The phonological system of the Hawaiian language is based on documentation from those who developed the Hawaiian alphabet during the 1820s as well as scholarly research conducted by lexicographers and linguists from 1949 to present.
Hawaiian has only eight consonant phonemes: /p, k ? t, ?, h, m, n, l ? ?, w ? v/. There is allophonic variation of [k] with [t], [w] with [v], and [l] with [?]. The [t]-[k] variation is highly unusual among the world's languages.
Hawaiian has either 5 or 25 vowel phonemes, depending on how long vowels and diphthongs are analyzed. If the long vowels and diphthongs are treated as two-phoneme sequences, the total of vowel phonemes is five. However, if the long vowels and diphthongs are treated as separate, unit phonemes, there are 25 vowel phonemes. The short vowel phonemes are /u, i, o, e, a/. If long vowels are counted separately, they are /u:, i:, o:, e:, a:/. If diphthongs are counted separately, they are /iu, ou, oi, eu, ei, au, ai, ao, ae, o:u, e:i, a:u, a:i, a:o, a:e/. There is some allophonic variation of the vowels, but it is much less dramatic than that of the consonants.
Hawaiian syllable structure is (C)V(V) where C is any consonant and V is any vowel. Double vowels (VV) may be long vowels or diphthongs. All CV(V) syllables occur except for w?, but wu occurs only in two words borrowed from English. Word stress is predictable in words of one to four syllables but not in words of five or more syllables. Phonological processes in Hawaiian include palatalization and deletion of consonants and the raising, diphthongization, deletion, and compensatory lengthening of vowels. Phonological reduction (or "decay") of consonant phonemes during the historical development of Hawaiian has resulted in the phonemic glottal stop. The ultimate loss (deletion) of intervocalic consonant phonemes has resulted in long vowels and diphthongs.
The following description of Hawaiian phonemes and their allophones is based on the experiences of the people who developed the Hawaiian alphabet, as described by Schütz, and on the descriptions of Hawaiian pronunciation and phonology made by Lyovin, and Elbert & Pukui. Some additional details on glottal consonants are found in Carter.
It is notable that Hawaiian does not distinguish between and . Few languages do not make this distinction, though several Polynesian languages have independently undergone the historical shift from /t/ to /k/ after the change of /k/ to /?/; Samoan is notable for using [t] in colloquial speech where [k] is used in formal speech. The American missionaries who developed written Hawaiian during the 1820s found that a [t] reflex was common at the Kauaʻi (Tauaʻi) end of the island chain, and a [k] reflex at the Big Island (island of Hawaiʻi) end. They decided to use ⟨k⟩ rather than ⟨t⟩ to represent this phoneme. However, that does not prevent anyone from using the t realization, in speaking or in writing, if they so desire. T is used more than k by speakers of Niʻihau Hawaiian.
The missionaries also found allophonic variation between and (written with d) and , between and , and between and .
There is basic free variation of and . However, since Hawaiian has no affricates, no fricative besides , and no other stops besides and , any non-labial and non-glottal stop, fricative, or affricate, can function as a /k/. In essence, , , , , , , , , , , , , , can all "work" as an allophone of /k/. Nevertheless, the main allophones noted by the missionaries in the 1820s, and by linguists, are [t] and [k]. Elbert & Pukui point out some instances of a allophone. Schütz conjectured that a t-dialect existed in the northwestern islands, and a k-dialect in the southeastern islands. Roberts documented a sound between that of English ⟨th⟩ and in free variation with [k] among elders from Oʻahu and Kauaʻi.[full ][clarification needed]
There is some evidence for instances of free variation between and .
There is also free variation between (lateral), (tap), and (approximant). Elbert & Pukui have pointed out some instances of and as allophones. Schütz has conjectured that [?] is prevalent in the northwestern islands and [l] is prevalent in the southeastern islands.
There is free variation of and . Pukui & Elbert (1986:xvii) have conjectured that there is conditioned variation of [w] and [v], but their use of "usually" makes their theory an admission of free variation. Schütz has conjectured that there was neither [w] nor [v], but rather "something between the two". This is most likely , a labiodental approximant (see also Schütz's (1994:113) quotes from letter of Artemas Bishop).
In Hawaiian, a phonemic glottal stop historically derives from an earlier consonant. A number of words have variant pronunciations between glottal (that is, both and ) and non-glottal consonants; it is conjectured that the forms with a non-glottal consonant are older and that this phenomenon is part of a process of consonant deletion.
