To the lowercase. It is nice to make special "small cap figures" for the caps, too… yet another figure option. But we've started to do this with some Linotype designs (Avenir Next, for instance). This means that an OpenType font has six figure styles, not including fractions, numerators, demoninators, etc.:

Proportional Lining
Tabular Lining
Proportional Small Cap Figures
Tabular Small Cap Figures
Proportional OsF
Tabular OsF

There isn't a pre-defined feature for SC figures, but you can write the small caps features so that if the number default is lining fuigures, and you turn on a small caps feature, the numbers will switch to SC figures when the letters switch to small caps.

i would agree with Dan and say with the lowercase as they are more correctly referred to as text figures with the assumption being that they are to be used in running text with the lower case. i also agree that if your smallcaps are taller, having a set of small cap figures would be useful. perhaps hybrid figures could be designed to work with the small caps?

It depends on the design, e.g. how much lowercase and smallcaps height differ. Sometimes, lowercase height osf look a bit small, so smallcaps height is not too bad, at the same time they work with small caps. Making a special set of small caps numerals of course is nice too, but there are already so many numerals to do ...

Good question. I don't think there is any rule about this. Adobe Garamond's OSF are between the x and the small-cap height. Karen Cheng gives examples of OSF both at and above x-height.

The best choice may well depend on the design. I'd be interested in people's views of the pros and cons of different choices.

I'd say trust your eyes not the x-height. Nothing worse than puny little old style numbers. If the typeface has a small x-height you'll probably want to go a bit highter and vice versa.

Quasi-simultaneous posting ... :)

Adobe Garamond’s OSF are between the x and the small-cap height.

In addition to my previous post: x-height for osf may work too -- in smallcaps context -- for numerals' ascenders compensate for their smaller x-height.

after reading more and really thinking about it, i'm agreeing that x-height might be too small depending on the design.

Another important question is the height (meaning vertical size, not position) of 3 5 7 9 and 6 8. Most designers simply shift cap-height numerals down to get the oldstyle 3 5 7 9. However, depending on the typeface they can be too tall and might descend even further than the descenders, which just looks wrong. This is a mistake seen quite frequently.

As a second step, I would always adjust (decrease) the size of 3 5 7 9. As a result, 6 and 8 also have to become smaller, making them smaller than the lining numbers.

Myriad and Minion are two examples where this is done particularly well.

This means that an OpenType font has six figure styles,

If the lining figures aren't full cap height, you also need another big set for the "case" (all caps) feature!

I don't really think tabular small cap figures serve any purpose, and how would they be supported in the application menu?

I propose that the logical disposition is to connect the "Small Caps" menu choice with proportional Old Style Figures, on the assumption that they will likely be used in text with U&lc -- whereas "All Small Caps" would bring up lining, proportional, small-cap height figures.

Or perhaps "Small Caps" should have no special provision for figures, but just leave the default or whatever other figure style is applied to the text.

But "All Small Caps" should definitely have lining, proportional figures, for use in places like a page folio, or in very small captions -- as a sort of optically scaled all-caps setting.

**

Back to the thread: my preference, where small caps are significantly larger than x-height, is for OSF with an "x-height" that matches small cap height. Otherwise things like postal codes will look scrappy; the proviso is that the form of "zero" be different than lower case "o".

I don’t really think tabular small cap figures serve any purpose, and how would they be supported in the application menu?

They're supported via the other tabular/proportional numeral options. So tabular smallcap figures are technically smallcap variants of tabular lining or oldstyle numerals, amd proportional smallcap figures are technically smallcap variants of proportional lining or oldstyle numerals.

But “All Small Caps” should definitely have lining, proportional figures, for use in places like a page folio, or in very small captions...

And in alphanumeric postal codes.

So tabular smallcap figures are technically smallcap variants of tabular lining or oldstyle numerals, amd proportional smallcap figures are technically smallcap variants of proportional lining or oldstyle numerals.

Dude, that is so featuritis.

Dude, whaddya call this:

"If the lining figures aren’t full cap height, you also need another big set for the “case” (all caps) feature!"

?

If the lining figures are not cap height, and you're proliferating figures to match every situation, then the lining figs should be cap height. You wouldn't need any 3/4 height figs.

In general, my answer to your original question is "x-height or lowercase", for the reasons named above. They are meant for use with lc, and they have extenders, which increases their apparent size, so they don't need to be as big as small caps. But I agree, if the lc is very tiny compared to SC then the osfs should be somewhere in the middle where they would work with both, or else make a set for every single alphabet. Keep in mind, though, figures aren't read like words, and they only need to blend into text so much, whatever height.

