video are less fatiguing? Did he have any theory about it?

"I think even if redge had to read a long book in 10pt Helvetica..."

Actually, I didn't say a single thing that suggests that I would do that.

In fact, I said a few days ago in another thread that I was adamant about using serif for the book that I'm working on despite suggestions from participants in this forum that I look at sans serif types.

The Straw Man makes a great character in The Wizard of Oz, but a very poor basis for an argument.

Cheers

>There may be an argument for this kind of research [on readability] as pure research, but as applied research it strikes me (with apologies to Luigi Pirandello) as research in search of a problem, and a non-existent problem at that.

>I just think that it is useful, if one wants to learn about and use type, to focus first on broader issues than legibility and readability. Why? Because I think that as a practical matter there are a lot of legible, readable type faces and there is more to be learned by paying attention to the aesthetic differences between them in relation to what one wants to communicate.

My example was to show that it is quite important practically how readable a typeface is--something I thought, perhaps mistakenly, you rejected in the above quotations.

A person designing a typeface intended for text should, I think, be extremely concerned about its readability. Otherwise it is going to be a bad text face. For the text type designer it should not be a 'non-existent problem'.

And similarly when a graphic designer sets extended text, he or she should be concerned that the typeface and the layout together are readable. And this involves looking at both the qualities of the typeface and the layout, both of which will contribute to readability or detract from it. It is not a trivial problem to produce readable text, I think, particularly when the designer also needs to be addressing the production and aesthetic demands of a text to be printed.

Aesthetics are interesting and important, but it seems to me readability is also at least equally worth focusing on for extended text. This you seemed explicitly to deny. Did I misread you?

Actually, I agree with everything that you are saying. If I wanted to quibble, I'd say only that aesthetics and readability, except where there is a deliberate decision to divorce the two, are co-incident.

I agree that a beginning designer needs to learn about what is legible and readable, although I'd also like to think that that is not a difficult thing to learn.

I also have no problem if someone wants to say that there is a holy grail of legibility and readability that, once observed, will leave the last few millenia in the dust.

That said, I don't believe that the holy grail, if there is one, will be found by studying bouma shape.

On the other hand, I do believe that there are developments in type that reflect cultural changes (New York subway grafitti is an obvious example), and that become art (the kind of thing that people like Basquiat did with subway art).

I guess that I just figure that I'll leave the search for holy grail, if it conceived as an empirical exercise, to someone else, especially since there is so much beauty (not to mention legibility and readability) in what we already have.

Cheers

>there is a holy grail of legibility and readability that, once observed, will leave the last few millenia in the dust.

Ok, now it's my turn. Where did I say or imply that?

It isn't a 'straw man', but it ain't me :)

My point was that I agree with you. I don't think that you have said anything that is controversial.

My comment is a reaction to some of the things said in this thread and in some of the threads that are linked.

I should add that I've edited my last post a bit, which may or may not clarify things and/or get me into more hot water :)

John,

I hope that there is a Cuban or Cuban expatriat who will respond to the thread in the design forum on revolutionary Cuba.

THAT is an interesting subject.

Any comments yourself?

Redge, in this thread, you asked: "What do discussions, definitinal and substantive, about legibility and readability tell me about whether I should set an edition of Moby Dick in Cheltenham, Gill Sans or Eaglefeather?" People have been discussing the unsuitability of sans fonts (such as Gill, and Helvetica) for setting long texts for ages, including on Typophile for 6 years. Many of us have done at least a fair job explaining why/that serifs help reading; certainly a much better job than the self-serving hackneyed droning of the Emigre mantra, or the cloying cultural over-valuation of poets.

> aesthetics and readability, except where there is a
> deliberate decision to divorce the two, are co-incident.

Poppycock.
Life isn't so one-dimensional.

> that is not a difficult thing to learn.

On the contrary, it's so difficult that even over
a century after Javal's discovery we're fumbling.

Holy Grail? No such thing. More like a taming of selfish artistry,
in order to help users, not humor one's own ego, or accomodate one's
intellectual lethargy.

--

BTW, editing is for the fearful.

hhp

> THAT is an interesting subject.

Boring as hell to me.
But I would be an idiot to ignore its relevance.

hhp

Hi hrant,

Thanks so much for explaning to me that I shouldn't set Moby Dick in Gill Sans. Until you came to my rescue, I was on the verge of making a big mistake.

How about if I set it in Eaglefeather? That way I could make an oblique reference to the close relationship between Herman Melville and Frank Lloyd Wright, which everyone knows about seein' as how Wright had a copy of Moby Dick in his library. Besides, they were both Americans.

If you don't like that idea, how about Cheltenham? I hear that it is even better than Times Roman. I also hear that the New York Times has recently decided to use Cheltenham, and New York isn't all that far from where Moby Dick starts off, although I'll admit that there aren't a lot of big white/albino killer whales in the Hudson River.

How much would it cost me to hire you to do some empirical research to figure out this problem, especially figuring out the most legible/readable type? I'm especially concerned because when Moby Dick got published, it was a good deal less successful than the Da Vinci Code, and I figure that you might be able to explain why Dan Brown's book is a hit and Melville's book was a bust.

Yours in anticipation, and with an open chequebook,

redge

PS. I agree that Cuba is boring, but as you point out it is relevant. Hemmingway's old man, unlike Melville's guy, managed to catch his fish. You can see what I am up against, trying to publish a book about a fisherman who can't catch worth a damn.

So, when you finally break down and reply to me directly (it didn't take much for your covert resolve* to wilt) you resort to tired Anglo humor to avoid actually making sense. I've seen this so many times, it's become utterly, predictably boring. Be forewarned however, when the humor fails even more clearly than the feeble attempts at logic, the next stage is often a frothing rage. But please, instead of cutting your ear off, go paint something. And keep it confined to your bedroom.

* Using the word lightly.

> I was on the verge of making a big mistake.

No, the mistake is in your distant past.
And the further it falls back, the harder
it will be to see and correct.

> you might be able to explain

No, I can't explain the behavior of masses
of peons, and I don't care to. I'm not into
fashion - my idea of culture is different.

BTW, Cuba isn't boring at all. But ignoring the needs of regular Cuban citizens in favor of masturbating about some semi-hallucinatory cultural insight into its past is not only boring, but vapid.

If you want to really understand type, get over your lack of interest in
functionality, and learn about how humans work. Otherwise you'll be
stuck in the dumb buxom-blonde world of display typography.

--

It's always the same with artistes butting into threads about functionality
simply to tell people to stop talking about it. Note that the other way around
almost never happens. Why? Maturity, objectivity, doubt. All anathema to painting pretty pictures.

hhp

Nick, that lay-out factors are equally or more important than type face does not contradict the claim that there are important differences in readability of typefaces, especially at text sizes.

Perhaps there are, but they can't be meaningfully measured as there are too many variables. With one set of fixed parameters, it may be possible to say that subject A preferred one typeface over another, when set at a certain point size, leading, line length etc, for text of a particular length, on a particular subject. But it's impossible in practice to vary all the parameters against one another to create a "readability index" for typefaces.

Hrant,

At the risk of seeming like an arteest, Don't you think there is a limit to how much we can discuss readability without additional work at developing typefaces specifically geared toward reading comfort, and also without mechanisms for testing them (and other, existing typefaces)? In other words, whenever someone asks about readability, there is a lot of discussion, but the final collection of data is always so impoverished. I have my theories, as well, but I can't spend time explaining them, if there are no prospects for new experiments that test the theories. I'd rather work on the experiments. What is the ideal testing environment? Can we propose some controls and methods to appropriate scientists?

Note that the other way around almost never happens.

Except when someone mentions Mrs Eves or Bodoni as a text face, and Hrant starts spouting his you-don't-know-readability mantra.

Carl:
Yes, I do think there's a limit. When we collectively
come close enough to it, I'll stop yakking, I promise. :-)

> I have my theories

Good. And lacking "proof" of them might limit how far you'd dare to apply your theories, but that can't stop you from applying them somehow. Otherwise you'd be guilty of hypocricy. Different people have different thresholds of crossing from assumption into action; the important thing is to care.

As for the issue of What Is A Good Test, that's exactly what needs to be figured out. I have heard some good ideas, including on Typophile by people like Peter E, but since actual practice in the field is so, so far from any ideal, this is actually an area where frankly I am hopeful of being paid to put some serious concrete thought into it by somebody or other eventually. What can I say, I'm an optimist. Now, I'm not the only person who can pull this off - but there certainly aren't many of us. So I'm saving myself so to speak. :-)

hhp

My eyes are bad, and getting worse, so my view of legibility would be anything 12 points and over. Not constructive I am afraid, but it does have something to do with the question

Nick, I'm not the only person who thinks Mrs Eaves* and/or Bodoni are flawed as text faces. Think of Robin Kinross for the former, and G W Ovink for the latter. You've been told this before, by people besides me, but somehow you manage to forget it. I understand, it's simply too inconvenient.

* BTW, if you think poor spacing has no effect on reading: please stop droning on about the Divinity of Even Color; and you might actually be in agreement with Kevin on something!

But anyway, I was actually talking about something else: how discussions of culture and aesthetics don't seem to get interrupted with "Don't you talk about that!" tantrums.

hhp

Nick, I’m not the only person who thinks Mrs Eaves* and/or Bodoni are flawed as text faces.

Their merit as text faces is not at issue: rather, your bitching about readability threads being hijacked by designers is, and saying it rarely goes the other way, when in fact you are always going on about your readability theories at the slightest opportunity. But of course, change the subject, a convenient distraction.

BTW, I don't know where you get the idea I think those faces are good for text. I've never used or recommended Mrs Eaves. I've tried all the Bodonis -- they're difficult to work with, but some versions are suitable for some text uses.

***

Perhaps the reason some designers and typographers are antipathetic to readability research is because they feel threatened by it. If one has spent many years acquiring a refined sensibility, and non-practitioners seek to quantify that in a banal manner, there's reason enough for hostility, protecting your turf. Especially when a company like Microsoft is involved -- your favourite bugbear, middle aged Anglo creatives, are all too familiar with the way hard-earned skills have been replaced by automated software.

What does the future hold in store? Already MS is touting the scientifically-proven readability benefits of its ClearType faces in marketing language, so who needs a typographer to spec a typeface when the engineer can tell the client what works best?

> you are always going on about your readability theories

But critically, not to the point of excluding the other half, is the point. Presenting opinions is one thing, trying to prevent others from doing so quite another. And it has seemed to me that those who prefer to focus more on aesthetics have a lesser tolerance of the other half than the other way around.

> they feel threatened by it.

Exactly. Weak.

> your favourite bugbear, middle aged Anglo creatives

Almost: remove "creatives" (such a dumb term anyway, as if they have the monopoly on creativity just because they've put themselves -or more often ended up- in a position of painting pretty pictures), and add "males". And it's not my imagination: the proportion of hostility towards me coming from that narrow demographic is extremely telling.

> hard-earned skills

Oh, but of course, the dumb, boring technical people have acquired their insights with no effort. Ask yourself: why do they seem to feel less threatened by flaming primadonna artists posing as designers who want to throw empiricism -not to mention plain old analysis- out the window without a second thought?

> who needs a typographer to spec a typeface

I wouldn't call the type of person you're talking about a "typographer", or even a designer of any sort. What you're talking about is a person primarily interested in expressing himself. Boring, and pretty useless. But luckily for them, there will always be people who fawn over the behavior and output of others, just like the masses of peons adulating celebrities. "Ooooh, your fonts are so pretttty! I don't care what kind of person you really are, just sleep with me... I'll even pay you!"

hhp

The subvisible subtlety and power of the text face cannot be
appreciated and certainly not leveraged by mere aesthetics alone.
To contrast it with the display face, I allude to this singular quote:

"The poster is to art what rape is to love."
- A M Cassandre

hhp

Good idea Bill - though perhaps Plantin Light could be substituted for Times (since PL has real serifs rather than just brackets).

