x-heights: a tale of two typefaces

Amado's picture

Really just a question about how this works. I'll pose some hypotheticals. And I don't know enough about how x-heights measurements are specified, so I'll do my best and just hope my hypotheticals make sense.

Say Type Designer A makes a typeface (let's call it SmallXTransitional (SmallXT)) with an x-height that is .7 of the point size. Type Designer B makes a very similar one (LargeXTransitional (LargeXT)) with an x-height of .8 the point size. Assume that SmallXT and LargeXT are quite similar, except for the x-height.

Years later, along comes a layout project (say, a 16-page booklet), and the client wants two different comps from two different typesetters. Typesetter X uses SmallXT at 16pt with a line-height of 20pt. Typesetter Y uses LargeXT at 14pt, again with a line-height of 20pt.

What will be the difference between the two? Am I wrong to suspect that the difference will only show up in the capitals? Or perhaps the ascenders/descenders? Will these two hypothetical settings be quite similar in character (all other things being equal), as I suspect?


finegrind's picture

They won't be identical. They will have the same, absolute, measured x height but SmallXT will appear slightly bolder. This is because the two faces would most likely be drawn to the same weight when at the same relative point size. Essentially you're still enlarging one more then the other.

Amado's picture

I wondered about weight. I almost threw in a bit about using SmallXT Book and SmallXT Semibold, versus LargeXT Medium and LargeXT Bold. That seemed a bit too fussy for a generic example.

But I s'pose what I'm driving at is: is it customary to take a typeface's x-height into consideration when choosing a line-height? Does a small x-height allow you to get away with tighter leading?

And, is there any way of determining a typeface's x-height short of busting out Illustrator and measuring it? And, is there a standard unit used to measure/specify x-height?

hrant's picture

Traditionally, an x-height measurement is most often specified as a percentage of the ascender height (less frequently as a percentage of the cap height). However this assumes certain value ranges for descender depth and talus, so it's not a robust measure.

In terms of a percentage of the Em, the typical x-height is way less than 70% - it's closer to 40%.

If your SmallXT and LargeXT only vary in x-height*, you have to realize that that affects the relative sizes of the ascenders versus descenders, a relationship that would get thrown off, possibly to the point of the descenders ending up longer than the ascenders (generally a design flaw, at least in a text face).

* BTW: http://typophile.com/node/81755

Similarly, in your typesetting scenario, besides Mark's observation that the 16 point text will look darker, you will also see a difference in terms of the dimensions besides the x-height. Now, with 4-6 points of leading descenders being too long for example is rather moot; but the texture will indeed have a different feel (especially at those large sizes) partly due to the caps as well.

> Does a small x-height allow you to get away with tighter leading?

Indeed, yes.

> is there any way of determining a typeface's x-height short of busting out Illustrator and measuring it?

Not necessarily Illustrator, but no, there isn't.
Hey, you can't even really tell the point size!

> is there a standard unit used to measure/specify x-height?


IIRC Jorge de Buen was working on such a measurement system.
There was at least one other guy too.


Amado's picture

My thanks to you both for the education. I'll go look up your references.

Is it commonly known that fontsquirrel.com has an auto-letterform-torturer? Under "Expert Options" at http://www.fontsquirrel.com/fontface/generator, there's an option that reads:

"X-height Matching [Resize to match the selected font's x-height]: None | Arial | Verdana | Trebuchet | Georgia | Times New Roman | Courier"

That *can't* be good, can it?!? Based on what you said, it must do very weird things to ascenders and descenders.

hrant's picture

I'm sure it scales all the metrics based on relative x-heights (that being the easiest way to approximate apparent size). If it scaled only the x-height it might actually be useful. :-/


Amado's picture

I'll test it, if I get a chance. It'll be a good learning opportunity for me.

Amado's picture

Right you are; it doesn't torture letterforms, it just scales the type and re-labels the different sizes. Above is Centaur, Georgia-ized. The x-heights match, but the cap height at "72" is obviously too big in the Centaur.

I overlayed normal Centaur with the one Font Squirrel x-height-matched with Georgia, and had to zoom normal Centaur 129.5% to approximate what Font Squirrel did.

So my x-height-matched Centaur, when I set it at 10px in my CSS, will "really" be 13px on the screen.

I suppose I'd better make a test HTML page to see whether the baselines got all screwed up too.

And when I'm done with all that, I'll be discarding my resized Centaur, since I'm doing this for testing purposes only and don't have a license agreement to use it on the web. Just, y'know, so I'm clear about that too.

Amado's picture

I didn't have to do my own sample. A pretty comprehensive one comes with the @font-face kit. In one block of text, they have the line-height at 1.1, and the descenders of one line overlap the ascenders of the next. The same is true at 1.25.

There are example text blocks elsewhere with line heights of 1.4444 - 1.5. The space between lines actually looks quite nice and balanced, but in places where descenders and ascenders line up between lines they *just* miss each other by an almost-acceptable amount. (Functionally, this is equivalent to "using less leading for typefaces with a small x-height.")

They line up the baselines and the "x-height line" (if that's the term for it).

