This article is an non-exhaustive list of rules that govern the Japanese pitch accent as spoken in Tokyo.
Particles and auxiliaries When a noun, verb or adjective precedes a or an , they may form together a minor phrase. Each particle/auxiliary or group of particles/auxiliaries has its own accentual behavior that can be classified as follows: Behavior type: Recessive: The particle or auxiliary does not alter the accent of the preceding word. The word determines the accent of the entire phrase, and if it is accented, the particle either has a low contour or forms its own minor phrase. Dominant: The particle or auxiliary determines the accent of the phrase, overriding whatever accent the preceding word has. : The auxiliary combines with a conjugational verbal form, typically the . This form is a bound morpheme, incapable of existing on its own, much less having accent. An auxiliary combining with it preserves the underlying accent type of the original verb. Accent type: Unaccented: The particle or auxiliary has no accent of its own. Accented: The particle or auxiliary has accent which surfaces only if it is dominant, or if it is recessive and the preceding word is unaccented.
An accented recessive particle or particle sequence can form its own downstepped minor phrase after an accented word, although this is less likely if the word is final-accented and the particle is initial-accented:
More details on the interactions between particles or auxiliaries and nouns, verbs and adjectives are listed below.
Particle + particle In general, when a particle follows another particle, they form together a single accented minor phrase, as exemplifed in the table above. The first particle's last mora becomes the nucleus. Some of these have become idiomatic conjunctions: etc. Some more examples:
The particles to, kiri, shika and dake alternatively does not add accent to their preceding particle:
The accent of the particles kurai / gurai, dokoro and bakari can alternatively surface:
Auxiliary + particle Verb-like and adjective-like auxiliaries simply follow the rules of verb and adjective accent before particles:
The particle to can remove accent from hortatives:
Particle + auxiliary In general, when an auxiliary follows another particle, they form together a single accented minor phrase. The particle's last mora becomes the nucleus. Some of these have become idiomatic conjunctions: etc. Some more examples:
The accent of the auxiliaries yō (da), rashī and gotoshi can alternatively surface:
Auxiliary + auxiliary Verb-like and adjective-like auxiliaries simply follow the rules of verb and adjective accent before auxiliaries:
Nouns Here are some rough percentage ratios in correspondence with the mora counts of nouns (0 means unaccented, +1 initial-accented, −1 final-accented, −2 penultimate-accented, −3 antepenultimate-accented and −4 preantepenultimate-accented): Unimorae: +1:0 → 70:30 Bimorae: +1:−1:0 → 60:20:20 Trimorae: +1:−2:−1:0 → 40:10:10:50 (10% is either 0 or +1) Quadrimorae: +1:−3:−2:0 → 10:10:10:70 Quinquemorae: −4:−3:−2:0 → 10:50:10:30 Sexamorae: −4:−3:−2:0 → 35:35:10:20
One of the principal rules that govern nouns is known as the "antepenultimate rule," which dictates a "default" location for the accent nucleus, which is either the actual antepenultimate mora (−3), or the preantepenultimate mora (−4) due to a leftward shift away from the deficient antepenultimate mora (see ). Even native speakers rely on this rule, among others, to guess how to pronounce newly borrowed or unfamiliar simplex or compound nouns, even completely made-up ones such as . As the above statistics show, the antepenultimate rule accounts for the high number of quinquemoraic or longer (pre)antepenultimate-accented nouns. It is also highly productive among recent loanwords of any length (although unimorae and bimorae do not have room for an antepenultimate mora and are thus necessarily initial-accented) and will continue to be so as more words are imported into the Japanese language.
Quadrimoraic nouns are predominantly (70%) unaccented. Unaccentedness is also typical in high-frequency words, and even a normally accented word can be deaccentuated by an individual if they use it frequently enough; for example, the word can be pronounced unaccented as by a guitar lover, instead of accented as . Quadrimoraic clippings are overwhelmingly unaccented and their number is on the increase.
Older speakers are more likely to have final-accented quinquemorae, and penultimate-accented sexamorae ending in special morae, both of which younger speakers are making (pre)antepenultimate-accented (the dagger marks obsolete or obsolescent accent):
Simplex Sino-Japanese nouns Over half of simplex Sino-Japanese nouns (spelt with a single kanji), especially infrequent ones, are initial-accented. The following are rough percentages in correspondence with mora counts, as well as some representative examples. Note that there are items with variant patterns, some of which may be used for different meanings:
Initial-accented Final-accented Unaccented Unimorae 68% 0% 32% , , etc. , , etc. Bimorae 77% 23% , , , , , , , , , , etc. , , , etc. , , , , , , , , , , , etc.
Modifier–head compound nouns This subsection deals with modifier–head compound nouns, with the modifier being the first component and the head being the second. Henceforth: A "short" component: If native, is bimoraic or shorter: , , etc. If Sino-Japanese, has one kanji (monomorphematic): , etc. A "long" component: If native, is trimoraic or longer: , , etc. If Sino-Japanese, has two or more kanji (dimorphematic or longer): , , , etc. A short compound consists of only short components, and a long compound contains at least one long component. Note that a short compound can become a long component in a long compound.
Short compounds Short compound nouns are accentually difficult to predict, as their accent correlates with the modifiers and/or the heads, in terms of mora counts and/or original accent.