This could become its own unhealthy obsession, however. Or at least delay your typeface release by months. "I've got superior and inferior figures for my demi-petite caps!"

demi-petite caps

hmmmm.... i LIKE it! >^p

If the lining figures are not cap height, and you’re proliferating figures to match every situation, then the lining figs should be cap height. You wouldn’t need any 3/4 height figs.

Unless you want to make these the default figures for applications that do not have OpenType Layout support or, as in the case of one multilingual typeface I'm working on, you want them as the default style because they match the height of the Hebrew letters.

> with the lowercase as they are more correctly referred to as text figures with the assumption being that they are to be used in running text with the lower case

Just for the record. If memory serves, in German the OsF are often called Minuskelziffern. Of course, that does not mean they must always be x-high.

I agree with William's earlier point; the osf x-height must be in context of the rest of the design.

I've just been browsing the Karen Cheng book (which has finally arrived in New Zealand, and is of course entertainingly discussed at http://www.typophile.com/node/18679) and she points to a number of possibilities here. There don't seem to be any hard-and-fast rules other than which osf descend and which ascend, and which go to a mean height that agrees with the lowercase and/or the small caps.

the proviso is that the form of “zero” be different than lower case “o”.

By the same token, the form of 'one' must be different from lowercase 'l' and uppercase 'I' – if you don't believe me, just have a look at Gill Sans.

There don’t seem to be any hard-and-fast rules other than which osf descend and which ascend...

With the caveat that in French Romantic types the 3 and 5 are often ascending.

John that really makes me giggle – those crazy French romantic types! Always up when they should be down...

I notice that 'four' looks quite massive compared to everything else on the line, but guess I need to amend that earlier statement to just 'there are NO hard-and-fast rules...' which doesn't get any closer to answering Nick's original question. um.

John, looking at this example, what would be the proportion of the number 2 to it's lining numerals counterpart?

Generally the same as the proportion of the lowercase x-height to the cap height, i.e. it will vary from typeface to typeface.

I was looking at a few type specimens to confirm the wisdom here; but while in Cloister Oldstyle the old style figures are larger than the x-height, in Caslon Oldstyle, they're smaller.

And what would you call the figures in Bitstream’s Venetian 301? ‘Scotch’ figures?! The latest, OpenType, version of Venetian 301 also has the cap-high lining figures, proportional and tabular. Those shown in the top two lines are tagged ‘oldstyle’. For comparison, this is what the old-style figures look like in the ‘Std’ digital version of the original design, Monotype’s Centaur. Their height is shorter than the l.c., let alone the s.c. B.t.w., the ‘demibold’ weight of Venetian 301 seems to work better in the body text.

And what would you call the figures in Bitstream’s Venetian 301? ‘Scotch’ figures?!

I like descriptive names for figure styles -- Scotch? that's a whisky --, so I would call the Venetian 301 numerals 'three-quarter (cap height) ranging'.

Since this post, I’ve taken to proliferating figure sets in my OpenType fonts, a practice which I had previously derided as featuritis; however, several other threads at Typophile pointed me in this direction, in particular one in which the need for tabular small-cap-height lining figures was convincingly argued, with layouts shown. However, I do vary the figure sets according to the typeface.

If the lining figures are not cap height, and you’re proliferating figures to match every situation, then the lining figs should be cap height. You wouldn’t need any 3/4 height figs.

As the Venetian 301 samples show, three-quarter figures (whether lining or ranging) are an alternative to old-style figures in a mixed-case setting. IMO both are required, as they cater to two distinct preferences amongst typographers (and their clients).

Two sets of lining figures are particularly necessary for grotesque faces, where the default figure style is lining. For these faces, three-quarter figures have always been a compromise: lining to sort-of match capitals and weighted to match lower case, but with height too short and stem width too thin for all-cap settings.

Hence this solution (Figgins Sans):

• Cap-height lining figures, with upper-case stem thickness, for use in all-caps text
• Three-quarter lining figures, with lower-case stem thickness, for use in U&lc text
• Small-cap-height lining figures, with small-cap stem thickness, and wider proportions, for use in all-small-cap text (c2sc feature)

I would call the Venetian 301 numerals ‘three-quarter (cap height) ranging’.

‘Three-quarter’ they may be but, John, they are not quite ranging. Nor can they be considered ‘old-style’:

I use the term 'ranging' very generally to refer to numerals that are not all aligned to a common height.