Writing for the ear is a different job

The first few hundred entries we wrote were good prose and bad listening, and it took us an embarrassingly long time to work out why.

The prose was fine on the page. Clauses balanced, references precise, the kind of writing that survives being read carefully by someone who can go back a line. Read it aloud and it fell over. Not in an obvious way. It just went slightly grey, and after four minutes you noticed you had stopped following it.

Here is what we learned, mostly by getting it wrong.

The reader can go back and the listener cannot

This is the whole thing. Every other rule is a consequence of it.

A reader who loses the story scans back half a paragraph and picks it up. It costs them a second and they barely register the cost. A listener who loses the story is just lost. They can rewind, and almost nobody does, because by the time they realise they are lost they no longer know where the loss started.

So a sentence for the ear has to be true on first hearing. Not clear on reflection. True immediately. That sounds like an argument for short sentences and it is not, exactly. It is an argument for sentences that resolve in the order the listener hears them.

Consider a sentence that opens with three subordinate clauses before the subject arrives. On the page that is elegant, a little architectural, and completely legible. Spoken, the listener is holding three things in memory with nothing to attach them to. By the time the subject shows up they have dropped the first clause. We wrote hundreds of those.

Pronouns are the main hazard

On the page, a "he" four sentences after the last named man is fine. The eye finds the antecedent without asking permission. In audio, four sentences is a long time, and if there were two men in the paragraph the listener has quietly attached the pronoun to the wrong one and is now following a different story than the one you are telling.

The fix is unglamorous. Say the name again. Say it more often than looks right on the page. In an entry about Anne Boleyn and Thomas Cromwell, we name both of them far more often than a printed essay would tolerate, and nobody listening has ever noticed, because the ear does not register repetition the way the eye does.

Numbers need a shape

"Between thirty and sixty percent" is a fact on the page. Spoken, it is a smear. The listener hears that a lot of something happened and retains nothing usable.

What works is giving the number a body. One in three. Twice the population of the city it replaced. Enough silver to pay an army for nine years. The precision does not drop. The number just gets somewhere to live.

The same is true of dates. "In the late fourteenth century" is a fog. "In 1381, in the summer" is a place you can stand.

Structure has to be audible

A printed piece can signal structure visually. Headings, white space, a paragraph break that tells you a turn is coming. None of that survives narration. The listener does not see the heading and does not hear the white space.

So the turns have to be said out loud. Not with signposting language, which gets tiresome fast, but with sentences that do the work of a heading. "That was the plan. What happened was different." The listener now knows a turn just happened and knows which direction it went.

We also found that an entry needs to restate its own spine about a third of the way in. Not a summary. One sentence that reminds the listener what the story is about, placed where attention naturally dips. On the page that reads as slightly redundant. In audio it is the thing that keeps people to the end.

Rhythm is not decoration

The last thing, and the one that took longest to accept. If a paragraph has no rhythm, a listener stops following it even when every sentence is clear.

Prose read aloud has a pulse whether you intend one or not, and a paragraph of nine sentences all roughly the same length flattens into a drone. Vary the length and the same information stays awake. A short sentence after three long ones lands like a door closing.

This is the part that resists rules. We do not have a metric for it. What we have is a habit, which is that every entry gets read aloud by a person before it gets narrated, and if the reader's attention drifts anywhere, that paragraph gets rewritten rather than defended.

What this means for what you hear

None of this is visible in the finished thing, which is the point. An entry that works sounds like someone who knows the subject telling you about it, which is the least writerly thing prose can sound like and by far the hardest to get to.

Anne Boleyn runs about sixteen minutes and went through this whole process. So did every other entry in the library. Whether it worked is not really ours to judge. Listen to it on a walk and see whether you are still with it at the end.

Follow the story.