Emphasis and focus.

The same sentence, two emphases

Read this pair of stories. The last sentence is identical in both; only the lead-up differs.

In the first story, we already know someone went overboard; what we do not know is who. In the second, we know Sam is the one in trouble; what is new is what happened to him. The emphasis lands on whatever the story has left unresolved.

The words have not changed. Only the context has — and that is enough to move the sentence’s main emphasis.

Swap the emphases and both sentences go slightly wrong — not ungrammatical, just off, like an answer to a question nobody asked. If your browser runs JavaScript, the button above will do the swapping for you; it is worth feeling rather than taking on trust.

New information takes the accent

This is focus: prominence marking the part of an utterance that carries the news. Speakers do it naturally, and listeners rely on it to interpret what is new. Information already established gets de-accented; what is new or contrastive gets the main accent (Halliday, 1967; Selkirk, 1986).

Newness is not all-or-nothing. It varies along a continuum of accessibility, and how a referent was last mentioned matters (Baumann & Grice, 2006). Mentioning someone in the immediately preceding sentence makes them old news; mentioning them several sentences earlier leaves them more newsworthy again. That is precisely the lever the canoe stories pull.

Focus can also be contrastive, and then it narrows sharply. Asked why you are not having French toast, the fitting reply puts the accent on the contrast: I’ve forgotten how to MAKE French toast (Ladd, 1980). French toast is given — it was in the question — so it steps out of the way.

Where the capitals come from

This site marks prominence with capitals, and that convention is borrowed rather than invented. Dictionaries use it in pronunciation guides: aioli, /ahy-OH-lee/. Comics use it for impact — POW!, OUCH!. And an email written in capitals reads as raised voice, which is why etiquette guides advise against it.

Those conventions make capitals useful as a research tool. A reader meeting SAM in print does not need to be taught what it means; the typographic habit is already there.

Even function words can carry it

Words like he, can, that, and a normally recede. They refer to established content, they are prosodically weak, and their vowels reduce. But when a function word is the news, it can take the main accent:

This has limits. Function-word focus is possible, not automatic: when speakers were given contexts that should have accented a preposition, they did so only about a third of the time, preferring to spread the accent across the whole phrase (German, Pierrehumbert, & Kaufmann, 2006). Prepositions embedded within larger phrases turn out to be relatively poor candidates for nuclear stress.

Do silent readers do this too?

Speakers and expressive readers plainly mark focus. Whether silent readers do the same is harder to establish, because the inner voice cannot be observed directly. The approach taken in the 2014 study from this project was to put capitals on the page and see whether readers preferred them where spoken focus would fall (Gross, Millett, Bartek, Bredell, & Winegard, 2014).

In the first experiment, readers judged how helpful the capitals were. They judged the capitalization more helpful when the emphasized word was the newsworthy one, and judged de-emphasized wording more helpful when the content was already given. On a five-point scale, congruent versions averaged 3.83 and 3.36 against 2.54 and 2.31 for the mismatched versions.

The second experiment replaced ratings with reaction time and added function words. Readers judged congruent items faster than mismatched ones — about 3,290 against 4,157 milliseconds for content words, and 3,558 against 4,399 for function words. Reaction time is harder to consciously manipulate than a preference judgment, which is the point of running both.

What the study does not show

It does not prove that anyone heard a voice. The paper is explicit about this: readers behaved as though guided by a prosodic inner voice. Some other linguistic feature that varied along with the manipulation could in principle produce the same pattern.

The accuracy data are inconsistent. Reaction times came out as predicted, but accuracy (proportion correct) did not follow suit — for function-word items it was actually higher on the mismatched trials. The paper reports this rather than setting it aside, and it is a reason to lean on the timing result rather than the accuracy result.

The readers were undergraduates. Skilled adult readers at one university, judging short constructed stories. Whether the same holds for developing readers, struggling readers, or people reading in a second language is a separate question.

Try it

Worth being straight about a limit here. Rhythm Reader and Rhythm Reader Pro model stress and rhythm — properties of words and phrases. Focus is different: it depends on what the surrounding discourse has already established, which the tools cannot see. Pro predicts the main accent of each phrase using the general tendency for English nuclear stress to fall late, and that will sometimes disagree with where focus actually belongs.

Which makes the tools useful in a particular way here. Paste in a sentence, look at where the beat is placed by default, then tap to move it where the context requires. The gap between the two is the work focus is doing.

Further reading

More Learn pages — intonation and implicit prosody — are in development.