“that the first thing is just to establish experimentally the validity of the distinction: to show that faces that are equally good when it comes to reading a word in a short flash are still different in readability.”

Bill, I think this might be very worthwhile and could definitively establish or unseat the presumption of ’generalizability’ or ’ecological validity’ across different testing paradigms, which both Hrant and I are concerned about. George, a full program of research would seek to elicit meaningful results against the kind of manipulations you describe. Kevin, did you ask because you see real prospects for carrying out studies based on the suggestions?

Are some of you arguing that a person is more likely to read War and Peace from beginning to end, or for that matter Catcher in the Rye, if there are improvements over existing type design?
Yes - or at least improvements over, say, Sabon Next. There are other reasons for failing to finish those two, of course: both are kiddies’ fables. How the Russians must chuckle at our misinterpretations of their literature: we even take that windmill of journalistic gesticulation Dostoevsky seriously!

If so, what are those proposed improvements and what concrete examples are there, as distinct from a lot of whinging and academic debate?
Well, my suggested (and unoriginal) improvements are given earlier in this very thread, and I’m sure lots of people here have their own pet predictions.

Specifically, is there a type design that will make it more likely that one will finish War and Peace (or if you want an Enfgish example, The Adventures of Augie March) and, if it exists, which one is it?
Thanks for providing an example of a long book in English. And the type you want is a decent Griffo revival - if only such a thing existed ;-)

Kevin, here is another area to test, which Peter Enneson and I have been e-mailing about.

The scrambled letter tests refute the idea that we read by the external shape of words only. However, Peter has argued that while we don’t simply read from the external outline, we do read from the whole pattern of salient letter characteristics in a word—joins, open and closed counters, ascenders, etc. This is different from a process of first identifying letters—even simultaneously in parallel—and then going from there to identify words.

Because we have evolved such a phenomenal ability to identify human faces, I find it plausible that at the very least we can recognize ‘th’ or ‘oo’ within a word and map that pattern—and perhaps even many whole words—directly to sound and meaning without going to the intermediate step of identifying the letters.

The interesting possibility here to me is that if Peter is on to something, mastering reading may involve a transition from letter by letter reading, to reading by groups of letter patterns, or whole word patterns. In other words, the mastery process would be not a matter of increased skill in processing already identified letters in a particular order, but the actual visual pattern made by the letters.

The problem is how Peter’s view could be possibly tested? One idea I have suggested to him is that the same test could be done on children with different degrees of mastery of reading. One idea is to increase letter spacing to the point that it is disruptive to expert readers. My idea is that this would not slow down the beginning reader significantly, as he or she would be doing the letter identification, but the expert reader would be doing less letter identification, and more pattern recognition, and so would be slowed down significantly.

The idea here is that the word as ‘pattern of letter features’ is dependent on tighter spacing than simple letter identification. And this test on different levels of mastery would reveal it. As I don’t know the literature on reading, I suspect that this test I am suggesting may well be too crude to be a crucial test. But I think that using different levels of reader skill, or those with reading disabilities, in tests would be the best shot at seeing whether pattern recognition beyond single letters is an important factor in reading expertise.

George,

I’d just appreciate it if one of the designers who criticise current type on this site on grounds of readability, and go on at length about imperfections, would say what they have designed that would be better for, for example, a serious novel.

I didn’t anticipate that someone would characterise War and Peace as a “kiddie’s book”, but if that is what you think, that’s fine. Just substitute a title by someone that you consider more adult than Tolstoy.

Ha! My jibe was immature, being no more than an expression of loyalty to a great Tolstoian, Nabokov. But W&P really was written for the “general reader”, and so has no artistic structure but lots of lovely morals. Anna Karenin is an altogether sleeker thing.

> ‘Readability’ is the ease with which the eye
> can absorb the message and move along the line.

There are at least two big problems with that:
1) It confounds perception and cognition.
2) It’s too short-range. Moving along a line is far
less important than performance in extended reading.

> For example, you could take the Sheedy study and do the same
> variation in spacing, and test this time not for speed, but
> for the fatigue factor.

1) Speed (over a long period) and fatigue come down to the same thing.
2) Studies like Sheedy’s will give you no insightful results because the subjects aren’t getting truly immersed - plain and simple. Essentially, they are confined to the legibility sphere.

> Are some of you arguing that a person is more likely to read War
> and Peace from beginning to end, or for that matter Catcher in
> the Rye, if there are improvements over existing type design?

As De Groot has wrily pointed out, when a reader is motivated it almost doesn’t matter how unreadable things are - he will complete the reading task.* The relevance of readability kicks in at the threshold: where a reader might give up reading something he has to read (sometimes before a deadline) because of fatigue.

* Although even in this case you could say a good designer should still feel responsible for not making things unnecessarily uncomfortable.

> The scrambled letter tests refute the idea
> that we read by the external shape of words only.

No test is required to proove something so obvious.

> map that pattern—and perhaps even many
> whole words—directly to sound and meaning

No sound.
But yes, as my TYPO article explains, we identify
chunks as wholes that are as large as possible.

> if Peter is on to something

Your crediting is wanting.

> reading may involve a transition from letter by letter reading,
> to reading by groups of letter patterns, or whole word patterns.

Yes, and the transition from letterwise to bouma reading typically
happens at around age 10 (see Tinker, and Watts & Nisbett). But there
is no further transition to whole words - that’s just a subset of bouma reading.

> One idea is to increase letter spacing to the point that it is disruptive to
> expert readers. My idea is that this would not slow down the beginning
> reader significantly

Again: you can devise all the smart tests you want, but if the subjects (even adults) are not getting fully immersed, you will get no solid data that disproves Kevin’s model. And note that Sheedy study expanded the spacing by as much as 2 points - more than enough to “disrupt expert readers”.

—

When I suggested to Kevin in Thessaloniki that varying letterspacing would give great clues into reading, I made sure to accompany it with a caution: he would only see variance in the results if he got his subjects truly immersed. Otherwise he would get the vacuous results that Sheedy did.

—

Testing something intelligent is only half the key.
The other half is getting experienced readers to
forget they’re being tested.

hhp

Again: you can devise all the smart tests you want, but if the subjects (even adults) are not getting fully immersed, you will get no solid data that disproves Kevin’s model.

So what you really need is a measure of immersiveness. Without that, you risk begging the question, because you’ll be able to say of any test that fails to disprove Kevin’s model that obviously the subjects were not getting fully immersed.

So you need a definition of immersive reading that includes measurable criteria.* You need to be able to establish immersion independent of the results of particular tests.

* Even a definition without measurable criteria would be useful progress, since the term immersive reading does not seem to feature in any of the cognitive psychology literature to date.

How about observing each reader’s response to a series of distractions, at first imperceptible but steadily increased, in order to asses the depth of immersion?

>Your crediting is wanting.

Hrant, I didn’t intend to ascribe the invention of the theory of reading by wholes to Peter. I know that it is a relative old theory—predating World War II at any rate—and that Bouma and Smith and others have contributed to it. I am not clear about what exactly you and Peter have added to these thinkers, and what exactly is the relationship of your and Peter’s ideas. I do know that you corresponded early on, and that you don’t fully agree now. I referred to Peter’s ideas because I corresponded with him and felt I had a partial understanding of these.

I will have to leave it to you two to explain your views and the influence of past views and your influence on one another, as I don’t know these things myself.

My contribution above, if it is indeed any contribution at all, is of an idea for testing the general idea of reading by word patterns; it is not a contribution to the theory itself.

I’m interested in hearing about the kinds of studies that people are interested in seeing, and am open to letting these ideas influence the directions I take. I am particularly interested in studies that would help people make practical decisions.

John is correct that immersive reading isn’t discussed among cognitive psychologists. I think the typical usage of immersion has little to do with typography or the reading process. Instead it is about how much you engaged with the content you were reading. Are there other definitions of immersive reading? George’s suggestion that we could measure reader response to distractions implicitly suggests that readers are immersed when they are not interrupted with external distractions. Is this true?

In an earlier post George suggested a variety of interesting type manipulations, but there are devils hiding in those details. For instance how do you test for optically corrected height. I assume that optical corrections have to be treated as a system – you wouldn’t change the height without also changing the counter-width and letterspacing? What size do you examine? Will people read novels or textbooks? What two (or more) reading materials are equal so reading times aren’t determined by the difficulty of the content. How will comprehension be measured. While there are a variety of ways that psychologists have measured reading speed and comprehension, I know of no accepted way for measuring “comfort”. Your point is well taken that there are more interesting typeface comparisons than Times v. Gigi.

William suggested that comfort can be measured with Sheedy’s threshold distance measure. I certainly do think this is an interesting measure, and Sheedy has demonstrated that it can successfully detect differences between fonts. I think this measure is interesting because it taxes the visual system and processes that must be involved in reading. Helvetica and Times do not measure equal either with a distance or time threshold measure. In Sheedy’s spacing study he tested both distance threshold (legibility) and reading speed (readability), but not a repeated measures on the legibility to see how that data changes. I’m a little uncomfortable naming this a fatigue or comfort measure because we don’t know that fatigue comes from the processes involved in this task.*

* There is some reason to believe that some kinds of eye fatigue are caused by subtle amount squinting and can be measured with the obicularius muscle. I was hoping to talk about this at ATypI this year, but the proposal wasn’t accepted.

On the proposal for testing the switch in children from letter-by-letter to word reading, I’m certain you see the problem for looking for a lack of difference between loose and tight letterspacing. Failing to detect a difference between two conditions does not mean that two conditions are equal – methodologies and statistics are set up to conservatively detect differences. There are no methodologies for detecting sameness. It would be a much stronger prediction to say that adult word readers will benefit from tight spacing while children letter-by-letter readers will benefit from loose letterspacing.

Cheers, Kevin

>William suggested that comfort can be measured with Sheedy’s threshold distance measure.

I’m afraid I wasn’t clear enough. My suggestion was to measure comfort *independently* by testing at what point comprehension decreases. For example, you could give a string of SAT type reading comprehension tests, and see how long it takes for people to get tired and their ability to answer correctly goes down. Once you have this for a standard layout, say in Times New Roman at a giving size and layout, then you could do variations and compare how the numbers answered correctly decline over time. Sheedy’s test on spacing—and I just saw the abstract, so I don’t know it—would just be one thing to try among many others. The point is having a decline in comprehension with time being a measure of reading comfort. This measure could be used in many tests. Is that clear?

>Failing to detect a difference between two conditions does not mean that two conditions are equal – methodologies and statistics are set up to conservatively detect differences.

Again, perhaps I wasn’t clear. There wouldn’t be a difference with the kids at an early learning stage—or a small difference, but there would be a difference with the expert adult readers. What would be significant is that the adults would have a big difference and the children little.

>It would be a much stronger prediction to say that adult word readers will benefit from tight spacing while children letter-by-letter readers will benefit from loose letterspacing.

Well that is a variation of what I was saying. I don’t see why no advantage to children vs big advantage to adults isn’t also a measurable distinction. Of course if children are advantaged at a certain point, that would be interesting too, but the results would be meaningful in either case.

Redge,

I haven’t been criticising typefaces. I’m disappointed with the way the tests were designed. The types have their merits in appropriate use. But many of the tests ask irrelevant questions, or compare apples and oranges, or fixate on micro results. Of those typefaces I’ve designed or am designing, only one would be suitable for any book work. That’s not the point though.

To answer Kevin: I want to test existing and new text typefaces in real-world, immersive reading situations to find out which ones allow fastest reading, AND least fatigue, AND best comprehension. Obviously it’s a difficult task. One, I hope it’s clear, that can’t be accomplished without an expert typographer, and most likely also a type designer.

