The Line Length Misconception
26 Comments
Classically-trained designers, and really every professional designer, should know the old adage that long line lengths can have a counterproductive impact on readability. The trusty copy of Bringhurst's authoritative The Elements of Typographic Style makes this rule fairly cut-and-dry:
Anything from 45 to 75 characters is widely-regarded as a satisfactory length of line for a single-column page set in a serifed text face in a text size. The 66-character line (counting both letters and spaces) is widely regarded as ideal.
So naturally, when designers work on the web, they're keeping this rule in mind. Consequently, it's become a best practice on the web to keep line lengths below 75 characters, and this best practice has been the source of dissent against movements for things like variable-width (fluid) layouts and the like.
But research doesn't support this claim, at least on the web. Instead, users are able to read significantly longer line lengths on the web, and it actually increases efficiency and comprehension.
The Research
The first research I could find relating to this topic went all the way back to 2005. The study,The Effects of Line Length on Reading Online News, looked at how well college students read news on the screen at different line lengths. They tested line lengths of 35, 55, 75, and 95 characters. The study came to two important conclusions:
- Reading speed was highest at 95cpl, and lowest at 35cpl on screen.
- Reading efficiency was again highest at 95cpl.
- They found that line length no impact on comprehension on the screen
These days, it would be a bit nicer to have a wider array of line lengths, perhaps going up to 115 or 135cpl, but this is a useful study.
Guesswork
In a very brief and informal search, I came across a healthy number of sites that had line lengths around or above 100 characters that seemed quite readable, including my own blog. So what makes the screen significantly different from print that would impact comprehension in this way?
A few guesses, with absolutely no basis in research, the third being my favorite:
- We've been conditioned to longer lengths by websites that have tended to stretch the top end of this limit, as opposed to newspapers, which tend to stretch the lower end.
- The active lighting of a computer display is more conducive to longer line lengths than passive paper.
- Perhaps line lengths that are out of proportion with their medium are more difficult to read. Print tends to have vertical layouts, which would naturally correspond to shorter lines, while we tend to use widescreen monitors these days. Perhaps the widescreen monitors are more conducive to the readability of a proportionally-longer line length.
If the range of line lengths had been wider in the study, and if we had an updated study since the widescreen monitors had become popular, we could really see what might be impacting the difference.
Best Practices
So should you really be limiting your line length to 75 characters? This research suggests you shouldn't. Users will be perfectly fine reading longer columns of text.
What is the new standard? Tough to say, but 100cpl seems to be within the range of feasibility. There may be a good opportunity for some new and more thorough research in this area that could offer some valuable new insight.
It’s very interesting to me that there haven’t been any conclusive findings in this regard. Looking through some of the older studies cited in the study, it looks like you could cherry pick studies to show just about anything you want.
I might be a little cautious about tossing out the perceived wisdom. It’s true that the line lengths increased reading speed, but it doesn’t look like there are any meaningful conclusions to be drawn about optimal lengths for comprehension, efficiency, or even user preference. Of course, there weren’t any significant negative effects found either.
In the end, it’s nice to know that I probably don’t need to be so anal about line lengths.
PS: I really wish they’d published their full data tables.
Thanks for questioning the standard wisdom of best line lengths. One of the interesting things, based on some research (http://www.humanfactors.com/downloads/feb03.asp), is that there may be a discrepancy between the line lengths that are most efficient for reading and the line lengths that people prefer - longer lines are more readable, but people prefer shorter lines. Do you know of any other research which tests user preferences for line lengths?
Very nice to know, Jackson. I’d always worried about that 75 character limit - glad to hear it’s not so relevant on the web.
I was always taught that the line-length rule was to help readers easy find their way back to the next line, and if this is indeed its purpose, a few other factors might come into play:
1) How long are your paragraphs? A strong visual cue for finding the “next line” is its distance from the beginning or end of a paragraph. The further you get from one end of a paragraph, the harder it seems to repeatedly find the next line. Inversely, it can be difficult to deal with lots of super-short paragraphs that stripe up the page. Stretching line-lengths can help keep long paragraphs to a healthy 4-8 lines, while reducing them might help bump up staccato ‘graphs to the same line number.
2) How are the lines spaced? Compare the reading experience on Victoria’s (excellent) link to an article on Ars Technica: I definitely find the latter ‘preferable’, and (anecdotally, of course) don’t notice a drop-off in reading speed or focus. Spacing could be a third factor that interacts with the other two.
It’d be fun to see some research done with mixed factors like these, I imagine there are some semi-ideal ratios out there to find…
Interesting topic. Personally, I find long line-lengths to be more intimidating, for lack of a better word. So, if I come to such a page, there’s a much better chance that I’ll bounce out and decide not to read it at all.
This is probably related to my reason for preferring shorter line-lengths: scanability, for lack of an actual word. I find it much easier to scan a page of content that has more narrow columns of content. I can jump around, re-find my place, and quickly pick out sections of content that are most important to me.
