Large X-Heights = More readable?

Dan Gayle's picture

Someone explain this to me:
Everywhere I look, people declare that typefaces with large x-heights are more readable than something with small x-heights. For instance, many people don't like Futura for body copy because of the long ascenders/low x-height (aside from the sans/serif debate). They say it breaks up the color too much. Same thing for other typefaces with low x-heights like Venetians.

But that doesn't make sense to me. Presumably, the larger x-height makes the letters appear larger. But that means that you need to increase the leading proportionately to compensate. If you used a typeface with a low x-height at a larger point size, wouldn't you basically have the same results?

William Berkson's picture

Interesting, Nick. Your font seems toward the lighter side, like the Carter font. It looks to me without measuring that your ascenders are a bit shorter than Carter's, but not much. Le Monde Journal's are much shorter--it was a design solution in his eyes at the time.

Your font is a bit wider than either of theirs. Did you find that the width was a gain in readability, more so than shortening the extenders further?

[edit: this was posted before your additional info--thanks. The question still stands...]

hrant's picture

I like that "y".

Concerning news faces (noting that the distinction between headline and text is critical), the trend for the last number of years has actually been to go larger in point size. That's one great reason to make the x-height smaller. Another reason is that the 70s have been over for a while now.


Nick Shinn's picture

I'm not sure where the theory leads: from my perspective, I work against the copyfitting benchmark of other news faces, and really have to see a press proof on newsprint to decide. At the begining of this project, for research I set the same piece of text to the same column depth, with the same 10 pt leading, in a variety of text faces. The sizes worked out as follows:

9.0 Utopia
9.4 Mercury
8.4 Nimrod
8.9 Worldwide
9.3 Charter
9.0 Farnham
9.6 Freight

Really, I think all these faces are excellent for news text, and the differences in their x-heights don't have much of a bearing. The key factor, IMO, is descender length. That's why Freight is the least flexible for the genre, as with its generous descenders it may require more leading.

I don't think cap height plays too much of role: Nimrod has ascender height caps, whereas Mercury's are much shorter. But in absolute value, they come out much closer due to the 9.4 vs 8.4 difference, and if Mercury's caps are still shorter, they have ample width.


Looking at the typical big-x-height ITC faces such as ITC Garamond and ITC Cheltenham, they appear stunted because (a) we are familiar with the originals, and (b) they have relatively long descenders, so the action in the x-height -to-cap region appears relatively crammed. In absolute terms, they are not much different in proportion from the above news faces, except for their "long" descenders. They are similar in proportion to Frutiger and Helvetica, too!

The really extreme "70s" faces are Usherwood, Eras and Antique Olive ('60s).


So to return to the original question, I think the comparison is really between three categories:
- Classic book faces with quite small x-heights and generous proportions all round
- Standard proportioned faces
- Extreme 60s/70s faces -- noting that the "chopped" ITC versions of the classics may actually have standard proportions.

Wasn't there an article in SOTA's "Interrobang" about a system of nomenclature for this?

hrant's picture

> the differences in their x-heights don’t have much of
> a bearing. The key factor, IMO, is descender length.

This is focusing too much on avoiding collisions between lines, which, in a properly designed face, is taken care of via the talus (internal leading)... assuming no negative leading! In reality: x-height, ascenders and descenders are best seen as one thing; and the x-height in fact has the most bearing on readability, although ignoring the rest is exactly what makes the 70s ITC junk so unpalatable.

> They are similar in proportion to Frutiger and Helvetica, too!

Which is one reason those two are also useless for immersive reading, the other main reason being lack of serifs, plus in Helvetica's case overly tight spacing.


Nick Shinn's picture

This is focusing too much on avoiding collisions between lines

That's very important for news text.

I don't think one can rely on internal leading -- which was what I was trying to demonstrate by comparing the different faces, above.

Left to right: Nimrod, New Shinn face, Utopia, ITC Charter, ITC Garamond, Mercury G3.

Top row, all same point size.
Bottom row, adjusted to equalize x-height.

Once the x-heights are equalized, the biggest difference is between the "old school" news face, Nimrod, with its short descender, and the "new school" news face Mercury, with its tall ascender.