Issues like disability, developmental stage, legibility, line spacing, line length, and other distractions need to be eliminated from such a study. Other variables and effects could be the subjects of other studies. There are so many useful, interesting topics (highway signage legibility, captioning quality, onscreen reading, wayfinding typefaces) I’d also like to see studied, that it’s frustrating to me to read the latest studies that don’t seem to do anything to even piece together the existing data into something we can use. By that I mean, something empirical that can be presented to the gigantic, new population of active type users, who are flying blind when it comes to producing functional documents and displays.

Kevin: George’s suggestion that we could measure reader response to distractions implicitly suggests that readers are immersed when they are not interrupted with external distractions. Is this true?

No. I’m quite capable of looking at a page of a book, even having my eyes perform reading actions, while a) thinking about something else completely and b) still not hearing what my wife is saying to me. :)

I agree that immersion must firstly imply immersion in content. I take it to mean something like complete unconsciousness of the act of reading, in which we are engaged only with what we are reading not with the processes of reading. That being the case, I suspect there are two different kinds of immersive reading: that in which we are also processing thoughts about the content, i.e. thinking about what we are reading analytically or critically, and that in which we are, for want of a better formulation, ’thinking the content’. The latter is most likely in the case of narrative texts such as novels or short stories with compelling storylines — the proverbial ’page turners’ —, which the latter is more common in non-fiction works. I suspect there is a difference in reading speed between the two, but I would consider both immersive in the sense of unconsciousness of the reading processes.

I’m not sure if immersion is something one could reliably measure, but I think one could identify kinds of content that subjects are more likely to read immersively than others. Having subjects describe their personal reading habits and patterns would be a useful first step in virtually any readability test.

I think immersive reading could be described by its duration as well as its intensity. Definitely, subjects need to think about the content, and never about the act of reading. If they do notice the type, that’s an indicator of low performance on the part of the typeface (where other variables are equal). But it has to be acknowledged in the study. That’s part of what I want to know about typefaces.

There is a thread in the Design Forum entitled Old Style Greek. This thread raises, in a concrete, practical way, the issues of legibility and readability.

The issues arise because the technology that governs film and video production and distribution has an impact on how images are made, including typographic images. The art of film/video, which governs title and end credit sequences, also has an impact.

In other words, in the film and video world, it is not possible to divorce the technological context and artistic choices from questions of legibility and readability.

That might be a minor matter, were it not for the pervasiveness of moving images.

Curious to know whether anyone has a comment on this.

There is some reason to believe that some kinds of eye fatigue are caused by subtle amount squinting and can be measured with the obicularius muscle. I was hoping to talk about this at ATypI this year, but the proposal wasn’t accepted.

Alas due to some confusion in the submissions: Kevin’s name also appeared on another proposal, and it would have been unfair to give him two slots, or so the programme committee thought. As it turns out, while he was a contributor to the research in the second proposed presentation, he was not going to be presenting it personally. If the proposals had been otherwise identified, I suspect Kevin’s proposal would have been accepted. Unfortunately, we found out about the confusion too late to change things, as all the speakers had been notified.

Redge, most of the reading we do in film and video consists of reading short bursts of text, often just a few words, often in the context of moving background visuals that also require our attention. This is a different kind of reading from the sort of thing we have been discussing here. It probably deserves its own thread.

As John points out, one can’t, from responses to distractions alone, distinguish between (some kind of) immersion in the content of the text and immersion in one’s own thoughts. But testing for comprehension of the text as well should allow that distinction, if only so unreliably as to require a large sample for decent results. That doesn’t answer the question of whether immersive reading is that kind of reading that reduces response to distraction. This guess is informed only by my own experience, and since introspection into absorption must terminate that very absorption, we may need a phenomenologist on the case :-)

Incidentally, Peter, I have now read some Merleau-Ponty - The Visible and the Invisible - but am unable to judge at any point whether he’s right or wrong. I also don’t know who to blame for that.

John Hudson wrote:
“It probably deserves its own thread.”

Yes, it would be interesting to see that discussion. Maybe it will emerge from the Greek old style thread.

“This is a different kind of reading from the sort of thing we have been discussing here.”

Assuming that that is true, and I am not sure of that, maybe the role of context, technological and artistic, is analogous. In any event, the issue, as it relates from everything to film to television to video games and the internet, is important.

But maybe I’m in for an education. So far, in this thread, I have been told that I am obviously Anglo (half wrong), that this is an ancestral defect resulting in an “Anglo” sense of humour, apparently a problem (well, I will admit that Saroyan and Monty Python or, to take a more contemporary example, Little Britain, portray the world differently), and that War and Peace is a “kiddie’s novel” with “no artistic structure”, written for the “general reader” (I think that’s me). Accepting, for the moment, that the people who run and participate in this forum apparently think that attacks based on perceived ethnicity are acceptable, I am ready for anything. The conclusions that I draw are, of course, a different matter :)

Assuming that that is true, and I am not sure of that...

Actually, I’m not sure of it either. Perhaps I could have better expressed the thought as ’different from the kind of reading we are assuming the existence of here’. :)

Hrant insists on something that he calls immersive reading as being different from something he has called contemplative reading, and his criticism of the results of most reading science is that he doesn’t believe it has tested immersive reading. Now I’m quite prepared to believe that we read at different speeds and with different qualities of comprehension based on a whole range of factors — of which quality of the type design and typography is certainly one —, and that both speed and comprehension are likely to be good when we are immersed in a text in the way that I described above. I’m not convinced though that there is a categorical difference between immersed reading and non-immersed reading, i.e. that our eyes and our minds are doing something different in these two states, which is what Hrant seems to be suggesting. It is a notion that raises some awkward questions. Is there, for instance, an in-between state? Is the transition from non-immersive to immersive reading sudden or gradual? Is a categorical difference necessary to explain the quantitative variations, or can they be explained as the same activity being performed more or less well, more or less efficiently?

Perhaps Kevin could tell us how the idea of immersive reading sounds to psychologists given what they know about the workings of eye and brain?

I assume that optical corrections have to be treated as a system – you wouldn’t change the height without also changing the counter-width and letterspacing?
Of course it would be best to limit variables, and I think one can. Say one wants to lower the tops of the bowls of b and d slightly below the top of o, and raise the bottoms of the bowls of p and q slightly above the bottom of o: it would be the lesser evil not to simultaneously alter bpdq’s counter-width, because relative character widths are probably more important to a type than either counter-shape or relative vertical alignment, and because, since letter-spacing corresponds to counter-width more than to counter-area, there would be little damage to the spacing so long as counter and therefore character widths were not altered.

What size do you examine?
The size of the type is less important than the size of the x-height; I think an x-height of about 6.5 pt would be about right for maximum comfort in people with normal vision.

Will people read novels or textbooks?
Novels, I reckon, of the kind they read for easy pleasure: you might consider giving each person a passage by an author he’s enjoyed in the past but from a novel he’s not yet read. Perhaps that’s ridiculously messy, but I worry that giving everyone the same text would be a spurious standardisation unless you’re sure that their interests and reading experience in the particular area were also adequately standardised.

The size of the type is less important than the size of the x-height; I think an x-height of about 6.5 pt would be about right for maximum comfort in people with normal vision.

Yes, I would second that. Especially in comparative tests of different typefaces, x-height should be the constant. Then variables such as cap height, ascender height, descender depth, stem weight, etc. should be expressed as ratios to the x-height. Given the almost complete meaninglessness of nominal point sizes in determining visual size, I’m often frustrated that reading researchers generally reference point size, especially when comparing types. For Latin and other European scripts, the x-height provides the best constant, since it contributes so much to the visual size of the type in normal text.

Sorry, my mental arithmetic wobbled: a 5 or 5.25 pt x-height would be about right. With x-height as 4 units, ascender as 3 and descender as 2*, a 5 pt x-height gives a body size of 11.25 pt, a 5.25 pt x-height a body size of 11.8125 pt.

*These proportions are arguable, but in the right ballpark for extended reading.

“I’m interested in hearing about the kinds of studies that people are interested in seeing,”

I’ve tried to suggest in numerous ways, ways to broaden the selections for users. Studies like we are seeing are what again? Not because the methods, materials or subjects are wrong, but rather as I’ve said because: a. Reading is disturbed by testing, if the test itself is not a disturbance, and b. the readers, scripts, documents and environments are “to the horizon” in all directions in the “Windows” market. What can you prove, e.g. by testing something users don’t normally do, like looking at one letter? How can you test how well they read two different fonts in two different texts, since the same text twice is not a test, and the same font with two texts is useless... :)

Here’s a plan.

1. Software first! develop a screen font tuner/recorder.
a. works on all scripts.

b. allows rendering options for
i. ClassicType (aliased),
ii. CraftType (integer spaced g.s.)
iii. CrapType (sub-pixel positioned g.s.)
iv. ClearType (your baby)
v. CanType (none rendering, like bitmaps or outline straight to bitmaps)

c. the user should be able to select any installed font (though all fonts should show an “optimal renderer” icon).

d. the user should be able to select any point size in 1 point increments, and specify dow to tenths is the want (though all fonts should show an “optimal size” value).

e. one key stroke in any app. should activate the tuner and allow the user to change the font, the size and/or rendering of any highlighted and all similar text, or change all text in the cursor location.

2. record! the other part of the Software first!
a. use millions of subjects, with at least their ages as data.

b. using a full range of options on their own machines in their own environments, collect their choices over any length of time to see what they see as best for themselves, by themselves. (Choices you can gather this way, are: script, font, size, renderer, age, application, document (if public), length of text, time of day, time from last size change, time from last font change, and number of font changes since leaving default, e.g.)

c. Publish results in a database available to all users and designers.

d. eventually, release the tuner to all users as their own personal “font caddie” allowing them to save and tune all their favorite fonts and/or settings from application to application, document to document, device to device, etc., without the diversely cumbersome, while still limited options available today. Give it away for free.

3. Based on a study of this data, research particular market and client needs, and then, design fonts.

4. Support the new w3c proposal for linkable fonts.

Can one read immersively from a screen? Surely any test should be from printed material. I was wondering about the test conditions though, as I understand it, stress reduces the amount of parafoveal vision available.
Tim

Kevin, I wonder if the claim that immersive reading isn’t discussed among cognitive psychologists is accurate.

To my mind Hrant’s ’immersive’ reading might be characterized as ’passage’ reading of extended text by experienced, motivated readers, when the intentional object of the engagement is sense-following.

Many theorists, including Hrant and you and I, make all kinds of claims about the passage reading of extended text by experienced motivated readers. So if ’immersive’ reading can mean something like what I think it means in Hrant’s usage, I think it is what is most discussed.

But if by immersion we mean: level of engagement with the argument in a discourse, or level of immersion in the fictional world of a narrative, you and John are probably right in saying it is not much discussed or tracked in tests.

Hrant, if bouma reading is learned early, becomes natural, and is robust, wouldn’t it be the first line of attack in any task where the stimulus material consists of words, even where artificial or threshold conditions prevail, and artificial tasks are imposed?

What I would most like to see is a team effort at 1) disentangling (through clarification and the development of collectively devised* criterial tests) the issues surrounding divergant opinions, such as Hrant’s, Kevin’s and my own, about how reading works and 2) addressing (by involving type designers and typographers in the design** and execution of tests) the issues surrounding performance and usability testing as raised by, for example Ole Lund, and developing better ways to provide meaningful data.

* to be meaningful the architects of opposing models must agree they might be criterial, or at least are able to extend our knowledge.

