Friday, January 18, 2013

The Serif Readability Myth

I've been involved in publishing all my life, and like many others I've always accepted as axiomatic the notion that typefaces with serifs (such as Times-Roman) are, in general, are more readable than non-serif typefaces (e.g., Helvetica). It never occurred to me that there was any doubt about the matter. Were the monks who invented serifs and other text ornamentations merely engaging in idle doodling? Weren't they consciously intending to increase the legibility of the important documents they were transcribing?

It turns out that, as with so many of the things we "know" are right, the idea that serif typefaces are more readable than non-serif typefaces simply isn't supported by the evidence.

At first, I scoffed at the idea that what everybody in the design world knows to be "obviously true" simply isn't. But then I happened upon the remarkable 1999 Ph.D. dissertation of Ole Lund (then of Høgskolen i Gjøvik), titled "Knowledge construction in typography: the case of legibility research and the legibility of sans serif typefaces" (download here).

It's impossible to do justice to Lund's stunningly thorough (and beautifully written) 287-page dissertation in a short space. You have to read it for yourself.

Lund undertakes an exceptionally detailed and critical review of 28 typeface legibility studies conducted between 1896 and 1997. He finds serious methodological problems in nearly all of them. Legibility itself is still poorly defined, even today, and is not well distinguished from readability. It turns out a surprising number of otherwise convincing "legibility studies" have been based on reading speed or reading comprehension, which have no bearing on glyph recognition per se. Reading speed is now known to be mainly a function of cognition speed, which varies considerably from individual to individual and is not related in any straightforward way (and possibly in no way) to typeface design. Reading comprehension is even further removed from type design.

Even if legibility is defined in terms of symbol recognition, one must decide how, exactly, such a thing is to be measured. Two common methodologies are variation of time of exposure (an attempt to measure speed of perception) and variation of distance ("perceptibility at a distance"). There are also methods based on type size. All have complicating factors. Harris [3] points to evidence showing that it is very likely that time of exposure methods as well as the variable distance method favor typefaces with relative large strokewidth, regardless of serifs. Type size is complicated by the fact that larger point-size fonts are not shaped the same as smaller point-size fonts, for a given font.

Designer George E. Mack, commenting on the concept of legibility in Communication Arts [5], said:

The basic concept is so tangled up in decipherability, pattern recognition, reading speed, retention, familiarity, visual grouping, aesthetic response, and real life vs. test conditions that contradictory results can be obtained for the same type faces under different test conditions.

Part of our "accepted wisdom" on the legibility of serif typefaces comes from research in cognitive psychology (most famously the work of Bouma[1]) around the notion that words are recognized not on a strict letter-by-letter basis but by the outlines or contours made around the word shape. This research has long since been shot down, as pointed out by Kevin Larson [4], who notes: "Word shape is no longer a viable model of word recognition. The bulk of scientific evidence says that we recognize a word’s component letters, then use that visual information to recognize a word."

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. However, Burt's statements about the supposed superiority of serif fonts turned out to be nothing more than idle conjecture dressed up to sound scientific. After his death in 1971, Burt's landmark work on the heritability of I.Q. was discredited (and his reputation destroyed) based on his use of nonexistent data and nonexistent coauthors. Rooum [7] and others found Burt's typeface research to be bogus as well (his coauthors on the 1955 typography paper seem to be fictitious). Today, anyone who cites Burt is citing discredited nonsense, basically.

So before you go around claiming that serif typefaces are easier to read than sans-serif typefaces, you might want to do a little checking around. The embarrassing truth is, there's no solid research to back up that claim. It's one of many myths you (and I) have accepted as true, that isn't.


References

1. Bouma, H. 1973. "Visual Interference in the Parafoveal Recognition of Initial and Final Letters of Words," Vision Research, 13, 762-782.

2. Cyril Burt, W.F. Cooper, and J.L. Martin. 1955. 'A psychological study of typography'.
The British Journal of Statistical Psychology, vol. 8, pt. 1, pp. 29-57.

