Well, no. Now we know what we don't know, rather than assuming we do know something. And when we specify one font over another we can say honestly "Because I like it" rather than dishonestly "Because science tells us it's best".
Shouldn't the burden of proof be the other way around? Otherwise it follows that goblins and fairies aren't necessarily false beliefs.
You could imagine anything and claim it is not necessarily a false belief until someone proves that it is. Worse, you could imagine something with properties such that it is impossible to prove that it doesn't exist. At this point we'd all have to go around accepting your imaginary world of "not necessarily false" beliefs because we can't prove them to be false. What a terribly unstable situation that would be. Imagine, millions of people believing in one not necessarily false set of beliefs, and millions of other people believing in a different set of not necessarily false beliefs, and both groups claiming that the other must be wrong, each without method of proving the other to be false. What a dangerous way to reason about things.
No, what he means is that the author should have said: "It is a myth without proof" but not "a myth which is not true", which is what the last sentence says.
Myths can be true, like, you know, the existence of Nineveh (yes, it was a "Myth" and then it was proved true).
I just flipped a coin at my desk, which I'm covering with my hand. There's no evidence to support that the coin came up heads, so saying that it's heads is a false belief. There's no evidence to say that it came up tails, so saying that it came up tails is a false belief. Since the beliefs that the coin came up either heads or tails are both false, we can determine, by process of elimination, the the coin must have landed on its side.
The myth is a false belief because there's no evidence to support it. However, being a false belief doesn't prove the statement false, simply undetermined.
A survey of preference may not say much, but it would be something. It would be more helpful to have data on Kindle page views per minute by text display choice, controlled for other factors like book, typeface size, etc.
My biggest problem with my Kindle is that I can't even get that sort of data for myself. I'd love to know when I read, how long, which topics, and so on.
I'm almost certain with the latest it ware update this data is available somehwhere. Kindle now estimates time left in a book as a function of your past reading speed.
I feel like it's easier to recognise simple shapes than shapes with lots of extra lines distracting the eye. But that could just be rationalisation for a learned preference.
Reading is quite a recent development in evolutionary terms - the first evidence of writing is from about 10k years ago. I wonder if different people might learn to do it using different neural mechanisms, some of them helped by serifs and others hindered.
I wasn't finding the sans-serif font in which most of the article is displayed to be particularly hard to read, but when I got to the short blockquote in a serif font, I found it noticeably easier to read. The effect is persistent for me. Have a look -- do you not have the same reaction?
Even the both examples on the wikipedia article are sans-serif! Serifs have very little to do with ink properties. Probably more to do with ancient romans and their chisels if you want a technological explanation for serifs.
I believe it will always comes down to context. Some people love pixel fonts (like Silkscreen), but they generally don't have a place in larger sizes or on paper. Likewise, browsers handling serifs' mix of thick and thin line weights can lead to unwanted results. When all of the lines are the same weight, you run into less problems with individual letters. That said, they can have a complete lack of personality. I rather prefer a hybrid typeface like the one Portland's TriMet uses - http://portlandafoot.org/documents/IdTrimetCasestudy.pdf (on the headers of the print pieces and the words "Trip Center" in the signage on page 3).
I've been involved in publishing too (LaTeX, QuarkXPress) and...
It is a fact that for short texts (eg titles) Sans-Serif is actually much easier to read. Studies have been conducted: this is precisely the reason city names and exit names on the highways are using sans-serif. Security (and lifes) are at work here: not just reading speed. The last thing you want is people reading Serif on the highways, being distracted by all the fancy curves, and realizing too late it's their exit and suddenly switching lanes in a hurry and hence potentially causing accidents.
But for long strectches of text, like inside books? I don't know. My take on it is that basically the reading speed is the same but that Serif looks nicer from a typographical standpoint. There's hardly anything that beats the beauty of the ligature of a Serif font between a 'f' and an 'i' (i.e. when the dot of the 'i' is connected to the 'f').
I do prefer: sans-serif for everything except serif for printed medium, where the resolution is at least 600 dpi (no, retina ain't close enough yet).