** this might include design-involvement manipulating stimulus material and stimulus material paramets.

involving type designers and typographers in the design** and execution of test

Or scientists who are better informed about typography, so we don’t end up with naive generalizations such as “10/13 is the best size” masquerading as scientifically meaningful insight.

immersive reading isn’t discussed among cognitive psychologists

Perhaps because it can’t be measured? The term has significant meaning for typographers, but that’s not to say that what we understand by it has a physiological, as opposed to a cultural, reality.

William: “I don’t see why no advantage to children vs big advantage to adults isn’t also a measurable distinction.”

The problem with is that you can’t measure no advantage. Lots of studies fail to detect real differences because of poor methodologies or poorly conducted studies. Psychologists joke about publishing these studies in the Journal of Null Results (psychologists are not a funny bunch). Note that psychologists would call these two reading speeds the same: 20 people reading 50 pages in an hour with a range of 40 minutes to 80 minutes and 20 different people reading 50 pages in an 70 minutes with a range of 40 minutes to 100 minutes. That might be a large real difference, but if the variances are large, psychologists are hesitant to declare a difference.

Kevin

I agree with John’s points on immersive reading. It would be quite surprising to find that we have two different reading mechanisms. It’s certainly true that we will be more engaged in materials that we find interesting, and we will likely read those materials. But I’ve never seen any evidence that reading processes dramatically change, only speed up or slow down. In addition it interest in the content, there are many variables that can speed up or slow reading. For instance a sentence in active voice is read faster than a sentence in passive voice. High frequency words are recognized faster than low frequency words. Tinker & Patterson have shown that most typographic variables have an effect on reading speed, and that most recommended values (e.g. Bringhurst recommendations) result in faster reading speeds.

Peter, I don’t think that Hrant uses immersive reading to mean passage readers by experienced, motivated readers.

Kevin

Peter: “involving type designers and typographers in the design** and execution of test”

I couldn’t agree with this more. I spend a lot of time evangelizing to psychologists about the importance of typography in reading, and want to see a closer collaboration between two communities that should be natural partners. I’ll do whatever I can to help this to happen.

>Lots of studies fail to detect real differences because of poor methodologies or poorly conducted studies.

However, wouldn’t the (hypothesized) existence of a big difference for adults validate that the methodology does indeed show potential differences? And thus the lack of difference for the children would show something real? I don’t see why a null result should be presumed invalid when you have validated the method independently.

In any case, if you must have differences for the children, you could certainly increase the spacing to the point where it would negatively affect children learning to read. Then you could compare the spacing needed to slow down the children with that needed to slow down the expert reader significantly.

My point is, such a test of different levels of reading skill would be worth doing, if it is indeed indicative of some kind of direct mapping of multi-letter groups or words to meanings. To me the first question is whether the results, if they did show differences, would be indicative of an important cognitive process going on, even if it does require further research to pinpoint it. It seems to me that it would indicate something interesting, and so would be worth doing. What do you think?

PS. On null results. The Michelson-Moreley experiment was a null result and was extremely important in the history of physics. The experiment of Galileo dropping heavier and lighter balls off the leaning Tower of Pisa, and having them land at nearly the same time was also a null result. (That the actual story may be false is beside the point here.)

From the point of view of scientific method in other fields, this prejudice against null results is not well founded. If the experiment is well designed and executed, the result will be significant even with a null result, as Galileo and Michelson-Morley indicate.

I look forward to your response on the readability issue. I think the main problem with reading research from a point of view of typography is—from what I hear—is that at least as far as print goes it hasn’t had good measures to test for what typographers instinctively regard as important.

Does differences in the rate decline of comprehension with time seem to you like an useful and valid measure of reading comfort? If the answer is ’yes’, then it should be possible to compare readability and legibility (quick discrimination of letters and words) of various typefaces and layouts. And there would be hope of getting better detection of the differences typographers see. And of course if they are indeed real, which I do think, then this could help guide tyopgraphy for both screen and print readers.

“I’ll do whatever I can to help this to happen.”

Then respond;) why do experiments need to corner the market into a, or a pair of rendering solutions? Why can’t point size be an adjustment readers can make with ease? Why can’t we agree that there are many, many different ways people compose different kinds of documents, and many, many different ways people read those many documents and so, there need to be many ways to display the amny fonts?

Or, do we want to keep poking pin-hole-size experiments in a huge sky and calling composition and reading “understood”? What is the motivation? :)

David, your suggestions for gathering data from user settings seem to assume a direct correlation between user preference and best readability. What grounds are there for such an assumption?

Sure, on the one hand people will make selections that they are able to read, but the range of what people can read is as immense as your huge sky, and finding out what people can read, or even what they think they can read well, isn’t of any value — in the context of research on legibility/readability, which is what is being discussed here — unless you can also test how well they actually read it.

3. Based on a study of this data, research particular market and client needs, and then, design fonts.

Yes, what you describe is a big market research project for font developers. I’m sure it would be fascinating and useful for the type business, but it wouldn’t tell us anything about readability.

[Aside:

4. Support the new w3c proposal for linkable fonts.

There is no W3C proposal for linkable fonts. There is a proposal from some individuals, who do not represent the W3C, to support web typography by directly serving freeware TTFs from servers. I fail to see how this is a good idea or how it solves any of the issues that have prevented widespread decent web typography.]

“David, your suggestions for gathering data from user settings seem to assume a direct correlation between user preference and best readability. What grounds are there for such an assumption?”

If an environment where all the major options (renderer, size, font et al) were available and user preferences were recorded, then what other ground would be closer to reality? Is listening to the opinions of someone who “reads” at 145 dpi closer to reality? Is measuring the recognizability of individual letters closer? or is measuring two unequally prepared sample fonts closer to reality?

“and finding out what people can read, or even what they think they can read well, isn’t of any value”

You’ve got “grounds for such an assumption” :) I mean really, what people think they can read well is NOT what they read well? Hmmm.

“unless you can also test how well they actually read it.”

Which you can’t, and NO ONE has said they actually can, here or anywhere. Have they?

4. Support the new w3c proposal for linkable fonts even if you think it’s just a bunch of freeware fonts involved.

“I fail to see how this is a good idea or how it solves any of the issues that have prevented widespread decent web typography”

Read the threads John, we need the font environment on the web that has improvements in numerous typographic directions. One of them, is that we need font selection to proceed beyond what MS can afford to distribute for free without fuss or muss.

“Yes, what you describe is a big market research project for font developers.”

:) no, what I describe is a responsible response to what Kevin asked. What Kevin asked, is how best to test the best way to serve 330 million users of a wide range of hardware, fonts, documents and scripts for the 1,000 type designers who must try to serve them.

And that is what I gave him.

William, I certainly think that the hypothesis that there is a qualitative difference between novice readers and experienced readers is interesting and worth studying. I don’t expect to find a qualitative difference* - I think the experience that comes with many thousands of reading hours improves the efficiency of the reading process in place in novice readers. But scientific progress comes with testing differences between competing hypothesis.

* The first 20 or so words that children learn are recognized by arbitrary, non-letter cues. By novice readers I mean after this stage when children are making letter to sound correspondences.

My point about null findings is that it’s an exceptional case when a null finding is meaningful. Much like iterating on a type design, psychologists iterate on study design in order to find statistically reliable results. First studies (frequently called pilot studies) usually find a null result because it takes time to figure out how to control the variation. For a study to find a reliable difference, the statistics need to say we are confident to 95% probability that the difference was not caused by chance. Everything else is a null finding.

Kevin

William: “Does differences in the rate decline of comprehension with time seem to you like an useful and valid measure of reading comfort?”

It might work, it’s hard to predict. It would work well if the most important factor driving fatigue onset is the typographic variable being study. If typography is a partial cause of fatigue but interest in the content is the first order factor, there might be too much variance.

I think discovering these new methods is one of the key challenges facing the study of typographic variables. Everyone is interested in testing nebulous factors like invitingness, comfort, ease, and immersion without defining what these terms mean in a way that they can be measured. Most of the work that I’ve been doing over the last three years is centered around developing new measures for measuring these kinds of qualities. I’m hopeful that we’ll be able do develop several interesting measures that get at qualities that we call comfort and immersion. I talked about two of these measures at ATypI last year: one based on mood and cognitive performance, and another based on direct measurements of mental effort (brain imaging). The other area we’re investigating is discovering the muscles associated with eye fatigue.

I’ll also mention that traditional measures of reading speed and comprehension also have their place. I think the Times v. Gigi study discussed here is methodologically sound. It successfully demonstrates that people read Times faster than Gigi. While this may have little practical or theoretical surprise, it shows that this method was able to detect that something about Times allows people to read it faster.

Kevin

Kevin: “I spend a lot of time evangelizing to psychologists about the importance of typography in reading, and want to see a closer collaboration between two communities that should be natural partners. I’ll do whatever I can to help this to happen.”

David: “What Kevin asked, is how best to test the best way to serve 330 million users of a wide range of hardware, fonts, documents and scripts for the 1,000 type designers who must try to serve them.”

David, your laundry list of tasks for me to accomplish is not the idea of collaboration I had in mind.

Kevin

>scientific progress comes with testing differences between competing hypothesis.

I certainly agree—in accord with my late teacher Popper. The suggestion about novice vs expert readers is designed to test such a difference. I would expect that there would be a difference between novice and expert readers on how they react to disrupted spacing. This would be a prediction of the ’word pattern’ processing view, which Peter and Hrant have been advocating. I am thinking that a null difference would be predicted by the parallel view. Am I right?

I don’t understand what you wrote about novice vs expert readers—I think there is a typo in what you wrote. Could you explain?

I’m delighted that you are looking for good measures; good luck!

My suggestion about using fatigue (as decline in comprehension) as a measure is based on my own experience. To me one the the main things about screen reading as opposed to book is that I don’t want to read long passages. Also a letter sized paper filled with Helvetica or Arial and little leading gives me the same feeling—I have to plough through it with effort, even though individual words are perfectly legible.

The idea of eye strain as a measure sounds excellent. But I would expect that it might be more for issues like excessive contrast in the typeface, small size, or shiny paper. It might also be a factor in screen vs print, which would be important if you could identify what is the eye-strain causing factor in the screen.

However, I suspect that for an issue like Helvetica vs Times New Roman that the decline in comprehension will be more significant than eye strain or even reading speed.

The Gigi vs Times is sound, I agree. We have to have measures that at least discriminate between obvious cases. But new measures, other than raw speed of short passages, are I think needed to distinguish more subtle but important differences.

The issue of interest is no doubt extremely important to fatigue. However, I would think that it would be easy to control for. For example, recent SAT tests would be equally boring or interesting for students, I would think.

I don’t know that decline in comprehension is an important factor, of course. But my ’type aware’ experience as a reader tells me that it has a good chance of being an important measure.

Kevin, when testing print reading, could you make sure that subjects are given quarto books with real if glued spines rather than A4? The object should provide the right triggers for extended reading.

Kevin” “I am particularly interested in studies that would help people make practical decisions.”

I gave you a path to the most effective way to do what you claimed you wanted to do, “help people make practical decisions.” What more practical way is there than to sit them down at their own machines, in their own reading environments, with their own documents, and record their choices as they practice reading? Really, what?

John: “finding out what people can read, or even what they think they can read well, isn’t of any value”

What basis do you have for this statement, John. Answer the question.

Research should be aimed at what people read best. While reading at machines might be easier*, the variables might render any results useless, even if one had a large enough sample of readers who are willing to practice reading, the average wouldn’t necessarily produce the best. Besides which I don’t believe that immersive reading is best tested on a screen, I, for one, find reading on screen much more fatiguing than a book.

*for a tester
Tim

“Research should be aimed at what people read best.[...] the average wouldn’t necessarily produce the best.”

Who’s looking for average? I’m looking for Wide Choice. Do you understand? I don’t think any study, properly designed, vetted and performed, can prove that one rendering method fit all users, scripts, hardware, fonts and documents. I don’t even think two methods are enough.