3. Harris, J. 1973. "Confusions in letter recognition." Professional Printer, vol. 17,
no. 2, pp. 29-34

4. Larson, Kevin. 2004. "The Science of Word Recognition."

5. Mack, George E. 1979. 'Opinion/Commentary'. Communication Arts, vol. 21,
pt. 2, May/June, pp. 96-97

6. Poole, Alex. 2012. "Fighting bad typography research."

7. Rooum, Donald. 1981. "Cyril Burt's 'A psychological study of typography': a reappraisal," Typos: a journal of typography, no. 4, pp. 37-40. London College of Printing

32 comments:

  1. Thank you for this! I've always had trouble reading Serif fonts. My eyes glaze after a few lines. Figured the problem was me since Serif fonts are supposed to be more readable.

    P.S: Discovered your blog through Alexis Grant.

    ReplyDelete
    Replies
    1. Ditto to this (except that I discovered your blog when you followed me on Twitter)! I've always thought that I was the one who was weird because san serif fonts are MUCH easier on my eyes. I might never have seen Lund's study if not for this post so thank you very much for doing it!

      Delete
  2. Interesting. I've believed the serif readability claim as long as I can remember. I'll have to rethink it; but it's so ingrained, it will be hard to let it go. Thanks for the blog and tweet.

    ReplyDelete
  3. The PhD thesis that you cite was done at the University of Reading. It's a shame that the fulltext seems to be behind a paywall.

    Alex Poole also has a good write-up on the matter: http://alexpoole.info/blog/which-are-more-legible-serif-or-sans-serif-typefaces/

    ReplyDelete
    Replies
    1. Richard1:12 PM

      You do have to register, but the download is free (I just downloaded it).

      Delete
    2. Am I the only one who thinks that it's hilarious that a PHD thesis on this topic was done at the University of Reading? I KNOW it's pronounced "Redding" but still...

      Delete
    3. Carmen, I thought it was hilarious.

      Although I thought this paragraph was satire and laughed a full minute:

      "It's impossible to do justice to Lund's stunningly thorough (and beautifully written) 287-page dissertation in a short space. You have to read it for yourself."

      Delete
  4. I have read (can't remember where) that serif is better for print and sans serif is better for screen. I find this true in my humble opinion.

    ReplyDelete
    Replies
    1. Ditto, on all points... By the way, the frikking captcha is frikking impossible to read...frik.

      Delete
  5. "Weren't they consciously intending to increase the legibility of the important documents they were transcribing?"

    As a calligrapher, I have two hands and several hundred years for you:
    Carolingian
    Textura Quadrata

    Carolingian was chosen and popularized for readability across the entirety of Charlemagne's empire. He wanted the former-Romans to be able to read it, and the monestary with the most legible writing's style was adopted as the standard, empire-wide.

    Go forward several hundred years, and we hit Textura Quadrata. It's compact. It gives a wonderful texture to the paper with all those equally spaced vertical lines. It's what most people know of as "gothic writing." It's nearly illegible when done properly. The spacing between letters and the spacing inside a letter (like n) is the same. The difference between a "u" and an "n" is whether there's a hairline gap between the diamonds at the tops or the bottoms of what could easily just be "ii" (except that has a gap at top and bottom). It gives a lovely effect from a few feet away. Those vertical lines lend themselves very well to tossing even more ligatures in (like æ and œ, but hey let's even make one for "de" or "do", those can share an edge too!).

    A more legible Textura was eventually made -- it has no serifs at the bottom of the letters and is called Textura Preciscus...because you can actually kind of read it.

    ReplyDelete
  6. Anonymous10:07 PM

    I did a survey of the arguments pro and con for a publication I was redesigning for the elderly. the material I reviewed went back to about the 1950's.

    What I found was that a number of tests were made to see what type reproduced well on a cathode ray tube. Times was compared to a sans face. Of course, the sans face was better displayed and read on a cathode ray tube. It's probably still better on current monitors. Sans is generally suited to rectangular pixels.

