[This is the first draft. As I said when it was originally posted on /gd/, it probably veers between being esoteric and patronising, hopefully that'll be corrected in time. Images that clarify and support things will be supplied.]

Specifying type involves answering a series of questions. Some designers might not ask these questions in depth, not all will need to, but in many these questions are answered through trial and error and asking the guy at the next desk. The first question is this:

Are you Massimo Vignelli?Edit

If you are then you already have a pool of six typefaces (Times New Roman, Helvetica, WTC Our Bodoni, ITC Garamond and one other that I think is Futura) picked out and will use these, and only these, for your projects. You will do so firm in the belief that these are appropriate for all uses.

If you are not: good, you had no hand in WTC Our Bodoni and tonight, when you lie in your bed, you can pull the duvet up to your chin, let a small smile play across your lips and rest easy. Your hands and conscience are clean.

That being said, you also have to answer more questions. Broadly these are:

  • What is the content?
  • What is the context?
  • How is that content best served?

They kinda interrelate so I'm going to break them down further:

How complex is the content?Edit

A poster's headline is fairly simple, you can pretty much use whatever, but an Icelandic/Polish dictionary is very complex. For the body copy alone, the dictionary will need a larger type family that has the accented and uncommon characters both languages use (Icelandic uses ð, Þ and ý, amongst others, Polish has other weird shit like the ę (which is usually very poorly drawn in Grotesks) and the ż). Over the course of a single entry, the reader has to be able to easily distinguish between two languages, pronunciation, definition, examples and etymological notes. To make this easy, you not only want to have a range of subtly different weights (a light, regular, semibold, bold might be enough), but for these weights to be have roman and italic characters, small caps, oldstyle and lining numerals in tabular and proportional forms, along with subscript and superscript numerals, symbols like the dagger, and maybe, just maybe, upright italics and petite caps.

Each of these has a function.

  • Oldstyle proportional numbers are best in text, lining proportional numerals are for folios or to sit amongst capitals. Tabular numbers are, as you guessed, for tables and that they're monospaced means they look so very neat in columns. Most of the time, tabular lining's set as the default.
  • Small caps should be used for sections of text that are all uppercase, acronyms and initialisms longer than three (generally speaking) letters long (US in caps. NATO and BBC small caps), or to indicate the name by which someone with a long formal name will be designated ('Mao' in Mao Zedong). They are not to be used for emphasis.

A poster, in short, can use Kada. A multilingual dictionary will want a typeface like David Březina's brilliant Skolar.

Is the content extended?Edit

If it is then you want a typeface that is actually brilliantly readable. If you believe myfonts or some type designers on behance, that's every typeface they're trying to sell, but they're either ignorant or talking bilge: Lobster should not be used to set running text. Even typefaces like Franklin Gothic (especially condensed) or Helvetica become problematical after several hundred words because they have small counters (the hole in o's and g's), small apertures (the gap between the top and bottom of an e), slightly too small characters or some of their characters look too alike at small sizes.

If you want a sans then humanist and humanist influenced sans like Syntax (Hans Eduard Meier did some brilliantly readable sans), Gill sans and Frutiger are generally safe choices; as is Legato. If you've got a good cut of it and you're a fairly knowledgeable typesetter, Futura can be very rewarding (Bobby Bringhurst recommends)

If it's a serif, then you're generally operating with more forgiving tools but —

Is the typeface appropriate for the size it's to be printed at?Edit

If I've remembered to do this while copying this vast sea of text, there should be a picture on this post which shows you five different versions of Garamond. The first four are from adobe's Garamond Premier Pro but the fifth is from the copy of Garamond that came packaged with my computer. The fifth version is fairly forgiving, sure, but Slimbach's Garamond Premier Pro is better. The optical adjustments that are made as the type gets smaller (the stroke contrast becoming less pronounced, the thin strokes and details becoming thicker or being done away with, the increase in the x-height and the looser kerning) fine tunes it for the intended sizes. Most typefaces don't have this attention to detail, or their digital version has been focussed toward a specific purpose from the get go, so they can suffer when printed at sizes they weren't intended to.

A good example of this is most versions of Didot: this version's been quite well hinted, but as it shrinks the thick strokes remain as the thin strokes begin to disappear. If it was printed at 7pt, the thin strokes would all but disappear and the type would be much harder to read.

In the case of sans serifs, again, the distinctions between a grotesk typeface like Helvetica, a humanist like Frutiger and a typeface that's been tailored to the task like Bell Centennial become important. Helvetica's design, (I shouldn't keep on picking on Helvetica, I'm sure you guys are getting bored of it and I don't actually totally hate it) specifically its slightly heavy weight and closed apertures, means that when it's printed at very small sizes (<7pt) the e's become indistinguishable from o's, the i's from l's and the numbers, already chancing it, are a bloody disaster. This is not an effect helped by the paper it's printed on absorbing ink. Frutiger weathers slightly better, characters like the open mouthed e remain distinct, although the i's still suffer.

Bell Centennial, however, thrives at small sizes. Everything that's weird or ugly suddenly makes sense when it's printed in a phone book and then magnified so you can contrast how the characters are printed with how they appear on the page. Which reminds me:

Where is the content?Edit

Is it on the web, is it on a pdf, is it on a passing motorway sign or on piss poor paper? Bell Centennial was built to be printed very small on piss poor paper and consequently the crotches of all the characters have quite pronounced ink traps bitten out of them. Transport was designed for motorway signage. It has to be read as you speed past the sign in poor weather so there's no need for ink traps but every need to make the characters individually distinct. The foot of the l stops it from being confused with the i, the open mouth of the e stops it from being confused with the o, and the e's quite low cross bar helps erase in potential confusion with the c. Don't quite know why they didn't make the numbers as distinct, but what the hell.

If it's to be used online, is it available through fontdeck, typekit, or fontface; or is it available on everyone's computer already? And as it's on screen, is it well hinted? Akkurat's hinting's okay, Akkurat Pro's is cleaner again, but when contrasted with a typeface like Lucida Grande, which was designed for the screen and was extensively hand hinted, it's left looking a little bit furry at small sizes. The bowls of characters like the o and the u seem a bit blurred and the apertures of e's and c's almost closed. Another example is contrasting an older cut of Times New Roman with Georgia - there's that same shift from slightly bold and dark in weird ways to crispness and clarity.

Is the content historical or does it have associations you should refer to?Edit

Is setting Dante's Inferno in Foundry Gridnik witty or too anachronistic? This partly depends on its context: if it's a staging of Dante's Inferno which uses Wim Crouwel's deteriorating relationship with the one client he never really got on with as a metaphor for the various rings of hell, then it might be quite appropriate. But if it's for a new translation being published by the Folio society, then a typeface that has its origins closer to fourteenth century could be more appropriate, Bembo for instance.

Using typeEdit

I was going to supplement this article with an in-depth series of barely coherent ramblings about the distinctions between real small caps and fake ones, but these two websites cover much of the same detail and it seems redundant to replicate them here.

It should be noted that while Butterick's Practical Typography covers a lot of detail about composing type; when to use an en dash over a hyphen; it's marred somewhat by his prejudice against system fonts (some of which, like Georgia and Verdana, are fantastic) and his pimping his own shit (which I'm not too sure about). That being said, these concerns are minor.