It's interesting that Nimrod, widely regarded as one of the best faces for immersive reading in newspapers, has quite similar proportions to the much derided ITC Garamond, at least, above the baseline.

Perhaps for text use, ITC Garamond Next would be a good idea, with the missing weight between Light and Book, and with slightly more robust hairlines -- but the same famous proportions.

hrant's picture

> That’s very important for news text.

I would say that the rough-and-tumble nature of newspaper reading
makes collisions more tolerable than in book typography for example.

> adjusted to equalize x-height.

That's not a good way to do it, because:

> Nimrod, widely regarded as one of the best faces for immersive
> reading in newspapers, has quite similar proportions to the
> much derided ITC Garamond

Only if you ignore half the Cartesian space. In the horizontal dimension the two are different in two huge ways: width, and spacing. Nimrod is narrower, and looser. Taking into account both dimensions prevents the comparison of fonts by a simple equalization of the x-heights; the only real way to do it is to equalize apparent sizes. For better of worse this cannot be done numerically - it has to be eyeballed - but anyway that should appeal to you, no? :-)

In fact many of us have long said that ITC Garamond (preferably a demi weight) could be used for small sizes... if it were spaced much looser. In fact smaller sizes than Nimrod, considering its massive width. It's a cornerstone of type design than all of a font's attributes have to be balanced with each other to yield a serviceable whole: a huge x-height can only work for text (and only small text) if the color is darkish, and the spacing is loose. Mrs Eaves is arguably the most prominent case of that balance thrown off: the x-height is small, which makes it ideal for somewhat large text, but the color is ideal for typical text sizes, and most of all the spacing (beside being bad - different story) is way loose, which is ideal for small sizes! The Next that we really need is Mrs Eaves, since otherwise it's such a beautiful design. Maybe call it Mr Richard. ;-)

It's tempting to think that a simple tracking adjustment in the layout app can usually correct for such an imbalance (assuming an adequate gradation of weights) but spacing boundary conditions (such as the right side of the "r") make that a half-baked solution.


hrant's picture

> one of the best faces for immersive reading in newspapers

BTW, according to you all reading is immersive anyway, right? Am I correct in assuming that you believe that the ideal vertical proportions for a news face for example are strictly a result of habit and culture?


Dan Gayle's picture

I love this. Somehow typography always comes down to two very precise and scientific measurements which are:

1) Somewhere in the middle
2) Eyeball it.

TBiddy's picture

It seems we're veering in two different directions:

1) X-heights for display type and
2) X-heights for extended passages of text

Both I think are quite different. While I don't have as much experience as Nick (or other Typophiles) would have in this, I do know that the context for reading a headline and reading copy are very different. The headline is "scanned" more than it is read. I'm of the opinion that a "lighter" face is more difficult to read (or should I say absorb) than a darker headline display face. The eye would tend to focus more on the white space the text creates than the actual words.

Carl pointed this out earlier as is difficult to discuss this without discussing some of the other variables in the equation. Such as some necessities for newspaper design. For one, column widths are an important factor to consider. While this might not increase legibilty, a condensed face (at least slightly condensed with reasonably short descenders) might be ideal for narrow column widths.

If you look at Nick's example, you can see that his face is certainly more condensed than that of ITC Garamond. Now for text faces...short descenders and condensed faces tend to be more difficult to read.

I love Fedra Serif and Lexicon for example, but I actually think the versions with shorter ascenders and descenders are more difficult to read. The larger x-height (at least to my eyes) actually looks and feels wrong. I'm not quite the scholar that many others on Typophile are, but I know Minion Pro is much easier on the eyes than Jenson Pro. Ripping off Nick :) here's an example of Adobe Jenson Pro, Minion Pro, and Adobe Garamond Pro all Roman weights.

Holy crap! Look at that x-height on Jenson Pro! Minion has a significantly larger x-height, and Garamond's width makes up for that x-height deficiency. Jenson doesn't work for me when I read text. Howeva...I do know someone who has vision problems who loves Jenson Pro. My theory is that while Minion or Garamond might be sweet for people with normal vision (because we look for letter familiarity) Jenson might work better for my friend because the character (personality/quirkiness) of the individual letters might aid in her letter recognition.

I have no theories or research to back any of my opinions up— I just know what works for me when I read.

