Stress.

One syllable carries the beat

Say these slowly and listen for where the weight falls:

This is lexical stress — stress stored as part of the word. Say PORcupine as porCUpine and it does not become a different word so much as a mispronounced one.

Same letters, different word

Sometimes the stress pattern is central to distinguishing two words spelled identically:

Word boundaries can hang on it too, and finer distinctions than two. BLUEbird is a species. blue BIRD is Big Bird after an accident with a can of spray paint. every blue BIRD is any bird that happens to be blue. HIStory is the past; his STORY is one person’s account of it. In speech, stress and phrasing tell those two apart; in print, the space does that work. Print marks some of what prosody marks — just not stress itself.

Print barely marks it

English spelling encodes consonants and vowels far more fully than it encodes prosody. Aside from limited punctuation, stress goes unwritten. Some orthographic patterns help — certain word-final letter sequences and morphological endings shift the odds — but they are probabilistic cues rather than instructions (Ševa et al., 2009; Treiman et al., 2020).

So readers infer stress, drawing on stored knowledge of words, of morphology, and of the surrounding sentence. Most of the time this is invisible and effortless. It becomes visible in long unfamiliar words, in technical vocabulary, and in the moment you meet a name in print and have to guess.

Stress beyond the word

Stress also operates above the word, marking which part of a sentence matters — the subject of the Emphasis and focus page. The same words, stressed differently, answer different questions:

Once stress is distributed across a whole stretch of text, the alternation of strong and weak syllables becomes rhythm:

Word-level stress supplies the beats; rhythm is the pattern they form.

What happens when stress and rhythm are marked

If print leaves stress out and readers have to supply it, a natural question follows: what if you put it in? Two experimental studies from this project have tested versions of that question.

The first ran two experiments in which readers rated how helpful capitalized stress marking was — on the last line of poems in three meters, and on stress-alternating heteronyms embedded in prose (Gross, Winegard, & Plotkowski, 2018). In both, marking that landed on the syllable the reader would have stressed anyway was rated more helpful than marking that landed elsewhere. Note what this design does and does not do: it asked about preference, not performance, and no training was involved.

Two subtler results came out of it. Readers were more sensitive to marking on a word’s first syllable than its second — which fits the fact that roughly 85% of English content words carry stress initially (Cutler & Carter, 1987), so the first syllable is where expectation is strongest. And congruent marking was rated more helpful early in a sentence than late, consistent with early reading leaning more on the letters in front of you and later reading leaning more on what you have already built.

The second asked whether training with combined orthographic stress and rhythm markers can strengthen prosodic sensitivity itself (Gross, Roldan, Kiessel, Esselink, & Dunlap, 2026). The two were taught together, so the study does not isolate stress marking as the active ingredient — a point worth holding onto when reading the result. Two independent replications of the same randomized active-control experiment, 528 undergraduates in total, compared brief orthographic prosody training — passages annotated with stress and rhythm markers, practised with feedback — against a time-matched control on spelling and word usage. Prosodic sensitivity was measured by forced choice: readers picked the metrically well-formed version of a short verse line. Training passages were heavily marked; posttest items carried only a single marker, so that success depended less on the scaffolding and more on what had been learned.

Across both samples the prosody group outperformed the control at posttest on scores adjusted for baseline, and the advantage held on the minimally marked items. The condition effect reached significance in each sample separately as well as combined (d = 0.25 and 0.34; 0.31 overall). Participants in the prosody condition also reported greater perceived improvement in their prosodic knowledge, clearer instructions, and a more positive overall experience.

What the study does not show

Seven limits are worth stating plainly, because they bound what the result means.

Mean percentage correct did not go up. It fell from pretest to posttest in both groups because the posttest was deliberately more difficult and was not a parallel form of the pretest. Raw scores did rise, but the posttest contained twenty items rather than ten, so raw scores are not directly comparable either. The finding is the baseline-adjusted difference between groups at posttest, not improvement within either one.

The effect is modest. Cohen’s d was 0.31. Statistically reliable and replicated, but small; for an educational intervention, larger effects would likely require longer or more intensive training. Reliability of the outcome measure was also modest (α = .40 at pretest, .43 at posttest), which cuts both ways: it adds noise that may have attenuated the observed effect, and it makes any single estimate less precise.

The 2018 design cannot separate help from hindrance. Those experiments had no plain-text control, because a helpfulness rating cannot be given for text with no marking in it. So the gap between conditions might mean congruent marking helped, or that incongruent marking got in the way, or both. The paper says so directly, and interference is a live possibility — rhythmic irregularity is known to be costly to language processing.

Durability is unknown. The posttest came immediately after training, with no delayed retention test. Whether any of it survives a week was not measured.

Expectancy cannot be ruled out. Participants knew which lessons they had received, and the prosody group reported greater perceived learning — a pattern consistent with motivation or expectation contributing to the gap. Against that interpretation, the control condition was not prosodically neutral: everyone completed the same prosody tutorial and the same beat-judgment tests. Those shared features weaken an expectancy explanation without eliminating it.

Transfer to unmarked text was not demonstrated. The posttest used fewer and less salient markers than the training passages, but it was not marker-free. Whether the learning reaches ordinary unenhanced print remains open, and it is the question that matters most for practice.

No reading outcomes were measured. Word reading, fluency, and comprehension were not assessed — by design, since a brief training is unlikely to move them. The theoretical case for downstream benefits is strong; the empirical link was not tested here. Participants were also predominantly skilled college readers, so the results do not speak directly to adults with reading difficulties, English learners, or younger students.

What the studies do support is narrower and still worth having: adult readers’ prosodic sensitivity appears malleable. Brief training with orthographic stress and rhythm markers produced a small, reliable advantage on an immediately administered, minimally marked meter-judgment task. The lessons were self-paced and computer-delivered, with recorded audio models and immediate feedback but no live instructor and no one-to-one coaching. That makes orthographic support for stress and rhythm a plausible target for reading-intervention research rather than a solved problem.

Try it

Rhythm Reader marks the stressed syllable of every word in text you paste in, in your choice of four display styles — capitals, bold, enlarged type, or beat dots. Three of those — capitals, bold, and enlarged type — are the marker types used in the training; beat dots are an additional way of visualizing the same pattern. Every syllable is tappable, so when the tool’s guess disagrees with your reading, you can move the beat.

Rhythm Reader Pro adds phrase-level analysis and marks the main accent of each phrase, which is where word stress and sentence stress meet.

Further reading

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