Two be or not 2 be

January 11, 2008 7:28 PM

How would you go about proving that numerals are more "readable" than numbers?

I'm looking for a way to prove that numerals ("2") are more readable than words ("two") in headlines and running text.

Yes, yes, it makes sense intuitively (at least, I think it does). But how do you prove it? How does 1 go about testing "readability" objectively? Is shorter necessarily more readable?

This is exclusively for Web writing, if it matters. Googling turns up style guides. I don't want style guides. We use CMS 15 at work and I want to subvert it, but not with another style guide.

posted by YamwotIam to writing & language (27 comments total)

Interesting question! Maybe you could set up a controlled experiment where volunteers read three paragraphs with numerals, "number words," and mixtures of the two, respectively. Mix up the order of these paragraphs so that you're not simply testing how tired people get of memorizing number amounts. After each paragraph, test the volunteers for their recollection of the material, and see which information comes back more easily.

The tricky part is making sure you design the experiment fairly. I might also completely be smoking crack on this one.
posted by Sticherbeast at 7:34 PM on January 11, 2008

Well, this may get deleted because it's not answering your question, but I find it less readable. A number where I am expecting a group of letters interrupts the flow for me.

Similar to reading all caps -- it's not what you're used to, so it's harder to read. It is of course "readable," but less so.

So, I wouldn't know how to prove the opposite. :)
posted by iguanapolitico at 7:35 PM on January 11, 2008 [2 favorites has favorites]

For instance, I triple-taked the "1" in your second to last paragraph, thinking it was an "I" and being unable to make sense out of the sentence.
posted by iguanapolitico at 7:35 PM on January 11, 2008

I'm sorry, I should be thinking more before I type. But I should also note that I prefer listening to music on CDs over mp3 players, and sometimes I even call them "albums." It could be that because I didn't grow up text messaging that I don't expect numbers in place of words in sentences. :) It may in fact be easier for, uh, younger folks to read numbers as you describe.

So, to test it, get a whole bunch of people in a room with some written text. Give your own reading comprehension test. Find out how many people respond as I did, and how many respond as if I'm a dinosaur. But get people from all demographics so you can report complete findings.
posted by iguanapolitico at 7:41 PM on January 11, 2008

You could flash headlines at test subjects very briefly, and ask them to write down the headline. If numerals are more "readable", then it would be reasonable to expect a better recall rate for headlines with numerals than the equivalent headline with spelled numbers.

But I agree with the others here-- at least for numbers less than 10, the spelled words are more readable. Seeing a single digit makes my brain attempt a context switch, which takes an extra split-second to resolve (e.g., "Are we enumerating things now? Is this point in a list? Is that the end-of-sentence? Etc.")
posted by qxntpqbbbqxl at 7:46 PM on January 11, 2008

Everything I've read has been the exact opposite.

(sorry, I'm too lazy to go digging through my books for support)
posted by Mick at 7:52 PM on January 11, 2008

This would be an excellent thing to run an experiment on, I think. Really, though, first you need to define what you mean by "readability". The first thing I thought of was how quickly one can read a sentence, but I notice others think it has to do with memory retention. So, define that first and then build an experiment out of it. Sticherbeast has a good suggestion, but I would add reading the entire thing aloud so an external observer can measure the results, especially if you're looking to see how quickly someone can read.
posted by backseatpilot at 8:01 PM on January 11, 2008

For me it depends:

37 is easier for me to parse than thirty-seven (though not much).

Eleven million, two-hundred two thousand and one is easier than 11,202,001.

When I see large numbers (millions, billions, trillions, etc.), I often have to count up, from right to left (ones, tens, hundreds, thousands...) before I can understand the magnitude.

Whatever test you do, I think it's import that you test each word and numeral separately. You're making an assumption that ALL numerals (or words) are easier. What if 2 is easier than two, but five is easier than 5? For me, zero is definitely easier than 0.
posted by grumblebee at 8:18 PM on January 11, 2008 [1 favorite has favorites]

I think you need to consider carefully what you mean by "readable" and how you're going to measure it. For instance, assuming same type size, a spelled out number is going to be larger/take more space, etc and generally be more visible from greater distances or when the chosen type size is particularly small.

When small enough or far enough away, 5 can pretty easily be mistaken for an S or an 8 for a B, but you'd probably have much less of a problem with "FIVE" or "EIGHT."
posted by juv3nal at 8:31 PM on January 11, 2008

It may just be different for different people, so a careful study of readers in your intended market might be in order.

