Reasons to NOT Standardize Cap-Height Across Font Library

Rishi Murugesan

(I'm still a student so apologies in advance if this is a basic question.)

If you're working with a 1000 UPM, I've noticed that the cap-heights for many typefaces seem to hover around the 700 mark (± 50ish units).

If you have multiple typefaces in your library (sans, serif, slab, etc.), is there a good reason to NOT standardize your typefaces across your library? (i.e., every typeface is set to say Cap-Height: 700?)

Wouldn't it possible to just modify your x-height/ascenders/descenders/zones to get the metrics you want? Is there a reason to set each typeface with a unique cap-height? 

Craig Eliason

Not sure I'm understanding your question, but: since most text is in mixed case which is dominated by lowercase, x-height has the largest (though not only) impact on perceived size at a given point-size. So I would think arbitrarily standardizing your cap height may make your large-x-height designs seem relatively too large. Or in other words, to meet expectations of how big a font should read at a given point-size, you will likely need to lower the cap height rather than just increase all the other metrics for designs in which that cap height is relatively smaller. 
If you're talking about a "superfamily" and not just a whole library of otherwise unrelated designs, then there would be more incentive to standardize the cap heights I would say. 

Thomas Phinney

So, first, Craig is right. Cap height is not usually the primary determinant of how large a font looks, to most eyes. This is why Apple rescaled Zapfino in between two OS releases, 2.5x as big as it had been before. So, clearly there are at least some reasons not to make all possible fonts have the same cap height in a library—in the case of Zapfino, because it had a tiny x-height.

That said, it is possible to standardize cap height across at least most/normal fonts in a font library. Even if you think it is a good feature, it is much more practical/feasible if you do it from the beginning. Revising an existing library is possible, but the new fonts would be incompatible with the old in a fairly major way (often causing reflow, not just very minor design tweaks). I for one would argue that it would not be worth the “incompatibility hit” to do so on an entire existing library. It seems very unlikely somebody would do so.

I have had occasion to look at a fair number of font metrics for legal reasons. The only foundry that I have run across that standardized their cap height was URW, for almost all their fonts—except for those that needed to be metrically compatible with something else. For the standard cap height they went with 2/3 of the em as their cap height.

So: yes there are reasons not to do it. Yes it is possible.

Dan Reynolds

Thomas Phinney said:

The only foundry that I have run across that standardized their cap height was URW, for almost all their fonts—except for those that needed to be metrically compatible with something else.

As far as I remember, almost all of the Linotype fonts mastered in the ’00s (and maybe even the ’10s) had a standardized cap height.

Berthold’s phototype fonts had a standardized cap height and I believe that they applied this to their PostScript fonts in the early 1990s, too. I don’t have any Berthold PostScript fonts, however, so I cannot verify this. It was certainly one of the reasons that Linotype was standardizing cap height when I was there, though. Our cap height, I believe, was slightly shorter than what Berthold had defined.

Helmut Wollmersdorfer

Most fonts for text have proportions somewhere around this (assuming 1000 units):

- x-height 500
- ascender length (h-height - x-height) 250
- descender length 250
- H-size (CAP-size) 700, if H is smaller than h

Keep in mind, that accents above capital letters need also space. 

Thomas Phinney

The most common ranges for:

• cap height: 57–75% of the em, typical 62–72%
• x-height: 40-52% of the em, typical 44-47%

Some very common fonts are actually among the relative outliers, relative to the em. Courier has an unusually small cap height and x-height. Arial and Helvetica have very large cap height and x-height.

Craig Eliason

More often than not I start designs using the Glyphs default x-height of 500 on a 1,000-unit em. And very frequently I wind up scaling the whole font down before I'm through because it winds up too big on the em. I should really try harder to develop the habit of moving that little line down a bit before drawing! 

John Hudson

Relative vertical proportions of x-height, cap height, and ascenders* are, historically, important idiomatic features of styles, and the rationalisation of these tends to dilute the stylistic differentiaton. The large x-height ITC advertising versions of classic types are perhaps the most obvious examples of this: they all end up feeling much the same, reflecting the common time and culture of their manufacture rather than the individual origins of the sources.

I typically begin a project by working out the relative proportions, and then apply these within the body height with a mind to intended use, with particular attention to targeted diacritic support.