Native heads Unimoraic modifier + native head The compound tends to have the same accent type as the modifier and retain its nucleus: Originally unaccented modifiers:
Originally accented modifiers:
Some compounds retain the bimoraic head's accent instead or in addition:
Bimoraic modifier + native head The compound tends to have the same accent type as the modifier, but the compound's nucleus tends to be the modifier's last mora regardless of the modifier's original nucleus: Originally unaccented modifiers:
Originally accented modifiers:
If the modifier's last mora is special, it's penultimate mora is accented instead:
If the head is bimoraic, the compound may also be unaccented:
Some compounds retain the modifier's initial accent, or delete its final accent: Originally initial-accented modifiers:
Originally final-accented modifiers:
Some compounds retain the bimoraic head's initial accent instead or in addition:
Sino-Japanese heads Unimoraic modifier + unimoraic Sino-Japanese head Almost 80% of these compounds are initial-accented: , , , , , , , , , , etc.
A devoiced nucleus may cause a rightward shift: , , , , , , , etc.
Some older compounds are unaccented or final-accented: Unaccented: , , , , , , , , , , , , , , , etc. Final-accented: , , , , , , etc.
Bimoraic modifier + unimoraic Sino-Japanese head Most of these compounds are initial-accented or unaccented. Semantically nominal compounds are mostly initial-accented, while semantically verbal compounds are mostly unaccented: Semantically nominal: , , , , , , , , , , , , etc. Semantically verbal: , , , , , , , , , , , , etc.
Note that some of the unaccented semantically verbal compounds above are not really modifier–head compounds, but rather verb–object compounds as per Chinese syntax (see ). For example, comes from in ("drinks") and shu ("booze"), hence "drinking booze."
A devoiced nucleus may cause a rightward shift: , , , , , , etc.
Some old trimoraic compounds may be final- or penultimate-accented: Final-accented: , , , , , , , , etc. Penultimate-accented: , , etc.
Some originally accented native modifiers may make its own last mora the nucleus, as they would if the head were also native:
The suffixal head often results in final-accented compounds: , , etc.
Unimoraic modifier + bimoraic Sino-Japanese head Some 74% of these compounds are unaccented. Semantically nominal ones are mostly initial-accented, and while semantically verbal ones mostly unaccented: , , , etc.
A devoiced nucleus may cause a rightward shift: , , , , , etc.
Some older compounds are final- or penultimate-accented: Final-accented: , , etc. Penultimate-accented: , , , , , , , , , , , etc. , , , etc.
Bimoraic modifier + bimoraic Sino-Japanese head Some 85% of these compounds, which are quadrimoraic, are unaccented. Some used to be initial-accented, but have become unaccented (the marks obsolete or obsolescent accent)ː , , , , etc. , , , , etc. , , , , , , , etc. , ,
Some older compounds remain initial-accented: , , , , , , , , , , , etc.
Some Sino-Japanese and most native modifiers make their own last mora the nucleus: , , , etc. , , , , , , , , , , , etc.
Long compounds Long compounds are more predictable, in that their accent largely correlates with their heads. Long accented compounds are generally penultimate-, antepenultimate- or preantepenultimate-accented (see also the statistics at #Nouns).
Short heads The head can be preaccenting, deaccenting or retain its original accent.
Preaccenting short heads:
Deaccenting short heads:
Other short heads can be either preaccenting or deaccenting:
Some short initial-accented native words retain their accent as heads:
Some exceptional short heads tends to be deaccenting or final-accented, but sometimes also initial-accented:
The suffixal head can be deaccenting, preaccenting or sometimes final-accented. When it is preaccenting, the accent does not usually shift even if it is on a special mora (see for more details): Deaccenting: , , , , , , etc. Preaccenting and accenting autonomous morae: , , , , , etc. Preaccenting and accenting special morae: , , , , etc. Deaccenting or preaccenting: , , , , , , etc. Final-accented: , , etc. Deaccenting, preaccenting or final-accented: , etc.
The suffixal heads and tend to be deaccenting or final-accented: , , , , , , , , etc. , , , , , etc. , , , etc. , , , , , etc. But: , , , , ,
The suffixal head may be deaccenting or final-accented, but the modifier, which would be a temple's actual name, may also bear accent: , , etc. , , , , , , , , , , etc. , , , etc. , etc. , etc. , , etc.
The suffixal head is usually preaccenting, but the modifier, which would be a mountain's actual name, may also bear accent: , , , , etc. , , etc.
The suffixal head is usually preaccenting, but sometimes also initial-accented after : , , , , , , etc. , , , , , , , , etc. But:
, , , ,
Long heads If the head is originally unaccented, or final- or initial-accented, or medially penultimate-accented with a special final mora, its first syllable tend to be accented in the compound: Originally unaccented long heads:
Originally final-accented long heads:
Originally medially penultimate-accented long heads with a special final mora:
Originally initial-accented long heads:
Other originally medial-accented long heads tend to retain their accent:
Some penultimate-accented long heads retain their accent despite ending in a special mora:
Some exceptional long heads Some heads become, or used to used to become, final-accented in long compounds. They have been regularized by later generations of speakers (the dagger marks obsolete or obsolescent accent):
Some bimoraic Sino-Japanese heads are deaccenting when combining with short modifiers. When they combine with long modifiers, however, they tend to be preaccenting or regularly initial-accented instead: