Superior/superscript, inferior/subscript and ordinals

Wei Huang · January 2015

tl;dr What factors (i.e. intended use, design style, etc) inform the design of these glyphs and are there any standards?

⁂ ⁂ ⁂

Long version (my apologies in advance for the length! Footnotes in the next post as post is too long)

Superior/superscript¹ figures

Why does the height alignment of superior figures vary² so much between typefaces — is it to do with the anticipated uses or stylistic preference? Or lack of standard practise? I've seen the top of superior figures commonly aligned:

Slightly above ascender height:
(FF Dax)
Between the caps height and ascender height:
(Canapé)
With the ascender height:
(Samo Sans)
With the lining figure height:
(Lucida Sans)

(There's more variation depending on whether numerals are ranging, lining, hybrid, slightly-less-than-cap-height and also whether the ascender height matches the cap-height, etc).

Ordinals / Superior letters

Furthermore, how should ordinals (superior letters) be designed/aligned? Most commonly, I've seen typefaces that align:

The x-height of ordinals lower than top of lining figures and baseline of ordinals below baseline of superscript figures³:
(PF Din Display)
X-height of ordinals with lining figure height (also cap height):
(FF Dax)

b. variation where x-height of ordinals is aligned to the ascender height, when ascender > cap height:
(Lucida Sans)
Cap-height (or ascender) of ordinals aligned to top of ascender:
(Brill)
Baseline of ordinals with baseline of superior figures:
(Substance)
The x-height of ordinals are also aligned to lining figures

Brill does this too quite neatly as it aligns baseline and cap-height of ordinals to fit the superscript figures.

(Many other variations exist where more than one of these methods are true, and different ways of defining the same kind of the alignment are possible too)

Inferior/subscript figures

I've only seen either the baseline of inferior figures aligned to the baseline, or just below.

⁂ ⁂ ⁂

Conclusion: What's typographically correct, or appropriate depending on the intended use? What's plain wrong?

Wei Huang · January 2015

¹ I understand that there technically is a difference but perhaps not so important here for the design considerations?

² Conflicting recommendations on superiors/superscripts:

³ There's an old Typophile thread by Dan Reynolds on the topic but the question wasn't comprehensively answered and I don't understand half of the thread (debate about OpenType/spec). Still wondering why ordinals (superior letters) are commonly placed lower than the superior figures — John Hudson's answer shed some light on mathematical typesetting where exponents (superior figures) must be higher than superior letters, but I can't imagine everyone decided that their typeface is likely to be used for mathematical typesetting. I speculate that perhaps in the cases where the top of superior letters are lower than or aligned to the top of lining figures, the designer has made it more suitable for ordinal use, i.e. after numerals. Therefore they're neatly aligned to the lining figures – the superscript figures then have been placed without necessarily optimising the placement with the superior letters.

Adam Twardoch pointed out that superior figures and letters are both used for referencing so should have a consistent baseline — he also gave recommendations on four different "small-sized 'alphabets' in a font".

attar · January 2015

I was wondering the same not too long ago.

The debate about the limitations of the OpenType spec (well, registry entries) lies on the fact that there are two subscript features (`subs` and `sinf` – scientific inferiors) but only one superscript feature (`sups`). You may have noticed that there is encoded superscript i and n for basic mathematical purposes. The other superscript lowercase presented in Unicode are in the Phonetic extensions plane so I don't use these encodings: they have distinct semantics purposes and are not guaranteed but have compatible metrics across typefaces — I know that at least Source Serif Pro uses these (if you sum all superscripts offerings across all of Unicode you can get a full lowercase alphabet except the q) but it's a hack as far as I'm concerned.
My approach for superiors was to use `sups` for encoded superscripts (numerals, parenthese, i, n, and +=-) plus some (unencoded) extras like brackets, comma and dot [by the way: some fonts have monetary symbols in super/subscript or fractions but I can't see any proper use for that]. And with `ordn` I added a simple substitution to A-ZÉÚ + lowercase in addition to the standard ordinals substitutions (N°, 2.° and so forth).
So that people willing to use superscripts in a context distinct from mathematics can have access to unencoded i and n (it is just a matter of features ordering).