    Times, I understand, was created to squeeze in the most amount of type on a newspaper page. It's almost, not quite, a condensed face. So, it can be hard to read for people whose eyesight is bad or deteriorating. People who may have to depend on the screens of electronic "readers". A bias in health care circles has therefore built up against serif faces based on a) old studies of cathode ray tube monitors and b) a poor choice for a serif face to begin with.

    Consequently, I designed my publication for the elderly to be set in a good serif face, at a decent size, with good leading. It was tested by the intended audience and passed with wide approval. In use now for over two years and no one has complained. In fact, my next publication will be for an Age Friendly cities project of WHO and will use the same type specs or very similar.

    Sans faces have their place, but not in large amounts of body type, IMHO.

    ReplyDelete
    Replies
    1. Anonymous11:41 AM

      I'm not sure how you got: 'sans better on monitors/CRT's', therefore 'serif better in print'.

      Your test didn't involve controls, it was simply an approval-test, not a preference-test.

      The most advanced tests, as far as I understand them, is that sans fonts are at least not worse in print than serif.

      Delete
  7. It'd be interesting to do a more thorough evaluation of the history of typeface design and why it changed over time. Here's a great article by Jim Felici about one step in that evolution:
    http://www.creativepro.com/article/paper-changed-type-design

    Practical concerns are intertwined with aesthetic concerns throughout the history of printing. Perhaps sarifs have something to do with how the ink lays on the sorts and transfers to paper? It might prevent loss of the letter shape if the lighter portions of a letter are poorly transferred. Just a guess.

    Could it have anything to do with the material the sorts are made from, and difficulties in carving them,
    or how they wear out, and how they tend to break? A sarif might add a bit of strength to an otherwise breakable tip of a letter. Just another guess.

    Or perhaps it was simply a matter of manuscript style being copied into blackletter fonts, and then over time they gradually had more and more ornamentation stripped away. Streamlining is a very common theme in the design of pretty much everything. Yet another guess.

    And I like the comment about aging readers, as I'm one myself. It's amazing how few people below the age of 40 are aware of what's going to happen to their vision in their 40s when the lens begins to harden. In my own experience sans serif letters are more likely to blur together than serif letters, especially the lower case L next to any other vertical lines (as in "billboard" which for me with a sans serif font looks like B-blur-B-O-A-R-D). Although this might also have to do with the particular axis of the astigmatism in one of my eyes. But perhaps a lot of professional type designers did their most influential work when they were older? So there's still another guess.

    My own imagination in these few minutes has probably barely scratched the surface on potential influences on font design over the ages. The point of all this is that font design in both the past and the present probably defies any simple one-dimensional analysis like "legibility."

    I'll also add that I don't think there's an easy answer on the best fonts for screens. The technical questions that spring to mind for me are: what's the resolution? Is anti-aliasing being used? Is the anti-aliasing technology good, or sucky?

    tom

    ReplyDelete
  8. Anonymous10:18 AM

    There's been a lot of research for legible road signs, e.g. the Clearview font(s).

    ReplyDelete
    Replies
    1. Anonymous12:08 PM

      As long as the roadsign is placed a goodly distance AHEAD of the turn-off, and not AT the turn-off, I don't care what tyeface is used. And make them big. We have bilingual street signs which can be a real PITA to read in the dark.

      Delete
  9. Anonymous10:20 AM

    Having just passed the Captcha test after the third try, it seems that Captcha-powered research into this topic would require only about three weeks programming and two days run time.

    ReplyDelete
  10. You lost me at the point where you decided to accept their measuring glyph recognition rather than actual reading speed. I understand the difficulties the other way -- but fast readers don't even see the individual glyphs. What matters is reading speed, and measuring something else just confuses the issue.

