Word stress · which syllable carries it The commonest habit among Mandarin speakers is not a wrong sound — it is giving every syllable the same weight. Mandarin is tonal, so every syllable carries a tone and stands on its own. Spanish gives each word exactly one stressed syllable and the rest must recede. This page measures how much stronger that one is.
Four steps to producing stress
Fix the position first (the rules decide it), then produce the difference (with duration), then hand it to the sentence.
Locate it with three rules
(1) A written accent marks the stressed syllable: papá, número, árbol. (2) No accent and the word ends in a vowel, n or s: stress the second-to-last syllable — hablo, examen, casas. (3) No accent and it ends in any other consonant: stress the last syllable — hablar, papel, ciudad. Together these cover nearly every word, and it is all readable from the spelling.
Use duration, not force
What this page measures: the stressed vowel is longer (median 1.27x) — not louder. The energy difference is around 1 dB, which does not read as stress. So the instruction is to stretch that vowel, not to shout the syllable.
Push the unstressed ones down
When stress does not come out, it is usually not that the stressed syllable is too weak — it is that the others are too clear. Shorten and loosen the unstressed vowels (an a drifting towards ə is fine) and the stress appears by itself.
Move it into sentences
Sentences add their own layer: the most important word in the phrase gets boosted. Practise the word first (the clips here are whole sentences — loop the target word), then the sentence. The two layers do not conflict, but mixing them at first is confusing.
Which one are you
The three most common cases.
Every syllable comes out equally heavy
The classic case. Mandarin syllables carry their own tone and even duration, and that habit transfers directly. The fix is mechanical: mark the stressed syllable with the three rules and stretch only that vowel.
I know where it goes but cannot make it audible
Knowing the position is step one. The difference has to come from duration (about 1.3x here), not volume. Shadow the clips and listen to how much longer the stressed syllable of the target word is.
Fine in one word, flat in a sentence
Expected. Sentence stress partly overrides word stress, and more so at speed. Get the word stable first; sentence-level contrast is a separate layer.
Corpus and acoustic measurements
Click any word in a sentence to play from that word. The pale band over the waveform marks the measured boundary of the target word.
Per-token data
Every number below is measured directly from the audio, not hand-annotated.
| Token | Accent | Sex | Target | Target word | Stressed vowel | Stressed / unstressed |
|---|
Where these numbers come from
This page uses human recordings, not speech synthesis.
Corpus: VoxForge Spanish, 12 recordings from 13 volunteers (prompts are public-domain text). 16 kHz mono, 5–12 seconds each. Each speaker's gender and dialect are labelled as recorded; accents were not flattened into one.
Why human recordings: we tried this page with synthetic speech first — the duration ratio was right in 1 of 5 minimal pairs and the energy ratio in 2 of 5, with one case where the stressed syllable came out quieter. The synthesis engine did not actually implement stress. In human speech the effect is real.
Boundaries: MMS_FA forced alignment, 2053 of 2053 words across 130 recordings came back char-anchored (boundaries on real character edges). Nothing is estimated.
Syllabification: maximum onset principle — a single consonant between vowels goes with the next syllable; with two consonants, a legal onset cluster such as pr / tr / bl goes with the next syllable whole, otherwise the first stays as a coda. A stressed weak vowel followed by another vowel is a hiatus: a-co-gí-an has four syllables, not three.
What the ratio compares: the vowels themselves, not whole syllables. Syllables also contain onsets and codas, and closed syllables or consonant clusters inflate them, drowning the stress signal. Measured on whole syllables the direction is right only 64% of the time; on vowels alone it is 78%. This matches the phonetics: Spanish stress is realised mainly as nucleus lengthening.
Whole-corpus result: across 580 polysyllabic words (3+ syllables), the median ratio of the stressed vowel to the mean of the other vowels is 1.27x, and 78% are above 1. By stress position: penultimate (llana) 1.27x, antepenultimate (esdrújula) 1.26x, final (aguda) 1.15x.
These 12 clips were selected: to make the difference audible we picked 12 words with clear ratios (1.50–2.80x) covering all three stress positions. Do not read those ratios as the typical value — the typical value is 1.27x.
The counter-examples are real: 78% means roughly one word in five does not follow the trend, because connected speech also carries sentence stress, tempo and position effects. So this page says this happens, about this much — not always.