* Less so descenders, which are more likely to be coordinated to the x-height to cap height distance regardless of style, which has the benefit of aligning ¿ and ¡ between x-height and descender depth.

John Savard

Back in the metal type era, it was considered an achievement when type foundries standardized the position of the baseline. Normally, there were three standard positions; one for titling faces, a "standard" line for workhorse faces with short descenders, and an "art line" for luxury faces or othentic old faces with long descenders.

Cap height was standardized as a result of this, because normally the cap height of a typeface extended up to very nearly the top of the slug.

So people are used to the 10 point size of a typeface very nearly filling a 10 point line.

However, digital practice seems to be very different, including what is considered a reasonable amount of leading in "10 point Times Roman" so that digital Times Roman is smaller than hot metal Times Roman.

Thomas Phinney

@John Savard That is an interesting mixture of true and false.

John Savard said:

Back in the metal type era, it was considered an achievement when type foundries standardized the position of the baseline. Normally, there were three standard positions; one for titling faces, a "standard" line for workhorse faces with short descenders, and an "art line" for luxury faces or othentic old faces with long descenders.

You are mostly right about the baseline, although even THAT was only standardized in the late 19th century along with the size of a point, and there were different point size standards for Anglos vs Europeans (and several in Europe in different countries, too), and there were the three baseline standards even in the Anglo realm. 

John Savard said:

Cap height was standardized as a result of this, because normally the cap height of a typeface extended up to very nearly the top of the slug.

No. Besides the many standards for baseline mentioned above causing further cap height variation, there were minor differences between different typefaces with the same baseline, or between different foundries, meaning there was effectively no standard at all for cap height. Cap height as a proportion of the em could also vary between sizes, for the same typeface.

John Savard said:

So people are used to the 10 point size of a typeface very nearly filling a 10 point line.

Utterly false. This was only ever true for caps-only titling faces in metal (a very tiny proportion of all typefaces), and later adaptations of those same typefaces into other technologies.

When phototype versions of metal typefaces were made, they generally adopted the same sense of point size relative to the em that the original metal type had, so people could spec the same size and get the same or a very similar result. (Because metal typefaces did not scale linearly, and phototype was generally used with a single master across a wider range of sizes, the size on the em wasn’t always exactly the same. But it often was exactly the same for at least some reference size such as 12 point, and reasonably close at many other sizes.)

John Savard said:

However, digital practice seems to be very different, including what is considered a reasonable amount of leading in "10 point Times Roman" so that digital Times Roman is smaller than hot metal Times Roman.

“However, digital practice seems to be very different”

Mostly not as far as point size goes. The whole reason digital point size is so frickin’ weird and hard to understand is primarily because it carries over from previous technologies and they did not standardize.

When type transitioned again from phototype to digital, existing designs used the existing point scale they already had. So the notion of an em square that was now purely virtual carried over, even if it was no longer so much of a limit on total vertical size.

“digital Times Roman is smaller than hot metal Times Roman”

No. I just now measured printed 14 point Times Roman metal caps as shown in Mac McGrew’s American Metal Typefaces of the Twentieth Century. I measured them at 0.133" in print. Current digital Times Roman and Times New Roman have caps about 66.21% of the point size, which at 14 point would be… 14/72 x 0.6621 or 0.129" in the master outlines. That is a 3% difference, which is completely consistent with the differences one sees at such sizes between the master font designs and printed output.

“including what is considered a reasonable amount of leading in "10 point Times Roman"”:

There is no single consistent practice for leading in usage of digital type. Pro graphics/publishing programs tend to default to 20% of the point size, which would tend to be about the same as the single most common leading in lead for 10 point type, and will result in less leading than was typical in metal at smaller sizes, and more than was usual in metal at larger sizes. Typical office apps such as Microsoft Word have the same practice of using a fixed proportion of the point size, but they tend to do that as a font-specific amount that varies between typefaces, and is ~ based on the font bounding box, which could be smaller than 20%, but is more often larger, sometimes by a considerable amount. (There is a complicated history here, where use of the font bounding box used to be absolutely true all the time, but more recently has a sort of escape hatch on the type design side, if the designer/foundry chooses to use it.)

But the important thing to keep in mind is that the leading is quite a separate thing from “how large is the cap height as a % of the em square?” I have no idea why one would bring leading up in this discussion. In both metal and digital type, it is an additional variable after the type is sized. It is arguably a good example of how digital type differs from metal, because there are defaults in the apps, which are not “set solid” (no leading).

