Serif

Every serif is the record of a tool. The capitals cut into Trajan’s Column end in flared terminations that one influential account explains as a brush stroke the chisel then followed, and the shape has outlived the technique by two thousand years. The word does double duty, which trips people up: a serif is one of those strokes, and a serif is also any typeface that has them, so “the serif on the R” and “set it in a serif” are both correct.

How the stroke meets the stem is most of what separates one serif face from another. A bracketed serif curves into the stem on a fillet, the way the old style faces of the sixteenth century do. An unbracketed serif meets it at a corner as a flat hairline, which is the signature of a Didone such as Bodoni. A slab serif carries the same weight as the stem and reads as a rectangle stuck on the end. The usual genre list (old style, transitional, Didone, slab) is really a sequence of answers to that one question, plus where the thin part of a curve falls and how much the strokes vary in weight.

“Roman” turns up as a loose synonym and means two different things. In its strict sense it names the upright form of a face, the one an italic is set against. In looser use it names the whole inherited tradition of serifed book faces, which is why a foundry can call a face “roman” and mean neither more nor less than “not sans”.

The claim that serifs guide the eye along the line, and therefore read faster, is repeated far more often than it is supported: controlled studies keep failing to find a reliable difference for body text at ordinary sizes. What did matter was resolution. On a 72 dpi screen a hairline serif had no pixels to live in, so it smeared or vanished, and interfaces settled on sans faces for a generation. That constraint is gone. Serif body text on a phone is now unremarkable, and the choice is back to being one of voice.