So, yes, for printed books I still prefer Serif.
As a side note: this effectively means I cannot yet get a PDF which I find pleasant to read both on screen and printed.
And of course I configure my browser to use sans-serif fonts instead of serif fonts wherever possible.
P.S: as to my coding / shell font(s)... Pixel perfect fonts of course. Like the Proggy family or the Terminus family. I hate being distracted by the blurriness of anti-aliasing / sub-pixel anti-aliasing (RGB decimation). But I've got a very good eyesight.
It's not sans-serif vs. serif that makes short texts more legible: it's the x-height of the letterforms that tend to make sans-serifs more legible. Automotive signage typefaces are usually designed with larger-than-normal x-heights, which has the effect of increasing the legibility of a typeface from a distance without increasing the font size (which is why, for example, Arial looks HUGE in Word at 12 pt whereas Times New Roman looks perfectly acceptable).
There could certainly be a serif typeface used for shorter texts. A good example would be TheSerif from the FF Thesis collection.
Another issue with serif fonts on screens that you touched on: most traditional serif fonts were simply not designed to be viewed on a screen. However, a good counterexample would be Georgia, a serif font designed by Matthew Carter that was made specifically for the screen and used by a large number of websites.
One last thing: a reason why serifs are often thought to be more legible in large bodies of text is due to designers being careless about leading. With sans-serif fonts, you generally need to let the leading out a bit in comparison to serif fonts. On the author's site, for example, having the font so big and the leading so small tended to make me lose my position while reading (remember that without serifs to 'anchor' your eye, you're basically limiting your recognizable shapes to lines and circles).
> That’s when I came up with the idea of the Legibility Test Tool, an OSX application I built that offers real-time simulations of different viewing conditions (distance, fog, halation, and positive/negative contrast) during the design stage.
Legibility and readability are intertwined, but still separate criteria.
Sans-serifs have better legibility, that's why they are used for all kinds of signage. Serifs are better for long-form text, they give visual cohesion to lines, making it easier to follow the breaks. That's what the 'myth' is about, not that serif is always better than sans.
Ironically, when I read a book on my phone, I think what happens is the really short lines means there's no visual cohesion to save since it's all broken up anyways, so serif ends up feeling easier to read (for me).
Except that that is exactly the thesis that the article has rather convincingly demonstrated to be completely unproven.
It may still be true, but we don't now. And I am not sure that Bodoni will beat Gill on the "visual cohesion to lines" metric, Bodoni is quite static and vertcal, while Gill is rather flowing.
> reading speed or reading comprehension, which have no bearing on glyph recognition per se
This post is focusing on legibility - how efficiently can we recognize the shape of letters and words. The studies on reading speed and comprehension, which he dismisses without any backing whatsoever, are the ones that surface the difference between serif vs sans.
His arguments are 1) "legibility is poorly defined" 2) "reading has no relation to legibility" (?) 3) "it's complicated" 4) "paper X turned out to be bogus" (because the author cheated in an unrelated study). That doesn't add any information to me, just controversy.
He also ignores the fact that sans-serif type is currently way better for reading on monitors due to their low resolution. Serifs look great on a >250 dpi display. How many books have you seen lately printed in sans that are not for kids?
Just because you can't prove it doesn't mean it's not true.
I use a serif font on my Kindle. I recently bought a book that was somehow locked to only use sans-serif. I got a few pages in, and had to stop, it was terrible. I wrote the author, who apologized and claimed he was trying to fix it.
There are going to be some sans-serif fonts that are more readable than some serifs, but in my 25 years as an amateur fontophile, that is a rare exception.
The author rightly asserts that legibility is a very difficult metric to evaluate, and there are a large number of other factors in play here like conditioning, but I think the market is evidence enough that serif fonts are still the choice of book designers and typesetters long after the initial technical reasons for doing so have become obsolete.
I don't think the article was saying that sans-serif is more legible than serif, but that the studies showing the serif is more readable have issues with methodology and their results are not valid.
Just because you can't prove it doesn't mean it's not true.