    ReplyDelete
  11. Anonymous1:37 PM

    Wouldn't be surprised to find differences in printing technologies on various papers and with various ink technologies cause readability issues, and effect outcomes. Goes without saying CRT, LCD, LED screens and other projection methods would change results.

    Implementation of each various display technique, as well as any bias, distortion and non linearity can alter recognition.

    The worst thing about web based design is exemplified by the NYT which has a hodgepodge of fonts which distracts and exhausts the reader. You'd think of all the folks the venerable NYT would know better.

    ReplyDelete
    Replies
    1. Agree with you on the NYT on line. Their new sans serif turns me off after the third headline.

      Delete
  12. Nice post, and thanks for the live links to the papers and blogs cited, which make for interesting reading too. As an advertising writer who has worked in direct mail for decades writing long four-page letters and text-heavy brochures, the subject of reading ease has always been of great interest.

    I had generally subscribed to the "serif is better" school of thought because (1) that's what I was told, (2) because I found I could more easily read longer text paragraphs when set in serif than sans—perhaps because of my preconceptions, and (3) because it seemed to be supported by the fact that most printed books containing primarily text—novels for example—were all set in serif. Yet I was always puzzled by the exceptions I ran across consistently: not short headlines, or road signs, but in solid text passages.

    I've come to realize that (for me) readability is much more complex an issue than serif vs. sans-serif. I believe that the most readable text is that which hits the sweet spot where font, kerning, line leading (very important) and point size all complement each other in harmony.

    Speaking of fonts, it's interesting that the "compose comment" box here uses Helvetica (or Microsoft's Arial) but the "Preview Your Comment" popup box displays the draft text in Trebuchet font.

    ReplyDelete
  13. I'd only been taught that we read serif fonts more easily because most people grow up reading newspapers throughout our lives, of which 98% are printed using serif fonts. Not sure if that was ever true, but now that more and more people get their daily information from computers, much of which is set in sans serif, that theory also goes out the window.

    I don't think I've ever seen a printed newspaper set in all or mostly sans serif type, but it would be interesting to see how such a printing might change that more traditional reading environment.

    Thanks for the article. Good food for thought.

    ReplyDelete
  14. Yup. I recently asked a bunch of dyslexia specialists who work with students in higher education and got pretty much the same answers "there's no clear evidence base" and indeed many of the people who'd wanted to research this had given up due to difficulties with methodology and overcoming "accepted wisdom".

    Apparently much research is also based on self-reporting and people often find familiarity a big factor which wasn't accounted for. In the 90s Courier was deemed readable cos it was one of the most prominent fonts, then came Arial, then Comic flipping Sans and so it continued.

    Line spacing and word spacing are bigger factors in readability and legibility than the font-face itself. I saw the preliminary research output from a study which also indicates font-size is much much much more of a thing than almost anything else for dyslexics at least, with 50% getting objectively better reading speeds at 18pt or higher. And given your remarks on reading speed and cognition it's evidently way more complex than even that.

    One thing which makes me especially angry is people selling "dyslexia friendly fonts" which again have only anecdatal evidence behind them which in many cases would detriment readability for visually impaired people whose needs are funnily enough different from those with specific learning difficulties like dyslexia or visual stress. How to screw over a majority for something that probably doesn't even work for a minority and claim you're doing access.

    Like many things in the "disability" market there's a lot of scammy nonsense and a lot of "rule of thumb" sold as FACT or TRUTH when in reality I think the priority should be flexibility and enabling empowering people to make their own decisions on access.

    I once wrote a document in Times New Roman 9pt and a dyslexic person who normally can't read TNR asked me what I'd done the document in. I'd used LaTeX which is a typesetting and publishing markup system which spaces out documents much more nicely than the usual word processors. She was amazed at the legibility of my document.

    ReplyDelete
  15. For the record, professor Ole Lund still has a considerable engagement with the Faculty of Computer Science and Media Technology at Gjøvik University College (Høgskolen i Gjøvik). Lund developed and leads a beautiful and now very well-renowned bachelor program in Media Design ( http://english.hig.no/study_programmes/bachelor/bmed ) and took the initiative to a new master program in Interaction Design (which Lund originally named "Master in User-Centered Media Design"). (For the sake of disclosure: I am the dean of the faculty).