If I’m committed to reading every word regardless of line-length (as was the assignment in the study you reference), I would expect that speed, efficiency, and comprehension would probably be about the same for me. But, frankly, I rarely ready every word—especially on the web—so I’d seek out a format that I can more easily scan.
Very interesting.
I’d love to see more research how leading affects readability and if the same results hold true for older generations. There are so many factors involved that I think it’s hard to make a blanket statement.
Along the lines of what Doug Avery mentioned above, I’ve studied reading and eye movement in cognitive science and design classes. The motions most often tied to line-length ideals are saccadic in nature: http://en.wikipedia.org/wiki/Eye_movement_in_language_reading give a better summary than I could, with graphics. For both leading and line-length, studies of eye movement and how we read word shapes indicates you should not interrupt saccades. Instead set type at saccadic groups, like three saccades across a line. That happens to correspond to the 64-84 character line length—roughly 6-8 saccadic movements. At line break, it’s one smooth saccadic movement to find the next line, based on perpheral v. foveal vision (which is also why justified text becomes difficult to read—no perihperal cues).
Lots of geek-out stuff in there, really—probably too much for sane design discussions; leave cones, rods and hot/cold color palettes for later, for sure. However, I do recall a lesson about short sentence lengths on the web changing things up a bit. Same with text legibility and contrast. Studies on the ACM somewhere (shot not look them up again) indicated that eyestrain and/or dyslexia is more than often mapped to irregular saccadic movements.
Our sources for line-length design conventions may lack guts under scrutiny, but understanding vision adds back some of that grounding.
It’s as much about comfort as it is about effiency and speed.
Lots of great things to respond to:
@Jackson: Yeah, definitely agreed that this is far from conclusive, but I think it’s more valuable to note that the traditional limits may be limiting without reason. Experimentation by designers in this area should be welcomed and embraced.
@Victoria: Obviously I should do a bit deeper dive into the books to see what’s out there. I’ll see what I can come up with.
@Brian: There’s probably (my hunch) a significant impact from the role of the text being set. Text to be scanned probably calls for more fragmentation (more, shorter paragraphs and a shorter line length), while text to be absorbed—like an essay—may benefit from these longer lengths. It’d be an interesting factor to bring in, and things like journalism kinda fall in-between.
@Tiff: There’s a reason you’re one of my favorite people. Good food for thought there. I’ll have to look this stuff up, but if three saccades is 64-84, why not four saccades (which napkin math suggests is 85-112)?
@Wolf: True, but there may be something to the notion that efficiency (especially when it comes to comprehension) is dependent on comfort, especially over a longer span of time. If you’re able to quickly comprehend something, does that imply a level of reading ease/comfort? If so, providing text that is most quickly and easily understood is tantamount to providing the most comfortable text setting.
Thanks for this piece!
Interestingly, Google Reader goes up to about 115cpl. And, for what is worth, I usually enjoy text in Google Reader more than I do at the original source.
My reading experience also agrees with what Doug says about paragraph length. Longer lines can make text look weird when paragraphs are short (and when paragraphs are separated by the full height of one line, as is the prevalent convention on the web).
Cheers!
Nice article! This is a topic I’ve been interested in for a long time and would love to see (or conduct) some more testing someday. As some have mentioned above, the toughest thing here is to assign any rigid rules because the topic is so heavily tied to the design of the text. Typeface, size, leading, color, all play a big role in making things legible.
One other thing I didn’t see mentioned that could be a possible reason people are able to read longer lines on screen without a drop in efficiency: distance to screen. We’re generally farther away from a screen than we would be holding a book at half-arm’s length, meaning that our eyes have less distance to travel (and thus take less time to travel) to the beginning of the next line.
@Jason: good call on the object distance point.
Based on my current understanding of the (limited) research I’ve read based on Tiff’s reply, though, the way we read would be impacted less favorably by the distance to the object than one might expect.
My understanding is that the time involved in the eye movements made while reading is fairly negligible, while the fixation of the eye on a set of characters is really what’s taking up the time.
Some interesting points here. I’ve often wondered about the line-length translation from print to web. Like other ‘rules’ of print design - e.g. consistent vertical rhythm - they are far less relevant (if at all) to the web due to other factors - the web is NOT the same as a book or a piece of paper, people!
For example, I can think of a couple of reasons why line-length MAY be allowed to get longer on the web:
a) Writing *tends* to be in shorter paragraphs, broken up by blockquotes, lists, and other visual clues. Links, bold/italic text, and other graphics alongside the copy can all aid the eye when returning to the beginning of the line.
b) For some readers, the distance between the top of the viewport (think of this like a ruler or bookmark) might be much closer to the text itself than the top edge of a book (if no other aid is available). This can aid visual location, and scrolling down a page is natural and easy whilst reading a long article.
Interesting article but I’m not convinced. The line length rule serves 3 purposes:
1. To help the reader easily find the next line of text.
2. To cause less distractions.
3. To not tire the reader.
All 3 improve readability. Speed and efficiency aren’t the main priorities of the rule. Speed actually doesn’t matter at all and faster reading could affect comprehension.