“Besides which I don’t believe that immersive reading is best tested on a screen, I, for one, find reading on screen much more fatiguing than a book.”

You for one are right! I’ve pointed out that establishment of a reader’s aptitude for reading print should precede any study of their reading online. I’ve pointed out that immersive reading online is but a small part of the reading being done there, and a shrinking part at that. But we are talking about reading online. And I am talking about users deciding what to read in Their Documents, on Their Equipment, with a wide choice of options for fonts, sizes and rendering techniques.

I thought, we were trying to save trees here. If one is actually trying to save trees by getting more people to read online, send a good group of engineers off to make it possible for a whole world of users to find their comfort spot. Seems far more likely to truly accomplish the goal than making the user assume some one else’s chair, in some foreign space and... Actually, those are the nicest parts of the studies I’ve read so far, from there the choices users have are so limited it’s not really possible to imagine any sentient being recognizing them as useful.

What do they do when they get home, or back to the office?!

Apologies, I didn’t realise that this thread was about reading on screen, I’m not aware of environmental impact studies on the benefits of reading screen as opposed to print I would have thought that each had their merits in terms of power used to create, use, recycle, sustainability etc.
I don’t want to be Big Brotherish but are readers capable of choosing for themselves or would they benefit from being offered a selection of rendering options/fonts/layouts, this might seem like not appreciating that readers are capable of making a choice but default settings will probably make up a significant part of the results, furthermore comprehension has got to be a part of the test surely?
Tim

David,

Consider that in testing environments, the questions being asked are not usually as sweeping as those you pose. Typically the questions are more manageable, due to the difficulty in establishing definitions, parameters, and controls. “these things must be done delicately”. Also, consider that Kevin has offered his assistance in this area, where no one else is even aware of our presence. Don’t beat him up because he doesn’t “get” your monumental proposals. You do have a very cryptic and glib writing style, even as I can see you are asking very incisive questions. Call him on the phone and discuss. Kevin offers us a terrific opportunity to communicate meaningful ideas about typography and reading to those who are poised to do research about them.

Conversely, I do suggest that Kevin and other researchers really try to hear and interpret questions such as David is asking, in order to design tests and projects that try earnestly and intelligently to answer them. I would prefer that testing make greater reaches with less predictable results, because, as it should be apparent, few of us feel we have the time to individually test micro elements of reading and type use, if those tests are not coordinated to fit into a larger inquiry. When new tests come out, I often feel they are not intended to build on, or add anything to, existing studies, and I suspect it’s because those conducting the tests are not focused on related questions. If a larger, overarching goal (such as David proposes) were used to structure individual tests (which researchers can reasonably hope to perform), I suspect we’d all be more intrigued and satisfied with the results.

>I didn’t realise that this thread was about reading on screen

There has been discussion of both screen and print on the thread, and the initial posting doesn’t specify which medium.

John: ’“finding out what people can read, or even what they think they can read well, isn’t of any value’

David: ’What basis do you have for this statement, John. Answer the question.’

Almost all of the physiological and cognitive activities involved in reading are unconscious acts. We perform them without conscious experience of the performance: we are not conscious of our saccadic eye movement, we are not conscious of the mechanism by which we recognise words, we are not conscious of how our brains put it all together. Instead, we have an experience of reading that is an involvement with the text (which is well and good). At best, we can extrapolate from our reading experience to a hypothesis about the quality of the reading performance, but this is really guesswork. If we enjoyed reading something, and it seemed to be quite effortless and to pass quickly, we can hypothesise that we were reading well, i.e. with a good correlation of speed and comprehension (two measurable and testable criteria). But what we have is an hypothesis, not scientific knowledge, and when you have an hypothesis you should try to find a way of testing it. So this is why I say that finding out what people think they read well isn’t of any value in determining what they actually read well, because only measurement of speed and comprehension can offer a reliable indicator of the latter. And we know that our conscious experience of reading is misleading about the mechanism of reading, because for centuries before Laval discovered saccades by measuring eye movement people assumed, based on the experience of reading, that the eye flowed along the line, and most people who have never heard of saccades still believe something like this because it is in accord with their conscious experience of reading. Just as we cannot accurately determine the mechanism of reading from our experience of reading, but must measure eye movement and consruct tests to try to determine how we recognise words, so I believe we cannot accurately determine what we read well except by similar efforts at measurement. Otherwise we’re just guessing and making assumptions.

John, John. “Almost all of the physiological and cognitive activities involved in reading are unconscious acts. We perform them without conscious experience of the performance: we are not conscious of our saccadic eye movement. . . (and etc.).

And? what on earth makes you think you can take people into a lab, give them limited and incomplete options, know nothing about their native environment, background or reading skills, and get a more accurate (not be confused with korrect), result versus the comfort and security of their own computer with hundreds of choices, and...unlimited time to make those choices, aka, real life?

“Otherwise we’re just guessing and making assumptions.” :) I know! Your argument is that you can, by creating the perfect study, (no free will involved), determine precisely what all 330 million users will want? What is the basis for this assumption?

Someone went to a mall and asked people to make a choice between aliased and anti-aliased ITALIC screen type, according to your rock hard opinion you’d promptly toss out the result, right? Not here, this CT-Only “hypothesis” was launched on the strength of just such a study. Did you support that? Did you believe that? Did you speak according to your beliefs then? John, you’re Hip, you draw good “o”s and you probably can crit with the best of them, but that does not mean you need to be a hypocrite.

I make text faces :) I make them good by listening, looking, and reading. I have a larger audience of immersive readers than you do, (maybe, more than even mighty MS :). In the name of oxygen production I’m begging you to destroy my market. But I’m sure, if you do destroy the whole world’s screen fonts options first, it’ll sadly ensure we have the time and income here to fix it.

And support the linking of fonts the w3c is asking about. :)

I recently tested a number of typefaces for legibility for use in a signage system for a transit agency. The method I used was to digitaly blur layouts of typical signs then get people to try and read the signs and evaluate their ablity to distiguish individual letters and (more subjectively) word shapes with a simple scoring system.

Basically, what I found concures with the list at the start of the thread, except for the extended width, which becomes a practical issue in signage, where you may have to sacrifice letter height to accomodate added length. The best typeface among the ones tested was TransitFont which is a little on the compact side. At the same letter height TransitFont was slightly better than APHont and Tiresias which were both designed specifically to be highly legible by the American Printing House for the Blind and the Royal National Institute For the Blind in Briton.

Russell: Road / performance testing, or testing of affordances under threshold conditions (low foreground / background contrast, blur, generation loss, extremes of size, tight durational constraints, unfamiliar location names in a non-native language) for specific contexts of use, stimulus sets and formatting parameters is a very useful exercise, especially when the studies are comparative. The question is: are results of these task- or case-, condition-, and parameter-specific tests directly generalizable to all conditions of typographic use such that universally applicable, qualitatively specific and quantitatively precise, categorical statements (about serifs, the relative sizes of ascenders and descenders, interletter and interline spacing, monocular versus binocular forms), valid across all conditions and tasks, can be made.

Russell, are you aware of the testing for Clearview Highway font? This seems to be the most extensive testing on legibility for highway signs.

“It would be quite surprising to find that we have two different reading mechanisms.”

Kevin wouldn’t a perceptual psychologist be inclined to think that the deliberative laborious, perhaps serial, decipherment or ’word reconstruction’ based on a ’synthetic combination of letter elements’ (’cues corresponding to sounds’) typical of begining readers, might entail a different perceptual processing mechanism at the neurological receptive field behavioural level than does the rapid automatic probability summation of an entire packet of stimulus elements whose aggregate identity the visual system has learned?

At the very least the eye-movement statistics must be radically different.

Neurobiologists are now begining to chart the changing roles of the different cortical regions involved in reading as a result of learning.

Adaptive systems theorists like Steven Grossberg, stress “how perilous it is to directly translate the distinctions of lay language into the definition of an underlying psychological [or neurobiological] process [like ’unitization’ (of component structures into a gestalt)]” (my additions in square brackets)

David: ’Your argument is that you can, by creating the perfect study, (no free will involved), determine precisely what all 330 million users will want?’

No, that is very much not my argument. The point I am making is that what 330 million users want isn’t necessarily what they will read best in terms of speed and accuracy. My point is that what people want shouldn’t be assumed to be equivalent to what they read best. My point is that people do not know what they read best, any more than they know how they read.

I’ve heard people raving about how much they love the type rendering in Mac Quartz, because they like the way it looks. This is the sort of basis on which people make preference choices about text on screen. People don’t sit down and time their reading in different rendering systems. They don’t have friends test their retension and comprehension after they finish reading something.

I happen to think that the experience of reading, understood as something distinct from the mechanics of reading, is very important, and making choices about what fonts to use and what rendering systems we prefer based on that experience is totally valid. What I’m saying is that if one is interested in understanding the mechanics of reading and being able to compare performance in reading, then the kind of preference choices that people make isn’t a reliable indicator of anything.

A metaphor: there are lots of foods that I like to eat. I know that I like to eat them because I enjoy the experience of tasting them. The fact that I like to eat them doesn’t tell me anything about their nutritional value. It doesn’t tell me anything about how my body deals with the chemical constituents of the food. It doesn’t tell me how much energy I get from the food, or how quickly this energy is released into my body, or how much of the food is going to be expelled as waste, or how much of it is going to clog my arteries. I am entirely dependent on information sources external to my experience of eating these foods for all knowledge about them other than how they taste. I’m reliant on the sciences of biology and nutrition to tell me these things. I certainly can’t assume that because I like to eat these foods they are good for me.

Now, I’m not the sort of person who is going to base all my choices on what I eat based on what science tells me is good for me, but nor do I think that my taste provides sufficient information about what is good for me. So it is with reading: at the end of the day, I am going to read what I want to read, and I hope everyone else will too, but I still want to know what science can tell me about how I read and, as a type designer, whether that science offers any insights that might be applicable to my job.

David: Someone went to a mall and asked people to make a choice between aliased and anti-aliased ITALIC screen type, according to your rock hard opinion you’d promptly toss out the result, right?

What was the point of the choice? If it was to simply determine which people preferred the look of, then I wouldn’t toss out the results, because a subjective preference can be determined by such a choice. But if the question put to the mall-strollers were ’Which of these do you think you would read most quickly and accurately?’ then you can be damn sure that I would toss out the results. Because reading speed and accuracy are not subjective preferences, they are measurable performance indicators. Indeed, not only are they measurable, they can only be determined by measuring.

Not here, this CT-Only “hypothesis” was launched on the strength of just such a study.

I don’t have a CT-only hypothesis.

Which study are you talking about? Yes, there have been tests of preference, but there have also been tests of performance. The latter tests are the interesting ones.

Have there been enough tests and the right kind of tests to make the kind of decisions MS is making about default rendering in the OS based on reading performance results? (as opposed basing them on engineering efficiency, maintenance cost, etc.) No, I don’t think there have been.

On the other hand, did Apple conduct any readability tests before deciding that full-fuzz rendering was going to be default in OS X, thereby ’destroying the screen font options’ of their users? Not to my knowledge. Has Adobe conducted any readability study of their Acrobat CoolType rendering? Or did they simply decide that it looked nice and smooth and left it at that?

First, I am aware of the ClearviewHwy testing.

It all started when I hired someone to critique our signage and he had some very severe comments about our use of Helvetica. As a commuter rail service most of our passengers are accessing the system before sun-up and after sun set for much of the year, and in all sorts of weather. And always before they have had enough coffee. :0) However, most passengers only need to figure out where they are going once, all be it, often while hurrying to catch a train in the dark before they are fully awake.