Mind you, the default leading is being done by the apps, not by the fonts. In fact, in the digital fonts themselves, things are indeed in close mimicry of metal type in this regard.

The big difference is that the vertical placement of the em box is effectively font-specific! Which is super weird compared to metal type, and leads to apps having to make tough decisions about vertical positioning of the first line of type in a text block—with pro publishing apps offering multiple options to control this.

Nick Shinn

I do standardize some vertical metrics between typefaces—but only when they are specifically designed as a suite (for instance Scotch Modern and Figgins Sans).

But in general they come out all over the place.
I think the reason, which I will attempt to explain with this visual analysis, is that I try to get a good presence on the page, when fonts are dropped into a layout application with default specs, “straight out of the box”: not too big, not too small, and suited to the genre of typeface.

In this comparison, two quite differently proportioned geometric typefaces are shown. Aptly is a condensed style with a big x-height. Neology is of normal width (and having circular o’s, fairly broad), with a smallish x-height.

When the styles are compared at the same type size, in mixed case, Aptly’s big x-height appears excessive, yet when all caps is compared, its taller cap height prevents its subordination. This was one reason it felt right to give Aptly a larger cap height (825) than Neology (717), as I wanted to give its all-cap setting sufficient heft. Also, in mixed case paragraph text, I wanted to make Aptly’s x-height large on the “M-square”, creating a somewhat unleaded look, to counteract the +20% linespacing that layout app designers provide as default, which has the unfortunate effect of making paragraphs of condensed type look less like blocks of text than widely spaced parallel grey bands.

Rishi Murugesan

Thank you all for the comments; this was very helpful! 

John Savard

Thomas Phinney said:

@John Savard That is an interesting mixture of true and false.

John Savard said:

So people are used to the 10 point size of a typeface very nearly filling a 10 point line.

Utterly false. This was only ever true for caps-only titling faces in metal (a very tiny proportion of all typefaces), and later adaptations of those same typefaces into other technologies.

I didn't mean the capital letters of a 10-point type would normally fill the line. I meant that the line would be nearly full once one included descenders and the extent to which ascenders went above the top of capital letters.

Thomas Phinney

John Savard said:

John Savard said:

So people are used to the 10 point size of a typeface very nearly filling a 10 point line.

Utterly false. This was only ever true for caps-only titling faces in metal (a very tiny proportion of all typefaces), and later adaptations of those same typefaces into other technologies.

I didn't mean the capital letters of a 10-point type would normally fill the line. I meant that the line would be nearly full once one included descenders and the extent to which ascenders went above the top of capital letters.

Oh! Of course, my apologies. Sure, that is true! Or at least it was in metal type, and in those typefaces descended from metal (including their phototype descendants).

However, it says little about the relationship between, or standardization of, cap height relative to point size.

I would also argue about the “people are used to” part: Calibri became a default font almost 17 years ago, now, and is in the midst of being replaced. It is only indirectly that the sizing of metal type still informs what people are most used to, today. There is now almost a generation that doesn’t remember digital devices before Calibri.

Ray Larabie

@”Thomas Phinney” Do you think increased language support has affected the height of capital letters in fonts? My cap heights are often determined by how much I can squeeze the Vietnamese stacked capital accents, unless I use the “past the line” approach.

Thomas Phinney

Yes, I think that.

As you say, Vietnamese support requires stacked diacritics, which can be a problem if cap heights are generous on the em. At the same time that Vietnamese support has become less of an afterthought, cap heights have decreased. When I compare the cap heights of new/custom system fonts such as Calibri (2007 release) with the previous generation such as Times and Arial. The newer stuff often  has a cap height more like 62–68% of the em rather than 66–72% of the old stuff.

It is hard for me to be really sure as the ranges overlap, and the choice of typefaces I have measured in the past was driven by various legal cases and not at all intended to be representative of all typefaces, or random, and only about 60 families.

But those caveats aside, there were a LOT of old classic fonts with cap heights around 72%, and in the particular samples I have looked at, it sure seems like the new typefaces never go that high.

I will ask a few relevant folks and see if they have any thoughts on whether that was the thinking in their particular font library development. I imagine that change happened in about 2002–2012 or thereabouts, for all-new designs.