I don't see a use for `sinf` though, `subs` is good enough as I don't need a full subscript alphabet.

I made all superscript glyphs the same height like John Hudson did. Sometimes superscript numerals and letters can be used in combination like some of your examples show. But some designers give less vertical offset to superior letters so that they flow better with the text in abbrevations like Mr/Mrs or things like that. In that case you probably want two distinct superiors alphabets, with one that has the same height as the superscript numerals so that your typeface works with more complex superscript needs (well in fact, whether you need an additional alphabet depends of the vertical metrics of your font and it's intended purpose).
For this reason in the typophile thread you mentioned people were asking for an hypothetic `ssup` feature (scientific superiors), which would probably be the proper way to go about it.

That is a tricky question, I hope others can step-in and give their view about it.

Wei Huang · January 2015

Thanks for your response and explanation Adrien. If it's not too complicated, do you mind expanding on what gets ordered and how? `sups` before `ordn`?

So that people willing to use superscripts in a context distinct from mathematics can have access to unencoded i and n (it is just a matter of features ordering).

Chris Lozos · January 2015

I align baseline of ordinals with baseline of superior figures but I know of no rule.

attar · January 2015

`ordn` prevails in my features ordering, so that one can use both the full unencoded superscript alphabets and numerals by activating both `ordn` and `sups`.
Quite simply in your fea files the order you write your features is the order in which they will be applied (e.g., here is the canonical features ordering as suggested by Tal Leming). Since you use Glyphs I think it has its has its default features ordering too, there may be a way to override it but I don't know, I don't really do this kind of thing in the program. Usually I'll just tweak the fea file exported inside the UFO at hand.

Wei Huang · March 2015

Also, is it bad practise to use the unicode encoded subscript numerals for the sinf feature as well?

attar · March 2015

Personally I only provide the codepoints when it comes to subscripts.

Bhikkhu Pesala · February 2017

Wei Huang said:

Also, is it bad practise to use the unicode encoded subscript numerals for the sinf feature as well?

I believe so. Subscripted numerals should sit on the baseline for use as denominators, whereas Scientific Inferiors should bisect the baseline for use in formulae like H20 or H2SO4.

Superscript figures may align with top of caps or top of ascenders (better IMO), as their primary use is for footnote references, or for mathematical formulae e.g. E=MC².

I design my superscripted letters to align with the baseline of superscripted figures. I then include a set of subscripted letters for use in a comprehensive fractions feature, that can also create fractions like 1/x or a/b.

Ordinals use the superscripted letters: s, t, n, d, r, h, è, ú. I use a chaining context so that they are used after figures, and can be permanently enabled for a paragraph style. Spanish Ordinals should not be included in the Ordinals feature, in my opinion.

Calibri and other fonts do not implement Ordinals as a chaining context, so users have to enable them just for 1st, 2nd, 3rd, etc. This is not convenient at all. The superscript feature serves that purpose.

Nick Shinn · February 2017

some fonts have monetary symbols in super/subscript or fractions but I can't see any proper use for that

Bhikkhu Pesala · February 2017

I found that I needed to include parenthesis superscripts in my fonts for footnote references(¹²³), and some fonts also include a good number of accented glyphs again for references.(Satipaṭṭhāna) I have yet to come across a need for currency symbols.

¹²³ this does not look right.
Satipaṭṭhāna with all but the accented glyphs in superscript would be very wrong.

Khaled Hosny · February 2017

In TeX and OpenType MATH table super and subscripts are drawn as a one set of full sized glyphs with the optical correction needed to look right when scaled down, and then the layout engine scales them down and calculate their vertical position (which is rather complex and depends on many factors). This have several advantages, like not having to provides different superscript and subscript glyphs, and also having better fallback for glyphs not covered as they wold be simply scaled down which will look OK most of time, and definitely better than just getting a full sized glyph.

I just wish regular OpenType super and subscripts were implemented in a similar way.