Voice
Every typeface class arrives with a history, and the history arrives whether you invited it or not. Set six words in eight faces and you get eight different speakers.
The same six words in all eight faces. Read each one and notice what kind of organisation you imagine saying it. Select a row to inspect it.
Source Serif 4 — transitional serif, 2014–2021, from 18th-century models
Bookish and even-tempered, an essayist's voice rather than an announcer's.
Belongs: Long-form reading, which is why it sets every paragraph on this site.
Where the associations come from
None of these effects are mystical. A didone reads as fashion because two centuries of fashion publishing chose didones, starting with the magazines that could afford the presswork its hairlines demanded. A grotesque reads as institutional because grotesques were the workhorses of nineteenth-century job printing — tickets, forms, signage — and the association stuck. A monospace reads as technical because for decades it was mechanically impossible for a computer to produce anything else. The class carries the residue of everything it has been used for.
This is why the voice of a typeface is not a property of its shapes so much as a property of its company. The reader has seen these forms thousands of times before arriving at your page, and every one of those encounters votes on what your sentence sounds like. You do not get to opt out of this system. You only get to choose your position in it.
Reading the specimen honestly
Go back to the instrument and read the sentence in Bodoni Moda. For most readers it becomes a promise made by a brand — expensive, slightly aloof, not entirely believable. Now read it in Public Sans, where it becomes a government service describing a process. In JetBrains Mono it stops being a promise at all and becomes a status message, something a system said rather than something a person meant. The words never changed. The speaker did.
One face in the list is doing double duty: Source Serif 4 is both the transitional-serif specimen and the face this paragraph is set in. That is a deliberate disclosure rather than an economy. If the lesson is that a face is a voice, the site should say plainly which voice it chose for itself, and why: a transitional serif is the essayist's register, unshowy and durable, which is the register this site wants.
Choosing, in practice
The practical method is to name the speaker before you audition a single face. Decide who is talking — an institution, a friend, a machine, a shop — and how they should sound, and only then open the specimen book. Choosing the face first and rationalising afterwards is how projects end up with a wedding-invitation didone labelling a settings panel. The face was beautiful. It was also the wrong speaker.
And when two candidates both fit the speaker, prefer the one with the better small sizes, the real italics, the tabular figures — the unglamorous equipment. The voice gets you the first impression. The equipment is what you live with for the rest of the project.
Choose the speaker first. The typeface is casting, not decoration.