Legibility

Why is a page set entirely in capitals harder to read than the same words in mixed case?

Capitals look authoritative, and designers reach for them when something must be noticed. Yet set three paragraphs entirely in capitals and readers slow down, lose their place, and quietly give up. The letters are larger and the strokes are no thinner, so nothing has been made less visible. Something other than visibility must be doing the work — which raises the real question: what is a reader actually using, that capitals take away?

Start with what reading physically is, because the intuition of a smooth glide along the line is wrong. The eye moves in jumps: it fixates for around a fifth of a second, jumps forward roughly the width of eight characters, fixates again, and every so often jumps backwards to recheck. Only a few characters around the fixation point are sharp; everything else is progressively blurrier and more crowded by its neighbours.

That has a consequence. During each fixation, the region just ahead of where you are looking is partly processed already — not read, but sampled for length, shape, the presence of a tall stroke or a descending tail. By the time the eye lands there, the word is half-guessed, and the fixation confirms rather than deciphers.

Now ask what capitals do to that. Two things, and the dull one matters more than people expect. Capitals are wider, so the same sentence occupies more space; more space means more fixations, more jumps, and more line breaks for the same content. Nothing subtle there — there is simply more page to traverse.

The second is about shape. In mixed case, a word has a silhouette: ascenders standing up in b, d, h, k, l, t, descenders hanging down in g, j, p, q, y, and a body of x-height letters between. "Ophthalmology" and "OPHTHALMOLOGY" carry the same letters, but only the first has an outline. Set everything in capitals and every word becomes the same featureless rectangle, differing only in length. The advance sampling still happens — but there is much less for it to find.

Does that mean we read by word shape? Here we should be careful, because this is where the popular account overreaches. The slowing is solid and has been measured repeatedly since the mid-twentieth century; continuous text in all capitals is read measurably slower than the same text in mixed case. The explanation is contested. A strong line of work argues that we identify letters in parallel and assemble the word from them, with the outline as a supporting cue at most, and that a large part of the capitals penalty is simply unfamiliarity: almost nothing we read at length is set that way, so we have far less practice.

The same caution applies to the claim you hear most often in design arguments — that serif type is read faster than sans-serif. Controlled studies of continuous reading have repeatedly failed to find a reliable difference. Whatever effect exists is small and confounded with what readers are used to.

So which typographic levers actually earn their reputation? The ones that change how much usable signal reaches the eye per fixation. Size relative to x-height, since two faces at the same point size can differ substantially in the height of their lowercase letters, and it is the x-height a reader effectively sees. Letter spacing, because letters that crowd their neighbours are harder to identify away from the fixation point. And line length and line spacing, because a long line makes the return sweep to the next line error-prone. These are unglamorous, and better supported than the arguments people prefer to have.

Two distinctions do most of the clarifying work here.

The first is between legibility and readability. Legibility is a property of characters — can this letter be told apart from that one at this size and distance. Readability is a property of a text as set — can this passage be read at length without fatigue. A face can be highly legible on a road sign and exhausting over ten pages, which is why arguments that mix the two never resolve.

The second is between an effect and its explanation. That all-capital text is read more slowly is about as settled as findings in this field get. Why is genuinely open, with word shape, sheer width, letter crowding and lack of practice all in contention. Much of this literature also involves modest effect sizes, laboratory tasks, and comparisons confounded with familiarity — which is exactly why the well-supported advice is boring and the confident advice usually is not.

None of this makes capitals wrong. For a single word, a road sign, or a label read in one glance, the width penalty barely applies and the uniform rectangle can be an advantage. The finding is about continuous reading, and applying it to a two-word heading is overreach in the other direction.

• How visual crowding limits letter identification away from the point of fixation.
• Why dyslexia-specific typefaces have generally failed to outperform ordinary well-set text in controlled trials.

Every term the collection defines is gathered in the glossary.