We had some pretty restrictive budget constraints, so we had to figure out how to come to a relatively objective decision in house. the testing it’s self was very rudimentary and relied heavily on existing research.

I did request supporting data and discussed my methodology with the designers of all the typefaces we tested - (except CleaviewHyw, simply because there is so much data readily available about their research.)

Sorry David,

I have been ignoring you because I find some of your statements about how to conduct research annoying. I’ve been thinking about how to move forward productively.

Clearly we have different interests: I’m interested in understanding how the reading process works with the intention of using that information to improve future reading environments. I think you are less interested in understanding the reading process and want to know about user preference among all available environments.

I find user preference to be a less compelling measure of the reading experience because I have seen enough instances where there is a mismatch between what users say and their actual behavior.
http://www.upassoc.org/upa_publications/upa_voice/volumes/2005/april/lis...

On the other hand, performance measures such as reading speed have been successfully used by psychologists for over a century. The results from such studies are very reproducible. 1) If two researchers carry out the same study, they almost always get the same results. 2) Individual performance is quite consistent – reader’s performance is the same in the lab as it is in “natural environments.” 3.) In reading, there is remarkably little variation across proficient readers – for example everyone shows the word frequency effect. This must be because the basic reading processes are the same for all proficient readers.

I think if your goal is to understand what choice a user will make, then preference is a fine measure. Here are two ways of measuring preference that are better than asking if you like A better than B: The first is using Likert scales. For each option give users a 5 or 7 point scale where 1 is strongly disagree and 7 is strongly agree and ask if they like condition A. The advantage of this method is that you can detect small versus large preference differences. A second method is newer method of measuring desirability based on understanding the positive and negative associations for a condition.
http://www.microsoft.com/usability/UEPostings/DesirabilityToolkit.doc

Btw, we have used Likert scales in many of our studies, and I’ve also used the desirability toolkit.

Cheers, Kevin

William: “I don’t understand what you wrote about novice vs expert readers—I think there is a typo in what you wrote. Could you explain?”

Peter: “At the very least the eye-movement statistics must be radically different.”

My above comments on novice versus expert readers are incoherent, sorry. I eluded to a study that showed that children learn a handful of words through arbitrary cues – such as the golden arches being recognized as the word McDonalds. By the time that children have learned 100 words they are using letter to sound correspondences to recognize words. I think the reading mechanism that these novice readers are using are in principle the same as adult readers. I would expect the same basic eye movements – that they make saccades that land in the middle of words - and the same areas of the brain to be activated. The process of becoming a proficient reader through thousands of hours of practice will make all the involved processes more efficient. There will be many fewer saccades and the cortical pathways will have been trimmed of unnecessary connections.

I would be very surprised to find that children’s eyes make smooth movements over a line of text or that parts of the brain that adults don’t use for reading (say the frontal lobe) are active.

There are clearly differences between novice and expert readers, but I think it’s a matter of efficiency, not a different set of processes.

Cheers, Kevin

Stepping sideways on the subject of reading, I discovered this book in a secondhand shop a couple of days ago and bought it: The ethnography of reading, edited by Jonthan Boyanin. I have only read parts of the introduction so far, but it looks like an interesting book and a timely reminder of the varieties of ways in which people engage with texts.

John, could you provide a synopsis when you are through or at least a list of contents?

[Kevin] “I would be very surprised to find [...] that parts of the brain that adults don’t use for reading (say the frontal lobe) are active.”

In a 2004 study “The Neurobiological Basis of Skilled and Impaired Reading: Recent Findings and New Directions,” Rebecca Sandak, Einar Mencl, Stephen J. Frost, and Kenneth R. Pugh identify 3 cortical systems or circuits involved in reading and reading aquisition: an anterior region, the temporo-parietal (dorsal) region and the occipito-temporal (ventral) region. The dorsal and anterior systems “predominate during initial reading acquisition in normally developing children,” with an increased ventral [visual word form area] response “as proficiency in word recognition increases. The switch seems to occur at about 10 1/2 years. Sandak et. al., speculate “that a begining reader on a successful trajectory uses a widely distributed cortical system for print processing including temporoparietal, frontal, and right hemisphere posterior areas. As reading skill increases, these regions play a diminished role, whereas left hemisphere ventral sites become more active and presumably, more central to the recognition of printed (word) stimuli.”

Pugh et. al., have little to say about activity in the visual cortex.
However at least two groups of neuroscientists are finding evidence of “top-down reorganization of activity in the visual pathway as a consequence of training” or evidence that perceptual learning “improves efficiency by retuning the decision ’template.’”

To my mind the central mystery in reading acquisition is encapsulated in this question articulated by Stephen Grossberg: ” How does a not-yet-established [pe: visual] word representation overcome the salience of well-established letter representations? How does unitization of unfamiliar lists of familiar items ever get of the ground?”

I have some ideas about this of course, but might suggest that rhythmic or typographic spacing, following the kind of protocols Noordzij recommends, catalyzes the retuning.

Very nice Peter!

Are you surprised yet Kevin?

Again about the ’null result issue.’ I asked my friend who has headed a cognitive science department, and she confirmed the skepticism in the field about null results. However, she also said that people are persuadable, as in the case of the historical examples I gave, that you can have significant null results.

In any case, I think the right way to look at the proficient vs relative novice is that a null result would be no difference between proficient and novice NOT that there would be no effect on the novice. As I said you could increase spacing to the point where the word and word space were confused, which would get an effect on the novice, but that I think is artificial.

The real question is whether the possible difference between novice readers at various stages and proficient readers is enough of a possibility to be worth testing vs the usual assumptions. I think the evidence Peter cites increases the argument for it being worthwhile.

William, I should mention 1) that the ’retuning’ and ’re-organizing’ researchers don’t use words as the basis for their demonstrations and data collection, and 2) the cortical circuitry analysts tend to think of ’visual word form’ in ’abstract orthographic shape’ rather than visual configural terms. So these researchers (or Kevin, for that matter) might not draw from their results the same provisional conclusions that I make to generate my hypotheses.

But stripping the interpretation of one set of results from the biases of the authors that report them, and extending into an adjacent domain the observations of a second group, makes an alternate point of view (with conventional typographical craft-wisdom resonances) plausible, if not attractive.

“Are you surprised yet Kevin?”

I am surprised to see an increase in activity in the occipito-temporal (ventral) region. I am not surprised that there are areas that become less involved with experience. I am also surprised that there are sudden changes. I would expect changes to happen gradually over time, not at a particular marker point. I’ll have to go read those papers. I have not been following the neuroimaging work as closely as Peter.

I suspect that Grossberg’s question was rhetorical. He is known for his work on self organizing neural networks – essentially taking undifferentiated input data and making sense of the data without being told what the output data should look like.

Cheers, Kevin

John, could you provide a synopsis when you are through or at least a list of contents?

Here is the list of articles in the Ethnography of reading book:

Boyarin, Daniel. ’Placing reading: ancient Israel and medieval Europe.’

Noakes, Susan. ’Gracious words: Luke’s Jesus and the reading of sacred poetry at the beginning of the christian era.’

Howe, Nicholas. ’The cultural construction of reading in Anglo-Saxon England.’

Fabian, Johannes. ’Keep listening: ethnography and reading.’

Baker, James N. ’The presence of the name: reading scripture in an Indonesian village.’

Digges, Diana & Rappaport, Joanne. ’Literacy, orality, and ritual practice in highland Colombia.’

Horton, H. Mark. ’Japanese spirit and Chinese learning: scribes and storytellers in pre-modern Japan.’

Long, Elizabeth. ’Textual interpretation as collective action.’

Boyarin, Jonathan. ’Voices around the text: the ethnography of reading at Mesivta Tifereth Jerusalem.’

Sarris, Greg. ’Keeping slug woman alive: the challenge of reading in a reservation classroom.’

John, thanks! This all sounds very interesting. If you tire of it, send it over.

Kevin, the Grossberg question is from a 1984 paper “Unitization, Automaticity, Temporal Order, and Word Recognition.” My interest in his work is aroused by the fact that he addresses neural dynamics of perceptual grouping and emergent segmentations. My sense is that most models of the perceptual processing component in reading make assumptions about perceptual grouping which might not be sustainable. Stanislas Dehaene, Laurent Cohen amd Mariano Sigman say (2005), “although many connectionist models of reading have been proposed, most have focussed on the interactions between orthography, phonology and semantics, ignoring the ’front end’ of visual word recognition ”. [my emphasis]

If our motivation in studying reading is to find reasons to believe that what we are inclined to think of as important in type design and typography is in fact important, than this is where the rubber hits the road, or so I think.

According to my scheme, once the process of accurate phonological mapping to orthographic sequences conjoined to meaning has stabalized, or been mastered, a perceptual remapping based on “encapsulating linguistic knowledge directly in highly efficient pattern-recognizing templates” (Pugh, Menel, Shawitz, et. al., 2001) can begin to take hold. The remapping happens, so it seems to me, at a more ’primitive’ level than the letter level.

This would be my interpretation of what must change for specificity for words in the ventral “visual wordform area” to emerge.

John: “But what we have is an hypothesis, not scientific knowledge, and when you have an hypothesis you should try to find a way of testing it.”

Please state the hypothesis.

“what people want shouldn’t be assumed to be equivalent to what they read best.” Oh? Please rephrase.

”... Mac Quartz,” Forget about Apple and Adobe. With all due respect, they are not significant to this issue. I will tell them they blew it when I get a chance.

“I don’t have a CT-only hypothesis.” :) someone does — it must be called the CT Tuner for a reason.

“if one is interested in understanding the mechanics of reading and being able to compare performance in reading, then the kind of preference choices that people make isn’t a reliable indicator of anything.”
Yes John, but “understanding the mechanics of reading”, and “implementing a ww reading Environment” can quite neatly be divided into two different things. Can’t they? One is an interesting quest that I’m proud to be witnessing and very grateful to learn about, and the other is somethin’ else, right?

“A metaphor: there are lots of foods that I like to eat... I’m reliant on the sciences of biology and nutrition to tell me these things.” First, YOU are not the issue: Rich, Educated, 148dpi, What’s the point? I don’t relish skewering this metaphor, but: A. digestion is a well understood process, B. 20th c. Science brought us the “food pyramid” before 21st c. genetics is proving a variety of pyramids are required, C. Despite the fact that Kevin doesn’t think so, Babies and Old Folk require special food, D. the human species didn’t need any nutritional science at all until the rise of the steamship, and didn’t need it this badly until the implementation of ww cargo container shipping.

If you understand my response, then you understand that the requirement of today’s ww base of readers is a genetically correct mother or granny, her kitchen, recipes and ingredients all installed in the home as opposed to any two meal options you can ever test and provide in the lab, even on a couple of million people. Get it?

John and the The Mall Test: “If it was to simply determine which people preferred the look of, then I wouldn’t toss out the results...” my point has been steadily focused on the fact that testing the “look of” something that is meant Only to be “read”, and then using the results to bolster supposed “Reading Capabilities”, as has been done in several tests, is... annoying.

Kevin: “I find user preference to be a less compelling measure of the reading experience because I have seen enough instances where there is a mismatch between what users say and their actual behavior.”

Kevin, I’m not interested in what they say either. My “laundry list” solves that by recording what they do (actual behavior)... for as long as it takes (actual time)...for as many kinds of type as they become used to using (actual use). (You could even feed your existing studies right into it, and see if you can refine your understanding of your own tests.

“your statements about how to conduct research [are] annoying”
Flies are annoying too;) I’m not trying to be annoying Kevin. I am interested in what they’ll buy. What you are telling me is they will only have a choice of OS bs (good bs, but bs nonetheless), so I’m here, attracted to the scent.

David: Please state the hypothesis.