For example, I am the exact opposite of Grumblebee, apparently. I prefer small numbers to be written out ("I invited five friends over for dinner,") but larger numbers in numerals ("Bill Gates spent $1,235,690 on Doritos last year.")

The context of the numbers may also be important. A lot of us are so used to digital clocks that reading "11:39 pm" is easier and faster than "eleven thirty-nine p.m." Keep context in mind when you're deciding which way to go!
posted by explosion at 8:41 PM on January 11, 2008

You're trying to prove something you are already convinced of. That's bad, no? If you want to do a study, you should be approaching it as "Which is easier to read: numerals or numbers written in text?"

For me, it's usually the opposite of your current belief and everything I've read on the topic agrees with me.
posted by dobbs at 9:01 PM on January 11, 2008

Just to be clear, you're not asking about using the numeral "2" as a replacement for "too" or "to" ("ya me 2", or "go 2 the store") but just the word "two"?
posted by jepler at 9:02 PM on January 11, 2008

Just to be clear

Right, right. "2" vs. "two" is the question.
posted by YamwotIam at 9:04 PM on January 11, 2008

The convention (as I know it in my native language) was always that one should write out small numbers (say, anything from zero to twenty) and use numerals for bigger (i.e, 21, 457) numbers. I stand by this in English as well, and in most intelligently written articles, books and letters I read, it's done this way.

I suspect that because this is the "accepted" convention and the most popular usage as well, that that's what people "expect." What deviates from it is probably much less "readable" because it slows us down. This is exactly what happened to me (as it did to iguanapolitico) when I read that
"1" in the next-to-last paragraph . . . it definitely slowed me down.

In any case, one could "prove" it either way. "11,250,305" is probably more "readable" than "eleven million two hundred fifty thousand three hundred and five," but it's got a lot to do with context. That "1" in the middle of a sentence causes pause when used in place of "one" because it just looks wrong. Hence, I believe, the convention of writing out smaller numbers and just using numbers themselves past a certain point. Intuitively, this makes sense, but doing it for every number simply doesn't. Credibility would be an issue for many people, as well.
posted by Dee Xtrovert at 9:05 PM on January 11, 2008

Show numbers as numerals when writing for online readers
posted by jepler at 9:07 PM on January 11, 2008

What a cool idea; someone must have done a Masters thesis on this. Whether the presentation if web based or not seems irrelevant. I can only offer this google search.

On preview, awesome relevant link by jepler!
posted by pgoes at 9:13 PM on January 11, 2008

Thank you jepler--eye-tracking data is exactly what I'm looking for. Is Jakob Nielsen generally considered the last word on usability among the Web-design savvy these days?
posted by YamwotIam at 9:13 PM on January 11, 2008

Careful about that; he appears to be saying not that numerals are more readable, but that people tend to fixate on them when skimming text. Not the same thing.
posted by ook at 9:25 PM on January 11, 2008

The people who commented on the "1'"s causing a pause are an example of Nielsen's point. Do you want people to stop reading and stare at the number, or do you want them to be able to read normally without a "forced" fixation on it?
posted by deeaytch at 9:36 PM on January 11, 2008

A friend of mine suffered a head injury which made his bad handwriting almost completely uninterpretable, but did not affect his ability to write numerals at all.

If a similar separation in brain areas involved in reading numerals and letters is common, and certainly the received view seems to be that reading and numerical calculation are in different places, you could argue that in order to use a written-out number in any kind of mental calculation, you have to translate it into numerals and this slows you down and potentially confuses you ('two plus two equals four' is much more difficult to read than 2+2=4 for me, for example).

There is also always the straightforward information theoretic argument, of course; representing most numbers in numerals requires fewer bits than writing them out, and it's reasonable to guess that this makes them more easily handled by the brain in the absence of other information.
posted by jamjam at 9:46 PM on January 11, 2008

Huh. Good points, ook and deeaytch. But for B2B Web writing, I wonder if "stopping and staring" and readability converge somewhat.
posted by YamwotIam at 9:48 PM on January 11, 2008

What most people find as more readable is likely to have a lot to do with their strong points in school. I know that myself, having been very strong in the sciences and maths, find reading numerals (1, 2, 3) a lot easier than reading numbers (one, two, three). I would speculate that someone with less strength in math, and more in language ("english" or literature) would find more comfort in the opposite. This second part is pure speculation, if there are any english majors out there that can confirm this, I too have become a little inquisitive on this one.
posted by mrw at 10:02 PM on January 11, 2008

Echoing that I had to double-take on your "1" to figure out the context. (Even more confusingly, you weren't using a number in that instance to express a quantity.)

