amadunia

Grammarsettled — decided September 3, 2026 · 1 open

Stress

Taught on the front page and in Lesson 01.

The rule

Stress falls on the second-to-last syllable of every word.

A one-syllable word carries it on the syllable it has. There is no exception, no marked word, and nothing to memorise per word.

DU-niaworld
a-ma-DU-niathe language’s own name
MER-sithank you
ka-LI-maword
sa-FA-rijourney
MII
ESis

What counts as a syllable

One syllable per vowel group. A vowel pair is one nucleus, not two.

This is the question that had to be settled first, and it is settled by the phonology rather than by choice: only five vowel pairs exist — ai, ao, au, ia, ua — three vowels in a row are forbidden, and those five are exactly the sequences the source languages pronounce as a single glide. Counting them as one nucleus makes syllable-counting mechanical: count the vowel groups.

Syllables
dunia2du-nia, not du-ni-a
akua2a-kua
mausim2mau-sim
familia3fa-mi-lia
uan1one vowel group

Under this definition the 300 roots are 57 of one syllable, 209 of two and 34 of three.

What stress does to the sound

It makes the syllable louder and a little higher. It does not change it.

No vowel reduces, none lengthens, and none shifts quality. The a’s in a-ma-DU-nia are the same a whether stressed or not, and dunia said with the beat on DU has exactly the sounds spelled in it. This is how Indonesian, Swahili and Japanese work, and it is the opposite of English, Russian and Portuguese, where unstressed vowels collapse toward a schwa.

That is not a stylistic preference. Design rule 2 says the language is written as it is spoken; a reduced vowel would break it the first time anyone said a word out loud, because the letter on the page would no longer be the sound in the air. Stress had to be the kind that leaves the spelling true, and this kind is. See pronunciation.md, where the vowels are.

The hyphen does not join words for stress

A hyphen joins two words for meaning — a plural, or a number. Each half keeps its own stress.

WrittenSaid
anak-anakA-nak A-nakchildren
mi-miboth halves stressed, each on its only syllablewe, not including you
des-uanDES U-aneleven

Why the second-to-last

It is what the largest source families already do. Indonesian and Malay, Swahili, Polish and Quechua stress the penultimate as a rule, and it is the default in Spanish and Italian. Austronesian is the family with the widest reach in this dictionary, and Niger-Congo is not far behind.

It is what the one constructed language that has actually been spoken for a century does. Esperanto has stressed the penultimate since 1887.

It is the most forgiving choice available. In 266 of the 300 roots — 89% — the second-to-last syllable is also the first, because 57 roots have one syllable and 209 have two. A learner whose own language stresses the beginning of a word is right about nine words in ten without being taught anything, and can only go wrong on the 34 three-syllable roots.

Rejected

Initial — always the first syllable. Czech, Hungarian, Finnish, Icelandic, Latvian. It agrees with the penultimate on 89% of the dictionary anyway, so it buys almost nothing, and it disagrees with Austronesian and Swahili — the two families this language leans on hardest — on the third where it differs.

Final — always the last. Turkish, Persian, French, Armenian, Hebrew. A real tradition and a large one, but it is the option that disagrees with the penultimate everywhere except one-syllable words, and it disagrees with Esperanto, Indonesian, Malay and Swahili at once. Choosing it would mean the widest-reach family in the dictionary is also the one whose speakers say every word wrong.

Free — no rule. What the language did until today, and the poem is why it could not stay that way: DIN lai and din LAI are different lines, so nothing could be sung, scanned or taught by ear.

What it unblocked, the same day

The poem was written while this was undecided and said so — being said aloud was “the thing the language is least ready for”. It was scanned within the hour, and two things came out of it that nobody could have seen before. Its paired lines match to the syllable, 6 and 6, 4 and 4, 6 and 6, 5 and 5, though the poem was built on parallel words and nothing counted beats. And its closing line, Kita ama dunia una, is the only regular line in it: X . X . X . X ., four trochees with nothing left over. That was luck rather than craft, because when the line was written no rule could say where its beats were.

Nothing was at risk — and what it cost anyway

No two words in Amadunia are told apart by stress, and none ever can be — that would be a minimal pair, which the language forbids. No word changed meaning, and no text needed rewriting.

But a rule that fits one tradition cannot fit them all, and it is worth saying plainly who pays. Taking the first source each entry names, and counting only roots of two syllables or more:

Roots
First source stresses the penultimate — Indonesian, Malay, Swahili, Spanish, Italian, Polish87, and they keep their beat
First source stresses the last — Turkish, Persian, French34, and their beat moves

anahtar is a-nah-TAR in Turkish and a-NAH-tar here. pencere is pen-ce-RE and becomes pen-CE-re. ates, yatak, korku, burun, bage, mersi all shift the same way. A Turkish or Persian speaker meets eleven per cent of the dictionary saying their own word with the beat in the wrong place.

That is the real price of the decision, it is not zero, and nothing makes it smaller: any single rule moves somebody’s beat. It is paid to the smaller side — 87 against 34, two and a half to one — and what is lost is recognition of rhythm, never meaning, because meaning never rode on stress here in the first place.

The ratio read nearly three to one until September 5, 2026. Eighty-seven to thirty-four is 2.56, which rounds to three only if you round it once; and the rounding went the way that made the price look smaller, on the one page whose stated job is to say plainly who pays.

Two thirty-fours, and they are not the same thirty-four. This page says a learner “can only go wrong on the 34 three-syllable roots”, and then that 34 roots have their beat moved. The two sets share four words — anahtar, pencere, restoran, telefon — and nothing else. Thirty of the roots whose beat moves have two syllables, which is to say they are among the roots the paragraph above has just promised an initial-stress learner cannot get wrong: that learner is right about where the beat falls and a Turkish speaker is wrong about the same word, because the two are being told different things. Both counts are correct and the collision is chance — 34 out of 300 twice over — but one page printing one number for two sets reads as one set. Found on September 5, 2026 by listing the members instead of comparing the totals, which is the only way it could have been found.

(French is counted as final-stress by simplification: it stresses the phrase rather than the word, so mersi has no fixed word beat to move.)

Open questions

  • Whether the short grammar words reduce in speech. The rule stresses every word, so Mi ama dunia is MI A-ma DU-nia, with three stresses in three words. Most languages let particles like es, no, aur, in, dari, por lean on the word beside them instead. Whether Amadunia does is a fact about how it is spoken, not about how it is written, and nothing here decides it.

Source: grammar/stress.md in the language repository.