Nick Shinn's picture

Hrant, you're right it's not just the x-height.

Equalizing x-heights is just a convenient method I've often used for comparing different text typefaces.

Another way is modifying point size to get them close to fitting a similar amount of copy, which is the experiment I described further above -- so what I could do for the comparative "I_thorn" diagram is to scale them according to the size ratios revealed in that test. Or maybe not--that's a bit too much maths right now.

There is a similarity in both comparison techniques, as larger x-heights take up more space, but as you point out, they're not the same.

So, there are three potential "controls" for comparing the differences in typographic quality between typefaces, in blocks of text:
1. Same point size
2. Same copy-fitting rate
3. Same x-height

Another way would be totally subjective -- to just pick the best setting for each typeface with no measurable commonalities. Nothing wrong with that.

What "control" do you prefer for comparing typefaces in text settings, Hrant?

BTW, according to you all reading is immersive anyway, right?


Am I correct in assuming that you believe that the ideal vertical proportions for a news face for example are strictly a result of habit and culture?

Speaking as a type designer, I'd say that the ideal proportions for a particular typeface emerge during its design, so yes, that is culturally specific.

As a typographer, I'd say that the design of the whole document informs the proportions of the typeface specified. The cultural circumstances surrounding the document, from the start, inform the design and typography. There's some feedback from the face along the way.

Dan Gayle's picture

Biddy: You got the order wrong there. You have Jenson>Minion>Garamond.

Making a distinction between display and body copy is very important because the purposes of both are different, like sport fishing and fishing for food.

Display text is sport fishing. Catch and release. It's designed to catch the eye first, then release to the content. Body copy on the other hand is like fishing for food. You want that fish to catch on, and hold on, until you've reeled it in. You want them to read the content until the message has been communicated.

So in that sense, is readability, as it concerns x-height, for display copy not as important as body copy? That's why you could get away with using Olive Antique as a headline, but never for body copy.

hrant's picture

I don't think a "design brief" can emerge from the inside.
If it does, I might call that an "art brief" instead.

> What “control” do you prefer for comparing typefaces in text settings, Hrant?

I think there are two dimensions at play: apparent-size vs. economy, and readability. The former can be controlled from one end or the other: you can equalize the apparent sizes (and apparent leadings) and see which uses up more space, or you can equalize the space used (while also minding the size/leading proportions) and see which one looks bigger. This dimension is basically about optimal legibility in the space. The latter dimension is more inexact and "unproven", but with a mature grasp of the mechanics of reading (including distinctions between deliberation and immersion, as well as optimal letterspacing) it's possible to do much better than just with the first dimension. For example a font with a very large x-height and very tight spacing will come out on top in the first dimension, but you can still say it's not good because the point size, reading environment, etc. are such that the extenders can play a greater role facilitating immersion, or because the spacing is so tight that notan is ruined, etc.

Practically, I tend to do this when comparing two text faces: choose a target apparent size, often simply a sensical point size in one of the fonts; equalize the apparent sizes and minimum apparent leadings; observe the "mechanical" difference between economy; observe the "inexact" difference between vertical proportions, spacing, color, letterform divergence, etc. with the purpose of gauging "non-scalar" readability; and finally try to decide which face strikes a better balance between economy and readability.

Seperate and equally important to this is the evaluation of which face has an atmosphere/mood more suitable to the subject, noting that often a face can sacrifice (sometimes, in fact preferably, quite consciously) apparent size, economy and/or readability for the sake of atmosphere (like how some glyphs in Patria are intentionally "too bulky").


dezcom's picture

There are some basic things that come into play for the person designing the document and choosing the face. The text is of a finite length and pages need be added in multiples of 4. With that as a background, I can't imagine making a decision without the copyfitting component considered. You can call this some degree of efficiency and this may be of greater or lesser importance depending on the job. Newspapers pack it in much more than fine books. Fore me, it is about finding the best range of "sweet-spot" combinations in a given type and applying those to the copyfitting needs. If a given font you may choose first does not copyfit well within the respectable sweetspot range for that job, you need look mfor another typeface. For me sweetspot takes in all the variables including the character of the writing to be typeset along with the typical parameters of xheight,leading, line width, etc..
The job does not end at the microcosmos of the typeface, it takes water to float a boat, wind to move the sails, and navigation to get you to your destination.


TBiddy's picture

Biddy: You got the order wrong there. You have Jenson>Minion>Garamond.

Good catch, its been corrected. Even when I proofed it I still "read it in the order my mind saw it". Why don't we make a typeface that reads your mind? :)

Dan Gayle's picture

Here's another one from Juan Pablo De Gregorio's blog: (Great blog by the way...)

5. x height
The area between baseline and x height contains most of the readable information (75% of the lower case letters). It is a very important area at the moment of reading text. Long ascenders and descenders require a small x hight. If we compare two types of designs, one with long ascenders and the other with short ascenders, we can see that the x height of the second one will be larger, so it will obviously be more legible. Look up for the difference between Times New Roman and Mrs. Eaves.

hrant's picture

> 75% of the lower case letters

This is not a useful statistic.

> Long ascenders and descenders require a small x hight.

This is not a good way to express it.

> it will obviously be more legible.

1) No, it depends (mostly on size).
2) Legibility vs readability makes all (or at least 50%) of the difference.


Dan Gayle's picture

You said the 75% stat wasn't very useful, but does the idea itself make sense? We've all seen the tests where they chop off either the bottom half or top half of a sentence, and we can all agree that we can read the top half better.

We do scan the top half of sentences, so does making the top half more level by increasing x–height play any part in increasing the readability?

hrant's picture

That 75% is not useful because it's based on the "pure" alphabetic set, which is out of context of real text. For example if you look at just the alphabet you might be tempted to make the descenders longer than the ascenders; but when you factor in actual letter frequencies you end up with the exact opposite conclusion.


crossgrove's picture

That graphic from deGregorio is probably the worst one in the series: to compare 2 designs with different x-heights, you don't want to scale one of them up; the apparent stem weight and serif weight are changed in his sample. Just having 2 faces with different x-heights doesn't compel you to set them the same size, with the same nominal leading. This simplistic conclusion ignores all the other factors we've been discussing. As Hrant points out, removing this from the context of a printed page voids the argument.

Dan, I would say that it is specifically NOT helpful to "level out" the top half of the line of type. Once again, the ascenders and descenders perform a function in reading, and it is not insignificant. Some have argued that this effect proves the opposite, that descenders are less important.

Why are people so eager to reduce this complex, interdependent process into some kind of flat, simplistic sound bite?

hrant's picture

Control. Monotheism. The Enlightenment. Modernism. Fear.


charles ellertson's picture

Just a note on "economy." I don't know web presses that well; I've always assumed the paper cost on them to be higher. Anyway, on a sheet-feed press, you buy a sheet. Depending on the size of the sheet & the book trim, you'll get 16- or 32- or even 64-page signatures. Paper isn't the only cost of course, so it may turn out that even with the charge for canceling 16 pages, printing, say, four 32s and a 16 is less than five 32s. But four 32, a 16, an 8 and a 4 is more expensive.

If a chapter begins a new page, the facing page (usually the last page of the previous chapter can run 5 lines long or full dept. If you are setting 38 lines per page and you want to know the "savings" for running 39 lines, you would need the chapter to be at least 38 pages long to be sure of saving a page. You might save a page with just a five-page chapter (or 4, if you'll run pages long), but it would be rare. If you don't save a page, you won't certainly won't save any money. And sometimes saving a page can cost you money (that extra 8- or 4-page signature).

Anyway, the point is that a lot of those people who feel type should be set smaller & leading tighter for economic reasons don't know how printing is paid for.

As far a legibility goes, that too is problematical. When my mother was in her late 70s, large-print books began to come out. They were far easier for her to read. Does that make them more legible, or just more legible for a specific audience?

I'm afraid I have to agree with Nick: Given a certain typeface at a certain size, there will be a word-spacing, leading and margin parameters that will just "click." Sometimes this is dependent on the text to boot (that influences how many word spaces occur in an average line).

dezcom's picture

"Given a certain typeface at a certain size, there will be a word-spacing, leading and margin parameters that will just “click.” "

This is also what I meant by the sweet spot for a type.


tiberiu's picture

You can always try the vicodin addiction typeface and create a new pattern that could prove to be something new and unseen.

Syndicate content Syndicate content