Randy, your subject line lead me to believe something different. :^P I thought you were going to go on and on about why there is no reason for serifs.

I don't have that book here at work, but doesn't it have a bibliography? That might be a good place to start. In fact, any of type book worth reading should have a good -- at least decent -- bibliography.

There are a number of good articles dealing with the role of serifs in Gunnar Swanson's collection of essays Graphic Design & Reading

...promote good word image by defining the space between letters and the counters within letters.

For open bowls (counters) yes. Either Fred's terminology is innacurate, or you may have paraphrased rather than quoting. Strictly speaking only fully closed bowls contain counters. But that's a hard-edged technical definition I suspect some tpyographers might dispute.

j a m e s

Tiffany: Yes is does have a bibliography of sorts (a "literature" section about cutting punches). I'll troll through to see if any would be worth following up on for my purposes.

James: Thanks for the book. Re: Terminology. I was paraphrasing, but he does call them open and closed counters for what it's worth. Regardless, we understand what he means.

Catich Edward - The Origin of the Serif (1968, 1991)

Randy,

Another good book on the subject is Walter Tracy's "Letters of Credit". It's out of print but you should be able to find it second hand. One of my top five books for long.

Cheers

Thanks for the references. It would be great if the books contained more than a history of the serif, but also it's influence on reading (readablility). -- Randy

Randy, The Serif is Dying thread at Typographica from 2002 offers clues and views from prominent contemporary typographers. Some of it might be of interest. Problem is the discussion starts off with serifs then morphs into a detailed dialog on readability, which is a subject in itself.

Thanks for clarifying Smeijers.

j a m e s

http://www.psych.ucalgary.ca/PACE/VA-Lab/gkconnol/Thesis.html

more?

Not to step on toes, by Tracy's book is available via Amazon.com.

Eric

David: More is good.

Side note: I'd be interested to see the Times Roman Sans in the linked study that was equally well received as Times Roman Serif. The only difference being the serifs trimmed off in the "sans" version.

Times Sans! That's Boschean.

Randy, the bulk of justifications of the serif, even by reputable and otherwise excellent sources is pretty hollow. Reasons like "they create train tracks for the eyes", "they define the line so that line-returns are easier", and "they define the ends of strokes better" are the sad norm - sad because these things don't hold up for 10 seconds once you really delve into reading.

Smeijers is probably the most reputable, mainstream author who has come closest to what I think makes sense - and that's pretty rare unfortunately. He's talking about notan, and that's the key. But he still doesn't go far enough. The articles by Peter and myself in Typo 13 try to go deeper, although they don't really focus on serifs. Serifs help essentially because they increase the convergence of letters into boumas. In comparison, the only way sans fonts can attempt this is by making the letters too close, and of course that backfires.

A paradoxical extension of this is that serifs help by making letters
less themselves (and more parts of boumas). It's possible to see this
as parallel to the belief that a real text font needs a certain ugliness.

BTW:
1) One thing that has yet to be really addressed is the semi-serif.
2) You might be interested to know that Bloemsma went through
a love/hate thing with serifs - he was conscientious enough that
he needed to justify them (as you seem to be trying to do) instead
of just leaning on them blindly or shunning them.

hhp

Smeijers is probably the most reputable, mainstream author
What he is saying makes sense. And I agree he is talking about serifs in the context of notan and boumas without the vocabulary. I'd like to read your articles in Typo 13. Can you send me the text? Or is it still available?

I wonder if there aren't at least several things fighting sans serif typography for text.
1. The word shapes break down into letter shapes when you space sans serif type optimally for text (by harmonizing the space between letters with the space inside letters). This is another way of saying what you said above.
2. In some sanserif type (and also in modern serif typography) the vertical stress accentuates the break down.
3. Defining word shape requires a solid perception of x-height and baseline. This is kind of like the traintrack idea, but more at a word level than a scentence level. Sans serif type really lacks some definition at the baseline, but also the x-height.

One thing that has yet to be really addressed is the semi-serif.
This is what got me started. I'm reworking a semi serif design that was laying around on my hardrive. What I would like to achieve is essentally "sans serif" for longER text, facilitated by the addition of serifs at critical locations. What I've found thus far is that it's very difficult not to allow stylistic choices to trump the schema that I'm trying to work out. I'm finding it very interesting.

I was messaging with Paul Hunt about opentype this afternoon and we stumbled on a further idea together. What about a semi-serif opentype face, that contextually places serifs based on neighboring glyphs? An interesting concept that could potentially allow a font to reflect this dialogue in a new way.

Thanks for your thoughts all and keep them coming. -- Randy

Typo 13: I guess the PDFs of back issues are not currently available.
Not sure if they'll be back or not - the "discussion" section of their
site seems to imply they will. But I can send you a print copy if you
give me your ground address - hpapazian thatsymbol gmail dot com.

> a semi-serif opentype face, that contextually
> places serifs based on neighboring glyphs?

I think that's exactly the type of thing that needs to be thought about.
Consider for example an "elaboration" on Legato, where the whites
inside and between letters depend on the letters.

hhp

My primary interest in the role of serifs in reading concerns their possible contribution to overcoming lateral masking in the parafovea. This is non-intuitive, but research has shown that adjacent forms that have connecting strokes are less affected by lateral masking that forms that are separated by gaps. Lateral masking is the phenomenon whereby which a form perceived in the parafovea that is adjacent to another form or between two other forms is harder to recognise than an isolated form. Obviously lateral masking has a huge impact on our ability to recognise any letter shapes in the parafovea, but some recent research indicates that serifs, by bridging the gap between the letters, may reduce the effect of lateral masking. As I say, this is counter intuitive, since it is the proximity of the letters that produces lateral masking in the first place. Note that this isn't counter to Hrant's comments about serifs helping convergence of letters into boumas: it explains in part how they do this in the parafovea, by overcoming the lateral masking that obscures the 'role architecture', to use Peter Enneson's phrase, of the constituent letters.

That would explain why a serif face seldom seems to be missing anything when it lacks vertical serifs such as on the s or c.

In 'Letters of Credit,' which you probably already know, Walter Tracy makes a number of interesting comments on the sans serif. One of these is that the sans needs to be more tightly spaced because of the absence of serifs.

John, which sans was tested in the research about lateral masking? The reason I ask is that I suspect the reason for the problems of letter recognnition in sans is that a balance between counter space and letter space is important for letter recognition. If the letters are wide--the 'o' more circular rather than oval--and the letters closely spaced, as in Helvetica, then I would expect the letters would be less easy to decipher. But when the letters are narrower--the 'o' is a narrower oval--then the narrower letter space of the sans would be less of a problem, and the 'lateral masking' less.

This is my effort to explain why I find Helvetica in small printed text much less readable than Gerard Unger's Vesta, or Meta for that matter. My explanation: the narrower counters balance the tighter spacing of sans faces.

I suspect that the 'knitting together' of words as units is a due to a different effect of serifs. I think the serifs help to achieve even color in the word, making it easier for the parafovea to see upcoming words as separate words, and plan for the next saccade. In other words, the parafovea can identify words as grey blocks separated by word spaces,even without being able to read the letters within the words. The set of grey blocks may be enough for planning, and highly functional.

Without the serifs, and with wide letter spacing, the grey word block is not as distinguishable from the word space. That is why, I think, you can't simply track Helvetica more widely and get a good readable text face from it. And it is why sans serifs need tighter spacing in the first place.

In other words, serifs serve two purposes: both knitting the words together visually as units and keeping the letters far enough apart in comparison to their inner spaces for easy letter recognition. In sans faces, these two different goals come into conflict, particularly for wide designs. I think even with narrow designs the horizontality of the serifs helps get better color to knit together words, but the advantage is less marked for designs with narrow counters.

Finally, I think the horizontal strokes help to tell the eye where the line of text is, so it is easier for the eye to 'hold the line', and not get confused with the next or previous line. I think that is a factor in why sans usually need more leading.

Now one of the reasons why all this is difficult to prove is that the brain can overcome all these obstacles that sans put up. But it is taxed by the effort. That is my explanation for why some faces seem friendlier and more inviting and easier to read. And also why I think Kevin Larson found that usable comprehenson was better with good typography. In other words comfort in reading--which I think you, John, were skeptical about once here--is a very real factor that makes an important functional difference, even though identifying it is quite subtle and hard to test for.

Incidentally, none of the effects I am hypothesizing here are dependent on the existence of a 'bouma' effect, such as Peter Enneson is hypothesizing--that is, a reading of whole words as patterns (including internal features of letters), rather than first identifying letters and then assembling them into words.

I think that’s exactly the type of thing that needs to be thought about. Consider for example an “elaboration” on Legato, where the whites inside and between letters depend on the letters.

This is what i was alluding to in my old thread the contextually sensitive notan

Perhaps we could wrap our discussion around a concrete example. Below are 4 variations on the letter n ranging from a "traditional semi serif" to a sans version. I've put them in an isolated context (a serif free zone if you will) -- the word "tone". Then I put it next to another potentially serifed letter -- in the word "tonne". Obviously these are simplified cases, but I'm curious what forms work in this specific context, based on people's notions about serifs and readability.

Note: The resolution of the image makes spacing problematic. I know this is a light weight which introduces it's own issues, but it's what i've got.

My observations:
1. The serifs do help to define the space between the n and rounds.
2. As a result tone #3 seems best to me (traditional semi serif placement)
3. One serif seems to be enough to define the space between the letters.
4. The space inside the n seems well enough defined without serifs inside.
5. It seems like the x height serifs are more critical than the baseline serifs.
6. As a result, tonne #3 seems best to me.
5. I'm making these observations as a designer in a meta context not a reader immersed.

Thoughts?

very interesting. i think you're quite right about the serifs on the inside being superfluous. i might have to steal some of your ideas for some changes to my interloc. i wanted to accomplish something like what you described with that face, but it seems you have it much more thought out! thanks for sharing your developments with us.

... right about the serifs on the inside being superfluous.
Maybe (and maybe only in this context and goal)

"where the whites inside and between letters depend on the letters."

This is always the case and is the real job of "Type Face" design vs. "Letter series" design. The Yin Yang of design where one needs the other so much that they help define each other. If you think about ambigrams, you get at the notion. You cannot ignore one in favor of the other.

ChrisL

"Perhaps we could wrap our discussion around a concrete example."

This is exactly what I was after last year when working on my Leporello type face. I was attempting to design a sans with enough appendages to solve some of the reading issues but never went as far as to make it a semiserif face. Most of the discussion in the critique was involving the stylistic quarks that come about by doing this. I wonder about the difference with showing the kind of face you are talking about to a naive audience rather than type Design people who perhaps know too much about how type has looked since its inception? I know that when I showed text set in Leporello to naive viewers, they never noticed the quirks.

ChrisL

Maybe (and maybe only in this context and goal)

well here's a quick experiment: is the second ton/tonne an improvement or not?

I think for my case, where mine is a serif (essentially a romanized italic) with some hacked off, i like the two serifs between the double n's.

The second tone tonne looks better to me at that size at least. I think size and quantity of text would certainly matter with this issue though.

ChrisL

I think, besides the serif, another element which contributes to readability is stroke modulation. High contrast is a problem that can't be forgotten. I should think. Helvetica and Didot both aren't friendly to the eyes.

"Helvetica and Didot both aren’t friendly to the eyes."

I asume you mean for different reasons? Didot being very high contrast and Helvetica being very low?

ChrisL

Yes. Thank you for reading my mind. :^)

I guess I should have said, any typeface which borders on total contrast or none at all.

Paul, your second ton and tonne look better, but here you are following the traditional way an italic would be done, so they conform more to reader expectations.

Another issue here is the narrowness of the letters and of the sidebearings.

Randy, on your example, one could get the total sans case to have more even color, similar to the more serifed version, by pulling in the side bearings appropriately. So the issue isn't just serifs but also spacing. If you start deleting or adding serifs depending on context, you would probably want also to play with the spacing according to context, when you add or substract the serif. This might get more complicated than it's worth. Still, I think it is an interesting exploration, and you may well come up with something very nice. Good luck!

Paul: My observations on my effort apply to yours too. I don't think you need the interior serif, but you will need to look at your glyph proportions. Hacking off a serif changes the size of the counter significantly. Spacing looks tight for text, but it's hard to say in this context. Pretty sure it will need to loosen. Your nn combo with two serifs may make sense in the context of your design (which is much closer to serif type than sans). In mine, i'm trying to add the minimum to maintain as much sans feel ... hence tonne #3

William: Removing a serif from the outside also shrinks the apparent size of the counter, though less than removing a serif inside (see above). In the sample, I did adjust the spacing (didn't simply whack off a serif), but I didn't slightly widen the sans version (which I should).

Chris: In Leporello you were approaching readablility by looking at glyph structure itself. What I hope to accomplish to keep glyph structure as consistent (normal?) as possible, and to isolate the effect of the serifs. This is for my own sanity. If you change too many things, you loose the cause of your effect.

Re: naive audience rather than type Design people

Since quirks will be limited to serif placement, how quirky the face appears will depend on the OT rules applied. The idea is to create n.full, n.xht, n.base, n.sans sets for the lower case (where applicable), and test various observations like I made above. So I'll be changing the rules instead of changing the glyphs themselves. Hopefully this will add a level of detatchment about normalness. In otherwords, I'll try to turn off the voice that says: "it's readable, but is it marketable?" To be honest, I don't think even the craziest placement would fall too far out of the mainstream.

Tiffany: The modulation here is looking more toward the helvetica end of the spectrum, though not as monoline or vertical. As the weights get heavier I plan to apply more contrast. The glyph structures in this light version should hint at what is to come though. There is a definate diagonal stress.

"...and to isolate the effect of the serifs. "

Randy,
It would be fantastic to isolate the variable. I hope you can do it successfully. This would be a prime candidate for a meaningful readability research experiment. By changing your parameters in your opentype feature, you could generate the same text with different serif removals and see what the effect would be to the reader. This would perhaps generate more insight on the serif's role in long duration reading. It would be nice if someone would give you a grant to test it. You could include some straight sans, serif, and semiserif faces as controls and measure your degree of readability enhancement with your experimental face.
BEST OF LUCK TO YOU on this!!!

ChrisL

William: John, which sans was tested in the research about lateral masking?

The lateral masking research to which I referred was not testing typefaces per se. Peter Enneson sent me the information, and I'll need to dig around in my notes to find it. The tests were using shapes of various kinds, first isolated, then adjacent to other shapes, and then with connecting strokes to adjacent shapes. Participants were better able to identify the shapes in the parafovea in the latter case, which surprised the researchers. My observations about serifs are an inference from this study, but obviously a lot more study needs to be done, and specific testing of typefaces should be done to confirm the inference.

I don't recall every being skeptical about the importance of comfort in reading. It seems to me very important. Eye-strain is one of the most easily quantifiable aspects of readability testing, because it is possible to measure the activity of the relevant muscles.

Your comments about the relationship of internal and interletter spacing in sans serifs are absolutely right (have you read The Stroke yet? this is illustrated very well). But this addresses the question of why one kind of sans serif, spaced in a particular way, may be more readable than another kind of sans serif spaced in a different way. It doesn't address the point of the serif.

You could include some straight sans, serif, and semiserif faces as controls
If I was a researcher, what I would do is start with Thesis (The Serif and The Sans and The Mix). Then modify The Mix as I've described and add the opentype. This would save a weary type designer a heck of a lot of time (but might violate a EULA). The benefit is that you have the same typeface, further reducing your variables.

As it is, I'd like to draw a nice reading sans (semi) using all the resources available.

John: I've not read The Stroke, but will order it up.

randy, i'm curious as to what will happen to your serifs as the weight increases. will they become big monstrous slabs or what? (i'm sure "or what" is the correct answer but what IS that what?)

> It doesn’t address the point of the serif.

I guess I wasn't clear enough. The serif reconciles the conflicting poles of even color and adequate letter spacing for readability. It also helps the eye hold the line of text.

Also you have the additional problem with sans that they are generally supposed to look linear, which makes it more difficult to get even color at the joins.

So the power of the serif, with modulated stroke, is that you have a powerful tool that solves a number of problems at once.

I was waiting till I didn't have to pay shipping from Europe for The Stroke. I'm ordering it now, and look forward to reading it.

I think Tiffany has an important point that is useful to consider: in developing a test typeface with enough variants to isolate the serif's effects, wouldn't it be useful to settle on a letterform that, at text weights, can accomodate sans as well as serif structure, i.e. a moderate-contrast thing like Schneidler, Chaparral or Joanna?

Then, too, there are other choices you would face, like, do I deny the possible importance of diagonal stress and make the thing all vertical like Baskerville or Cheltenham? Throw out the ideas that Bloemsma was getting at?

Then, how to filter out the variable of width? It's true that not only spacing but letter widths should change as well depending on serif presence.

But also, proportions between letters must have an effect on reading comfort; surely monospaced type is not optimal, but then too much width contrast also hinders reading comfortably.

Rather than enumerate the variables, I guess I'm saying that if you want to test the usefulness of serifs, it may require more complex, interlinking factors than simply which serifs are present. What assumptions do you start with? Obviously Helvetica and Didot are not candidates. So many things affect readability that it might be necessary to test existing faces that are related but differ in one aspect, like contrast, and come to some conclusion about the ideal contrast level. Then, test faces with the same contrast level and differing amounts of diagonal stress. Determine the optimal diagonal stress. Then test faces with optimal stress and contrast with different widths.....

Of course if you start that kind of elimination with serif faces, find "the one" most readable, and chop off a few serifs, it might fall apart..... Back to square one?

what will happen to your serifs as the weight increases
As the weight increased the contrast will also increase. Light to bold, they will keep the wedge shape, and be analagous to the thins (slightly thinner). Above that we are not really talking about extended text so It might get funky. I'm not sure. I can say it won't be a straight interpolation light to überblack, with correspondingly übermonster case of the slabbies.

I can say it won’t be a straight interpolation light to überblack

superpolation?

p.s. i like überfett better.

Maybe Überfëtt is the new band name....

I liked Die Schrïften! better, although i'm sure my German's all wrong.

back on topic now...

Re: Überfëtt
No relation to Böbafëtt.

Carl: I think using a font with moderate contrast is a good idea. And I agree about reducing as many variables as possible. Hence my above example of Thesis. While you may not see many books set in The Serif, you could probably still find out about how serifs are functioning using a design normalized by a skilled typeface designer across styles such as Thesis. Certainly with better results than a researcher-turned-typedesigner's attempts at Times Roman Sans mentioned above. (Chris: This is the original Sans of Times!)

"(Chris: This is the original Sans of Times!)"

LOL!!! Love it!!!

ChrisL

Sorry, have'nt read every post here..

Anyway. My teachers allways said that the serif's draws a baseline, and that makes it easier to read in long texts. And that's also what some books say.
But in my point of view the thing making it easyer to read is that the letters become more equal in width with serif's.
The i, l, j and so on become wider. But letters like o stays the same.

So "ioi" looks better in serifs than in sans. Serifs evens out the white space.

I'm pretty new to typography. But anyway, that's how i see it.

Lasse, interesting observation about the ilj. These three letters are given serifs in FF Info Display. I think it idea was to make them easy to differentiate from each other and the cap I, but this would be an additional benefit.

John, I finally got my (cheap) Amazon USA copy of 'The Stroke'. The illustration you referred to, on p. 42-43, is just brilliant. It really settles the argument as to the importance of the ratio of interletter space to counter space. An additional factor he doesn't go into, but which I think is important is the influence of the weight of the letters on ideal spacing. These all interact with the influence of serifs.

The interaction of different factors is I think why Randy's inquiry is tough. It isn't only the presence or absence of serifs, but also the *ratios* of stroke weight, size of counters and inter-letter spacing. And Lasse's insight raises the question of whether there are special issues for narrow letters.

> the letters become more equal in width with serif’s.

Except you want divergence of widths
for optimal readability, not convergence. :-/

On the other hand, your point about serifs making
some characters wider and thus giving them more
of a chance to stand out is a very good one.

> It really settles the argument as to the importance of
> the ratio of interletter space to counter space.

It's just a damn shame that in the
end that importance is not properly
addressed via chirography...

hhp

Hi Hrant, don't you think that it's divergence (of width in this case) within a limited spectrum that's most likely to be easily processed as pattern? Such that widening i and l with serifs brings outliers into a width-range over which differences are more readily picked up?

James also mentions above the Swanson book, which I've read; apart from Rolf Rehe's and your essays (and perhaps one or two others I've forgotten) it's the most astonishing rubbish, semi-literate garbling of continental philosophy. What were you thinking in associating yourself with these dismal frauds?

>divergence (of width in this case) within a limited spectrum that’s most likely to be easily picked up as pattern?

Well put, George. I've been arguing with Hrant that divergence has to be within the constraints of even of color and good rhythm, but you put the case much better here.

There certainly are limits. But the main
thing is to see that divergence provides
the information, not convergence.

> within the constraints of even of color and good rhythm

Even color: Here again I have to point out that even color,
perhaps only because of the wording, tends to convey the need
for total uniformity, which is something we can hopefully agree
is a Bad Thing. So when you say "within the constraints", I hope
it's clear that even color is just another factor that needs to be
balanced, not some sort of absolute requirement.

Rhythm: I'll kindly let that slide this time. :-)

--

Graphic Design & Reading: Well, I was, and still am, actually
flattered to have been asked by Gunnar to contribute. I'm also
grateful in a practical sense, because for some reason it seems
I only write articles when somebody asks me to, and then I have
something useful to brandish. The pseudo-philosophy you complain
about was the hallmark of the 90s... and GD&R came out towards
the end of that era. Actually, I didn't know what else would end
up in that book; but even if I did I would have gone ahead with it:
it was a great opportunity to reach a lot of people; in fact a quarter-
length version of that essay (with a really good Spanish translation
by Jorge de Buen) was subsequenly published in tipoGrafica. BTW, a
lot of people think MINE was the rubbish part! :-)

hhp

a lot of people think MINE was the rubbish part!
It's a superb piece, and I'd love to see whether divergence in incunable-style micro-properties can work as well as the macro-divergence of your example lowercase (on the grounds that the reduced clarity of vision in the parafovea isn't really gaussian - there's no certainty that different architectural forms are more clearly perceived than things like different lengths of serif or different relationships between internal and external axes). As for colour, I do find consistently divergent colour very off-putting in, say, the light s of the original California Old Style or the light h of digital Poliphilus, but perhaps if divergent colour were spread more evenly across the alphabet it wouldn't matter so much.

Thanks.

Color spikes: I think if they're mild enough (and yes, spreading
the divergence around is crucial*) so as not to trigger errant
saccades (saccades from anything except the previous fixation)
then they can only help.

* Could one say that the mathematical derivative
of the color map should be totally even? Hmmm...

hhp

Hrant: "divergence provides the information"

Yes! but if my fourier transforms say what I think they do, too much hurts readability by inhibiting bouma formation, i.e., has an adverse affect on what I called in my Typo#13 contribution a wholistic (rather than letter slot-processing skewed) response bias.

George: "reduced clarity of vision in the parafovea isn’t really gaussian"

Yes! brilliantly put! the increasingly well-understood effects of lateral interference on the processing routines applying to processing in parafoveal vision of letter-clusteral material are not, to my mind at least, effectively integrated into Hrant's scheme.

> too much hurts readability by inhibiting bouma formation

Interesting.
The question is: Is this within errant-saccade tolerances?
Because if not, it's largely moot.

> not ... effectively integrated into Hrant’s scheme.

Admitted.

hhp

"The question is: Is [...]"

Could you elaborate: I don't understand the question.

If the color divergence necessary to inhibit bouma formation is
greater than the divergence necessary to cause a saccade (to fixate
on the color spike) from a fixation other than the "correct" (previous)
one then it's moot because reading is disrupted so globally that optimal
boumas never have a chance anyway.

hhp

I would think the level of divergence that threatens effective bouma formation would threaten efficient saccading as well. We could call the 'rank' of these types of issues the molecular 'rank'. Disturbances that threaten functioning at the molecular rank would also create problems of colour at the text-block rank.

Re: divergence, and your fondness for Fleischman 65, I produced a "generic 65" by modifying the most vanilla Dutch baroque style face I could find - Bitstream Aldine 721, actually a narrowing of Plantin but so far from Granjon that its ultimate origin is irrelevant - to reduce the height of bowls, verticalise o, Jannonise the head serifs and Fournish the feet. I think it works: I can try to do it again rather better if you'd like to see the results.

"Rank": wouldn't fixation-scope parafoveal issues be of a different
rank than broader, non-content-relevant peripheral issues?

> "generic 65"

Sounds interesting!
Got a visual?

hhp

I've posted a PDF to the Serif Crit forum.

"wouldn’t fixation-scope parafoveal issues be of a different
rank than broader, non-content-relevant peripheral issues?"

I'd see fixation-scope, and the parafoveal issues you like to foreground when you talk about reading as bouma-based, as relevant to the molecular level.

Starting with:

Patterson, D. G., Tinker, M. A. (1932a). Studies of typographical factors influencing speed of reading: X. Style of type face. Journal of Applied Psychology, 16(6), 605–613.

I find that they reference:

Roethlin, B. E. (1912). The Relative Legibility of Different Faces of Printing Types. American Journal of Psychology, 23, 1–36.

Who in turn references:

Koopman, H. L. (1909). Scientific Tests of Types. The Printing Art, XIII, 81-83.

Has anyone found a primary source for the typographic convention which claims "serifs are easier to read than sans" on which these studies are based?

There is no strong empirical evidence yet. Kevin has to assume for the sake of argument that serifs help (because of the persistent anecdotal evidence) and then devise rock-solid testing that proves or disproves it.

Nobody besides Kevin has the combination of resources and insight.

----

Today, learn about the Armenian Genocide.

hhp

Perhaps a piece of typographic history akin to quoting Frederic Goudy for saying "Anyone who would letterspace blackletter would steal sheep."

I'm not aware of a primary source for this claim, Christopher.
___

Have you read Mark van Rossum's paper ‘A new test of legibility’ (Quærendo 27/2, 1997)? A report on his study is available on his website. See the last page of the PDF for the illustration of the possible serif effect in the parafovea (I say possible because what the illustration shows is a gaussian simulation of parafoveal blur, so the actual effect might be different).

Thanks for the reference John. I have not read this one yet.

(I'm pretty sure I have asked this before, but what is the syntax to make it so that a link downloads a PDF the way you did?)

I just did a straight href link to the PDF file. I think how that is interpreted depends on your browser and on Acrobat. When I click the link, the PDF opens in my browser.

Is this some kind of game, where we find a four year old thread and suddenly revive it? Cause I really don't want to play.

There is no expiration date on quality discussion.
And closure is a cozy, over-rated illusion.

Anyway, I have to ask: who signed you up?

hhp

>Serifs. What's the point?

Succinctly put, thanks.

They have data for us.

Cheers!

I believe there is a personality difference between serif and sans serif typefaces, but not an inherent legibility difference. In several legibility studies that look at a variety of typeface styles, we don’t find that all the serif typefaces or all the sans serif typefaces rise to the top. Type design is more interesting and complicated than that.

When Dawn Shaikh measured the font personality of dozens of typefaces, she found that people consistently characterize serif typefaces differently from sans serif typefaces. Serif typefaces are more formal and assertive, while sans serif typefaces are more youthful and casual. The personality differences can make one style more appropriate for a particular document. This was the topic of my ATypI talk in Mexico City.

Cheers, Kevin

One notable contribution to the serif versus sans serif debate involved Chuck Bigelow creating a font that had equally competent serif and sans serif forms. The title of the paper is quite informative:

Morris, Robert A. Kathy Aquilante, Charles Bigelow, and Dean Yager. 2002. Serifs slow rsvp reading at very small sizes, but don’t matter at larger sizes. Society for Information Display International Symposium Digest of Technical Papers, P-13.

...personality differences can make one style more appropriate for a particular document...

That's almost like... graphic design.

**

Aren't you making rather a large assumption that results from RSVP reading tests have any great relevance to normal reading?
That seems to be the inference.
After all, as you yourself said, "type design is more interesting and complicated than that".

As I have written before, the studies I have heard about I believe do not take into account the difference between readability and legibility. I just noticed that Walter Tracy's wonderful Letters of Credit includes a section—p. 30-32—discussing the distinction, in exactly the same terms as I have done here. Many of the best type designers, including Tracy—whose book summarizes centuries of type lore—routinely operate with this distinction.

To test for readability, ease of reading extended text with comfort and comprehension, it is necessary to test reading for extended periods. I have suggested using timed SAT reading exams in various settings, and seeing how comprehension and speed and self-reported fatigue vary.

In typography literature, it is common held that sans serifs need more leading for optimal setting. And this is probably because serifs act to define the base line to the eye more easily. This would be one advantage, but I think there are others. But I think we will only get the real information if the distinction between readability and legibility is acknowledged, and tests are done on reading extended text.

This is a big gap in testing, I think. Kevin has referred to a 1948 book that is I believe the only one to test with this kind of goal. But it has never been repeated, and I strongly suspect that its methodology was flawed. It is long past time to do some good testing on extended text, on paper.

"Kevin has referred to a 1948 book…"

Reference? I can't seem to find him mention this within this thread.

> … serifs act to define the base line to the eye more easily.

I think this is an interesting hypothesis that I think should be investigated. If a student is interested in picking this up as a masters or PhD project, I would be happy to partner with them on this. This is the only argument I’ve heard that predicts a difference between legibility and readability.

To my knowledge there is no evidence that there is any difference between legibility and readability. Models of reading assume the word recognition is serial. We read one word, then our attention and sometimes eye fixation moves to the next word, and we read that word. If we read words one at a time with no interaction from the previous word, then legibility and readability are the same thing. If serifs can act to improve the saccade accuracy, then there is a clear interaction between the previous word and the word that you are currently reading.

Another one for the list:

De Lange, R. W., Esterhuizen, H. L. & Beatty, D. (1993). Performance differences between Times and Helvetica in a reading task. Electronic Publishing, 6(3), 241–248.

(with an unfortunate failure to counterbalance the order in which Ss were exposed to the two different typefaces)

Kevin: If serifs can act to improve the saccade accuracy, then there is a clear interaction between the previous word and the word that you are currently reading.

What do you make of Mark van Rossum's study, Kevin? It suggests a role for serifs in strengthening the terminals of extenders in the parafovea, illustrating how in a blurred state the extenders of sans serif letters are less well defined. This implies a possible role for serifs in saccade accuracy and also, I would think, for word recognition at the edges of a fixation and, hence, a reduction of regressions.

> Type design is more interesting and complicated than that.

Not least because so is readability.

And as a result, so must the testing also be.

> serifs act to define the base line to the eye more easily.

This cannot hold water. Linespacing already does that. If the linespacing is inadequate, no length of serif will help. The baseline is purely aesthetic.

hhp

Morris, Robert A. Kathy Aquilante, Charles Bigelow, and Dean Yager. 2002. Serifs slow rsvp reading at very small sizes, but don’t matter at larger sizes. Society for Information Display International Symposium Digest of Technical Papers, P-13.

On this account serifs affect performance at thresholds of performance differently than they affect performance on plateaus. If we were to relate legibility to performance at thresholds and readability to performance on plateaus, the above might be enough to motivate the making of a distinction, and an investigation of how serifs effect the size of the plateua, especially under manipulations of spacing.

Test results are moot when the testing is fatally flawed.

And RSVP?! Let's get real, guys.

hhp

Did that mean they have more data for us? ;)

Cheers!

Hrant: This cannot hold water. Linespacing already does that. If the linespacing is inadequate, no length of serif will help.

But Bill's larger hypothesis is that adequate linespacing is different for serif type than for sans serif type. That hypothesis might be true even if his comment re. ‘defining the baseline’ is not.

Personally, I don't think it's about the baseline, but about what happens at the end of extenders, and I can imagine how the presence of horizontal serif elements better define the upper and lower limits of a line of text, allowing for tighter adequate linespacing. [Adequate linespacing means, to me, linespacing that prevents one line of text optically interfering with another during reading.]

> adequate linespacing is different for serif type than for sans serif type.

If this is true, I think it must be due to serifs on extenders helping to form better boumas (as opposed to better defining virtual horizontal lines).

hhp

> What do you make of Mark van Rossum's study, Kevin?

I like that van Rossum starts from knowledge of the visual system and attempts to build a model. I’ve been working with the S-CIELAB model of the human visual system to do something similar. An improvement of the S-CIELAB model over using Gaussian blur is that the human visual system is more sensitive to some spatial frequencies that others. We are particularly sensitive to 3-5 cycles per degree. If this improvement was added to van Rossum’s blurring, I believe that you would find that small features like serifs disappear more quickly in the parafovea.

This could also be tested more directly by conducting a letter legibility test in the fovea and parafovea. If serifs strengthen the terminals of extenders in the parafovea than we should expect that the serifs perform relatively better in the parafovea than in the fovea. It would be better to use a serif and sans serif face that match like in the Bigelow study rather than comparing four typefaces that differ on many dimensions.

> I like that van Rossum starts from knowledge of the visual system

The best way would be to start from a knowledge of the visual system coupled to anecdotal evidence. Ignoring anecdotal evidence is unscientific.

> small features like serifs disappear more quickly in the parafovea.

Which is why they need to be big enough to "bind" letters together into boumas.

> letter legibility test

If you mean testing the legibility of individual letters in both the fovea and parafovea:
1) I'm pretty sure that's been done, IIRC by Herman Bouma himself.
2) It's largely moot because: that's not what we really read; and serifs cannot play their role without multiple things to bind.

> It would be better to use a serif and sans serif face that match

Indeed. But determining proper letterspacing compensation for lack/presence of serifs is non-trivial.

hhp

I don't have time to do this discussion justice, but just a point of clarification.

I believe that a problem with too little leading is that the eye has problems in holding to the line, and/or in jumping correctly from one line to the next. That is, with too little leading the eye will mistakenly move to the wrong line, either while reading a line or when moving from one line to the next line. It may also tax the reader to hold his or her eye on the line.

If this is correct, it should be possible to detect with eye tracking equipment, if I understand what that does.

The view about leading with sans, which I have read in many places and is not my own—though I accept it—is that "too little" leading happens at a greater distance between lines than it would with a comparable serif.

I think that this is by no means the only advantage of serifs. Another important one has to do with maintaining both even color and good spacing between letters—which in tern affect ease of visual word formation.

>To my knowledge there is no evidence that there is any difference between legibility and readability.

That's because the testing on extended reading hasn't been done. Not a good reason not to try.

>But determining proper letterspacing compensation for lack/presence of serifs is non-trivial.

Maybe, but we do it every single day along with determining proper height, weight and other compensations not only for lack/presence of serifs, but everything in between serif and sans.

Cheers!

>> To my knowledge there is no evidence that there is
>> any difference between legibility and readability.
> That's because the testing on extended reading hasn't been done.

Bingo.

David: Sure, but [somebody like] Kevin would have to do it empirically. So I guess that means optimal letterspacing would have to be empirically established for a given font, and then that font could be compared to a sans/serif "counterpart" (which has also been letterspacing-optimized).

hhp

A literature review by Alex Poole.

Kevin: If this improvement was added to van Rossum’s blurring, I believe that you would find that small features like serifs disappear more quickly in the parafovea.

The serifs disappear as features in van Rossum's gaussian blur too, but what is interesting is what they leave behind in terms of a more strongly defined extender. In the sans serif example, the extenders are badly compromised by the blur, whereas in the serif example the serifs are lost but the overall extender remains much stronger than in the sans serif. The more you have to lose, the more remains. I'd like to see the same kind of comparision with your S-CIELAB model

On a practical design level, I've been experimenting with ways to visually strengthen the ends of extenders other than with serifs.

Hrant: Ignoring anecdotal evidence is unscientific.

What do you understand to be meant by ‘anecdotal evidence’?

It's a phrase that we throw around pretty loosely here -- me included --, but if we're going to start making statements like this one of yours, then we need to be more rigorous in our usage.

"A literature review by Alex Poole."

Thanks, Christopher!

I try to be pretty rigorous. To me "anecdotal evidence" here specifically refers to the belief -resulting from trying and failing- that setting a book in a sans makes it harder to read.

hhp

Ah. Maybe you want a different term then. Anecdotal evidence usually refers to something that is either unverifiable or which cannot be be used to logically reach the conclusion presented. As such, true anecdotal evidence should not only be ignored by science but needs to be systematically excluded from scientific experiment and from conclusions based on experiment.

On a practical design level, I've been experimenting with ways to visually strengthen the ends of extenders other than with serifs.

Neo-cuneiform!

John, I don't see "anecdotal" as being so negative, and now that I double-checked it really doesn't have to equate to unverifiable* and certainly not untenable. But anyway, now that I've explained what I mean: I hope you agree; and feel free to suggest a term you consider better.

* Very difficult to verify, sure.

hhp

Hrant, the term "anecdotal evidence" is generally used as a term of disapproval for evidence that is of dubious value, or used improperly, as the wikipedia article on it explains. However, another thing is "observation reports," which science actually is obliged to account for in one way or another.

The fact that many expert typographers see sans as needing more leading is something that any complete reading theory should account for, eventually. To blow it off as "anecdotal" would be unscientific, in my view. Because you can show many people the same examples of leading, and get the same kind of reaction as to which is best.

This doesn't mean that people can't interpret their observations wrongly, but if so the theory needs to explain why and how the mistake is made systematically.

It looks like the construct "anecdotal evidence" is indeed negative. But "anecdotal" is fine, and very pertinent. So maybe "anecdotal reporting"?

BTW, it's very unfortunate -but telling- that Formal Science has driven the concept of the anecdotal into such a corner. I guess it's just not convenient to be reminded how narrow and limited one's research is, so Formal Science makes all anecdotalism negative...

hhp

The standard in science is of "inter-subjective testability" (Popper). In the case of experimental results, it means that others can and do repeat the experiment and get the same result. That has been around as a standard of scientific method since Robert Boyle in the 17th century. In the case of a one time observation, it is helpful if many people see it, if they are reliable witnesses, if someone takes a picture, etc, as evidence would be tested in a court of law.

In the case of this typographic stuff, it is all repeatable, and repeated. "Reports of expert observers" should work.

Hrant, I understand the feeling that science somehow misses things that might seem obvious to experience, but this is not a fault in science, though; rather, it is what distinguishes science as a method. I've often found myself thinking this or that experiment is failing to take into account something that seems to me important, but I'm usually at a loss to consider how the experiment could have taken it into account. Science has to proceed in its method, and while we can and should acknowledge that there are things about which science might not be able to enlighten us, we shouldn't expect science to be other than it is.

Regarding whether extended sans serif text is harder to read than serif, I think it will be difficult to first get past the issue people may not want to read extended sans serif text.

>this is not a fault in science

No, it is a fault in the science if it fails to account for observation. That's what is the task of science. If it plods along not addressing these issues, that's weak science. Don't blame scientific method. The real thing has imagination and verve.

“Reports of expert observers” should work. [Bill]

Maybe for reading it should be “experiential reports of engaged and motivated readers.”

This is different from “typographical craft wisdom.”

The trouble with both is we don't know how much of it is conditioning.

The other side of the conditioning issue is to what degree is typographical and type-design history guided by the the hidden hand of ‘what the reading body needs’ in material or hard-wired terms.

We are still, I believe, a long ways from a grounded understanding of the perceptual processing impacts of manipulating type-form micro-variables. Are our current measures on the order of trying to weigh gemstones on a bathroom scale?

>Are our current measures on the order of trying to weigh gemstones on a bathroom scale?

Generous of you. Helium atoms and truck scales come to mind.

Cheers!

>we don't know how much of it is conditioning

Nobody said science is easy. It's job is to sort out the causes. And yes, I also agree the proper measures—I would say tests—haven't been there. I've proposed some other tests, such as level of proof reading mistakes, and the extended reading tests with speed comprehension comfort variables that I mentioned above, but they haven't been tried.

Bill, I'm not blaming scientific method for anything, I'm defending it against intrusion of anecdotal evidence. There are things that science seeks deliberately to exclude, because including these things will distort results and make conclusions logically untenable. It isn't a fault of science to sweep anecdotal evidence into a corner, as Hrant suggested. Of course it is a fault if science excludes something that is verifiable observation, but that's not anecdotal evidence.

Nobody said science is easy.

No, that would typography.
Tests have proven that Arial outperforms other fonts, for fresher, cleaner pages, at less cost.

The term anecdotal evidence gives me the heeby jeebys. I most strongly associate it with the wide variety of snake oil medicine sold in the United States. All of these snake oils (airborne, homeopathy, acupuncture, chelaton therapy) have anecdotal evidence that supports that their treatment works. People try something, feel better, and report that the medicine works. Of course some people didn’t feel better, and we don’t report that. And some of the people that did feel better might have felt better had they done nothing. None of these treatments are more effective than a placebo under controlled tests.