I did, in the same message from which you extracted my quote, in the sentence before the one you quoted:

At best, we can extrapolate from our reading experience to a hypothesis about the quality of the reading performance, but this is really guesswork. If we enjoyed reading something, and it seemed to be quite effortless and to pass quickly, we can hypothesise that we were reading well, i.e. with a good correlation of speed and comprehension (two measurable and testable criteria).

The hypothesis is not mine, it is the one you seem to be making: that there is a correlation between the fonts, and text size and rendering options that people choose to use and what they read best. It is a very reasonable hypothesis, but it is still a hypothesis, and in order to be confirmed it needs to be tested against something other than your very good ’laundry list’.

If the vast majority of users selected font X at size a using rendering option 1, that would be a great indicator of what they would buy, and of their preference, but they might still read font Y at size b using rendering option 2 more quickly, with greater accuracy, with greater comprehension and retention of information, and with less fatigue.

As you say, you are interested in what people will buy, but this was a thread about legibility and readability research, so the question of what people actually read best, and whether this might differ from what they prefer, is surely relevant to the discussion. It might not, ultimately, be relevant to the fonts that we make, especially if there is a difference between preference and performance, but I’d still like to know.

Thanks John. I would not state it that way though. . .

The quality of the reading experience is dependant on a large number of diverse variables from lighting, text matter and weather, to a readers frame of mind. It has not ever been quantified over a million-strong sample of users, as a science, and even under the most tightly controlled study, saying one true thing about all one million users would be impossible. In the area of world wide screen reading, where much larger populations exist, where users make their own environments, and under the pressure of diverse technologies and document types, science is useless. But, in the midst of these variables are typographic variables including Em interpretation, script, size, style and rendering. As long as the management of these and the other typographic variables is performed with traditional typographic care, fonts can be made forever in a straightforward relationship between the designer of the font and the user. But if the management of the typographic variables is not done with traditional typographic care, e.g. if the em interpretation is false, or unequal treatment and/or rendering of x and y dimensions occurs, the straightforward relationship between the designer and user is broken. Then, only with long observation of the new options, ideally added to the old options, and under the personally selected environmental circumstances of many users, can the type designer begin to resolve user group issues. And, these issues will continue to surface and evolve over the time it takes for the newly mismanaged typographic variables to become a new tradition.

“Yes John, but “understanding the mechanics of reading”, and “implementing a ww reading Environment” can quite neatly be divided into two different things. Can’t they?” Can’t they?

“testing the “look of” something that is meant Only to be “read”, and then using the results to bolster supposed “Reading Capabilities”, as has been done in several tests,...” Right? Right?

“you are interested in what people will buy, but this was a thread about legibility and readability research, so the question of what people actually read best, and whether this might differ from what they prefer, is surely relevant to the discussion.”

Yeah, but what people actually “read best”, and whether this might “differ” from what they “prefer” is (in the “new tradition”), a hostage of the options, as is the research itself, as I’ve seen it.

David: “even under the most tightly controlled study, saying one true thing about all one million users would be impossible”

Saying something that is true about one billion users is exactly my goal. Reading is not a process that varies greatly from individual to individual. Like digestion or walking, reading happens in about the same way for everybody on the planet. The brain is similar to other organs in that it is essentially the same in every person. I will quickly prove your above statement wrong by saying that EVERY reader can read high frequency words faster than low frequency words.

It is my contention that that if we can learn more about the reading process that typographers will be able to use this information to improve the reading experience for every reader.

David: “Despite the fact that Kevin doesn’t think so, Babies and Old Folk require special food”

While I do think the reading process is essentially the same (you don’t have a separate digestive track or another set of legs as a child), it’s a silly statement to say that I don’t recognize life stages. My Ph.D. dissertation was on reading acquisition. I developed tools to make it easier for children to learn to read.

Cheers, Kevin

“Saying something that is true about one billion users is exactly my goal.”

” EVERY reader can read high frequency words faster than low frequency words.”

Nice: You’ll confirm to me when you’ve proved this on every one of any number of 7 year-old Chinese readers then? But at this point, the only thing you can say about a billion users is that they lack an OS competition to profit from in their reading options. Wonder how that happened, oh yeah, now I remember.

“(you don’t have a separate digestive track or another set of legs as a child)”

Not separate, but not equal. Are you not aware of breast milk? You are? How about PKU? and the adult carrying a child thing is another set of legs. I guess in your world: legs are all legs, eyes are all eyes and fonts are just “a” font. Should I go on to old age where, the digestive system becomes unable to handle garlic or gin and the legs become wheels?

I’m afraid if you ever really do recognize life stages, differences in taste or the turn of the 21st century, I’ll most likely be waiting to welcome you thru pearly gate #72.289. If not, I guess I’ll be sitting between Norton and Goudy looking down on a group roast. ;)

“It is my contention that if we can learn more about the reading process that typographers will be able to use this information to improve the reading experience for every reader.”

Can you give us an example of a study you’ve run that changed the way you make type?

John, Kevin: you still have some questions I would like to know the answers to:

“Yes John, but “understanding the mechanics of reading”, and “implementing a ww reading Environment” can quite neatly be divided into two different things. Can’t they? I mean, even if it could be proven that there was one unifying theory of reading for all the world, then would all the people in the world need one perfect font (per script), one size, one rendering technology? I think you’ve gotten onto a very partisan track here that is pretty clearly dictated by a previous investment in lack of knowledge.

“testing the “look of” something that is meant Only to be “read”, and then using the results to bolster supposed “Reading Capabilities”, as has been done in several tests,…” is this proper scientific process?

David, Kevin’s statement that there are universal scientific laws concerning reading is not inconsistent with the need for variation, both for individuals and for different situations, as you emphasize. You are just looking at different aspects of the problem.

To make an analogy, there is one law of gravity for everyone, but different people deal with gravity differently. Most use two legs to get around inspite of gravity pulling them down. But babies crawl on all fours and some whose legs are hurt need canes or wheels.

Furthermore, there are different shoes good for different situations, and looks are important also, though all shoes need to protect against sharp things brusing or cutting us when our weight (gravity) presses against them.

Knowing what those laws of reading are—and we only have a glimpse of them now, it seems to me from what I hear—will help greatly. Dealing with variation will still be a challenge, but knowing more of the laws should help a lot.

Oh, and you don’t have to test a purported law for everybody. After repeatedly getting no refutations in spite of looking for conditions that have the best shot at refuting the purported law, it is just accepted. You don’t have to check whether the law of gravity applies to all Tibetan childen. It does.

Yes John, but “understanding the mechanics of reading”, and “implementing a ww reading Environment” can quite neatly be divided into two different things.

Regardless of whether they can be neatly divided, I thought this was a thread about the former, ’research about legibility’, not a thread about screen reading environments, and that is the basis for all my comments. If these things can be neatly divided into two different things, why do you want to talk about the latter in the middle of a conversation about the former? :)

Some of Kevin’s work is specifically involved with ClearType and some of that work is specifically involved with readability. He’s also involved in studies on readability that are not specifically related to ClearType (what I thought this thread was about) and work that is not related to readability but to preference and ’personality’ in type selection. Since I suspect Microsoft decisions about their default rendering system are primarily based on engineering decisions, I’m actually rather more interested in Kevin’s general readability research than anything ClearType specific.

David: “You’ll confirm to me when you’ve proved this on every one of any number of 7 year-old Chinese readers then?”

This paper has the data you’re looking for:
Chiang, Wen-Chi (2003). The Basic Orthographic Units in Children’s Recognition of Chinese Characters, Chinese Journal of Psychology. 45(1), 47-74.

David: “I guess I’ll be sitting between Norton and Goudy looking down on a group roast.”

That is the nicest possible call to Beelzebub I’ve ever received. I’m only sorry to hear that you’re already planning your trip to heaven. You must not look your age.

David: “Can you give us an example of a study you’ve run that changed the way you make type?”

I’m hopeful that my paper on word recognition has convinced some people that we do not recognize words as whole shape patterns. That alone should have some impact on how a typeface is designed. It would at least argue for not spacing letters as tightly as possible. I’m also hopeful that I still have more to contribute.

David: “if it could be proven that there was one unifying theory of reading for all the world, then would all the people in the world need one perfect font (per script), one size, one rendering technology?”

I’m not of this opinion at all. While I think that there are features that can make letters more or less legible, I also think that typefaces can convey additional information through their personality that contributes to communication. This is also an interesting area to study.

Cheers, Kevin

Kevin “Chiang, Wen-Chi” Thanks, I’ll check it out, but the thing I’m saying about readability research considering the context and purported purpose of your work, is to agree that a far more important thing you can say about all users is that they read “where they read” more, and more naturally, than they read “where and when you tell them to read.” Research anyone?

“[Kevin’s research] has convinced some people that we do not recognize words as whole shape patterns.” As I’ve always said, with the exception of “I”, “a”, Chinese, and more, I’ve never disputed this and I consider the entire issue a puff of fog intended, or not, to part us from the consistent letter spacing of print, and the integer spacing of screen fonts. An option I think users should have in conjunction with some form of anti-aliased letterform.

“That alone should have some impact on how a typeface is designed.”
Great impact. But I...am looking for improvements.

“It would at least argue for not spacing letters as tightly as possible.”
With all due respect, No text face, assuming we’re still in readability, has ever succeeded by being spaced as tightly as possible, or approaching such a state of spacing. Only Beverly Readahillbillies would need research into this. :)

John: “this was a thread about the former, ‘research about legibility’, not a thread about screen reading environments”
I....tink...de...r...elated. I get the feeling Mr. 145 dpi just don’t want to connect the dots.

“Microsoft decisions about their default rendering system are primarily based on engineering decisions, I’m actually rather more interested in Kevin’s general readability research than anything ClearType specific.”

So why’s it seem so hard for you to accept my point that all this research as pure marketing fog after the fact. With the fact being that engineering decisions are made, and then the psych and type departments react. In essence typographic “Fire, Aim, Ready” at the ultimate cost of a proper set of readability options for the user? And, this is not the first “Fire, Aim, Ready” we’ve gotten from your patient, Dr. John. ;)

I….tink…de…r…elated.

So I respond to a post in which you say that they ’can quite neatly be divided into two different things’, and then you come back at me insisting that they are related.

So why’s it seem so hard for you to accept my point that all this research as pure marketing fog after the fact.

Who said it is hard for me to accept that point? I don’t entirely agree with it, but yes Microsoft does employ the research in a marketing way, not least in internal marketing to persuade other groups within Microsoft to pay attention to typography (not a bad thing, that). I’m not actually that interested in how MS uses that research, since I’m going to review the published papers myself and draw my own conclusions, which is the nice thing about having the research done by academics and published in peer reviewed journals: you don’t have to take Microsoft’s interpretation of the results if you suspect them of being ’spun’ for marketing purposes. Some of the studies I think are pretty silly — the ’personality association of typefaces’ study, for example, which seems to me to beg the question — and I can sometimes spot things in experiment design that make me question whether the results would be different if the experiment had taken some additional factors into account. As far as I recall, I have only ever cited one study that related directly to ClearType.

As I wrote above, I’m primarily interested research that tells us something about how we read, because I’d really like to know, and for the most part this is not the sort of research Kevin his doing. I’m not actually seeking research that will tell me how to design a typeface or what screen rendering is best. Very little of the information I’ve read about reading seems practically applicable to type design in the sense of doing something different from what I’m already doing. Sometimes I read something that gives me an idea or two to explore in a new typeface, but I think that’s about the limit of the practical application.

“Sometimes I read something that gives me an idea or two to explore in a new typeface [...]

John, what would be an example of this?

