Mora Theory

What a mora actually is

Imagine a metronome ticking under speech. In English, the metronome catches stresses — "the cat sat on the mat" places two strong beats with whatever unstressed material falls between them, however many syllables that turns out to be. In Japanese, the metronome catches something smaller: a unit roughly the duration of one short CV syllable. Every long vowel, every coda /n/, every geminate consonant adds an independent tick. That tick is the mora.

Formally, a syllable is built out of a one or two moras. A light syllable contains one mora — a short vowel with at most an onset (CV or V). A heavy syllable contains two moras — a long vowel (CVV), a diphthong (CVV), or a coda consonant (CVC). The mora count, not the syllable count, drives stress placement, prosodic phrasing, and metrical poetry in mora-counting languages.

Counting Japanese moras

Every kana symbol in Japanese hiragana or katakana corresponds to exactly one mora. That includes a few special cases that look counterintuitive to English readers:

• CV moras. A consonant + vowel: ka, ki, ku, sa, shi, su... one mora each.
• Long vowels are two moras. Tōkyō (東京) is to + o + kyo + o = 4 moras. The ō macron just signals a doubled vowel.
• Moraic /n/. The syllable-final ん is its own mora, even though it has no vowel. Honda is ho + n + da = 3 moras; shimbun "newspaper" is shi + n + bu + n = 4.
• Geminates count. The "small っ" doubles a following consonant and adds a mora. Kita "came" (ki+ta = 2 moras) ≠ Kitta "cut" (ki+t+ta = 3 moras).
• Glide combinations are one mora. kya, kyu, kyo each count as one — written with a small ゃ, ゅ, ょ.

Native speakers count moras effortlessly. Ask any Japanese person how long Tōkyō is and they will say four. The intuition is so robust that schoolchildren learn the haiku 5-7-5 pattern by clapping moras, not by counting syllables.

Haiku as a mora pattern

The classical Bashō haiku furuike ya / kawazu tobikomu / mizu no oto ("the old pond / a frog jumps in / the sound of water") is a textbook 5-7-5:

Line	Japanese	Mora breakdown	Mora count	Syllable count
1	furuike ya	fu-ru-i-ke-ya	5	4
2	kawazu tobikomu	ka-wa-zu-to-bi-ko-mu	7	7
3	mizu no oto	mi-zu-no-o-to	5	5
—	Tōkyō ni	to-o-kyo-o-ni	5	3
—	Hokkaidō	ho-k-ka-i-do-o	6	3
—	shimbun da	shi-n-bu-n-da	5	3

The classical 5-7-5 form was never a syllable count. English haiku translators who treat it as one usually cram far more material into each line than Bashō ever intended — Japanese moras are short and uniform, English syllables are long and variable. A more faithful English render typically uses 11-13 total syllables, not 17.

Speech rhythm typology

Languages can be roughly grouped by what their rhythm "ticks" align with:

Rhythm class	Unit timed	Example	Sample word/phrase	Effect
Stress-timed	Stressed syllables roughly equal in spacing	English	"The cat sat on the mat"	Vowel reduction in unstressed syllables ([ə])
Stress-timed	Stress isochrony	German, Russian, Arabic	German "Apfelkuchen"	Compressed unstressed material
Syllable-timed	Each syllable roughly equal	French	"U-ne-pe-ti-te-mai-son"	Even pacing; less reduction
Syllable-timed	Syllable isochrony	Spanish, Italian, Mandarin	Spanish "ca-sa-blan-ca"	Clear vowel quality preserved
Mora-timed	Each mora roughly equal	Japanese	To-o-kyo-o (4 ticks)	Long vowels and geminates literally double
Mora-timed	Mora isochrony	Hawaiian	Ka-me-ha-me-ha (5 moras)	Every mora gets equal time

The classification (Pike 1945, refined by Ladefoged) is somewhat idealised — phonetic measurements show all languages cluster on a continuum rather than three crisp categories. But the perceptual difference is strong enough that listeners can identify the rhythm class of an unfamiliar language from low-pass-filtered speech.

Geminates as moraic structure

A geminate consonant is a single consonant pronounced about twice as long as its singleton counterpart. In mora theory, the first half of a geminate occupies the coda mora of the preceding syllable, and the second half acts as the onset of the following one. Japanese minimal pairs make the structure visible:

kita    ki-ta       2 moras    "came"
kitta   kit-ta      3 moras    "cut"
saka    sa-ka       2 moras    "slope"
sakka   sak-ka      3 moras    "writer"
oto     o-to        2 moras    "sound"
otto    ot-to       3 moras    "husband"

Italian, Finnish, Arabic, and Estonian have parallel systems. Italian fato "fate" (2 syllables, 2 moras) versus fatto "done" (2 syllables, 3 moras): the geminate /tt/ adds duration without adding a syllable. Mora theory captures this elegantly — the first /t/ is a coda mora, the second is an onset, and the speaker hears the timing increase.