    ReplyDelete
  16. "Burt's statements about the supposed superiority of serif fonts turned out to be nothing more than idle conjecture dressed up to sound scientific."

    I never believed the readability claims of pro-serif folk. It struck me long ago that “comic” was the most readable font in the world and I create all my course material in that font. My students mock me.

    What would be a valid test of font readability? Data gained from dyslexic studies might be applicable only to dyslexics. It might be that for 90% of the population, Times New Roman (or some such serif font) is as easy to read as Ariel (or some such sans-serif font). Could we measure how much brain activity it takes or how many calories are burned? Could one font lead to more fatigue?

    "Word shape is no longer a viable model of word recognition. The bulk of scientific evidence says that we recognize a word’s component letters, then use that visual information to recognize a word."

    When we are trained to read in our native language, most reading is essentially sight reading or memorization of words. After sufficient reading, we grow to recognize most words. My guess is that font has no impact on readability, but some fonts might take slightly more work, some slightly less.

    ReplyDelete
  17. Anonymous12:19 PM

    What about the other accepted wisdom, that sans-serif is more readable on screen? Is there evidence to back it up?

    ReplyDelete
  18. Nothing in this post shows that it's untrue that serif fonts are 'more readable' (in whatever circumstances we choose to care about). The proposition 'all the scientific studies supposedly showing that X is true are flawed' doesn't imply 'X is untrue'.

    I guess if you believe there is no such thing as readability of fonts, by definition any proposition concerning readability is untrue. But some fonts clearly *are* hard to read so it's silly to discard the whole concept.

    Extrapolating from my experience, a 500-page novel printed in small sans would be a lot less readable than small Garamond...

    ReplyDelete
    Replies
    1. Thanks for this comment, this is exactly what came to my mind!

      Delete
    2. Anonymous11:52 AM

      If we're talking about small sizes, the 'serifs-guide-the-eye' argument from days of yore may indeed carry some weight.

      Otherwise...

      Delete
  19. Did the study factor in/out the use of mixed case vs. all upper case?

    ReplyDelete
  20. Thanks for another article on this. Note the date on the research. It's been around, but even I was shocked to find it out a few years back, and made a (small) point of mentioning it in my book.

    http://4ourth.com/wiki/Readability%20and%20Legibility%20Guidelines

    However, as some have pointed out, the data is more fuzzy than denying serifs have value. I am still comfortable saying "Serifs may or may not help with readability," and letting other factors be considered first. I'd hate to see this get overblown and now everyone says Helvetica is the most readable typeface.

    P.S. The blog post itself seems to conflate readability with legibility. They are different and that difference can matter.

    ReplyDelete
  21. I appreciate your overall thesis, that we should watch out for group think in published research. However, as with your attacks on Ray Kurzweil, you seem to be creating an argument to fit what you want to say. If you want to really change minds, as opposed to just flaming, then it's important to build up an element of trust among your readers. They should be able to trust that basic elements of your arguments will hold up under scrutiny.

    For example, in this article, you spend considerable prose arguing that choice of font simply doesn't matter. For example, you write: "Reading speed is now known to be mainly a function of cognition speed, which varies considerably from individual to individual and is not related in any straightforward way (and possibly in no way) to typeface design."

    At the extremes, that's simply not true. Mackenzie pointed out Gothic fonts. Others pointed out comic sans and small sans. I would add small caps to the list. There are a host of cases where the evidence is overwhelming that there are attractive fonts that are hard to read.

    By Occam's Razor, we should assume that things true at the extremes are also true at milder degrees. It's the simpler theory. For example, I expect there is a gravitational attraction between my chair and my desk. Yet, the effect is so small that there will never be a direct measurement of it.

    ReplyDelete

Add a comment.