No, but the body of evidence suggesting that there is little if any difference might. The best starting point I know if you want to explore it is Alex Poole's site[1], which cites a large number of sources and some critical commentary similar to the blog post introducing this HN discussion but with much more material.
... I think the market is evidence enough that ...
The "market" used to think the world was flat and if you didn't drown you were a witch. Common perception is a lousy substitute for properly conducted experiments, particularly in any usability-related field, where it's been found time and again that self-reported preferences don't necessarily correlate with measured performance.
> Just because you can't prove it doesn't mean it's not true.
It does mean you can't show it to be true, which means it's entirely rational not to accept it as true until that kind of evidence is presented.
What is really going on here is that you prefer serifs, without this being a global truth for everyone. That preference would continue to be a thing even if it were found that sans-serif were 1% more readable or something. Some people will still prefer serifs and be upset if they don't have them and in cases like ebook readers, that may be important to consider.
It isn't necessary to stretch to claims that serif fonts are objectively better in some way, in order to have a reason to use them or make them available.
And just because one particular serif font looks better to one particular person than one particular sans-serif font on one particular medium quite obviously does not mean it is true.
Nor do your amateur fontophile gut feelings matter--and, I should add, nor do mine, or anyone's.
> but I think the market is evidence enough that serif fonts are still the choice of book designers and typesetters long after the initial technical reasons for doing so have become obsolete.
Trivially true, but that's not at all evidence that book designers and typesetters pick them due to their readability[0], rather than, say, cargo-cult psychology.
For my part, I'd rather wait until there's evidence either way, rather than presuming that my personal preferences (which probably align mostly with yours, FWIW) are due to objective factors. That's what science and evidence are for, after all: to prevent us from convincing ourselves of things that really seem true, but aren't.
[0] Of course they pick fonts based on readability--they wouldn't use Zapf Chancery for Moby Dick--but I mean that they don't necessarily eliminate sans-serifs for such reasons.
One may be dismayed to find out that most things in design haven't been proved conclusively (in a scientific way). There are many reasons for this, but I think it all boils down to the difficulty of measuring how the human mind reacts when interacting with something.
What we have nailed down is the effects of those interactions: clickthroughs, purchases, and such. So, when you try to perform an experiment about legibility you usually apply a comprehension test and measure the speed at which the text was read.
In science you should isolate as many variables as possible, but I think in the case of typography there's just too many factors at play: the content, the medium and the reader introduce too much noise into the experiments.
I think that in the future brain imaging will advance enough to study how much people are able to concentrate with different typefaces. A researcher might give you an essay written in Comic Sans and see your mind wander more often than other subjects reading the same essay in Times.
If most things in design haven't been proved conclusively, that would just mean that design as a field doesn't take research seriously and doesn't insist on a strong empirical basis.
You don't need to wait for unspecified advances in neuroimaging. The basic research can be done 50 years ago. You measure the mind with behavior, end of story. That part is as easy as A/B testing. Just do it. If you go to neuroimaging, you should first know why it is relevant and worth spending the $$$.
And your feeling that the factors you suppose to be important are too many shouldn't ever stop us from looking into it. When we really get down to brass tacks on a subject, it's common to find that a lot of the factors we suppose to be important contribute negligibly to the final outcome. You just don't know which these are until you look.
I thought the "conventional wisdom" was that serif fonts were best for print, while sans-serif fonts were best for computer screens. Honestly, I've never heard anyone claim that serif fonts were all-around more readable.
One of the most-cited "authorities" on serif legibility is Cyril Burt, whose 1955 article [2] in The British Journal of Statistical Psychology (a journal he was the editor of) seemed to end the debate on whether serif typefaces are more readable than non-serif typefaces.
I think this article is referring specifically to print given the dates of the research.
I actually did a project on something similar in high school. I wrote a little program that would show people two passages in a random order, timing how long it took them to read the passages, alternating between one of four random typefaces. The four faces had two sans-serif faces and two serif faces, but I was interested in a different quality: whether monospace was easier to read.
Unfortunately, I ran into serious methodological issues which rendered the data completely inconclusive, but it was a fun experiment nonetheless.
The main thing I learned was how awesome computers are at saving me effort: I got all my data in one go while others had to conduct their experiments manually, one-by-one. Even counting how long it took me to hack together the program--and even in high school, using Java Swing, it did not take long--I still won out time-wise. More importantly, I traded boring data gathering for fun programming.
IN A SIMILAR VEIN, TEXT WRITTEN IN ALL-CAPS ISN'T ANY LESS LEGIBLE THAN TEXT WRITTEN IN lowercase; however, I'm sure you found the beginning of this sentence much more irritating to read than its ending.
It's about comfortableness, normality, custom. Serif evokes a relaxed, readable mood. Sans-serif is for quick points of information and eye-catchers. Capitals are for critical information. (And for us coders: mono-space is for code.)
There's nothing innate to the form to justify this is so, it's just how we've been taught/grown to interpret their presence in text. If the form doesn't fit the usual function, we feel a little uneasy (even if we don't realize it). I'm sure people who haven't been taught the distinction won't have this uneasiness... A bit related: https://xkcd.com/1015/
My opinion: People don't read individual glyphs, they read words. The amount of variety between different glyphs and other factors such as kerning which help the letters combine to make an elegant whole word. A classic example of how not to do this is blackletter; all the glyphs are so similar that at a glance, the text looks like variable length blocks of repeated characters. It is also very hard to read.
In my experience, many computer fonts also fall into this trap. By trying to establish a consistent theme or feel to the glyphs, they iron out all the letters' individual quirks that make them easy to distinguish and recognise, thus readability suffers.
I think in terms of readability, the sans/serif issue is largely irrelevant (as the article might suggest).
46 comments
[ 2.9 ms ] story [ 88.2 ms ] thread> "So before you go around claiming that serif typefaces are easier to read... It's one of many myths you (and I) have accepted as true, that isn't."
It isn't true? Prove it. Until then, you don't know whether the myth is true, and it is not necessarily a false belief.
The bit you elided said this:
"The embarrassing truth is, there's no solid research to back up that claim."
Shouldn't the burden of proof be the other way around? Otherwise it follows that goblins and fairies aren't necessarily false beliefs.
You could imagine anything and claim it is not necessarily a false belief until someone proves that it is. Worse, you could imagine something with properties such that it is impossible to prove that it doesn't exist. At this point we'd all have to go around accepting your imaginary world of "not necessarily false" beliefs because we can't prove them to be false. What a terribly unstable situation that would be. Imagine, millions of people believing in one not necessarily false set of beliefs, and millions of other people believing in a different set of not necessarily false beliefs, and both groups claiming that the other must be wrong, each without method of proving the other to be false. What a dangerous way to reason about things.
Myths can be true, like, you know, the existence of Nineveh (yes, it was a "Myth" and then it was proved true).
The myth is a false belief because there's no evidence to support it. However, being a false belief doesn't prove the statement false, simply undetermined.
My biggest problem with my Kindle is that I can't even get that sort of data for myself. I'd love to know when I read, how long, which topics, and so on.
I would love to see what others think over this. Which font do you prefer?
I feel like it's easier to recognise simple shapes than shapes with lots of extra lines distracting the eye. But that could just be rationalisation for a learned preference.
Reading is quite a recent development in evolutionary terms - the first evidence of writing is from about 10k years ago. I wonder if different people might learn to do it using different neural mechanisms, some of them helped by serifs and others hindered.
But when used with proper spacing, I like Sans Serif anywhere. Even HN is using Sans Serif and its quite usable.
Related info: http://en.wikipedia.org/wiki/Ink_trap
It is a fact that for short texts (eg titles) Sans-Serif is actually much easier to read. Studies have been conducted: this is precisely the reason city names and exit names on the highways are using sans-serif. Security (and lifes) are at work here: not just reading speed. The last thing you want is people reading Serif on the highways, being distracted by all the fancy curves, and realizing too late it's their exit and suddenly switching lanes in a hurry and hence potentially causing accidents.
But for long strectches of text, like inside books? I don't know. My take on it is that basically the reading speed is the same but that Serif looks nicer from a typographical standpoint. There's hardly anything that beats the beauty of the ligature of a Serif font between a 'f' and an 'i' (i.e. when the dot of the 'i' is connected to the 'f').
I do prefer: sans-serif for everything except serif for printed medium, where the resolution is at least 600 dpi (no, retina ain't close enough yet).
So, yes, for printed books I still prefer Serif.
As a side note: this effectively means I cannot yet get a PDF which I find pleasant to read both on screen and printed.
And of course I configure my browser to use sans-serif fonts instead of serif fonts wherever possible.
P.S: as to my coding / shell font(s)... Pixel perfect fonts of course. Like the Proggy family or the Terminus family. I hate being distracted by the blurriness of anti-aliasing / sub-pixel anti-aliasing (RGB decimation). But I've got a very good eyesight.
Please cite them. Good quality research on typography is unfortunately rather hard to find, and I've never encountered any study of that type.
It's not sans-serif vs. serif that makes short texts more legible: it's the x-height of the letterforms that tend to make sans-serifs more legible. Automotive signage typefaces are usually designed with larger-than-normal x-heights, which has the effect of increasing the legibility of a typeface from a distance without increasing the font size (which is why, for example, Arial looks HUGE in Word at 12 pt whereas Times New Roman looks perfectly acceptable).
There could certainly be a serif typeface used for shorter texts. A good example would be TheSerif from the FF Thesis collection.
Another issue with serif fonts on screens that you touched on: most traditional serif fonts were simply not designed to be viewed on a screen. However, a good counterexample would be Georgia, a serif font designed by Matthew Carter that was made specifically for the screen and used by a large number of websites.
One last thing: a reason why serifs are often thought to be more legible in large bodies of text is due to designers being careless about leading. With sans-serif fonts, you generally need to let the leading out a bit in comparison to serif fonts. On the author's site, for example, having the font so big and the leading so small tended to make me lose my position while reading (remember that without serifs to 'anchor' your eye, you're basically limiting your recognizable shapes to lines and circles).
http://ilovetypography.com/2012/04/19/the-design-of-a-signag...
Excerpt from above link:
> That’s when I came up with the idea of the Legibility Test Tool, an OSX application I built that offers real-time simulations of different viewing conditions (distance, fog, halation, and positive/negative contrast) during the design stage.
HN discussion:
http://news.ycombinator.com/item?id=3870237
Sans-serifs have better legibility, that's why they are used for all kinds of signage. Serifs are better for long-form text, they give visual cohesion to lines, making it easier to follow the breaks. That's what the 'myth' is about, not that serif is always better than sans.
> reading speed or reading comprehension, which have no bearing on glyph recognition per se
This post is focusing on legibility - how efficiently can we recognize the shape of letters and words. The studies on reading speed and comprehension, which he dismisses without any backing whatsoever, are the ones that surface the difference between serif vs sans.
His arguments are 1) "legibility is poorly defined" 2) "reading has no relation to legibility" (?) 3) "it's complicated" 4) "paper X turned out to be bogus" (because the author cheated in an unrelated study). That doesn't add any information to me, just controversy.
He also ignores the fact that sans-serif type is currently way better for reading on monitors due to their low resolution. Serifs look great on a >250 dpi display. How many books have you seen lately printed in sans that are not for kids?
It is called 'roman' for something which has nothing to do with monks or the middle ages.
Sorry: the fact that it has not been proved true does not mean it is not. Caretul. (He says so in the very last sentence...)
I use a serif font on my Kindle. I recently bought a book that was somehow locked to only use sans-serif. I got a few pages in, and had to stop, it was terrible. I wrote the author, who apologized and claimed he was trying to fix it.
There are going to be some sans-serif fonts that are more readable than some serifs, but in my 25 years as an amateur fontophile, that is a rare exception.
The author rightly asserts that legibility is a very difficult metric to evaluate, and there are a large number of other factors in play here like conditioning, but I think the market is evidence enough that serif fonts are still the choice of book designers and typesetters long after the initial technical reasons for doing so have become obsolete.
Meaning, the debate is still ongoing.
No, but the body of evidence suggesting that there is little if any difference might. The best starting point I know if you want to explore it is Alex Poole's site[1], which cites a large number of sources and some critical commentary similar to the blog post introducing this HN discussion but with much more material.
[1] http://alexpoole.info/blog/which-are-more-legible-serif-or-s...
... I think the market is evidence enough that ...
The "market" used to think the world was flat and if you didn't drown you were a witch. Common perception is a lousy substitute for properly conducted experiments, particularly in any usability-related field, where it's been found time and again that self-reported preferences don't necessarily correlate with measured performance.
It does mean you can't show it to be true, which means it's entirely rational not to accept it as true until that kind of evidence is presented.
What is really going on here is that you prefer serifs, without this being a global truth for everyone. That preference would continue to be a thing even if it were found that sans-serif were 1% more readable or something. Some people will still prefer serifs and be upset if they don't have them and in cases like ebook readers, that may be important to consider.
It isn't necessary to stretch to claims that serif fonts are objectively better in some way, in order to have a reason to use them or make them available.
Nor do your amateur fontophile gut feelings matter--and, I should add, nor do mine, or anyone's.
> but I think the market is evidence enough that serif fonts are still the choice of book designers and typesetters long after the initial technical reasons for doing so have become obsolete.
Trivially true, but that's not at all evidence that book designers and typesetters pick them due to their readability[0], rather than, say, cargo-cult psychology.
For my part, I'd rather wait until there's evidence either way, rather than presuming that my personal preferences (which probably align mostly with yours, FWIW) are due to objective factors. That's what science and evidence are for, after all: to prevent us from convincing ourselves of things that really seem true, but aren't.
[0] Of course they pick fonts based on readability--they wouldn't use Zapf Chancery for Moby Dick--but I mean that they don't necessarily eliminate sans-serifs for such reasons.
What we have nailed down is the effects of those interactions: clickthroughs, purchases, and such. So, when you try to perform an experiment about legibility you usually apply a comprehension test and measure the speed at which the text was read.
In science you should isolate as many variables as possible, but I think in the case of typography there's just too many factors at play: the content, the medium and the reader introduce too much noise into the experiments.
I think that in the future brain imaging will advance enough to study how much people are able to concentrate with different typefaces. A researcher might give you an essay written in Comic Sans and see your mind wander more often than other subjects reading the same essay in Times.
You don't need to wait for unspecified advances in neuroimaging. The basic research can be done 50 years ago. You measure the mind with behavior, end of story. That part is as easy as A/B testing. Just do it. If you go to neuroimaging, you should first know why it is relevant and worth spending the $$$.
And your feeling that the factors you suppose to be important are too many shouldn't ever stop us from looking into it. When we really get down to brass tacks on a subject, it's common to find that a lot of the factors we suppose to be important contribute negligibly to the final outcome. You just don't know which these are until you look.
Unfortunately, I ran into serious methodological issues which rendered the data completely inconclusive, but it was a fun experiment nonetheless.
The main thing I learned was how awesome computers are at saving me effort: I got all my data in one go while others had to conduct their experiments manually, one-by-one. Even counting how long it took me to hack together the program--and even in high school, using Java Swing, it did not take long--I still won out time-wise. More importantly, I traded boring data gathering for fun programming.
It's about comfortableness, normality, custom. Serif evokes a relaxed, readable mood. Sans-serif is for quick points of information and eye-catchers. Capitals are for critical information. (And for us coders: mono-space is for code.)
There's nothing innate to the form to justify this is so, it's just how we've been taught/grown to interpret their presence in text. If the form doesn't fit the usual function, we feel a little uneasy (even if we don't realize it). I'm sure people who haven't been taught the distinction won't have this uneasiness... A bit related: https://xkcd.com/1015/
In my experience, many computer fonts also fall into this trap. By trying to establish a consistent theme or feel to the glyphs, they iron out all the letters' individual quirks that make them easy to distinguish and recognise, thus readability suffers.
I think in terms of readability, the sans/serif issue is largely irrelevant (as the article might suggest).