Of course there are other rendering problems, because Centaur has all those little details and graceful curves. And at "12 pt", the /o/ is comically way too large. I only picked Centaur because I knew it had a tiny x-height and would make a good contrast to Georgia; it's clearly not made for low-res screens.

So, via a circuitous route, I learned what I wanted to.

Maxim Zhukov's picture

the “x-height line” (if that’s the term for it).

I thought the term for it was mean line.

hrant's picture

I've also heard it called "x-line".


Amado's picture

Aha! Now I've got questions about the graphic that Maxim uploaded. Just two, initially.

  • What is the relationship between a typeface's Point Size and the distance between the Bottom Bearing Line and the Top Bearing Line?
  • Is there any standardization in where the Baseline is situated in the space between the Bottom Bearing Line and the Top Bearing Line?

If there's a book (or website) I ought to be reading, instead of bugging y'all, please feel free to direct me.

hrant's picture

Top to bottom is the point size.

There's no standard baseline placement (well, ATF had one, but that was a different era and a different technology) but there are ballparks (as well as exceptions, especially for script fonts). That diagram actually seems to fall within the ballpark - except the taluses are exaggerated (probably simply to fit the text in there).

A book? Those things only have one author. :-)


kentlew's picture

For most of the 20th century, in metal there were a few standardized schemes for the placement of the baseline on the body. Standard Line (also known as Point Line) was probably the most common. Art Line was another.

Standardization was desirable because mixing types on a single line was driven by the physical relationships of top/bottom, and they would only line up if there was a common placement of the baseline.

Nowadays, software takes the baseline as the reference point, so top and bottom are virtual and related to the origin/baseline, rather than the other way around. If that makes sense; might need visuals.

Amado's picture

Top to bottom is the point size.

Then there's something happening in Illustrator that I don't get.

I type a line of (say) Palatino Linotype at 80pt, and the distance from the descender line to the ascender line is 80px. When I select the object, a box outlines the area the text is in, a hair below the descender line and 10px-11px above the ascender line. The "transform" tool tells me that this box is about 92px tall.

I type the same text in another font at 80pt, I get 74px from descender to ascender. The bounding box is maybe 5.5px below the descender and 8.5px above the ascender, and the transform tool says that the object is 89px tall.

Obviously these bounding boxes in Illustrator have little, if anything, to do with the Bottom Bearing Line and the Top Bearing Line. I presume that these lines originate from: they were the physical edges of the piece of metal that the letterforms were carved into.

Surely I'm not alone in finding this perplexing? I'd think that some standard for the vertical positioning of the letters would have evolved organically in the industry. And I'd think that design tools would make the bounding boxes of objects have some relation to that standard. What am I failing to understand here?

hrant's picture

From what -I think- you're describing, the talus is kicking in.

Should there be a standard baseline position? I personally can't see why (in digital type). Certainly not a standard that extends beyond the display-text threshold.


kentlew's picture

The bounding box surrounding a highlighted letter may differ in different applications depending upon which vertical metrics the programmers decided to interpret. There are now several different metric fields in the font meta data. None of which is absolutely required to coincide with what used to be considered the body of the type, nor the virtual lines identified in Maxim’s diagram as Top/Bottom Bearing Line.

I don’t know which Adobe Illustrator is following.

In the Postscript format, the ascent + descent were obliged to sum to the em. But digital type is virtual. These various metrics have drifted further and further away from any physical absolutes, for various reasons.

Amado's picture

Should there be a standard baseline position?

Oh, I didn't say that there should be one. If there was one, I'd automatically take more interest in type designers who gleefully bucked the standard. But I am surprised that there isn't one.

And I'll just take anything that my Adobe products say with a grain of salt. Maxim's picture tells the tale.

I don't think I need the visuals, Kent; I see what you mean. Rather than aligning the Top/Bottom Bearing Line and letting the Baseline fly where it may, we now align the Baseline and let the (invisible) Top/Bottom Bearing Lines fly.

I thank y'all for taking the time and the interest! I learned stuff.

quadibloc's picture

But I s'pose what I'm driving at is: is it customary to take a typeface's x-height into consideration when choosing a line-height? Does a small x-height allow you to get away with tighter leading?

I see Hrant did already answer this one with a resounding yes, but I had to look carefully to find that answer.

Typefaces with a small x-height, such as the more authentic old-style faces, are commonly used without any leading at all.

Faces with large x-heights can be used without leading too - this happens when space is at a premium, and a cramped appearance is accepted. But if the intent is to get an appearance that is good by the standards of traditional book typography, leading will be used.

So the advantage of a large x-height is greater flexibility. A face like Times Roman can be used for book typography - the fact that there will be wasted space between the lines that could have been used by longer ascenders and descenders is relatively minor. But unlike Caslon Old Style or Cloister Oldstyle, Times Roman (or, for that matter, Century Expanded) could also be used without leading to achieve a high density of text.

So a printer that can afford only one text typeface probably is better off getting Century Expanded instead of Caslon or Garamond. (Of course, there is Caslon 540 which does have stubby descenders...)

Syndicate content Syndicate content