I would think that you would want to test reading comprehension?
posted by desuetude at 11:36 PM on January 11, 2008

IANA psycholinguist or a statistician, but I am studying both.

Spelled numbers have more distinctive shapes than numerals, because of ascenders and descenders, and so they are easier to recognize provided the whole number can still be read at one glance. The average reading saccade is 7-9 letters, so I'd guess that's about the width of an eyeful, and that any number under ten letters is better spelled out.

That said, 'readable' is different in different settings, and not always the same thing as 'legible' either. Are you trying to optimize through-reading or scanning? If there is text that needs to be with the numbers, you're looking at through-reading; if there isn't, IMO, the numbers should be in a table, and not in running text at all.

Anyway, here's a sketch of an experimental design you could use. Give your experimental subjects short paragraphs to read with two or three numeric quantities in each, and tell them you'll check their comprehension with a question or two after each. Turns out some of your questions will involve the numbers, but some should not or they'll give up through-reading and start scanning anyway. They also don't know that you've got a spelled version and a numeral version of each paragraph, which you've presented randomly. I'm not sure whether they should know that you're timing them.

Now your analysis needs to check whether the differences in accuracy and reading rates are significant. In particular, this means knowing what the center and spread of of these are are (for a paragraph taking all respondents and for a respondent taking all paragraphs). Then you calculate how much the center changed across the spelled/numeral pairs, and see whether each falls outside the spread of the other (there are standard statistical techniques for all of this; consult a statistician). If so, you have support for a claim; if not, you have bupkis. It's a lot of work, but, well, knowing what you're talking about always is.

Or you could, you know,

• ignore the difference between scanning and through-reading,
• ignore the principles that dictate two different, good document designs for two different informative purposes,
• ignore the nagging doubt that says writing text that you mean people not to read is bad design,
• ignore the overwhelming consensus of MeFites, stylists, and editors, and
• try to make your point by appeals to the authority of some random web page.
• You should also be prepared to ignore the ridicule, when someone points out that this alleged expert on readability has used sans-serif for running text and black on white for on-screen display, with lines too wide and leading too small for an effective, accurate eye return at line breaks.
• And if his point is scanning, he should have used a smaller face so that each eyeful and each screenful takes more in; you'll also have to ignore that.

Probably easier that way. Just depends whether the 'argument' you're after is 'a connected series of statements intended to establish a proposition,' or 'just the automatic gainsaying of any statement the other person makes' (Python 1972). Up to you.
posted by eritain at 1:30 AM on January 12, 2008 [1 favorite has favorites]

Well. I thought you meant numerals in place of their phonemes! (Well, that's how you used them twice in your post. Spelling out "two" was in the minority and I thought you were just choosing any random phoneme. :) So my early comments can officially go ahead and delete themselves now, because we were talking about different things.

So just to have a do-over on my prefs, easiest to read are numerals for numbers over ten (or 10 and over ... I can never remember and am personally undecided) and words for numbers under that. But again, it's what we *expect* because that's the basic writing guideline and has been for a long time, so that makes it easier to read.

jepler's link is fantastic. Scanability is a real issue on the internets.
posted by iguanapolitico at 6:16 AM on January 12, 2008

(Ugh, my reading comprehension officially sucks this week: it was only once that you used "2" in place of a non-number, "to," and twice that you used numbers in place of actual numbers. Still, sorry about misunderstanding the question.)

Nielsen is basically the most *famous* web usability guy, but his opinions shouldn't be taken as gospel. Plenty of web developers and users disagree with his strict rules. (I personally love rules and do worship him, though, yes.)
posted by iguanapolitico at 6:20 AM on January 12, 2008

eritain has it.

Another question you need to ask is what impression you create. When you wrote "one" as "1" in the question, I had to stare at that for a few seconds, and frankly, thought "this person is a quasi-literate raised on SMSing." I'm conservative with my use of language, so if you were writing for an audience of me, you'd want to avoid that.
posted by adamrice at 12:50 PM on January 12, 2008