Worked example: Latin dactylic hexameter

Virgil's Aeneid opens with one of the most famous Latin lines:

arma virumque cano, Troiae qui primus ab oris
ar-ma | vi-rum-que | ca-no, | Troi-ae | qui pri-mus | ab o-ris
–  ⏑⏑ | –   ⏑   ⏑ | –  – | –  – | –   –  – | –   –  –

The metre is dactylic hexameter — six feet, each foot either a dactyl (long-short-short, 2+1+1 = 4 moras) or a spondee (long-long, 2+2 = 4 moras). The metric "long" is a heavy syllable: long vowel, diphthong, or short vowel followed by two consonants. Stress played no role at all. Roman schoolchildren scanned poetry by mora count, just as Japanese children clap haiku.

The same machinery underlies Greek hexameter (Homer), Sanskrit śloka, and the Vedic gāyatrī. The proto-Indo-European poetic tradition was thoroughly mora-based, a thread that English Germanic alliterative verse (counting only stresses) has long since cut.

Worked example: Sanskrit vṛddhi grades

Sanskrit ablaut runs across three grades — zero, guṇa, and vṛddhi — each adding more mora weight to the vowel slot:

Root	Zero grade	Guṇa grade	Vṛddhi grade	Mora pattern
i / ī	i (1μ)	e (2μ)	ai (2μ)	1 → 2 → 2
u / ū	u (1μ)	o (2μ)	au (2μ)	1 → 2 → 2
r̥ / r̥̄	r̥ (1μ)	ar (2μ)	ār (3μ)	1 → 2 → 3
l̥	l̥ (1μ)	al (2μ)	āl (3μ)	1 → 2 → 3
a	— (0μ)	a (1μ)	ā (2μ)	0 → 1 → 2

The five-vṛddhi system maps neatly onto a mora-graded ladder: each ablaut step adds approximately one mora of phonological weight. Pāṇini's Aṣṭādhyāyī (~5th c. BCE) describes the alternations in terms that anticipate modern moraic phonology by 2400 years. The grades drove Sanskrit verbal and nominal morphology — perfect tense, causatives, past participles, derivational nouns — far more than mere vowel quality could.

Moras drive English stress too

Even in stress-timed English, the assignment of stress is partly mora-sensitive — a fact most visible in Latinate vocabulary. The classic Latin stress rule, which English largely inherits in its loans:

• If the penultimate (second-to-last) syllable is heavy (long vowel or closed by a consonant) → stress it. aGENda, comPUta, aROma.
• If the penultimate is light (short vowel, open syllable) → stress the antepenultimate (third-from-last). CINema, MEDicine, SYLlabus.

The diagnostic is mora weight: heavy = 2 moras, light = 1. English speakers apply this rule unconsciously when they pronounce new Latin-Greek scientific words for the first time. Mora theory is alive in English even though English doesn't formally count moras.

Adjacent phenomena

• Bimoraic minimum. Many languages require content words to weigh at least 2 moras. Japanese loanword basu "bus" added a vowel to make the original /bʌs/ into a 2-mora word; jiisan "grandpa" lengthens the vowel to satisfy the same constraint.
• Mora-trochee foot. A common metrical foot consisting of two moras that may or may not coincide with a syllable. Drives stress in Hopi, Yidiny, and Latin.
• Mora as tone-bearing unit. In Tokyo Japanese, pitch accent docks to specific moras: haśi "chopsticks" (HL) vs haśí "bridge" (LH) vs hashi "edge" (LH but no fall). Each mora can carry an independent tonal target.
• Compensatory lengthening. When a coda consonant deletes, the preceding vowel lengthens to preserve mora count. Latin nokt- > Italian not- with optional lengthening; Old English tōþ "tooth" from earlier tanþ-.
• Mora vs prosodic word. The mora is the smallest weight unit; the prosodic word is the largest. Stress and accent rules typically reference both.

Why mora theory matters

• Phonological theory. Mora theory replaced "long vowel = single segment" analyses in the 1980s (Hyman, Hayes) and is now mainstream.
• Japanese pedagogy. Foreign learners of Japanese who count syllables instead of moras systematically mispronounce Tōkyō, Ōsaka, and any name with a long vowel.
• Speech synthesis. Japanese TTS engines must explicitly model mora durations or output sounds clipped and unnatural.
• Classical philology. Reading Virgil, Homer, and the Vedas requires fluency in mora-based scansion.
• Comparative typology. Mora-timing distinguishes Japanese phonology from Korean and Mandarin in ways that surface transcription hides.
• Music and language. Japanese composers align lyrical syllables to mora-equal time slots; Western settings of haiku often distort the metric.

Common pitfalls

• Treating mora as a synonym for syllable. They overlap when every syllable is light, but diverge whenever long vowels, codas, or geminates appear.
• Confusing mora-timing with monotone delivery. Mora-timing is about durational regularity, not pitch flatness. Japanese has rich pitch accent superimposed on its mora rhythm.
• Counting Japanese moras from English transcription. Romanised Tokyo looks like 3 syllables / 6 letters; the underlying form is 4 moras: to-o-kyo-o.
• Assuming all languages count moras. Most don't. Mora-counting is a real but minority strategy; stress-timing and syllable-timing dominate.
• Reading Latin metre as English-style stress patterns. Dactylic hexameter is mora quantity, not stress alternation. English imitations (Longfellow's Evangeline) substitute stress and lose the underlying counting principle.
• Treating Korean as mora-timed. Korean is sometimes classed as syllable-timed; it is not mora-timed. Modern Mandarin similarly counts syllables, not moras, despite tonal complexity.