The ideas are often tangential to the research, which acts more as an inspiration for a line of thought than suggesting a specific concept to be applied in the design. The most significant bit of readability research for me is that which relates to the importance of high contrast typeforms with consistent stroke density, which is a predictable result when one thinks about reading in terms of ink on paper and just what makes that such a superior reading experience, but which some people don’t seem to want to admit to having implications for greyscale screen rendering. But it also has implications in type design for print, and I’m a lot more conscious now of designing details of type, especially things lik connections between stems and bowls, in terms not merely of avoiding unwanted heaviness but of achieving a specific level of overall density. I think a lot of type design has historically been aimed at avoiding things looking too dark or too light, i.e. a kind of negative, usually iterative approach, rather than trying to determine positively what is ’just right’ and analysing shapes and their relationships in terms of nailing that. So there is an example of an idea coming from research and influencing not simply a type design but a general approach to type design.

Related to this are some ideas I have for small type, especially in non-Latin scripts. I’m working out ideas for a 6pt Devanagari type, not because I think there is a huge untapped market for such a thing, but because I’m interested in how it could be done.

The counter-prediction results of the lateral masking study you sent me, Peter, continue to intrigue me, and have me thinking about serifs a little differently than I used to. I’m not sure yet how this might translate in design terms.

A good question, Peter, that makes me think. I know that readability research has inspired some ideas, but I find I can’t easily point to a specific relationship between some test result and a design decision.

David, you claim that my studies aren’t meaningful because I test readers in labs rather than following them home to see what they want to read.

Yes, I use a lab and establish specific reading conditions in a study. That’s how science works. I can’t compare Person X in Environment X with Reading Condition X to Person Y in Environment Y with Reading Condition Y because I won’t be able to tell if differences are due to the environment or the reading condition. It’s an involved process to design a study that only measures the variable you’re investigating.

This is a useful way to conduct research because I know that people use the same reading processes at home or in a lab. My point with the word frequency example is that there are universal processes in reading. I can expect to find a word frequency effect if I test a 7 year old Chinese child at home or if I were to test you in my lab. There are many other universal variables that can make reading easier or more difficult. For example I can read faster if my environment has 50 foot-candles of levels then if I’m reading by 10 foot-candles of light. I can also read faster with +2 points of leading than when text is set solid. I would find these differences in someone’s home or in my lab. Are these differences obvious? Sure, once you know the results.

David: “Only Beverly Readahillbillies would need research into this.”

Your response shows a lack of curiosity. I set up a straw man that text shouldn’t be set as tightly as possible and you responded that everyone knows that. I would like to hear your thoughts about how to define the amount of letterspace necessary for an optimal reading experience. It’s not clear to me only Beverly Readhillbillies need research into this. Tracy, Kindersley, or Noordzij’s rules of letterspacing do not come to the same conclusions. I think this is a very interesting area to study. Are your fonts well designed?

Cheers, Kevin

“I set up a straw man that text shouldn’t be set as tightly as possible [...]”

Kevin, how is this a straw man in terms of your claim that we don’t recognize words as whole shape patterns.

Kevin, I am puzzled. I noted above (second post of May 19) that the idea of reading by the gross external word shape is refuted. But as I said, refuting the idea that we read by word patterns, where the pattern includes salient internal features of the letters, seems to me not so easy. Do you think that this view—which Peter and I believe Hrant advocates—is refuted by your paper?

This view is the one I was recommending that you try to test by comparing the effect of disrupted spacing on novice vs proficient readers, and possibly also dyslexic vs proficient readers.

Incidentally, while Hrant has I believe regularly advocated tight spacing (and of course he will speak for himself), I don’t think Peter has. I don’t think this view that we read by ’word pattern of salient letter features’ necessarily dictates tight or loose spacing in itself. It would depend on what most helps the identification of the salient letter features. Personally I think that a lot of spacing of digital text fonts is a bit on the tighter than ideal—ideal for the sake of readability I mean.

[William] “I don’t think Peter has.”

I borrow my spacing principle from Noordzij. The Noordzij principle relates the interletter spacing to the size of the spaces inside the letters with closed or open counters. These inter-letter and intra-letter spaces must be in synch, that is, rhythmically co-ordinated. In a font with a high stroke counterpoint (stroke width) to x-height ratio, or a high stroke counterpoint to horizontal glyph width ratio, the counters will be smallish in size so the spacing can be tight.

(Does Hrant really advocate the view of reading implicit in your statement? Internal features are masked in parafoveal vision, which for Hrant is the site of the hidden core of reading.)

>relates the interletter spacing to the size of the spaces inside the letters

Tracy’s method also does this. I assume that this is a traditional typefounder’s view.

As I said earlier, I don’t really understand Hrant’s ideas aboout ’boumas’, so he can speak for himself as to how much he agrees with you or not.

Peter: “how is this a straw man in terms of your claim that we don’t recognize words as whole shape patterns.”

My claim that we shouldn’t space letters as tight as possible is a straw man because it is underdefined. I imagine that everyone agrees that letters in a word shouldn’t overlap by half. The parallel letter recognition model does predict that each letter needs some amount of separation to make each letter recognizable, but doesn’t go far enough to make testable predictions about spacing.

I’m not ruling out Tracy’s or Noordzij‘s model of spacing. They are the only real theories of letter spacing that we currently have.

William: “refuting the idea that we read by word patterns, where the pattern includes salient internal features of the letters, seems to me not so easy.”

I think internal and external letter features are critical for recognizing letters, but I find it highly improbable that whole words are recognized by internal letter features. There is less direct evidence about this theory, but the moving window study described in Typo 13 would say that any formulation of a whole word approach is wrong.

Cheers, Kevin

John: “So I respond to a post in which you say that they ‘can quite neatly be divided into two different things’, and then you come back at me insisting that they are related.”

:) You lost you. I said “understanding the mechanics of reading”, and “implementing a ww reading Environment” can quite neatly be divided into two different things. On this otherwise pure thread on this otherwise pure threaded forum, I apologize.

“the importance of high contrast typeforms with consistent stroke density ...some people don’t seem to want to admit to having implications for greyscale screen rendering.”

And are these “implications” different for people who don’t seem to want to admit to having implications for fonts dipped in CT?

Kevin: “David, you claim that my studies aren’t meaningful because I test readers in labs rather than following them home to see what they want to read.”

I don’t consider your studies meaningless because you test readers in labs, I consider your studies to be meaningfully matched to the problem they claim to address, and to the claims made in relation to some of your studies. I just think the meaning is wrong and the claims misguided. And, readers’ll be plenty scared about this at the “office”. Following them home is probably not a good idea at this point...sending software, however...

Also, I appreciate scientific method, but if you’ve a critical variable in the center of the issue, (what text’s being read), that Cannot be Repeated, you have uncertainty as a result of being forced to change 2 variables at once even if you don’t want to . . . Don’t ’cha?

“I would like to hear your thoughts about how to define the amount of letterspace necessary for an optimal reading experience.”

I would like to hear Napoleon’s thoughts on how to define the amount of salt necessary for an optimal eating experience, I would like to hear Georgia O’Keefe’s thoughts on how to define the amount of white necessary for an optimal art experience and I would like to hear Lenny Bruce’s thoughts on how to define the amount of silence necessary for an optimal social experience. What’s the rest of the food, and who’s eating? What’s the rest of the painting, and who’s looking? Is it a wedding, a funeral or what? and who are the people? An optimal reading experience is not a single thing to all people, is it?

The word space, and leading control, along with the availability or lack of pair kerning are where any serious discussion of letterspacing begins. But, before you look at them for an optimal reading experience, you (not normally, just here), have to trace the whole system back to its “root” interpretation of the em to see what you’ve got to work with. In this case of Windows, one begins with a redefinition of points per inch. In this shrinking of the size of all type, a threshold is crossed that makes it advantageous, in the pursuit of optimal reading experience, to increase the x-ht, basically, altering the proportions of the font away from a master capable of performing outside of the text range. This is fine, but if you do this, I don’t think you need to experiment with the letterspacing or consult any grand theories (all about print anyway), it should just open it up at a rate slightly greater than the rate of the lowercase’s already expanded body. And, if you plan on preparing the fonts for a system that lacks grid fitting control in X, add a little more space onto that, and you’ve got a 8 pt. design, with a label that says 12 point. Good?

After print (right now), is similar to several other points in history, though it’s been a lot worse. Post Gutenberg there was a bonanza of bad letterspacing from the unskilled use of truncated cases. The fledgling type biz had just eliminated the scriptorium, but not yet invested in properly compromised letterforms. We also saw letterspacing fall into a long dark interim between the invention of the typewriter, (an obvious letterspacing blight) and the second generation of personal computers (letterspacing without kerning). In both cases, the type designer was not entirely unlike Goldfish vs. Firehose. And, in those sad times the letterspace optimizer was/is ye olde large word space :)

[Kevin] “I’m not ruling out Tracy’s or Noordzij‘s model of spacing. They are the only real theories of letter spacing that we currently have.

I suggest we specify a benchmark for testing based on fourier transforms. I base this on the observation that spacing which meets Noordzij’s standard leads to a fourier transform profile in the frequency domain that shows a marked clustering of vertical components around—but not uniformly on—a phasal mean, with no clear competing harmonics. It should be possible to find this for every typeface tested. This distinctive profile breaks down when spacing is varied beyond a given range.

The goal than would be to require the parties of the debate about models of reading to make predictions, relative to this benchmark, on the basis of a fully specified theory about perceptual processing relative to phasally rhythmic arrays. Fully specified would mean neurophysical modelling, given what is known about colunms and layers, transduction, summation, lateral inhibition, receptive field behaviour and feature binding in the visual cortex.

David: Sorry I’m not responding to your often interesting posts. I’m having trouble compiling a clear image of what you’re after or about. (Others might say the same about mine)

By the way, I’ll be in the NYC area for a few days. Anybody I should meet?

[Kevin] “[...] the moving window study described in Typo 13 would say that any formulation of a whole word approach is wrong.”

If, as I claim, the decisive part of visual wordform resolution happens in the form resolving field just around the point of fixation and indexes the entire word as the unit of perception (when the word is of a modest length), the moving window study would not directly touch my claims about how the foveal part of reading works. The moving study would only assist me in guaging the extent and coarseness of the parafoveal preview benefit.

I would claim parafoveal vision assembles some level of ’ensemble statistics’ about what lies ahead in the line of reading, and these are sufficient at least for ’indexing’ or pretuning the foveal attentional window as well as for deciding a strategic landing site for the next fixation. But exact specification of the nature and coarseness of the ensemble statistics and how far xxx’s go in confounding them is difficult.

Xxxxing out some letters may not be fatal to the benefit (although it could be detrimental to some minor degree, but Xing in some spaces even if all the letters are in place might be more detrimental.

>the moving window study described in Typo 13 would say that any formulation of a whole word approach is wrong.

Kevin, at the time I read that paper, I pretty much dismissed any form of the ’whole word’ view, so, agreeing with you, I didn’t look at your arguments closely. Only after discussing Peter’s views with him—after being critical of his rhetorical style!—I began to understand them, and to see that he might be on to something.

Now rereading that portion of your paper on the ’window study’ method, I do agree that it calls any form of the ’whole word’ view into question.

However, it doesn’t seem conclusive to me. Part of the problem is that the difference between the number of letters and whole words is very little. Therefore, I wonder how many times when the person looks at a group of letters he is actually catching the same number of words under both regimes—particularly as the saccades seem to be planned ahead to catch words.

Another issue is that I suspect we are able to identify pieces of words, especially phonically meaningful pieces, like th and oo, as wholes. And then we can guess words from these, and not only letter identification, or whole word identification. In other words, it might be that every method may be tried, but the whole word pattern first and most efficiently.

To make more sensible comments on this I will obviously have to read the paper and study its methodology and results. However, at this point it still seems to me worthwhile to try to devise a more crucial test, either using the moving window type of test, or the spacing disruption I mentioned above.