I think test is flawed because it’s measure the wrong attributes.
I’d have to agree with Antonio on this one. The tests from provided links were looking at how fast one can read through text, not what for readability or eye strain.
That said, this is an area that could probably use some more research/study. The medium (paper vs lcd) may affect this rule, but only in the sense that the light coming off the screen will burn your eyes out.
Plus research should never stop.
I can’t help but wonder as well if there is a difference between serif and sans-serif typefaces. Even Bringhurst’s book specifically mentions serif type in the rule.
I did my undergrad thesis on a closely related topic, and all of the serious research on this shows a clear, unquestionable relationship between line-length and line-height / spacing.
You really can’t have a productive discussion of line length without discussing height and spacing. This topic has been explored so extensively in real, peer-reviewed and published academic research studies that you can’t just say “Users will be perfectly fine reading longer columns of text.” It’s like saying “it’s perfectly fine to use Flash on any site you want” and leaving it as a blanket statement without even mentioning accessibility, seo, penetration, etc. It’s so vague and incomplete as to be misleading and effectively would be bad advice if followed literally.
To readers who want to know more, get on an academic research database and search for ‘readability’. There’s plenty of information out there waiting for those willing to search for it.
Additionally, it probably doesn’t make sense to base your typographic decisions on a target number of saccades. The amount of information that can be taken in between each saccade varies so widely from person to person and across age, reading experience, lexile level of the text, etc, that it’s not really plannable. It’s useful to understand the concept of saccades as the physical explanation for a lot of typography/readability research findings, but in practice you should step up one level and use the findings rather than the underlying explanation.
@Anon Re: height and spacing, I’m with you. While undoubtedly it’s true that the online reading experience differs dramatically from reading printed text, I still find that longer lengths confuse and/or tire me. The ideal length is still probably related to both the type size and the leading, so that given those parameters on a website, there will be a length that works for most people.
I imagine there’s some kind of “golden proportion” involved in there somewhere…
Unfortunately, most of the real reseach on this topic isn’t on blogs and isn’t Googlable; you have to read actual published papers, which requires a trip to the library and the use of a photocopier.
In other words, this post’s reliance on a single study is unwise.
Joe: Absolutely, but you’re missing that the purpose of this post was to expose that there is dissent on the issue, not to make any hard claims.
However, it’s a bit unsatisfactory that the relevant research isn’t terribly accessible to the people who need it most.
So today, we opened a bit of a research project that might help the situation a bit. Head on over to http://readability.viget.com where we have a timed reading sample and some straight-forward comprehension questions.
Maybe we can come closer to some decent conclusions.
“People who need it most”? You make it sound like emergency relief for disaster victims.
I don’t find it “unsatisfactory” that peer-reviewed research is published in actual journals. You seem to think everything should be Googlable. What if not everything should?
Really, you seem to be insisting that “dissent” disproves reputable scientific claims because you can’t search for and read those claims from your iPhone.
I’m not even going to bother with your pointless little demo. Hire a Ph.D. to run a real controlled experiment and then we’ll talk.
Joe, It’s unfortunate you chose to pepper your remarks with ad hominem, straw man, and red herring arguments. I won’t reciprocate.
It’s certainly a belief of mine that research should be readily available to the industries it impacts as much as possible. For those who work on the web, “readily-available” is largely synonymous with “googleable.” If we were talking about a field dominated by academia, it would, I argue, be a different story.
I assert that a decent estimate you can put your hands on, as a practitioner in this field, is far better than the best-laid research you can’t. What we’re offering is an attempt at a decent estimate, and that’s all it claims to be. I welcome your dissent as well as your support.
Really interesting read, I agree with Antonio above though on his three points the test should be about comprehension not speed.
While research may indicate that users can read long line-lengths, it doesn’t mean that they necessarily enjoy reading long line-lengths.
I can read small print on a cereal box, but I sure don’t enjoy it. (But alas, I have nothing better to do at the breakfast table :)
Actually, there was some research done by a UK institution years ago showing that line lengths below 62 and over 110 didn’t perform as bad (can’t recall the resource off the top of my head). It found that 72-96 was optimal for speed and retention. The reason I recall the study is that it was subject for debate during a project I worked on at Cornell. So, we decided to test it ourselves.
In our testing, which was done originally in 2002, we found that 72-96 had the best reading and retention rates. We saw reading and retention drop off around 110-112 characters per line (12-14px font size 1.125-1.5 line height). We’ve done a number of studies since, where line length wasn’t the focus of the study, but a secondary measurement, and we continue to see the same results today.
The crux of the issue is that people say they “prefer” to fill their browser windows, but when line lengths get over 110, they’re actually doing themselves harm, or creating sub-optimal situations.
Next entry: Sketching in the Kickoff Meeting
Previous entry: Prototyping with Production Purposes in Mind
Recent Comments