Observations about the world are the first step of the scientific method. Of course observations should be used to guide tests, but the scientific method is a far more powerful truth finder than anecdotal evidence. Evidence about reading performance from measuring 30 people reading is far more likely to accurately represent reading performance than one person’s observation.

Btw, what is the anecdotal evidence about the point of serifs? It’s not clear to me if there is one consensus about the purpose of serifs, or if every typophiler has different anecdotal evidence?

Catich in his book "Origin of the Serif" gives a compelling case for how the serif came about. It had nothing to do with readability. Every other thing written in the "helps carry the line" venture is pure conjecture. The serif may or may not play a role in readability but I have seen absolutely nothing that convinces me either way.

http://www.amazon.com/Origin-Serif-Brush-Writing-Letters/dp/0962974005

Kevin: None of these treatments are more effective than a placebo under controlled tests.

Although apparently one study found that fake acupucture, in which the subject believed needles were being inserted into his back when, in fact, they were not, is more effective than a placebo pill. As a former naturopath whom I know put it: not acupuncture is more effective than not medicine.

Chris, the origin of the serif is independnet of any function(s) it might have in reading.

As a parallel: the contrast of thin and thick strokes in type has its origin in the characteristics of the broad nib pen, but that's independent of the effect of such contrast on spatial frequency.

>Observations about the world are the first step of the scientific method.

According to Popper the first step of scientific method is not to observe, but to recognize a problem, something to be explained. Then come tentative theories to explain what is unexplained. Then come observations and experiments to test the theories. The most powerful observations and experiments distinguish between competing theories, corroborating some and contradicting others. The new observations and experiments lead to new problems, and so on.

The difference in starting point is not trivial. Plenty of science done today is too "inductive" in method, as Popper calls it, thinking that accumulating observations without a strong focus of problems and theories will help. This is particularly notable in the social sciences, but also in psychology. It leads to a confusing welter of data that nobody can make sense of.

"the origin of the serif is independnet of any function(s) it might have"

I was not concluding otherwise. I was just trying to answer Kevin's question: "It’s not clear to me if there is one consensus about the purpose of serifs"

Purpose is viewable in many ways. If I say the purpose of the serif was for the expedience of the ancient Roman craftsman in painting or cutting letters, I am still answering his question. My point is that no one set out for the purpose of serifs to aid in readability. Their use has come about in history for other reasons so why would anyone assume that they surely must aid in reading text? There is no reason to assume serifs have this function and there is no clear evidence to show that they help readability even unknowingly. The tests have some minimal data that may make it possible to say certain roman typefaces may perform better at sustained reading than sans serifs but it has not been proven that the deciding factor is the serifs or some other difference between the typefaces such as contrast may be the reason.

>no clear evidence to show that they help readability even unknowingly.

I think there is evidence that points to the superior readability of serifs in extended printed text, even though it is not conclusive and is arguable.

We have had sans around for over 150 years, and for 50 in widespread use for text. Yet print publishers and readers have stubbornly resisted sans for extended text. You really have to look to find a newspaper or novel or even textbook set in sans. A it is not like it hasn't been tried, repeatedly. For example in the early 20th century a lot of Germans wanted to go to sans, rather than either to stick with blackletter or go to "antiqua" serifs. There was a whole modernist push for sans, and it has been accepted for display and short text, but not for extended text.

Now you might just put this down to the conservatism of readers, but this has been true across diverse cultures, changing fashions, and a long period of time. To me, this argues for a perceptual basis. I have mentioned some of the theories, which to me are plausible, but not proven.

And I do think that the right tests—and those I have proposed might well work—would reveal the superiority.

I could be wrong, of course, but it's not like there's no prima face case here. There is.

Bill, when Kevin wrote that ‘observations about the world are the first step of the scientific method’ I took him to mean something akin to Popper's ‘recognition of a problem, something to be explained’, i.e. that he wasn't referring to observation in the in-method sense, but in the more general sense of looking at the world -- reality external to the mind -- and seeing things that want explaining.

“Btw, what is the anecdotal evidence about the point of serifs?” [Kevin]

To my mind anecdotal evidence is things that get passed along in stories or short reports based on lived or schooled (conditioned) experience.

A Noordzijian might say serifs make an important contribution to the consolidation of the word image. Hrant might say serifs play an important role in the formation of robust and distinctive boumas. I might say serifs frustrate recognition-by-parts in word reading, where recognition by parts is seen. [Recognition-by-parts is a well-specified approach to recognition which contrasts recognition “through a single combining of features over the whole object” with a “separate combining [of features] over each of several regions of the object.” {Denis Pelli] In the case of words the regions are defined by letters.]

Of course serifs also affect what Kevin and others call personality, and I define as a font’s gestural-atmospheric niche.

Kevin, is this something you can work with in science?

>Btw, what is the anecdotal evidence about the point of serifs? It’s not clear to me if there is one consensus about the purpose of serifs, or if every typophiler has different anecdotal evidence?

The consensus of anecdotal horrors points to the fact that serif fonts contain more information than san serif.
But, apparently, one could much more easily answer the question by asking why sans is used at all.

Cheers!

I agree with Peter's theory that serifs help promote recognition by the assembly of parts over the whole word, rather than letter by letter, with a look up.

However, because serifs affect spacing and color, one can argue that they help put the letter in a better place for rhythm in the word, and so would help letter recognition even in a letter-first theory. How tight the rhythm is can be detected in Fourier transform pictures, so the rhythm in sans and in serifs could be compared this way. That test could be done now.

I have not heard of Fourier transform pictures so I gave them a brief search and came up with several different results. For the discussion, can you please define what you mean by "Fourier transform pictures" and "rhythm."

I am intrigued by your statement "That test could be done now."

Christopher, Peter Enneson who I've learned about this from, so perhaps he can give some links and/or pictures.

By the way, the other texts, of extended reading using timed SAT reading tests with different fonts and layouts can also be done, now.

Actually, I've got a lot more, now :)

ps What the Fourier transforms might identify are differences in rhythm between sans and serifs. It wouldn't in itself say that there are readability differences. I think that narrower sans will be more similar in rhythm to serifs than are wide ones.