This can still be seen in the historical development of the dual personal pronouns. This is exhibited in the suffixes for dual and plural number, which come from lua ('two') and kolu ('three') respectively.
|1st person exclusive||1st person inclusive||2nd person||3rd person|
|Meaning||we two||we two||you two||they two|
|Old form||/ma: + lua/||/ka: + lua/||/?o + lua/||/la: + lua/|
|Glottal form||/ma: + ?ua/||/ka: + ?ua/||/la: + ?ua/|
|New form||/ma: + ua/||/ka: + ua/||/la: + ua/|
The /l/ of /-lua/ in the first and third person forms has "died" or "vanished", resulting in the modern forms m?ua, k?ua, and l?ua. The presence of the glottal stop marks the absence of a "phonetically fuller" consonant. The second person form, ʻolua, contains a glottal stop, implying that the /l/ used to be there and /?/ still exists in place of /l/ in the intermediate forms, /ma: + ?ua/, /ka: + ?ua/, and /la: + ?ua/.
A Hawaiian glottal stop thus represents the maximal phonetic reduction of other consonants in centuries past.
Elbert & Pukui have shown instances of k ? ?, and l ? ?, such as mukumuku ? muʻumuʻu ('cut'), and p?liki ? p?ʻiki ('embrace'). Carter (1996:373-374) has shown examples of all seven of the (other) Hawaiian consonants alternating synchronically with glottal stop:
|/C ? ?/||/p/||/k/||/h/||/m/||/n/||/l/||/w/|
Depending on how one analyzes the inventory of Hawaiian vowel phonemes, it has either 5 or 25 phonemes. The minimum figure of 5 is reached by counting only , , , , and /a/ as phonemes. Diphthongs and long vowels are analyzed as being sequences of two vowels. For example, the written form ⟨au⟩ is phonemically /au/, and the written form ⟨?⟩ is phonemically /aa/. The maximum figure of 25 is reached by counting separately the 5 short vowels, the 5 long vowels, the 9 short diphthongs, and the 6 long diphthongs. A reason given to support this analysis is that the diphthongs "act as unit phonemes in regard to stress."
It is not necessary to postulate that the long vowels and diphthongs should be counted as separate single phonemes, because they can be treated as sequences of two vowels. They are in fact historically derived from two-syllable sequences. This is easily seen in the synchronic co-existence of allomorphic pairs of Hawaiian forms such as kolu with -kou, both meaning 'three'.
The example can be analyzed as a four-phoneme CVCV sequence alternating with a three-phoneme CVV sequence, where the CVV form is derived from the CVCV form through loss of the second consonant. In other words, /kolu/ loses the /l/, resulting in /kou/. Kolu is a root form, while -kou is found in the plural personal pronouns (indicating three or more referents) m?kou, k?kou, ʻoukou, and l?kou.
The vowel phonemes are shown in the following tables. The information given on allophones constitutes a basic description, not exception-free laws. Native speakers of any language can get away with tweaking their own personal pronunciation.
Vowel quality is the same for long and short vowels, except for /e:/ vs. /?/, and /a:/ vs. /?/:
One might argue for free variation of [a] and [?] for stressed short /a/. However, Elbert & Pukui (1979:22-24) have made citations to Kinney (1956) and Newbrand (1951), based on tape-recorded evidence, which specifically noted [?], but not [a]. Even so, the pronunciations [?a'?oli] and [?a'?ohi] exhibited above, show that there are at least a couple of forms where /a/ is realized as [a].
|Ending with /u/||Ending with /i/||Ending with /o/||Ending with /e/|
|Starting with /i/||i?u|
|Starting with /o/||ou?||o?i|
|Starting with /e/||eu?||ei?|
|Starting with /a/||au?||ai?||ao?||ae?|
Unlike the other falling diphthongs (that is, falling from higher to lower sonority), /iu/ and /oi/ are rising diphthongs. For example, kiu ('spy') is pronounced [kju], like the English name for the letter Q.
As with its constituent vowels, diphthongs with short and are subject to the same free variation described above. In rapid speech, /ai/ and /au/ can become [ei] and [ou] respectively.
|Ending with /u/||Ending with /i/||Ending with /o/||Ending with /e/|
|Starting with /o:/||o:u?|
|Starting with /e:/||e:i?|
|Starting with /a:/||a:u?||a:i?||a:o?||a:e?|
All long diphthongs are falling.
Hawaiian syllables may contain one consonant in the onset, or there is no onset. Syllables with no onset contrast with syllables beginning with the glottal stop: /alo/ ('front') contrasts with /?alo/ ('to dodge'). Codas and consonant clusters are prohibited in the phonotactics of Hawaiian words of Austronesian origin. However, the borrowed word Kristo is pronounced ['kri.sto].
The syllable has a minimum of one vowel, and a maximum of two. A one-vowel syllable has any one of the short or long vowels. A two-vowel syllable has any one of the diphthongs.
The structure of the Hawaiian syllable can be represented as being (C)V(V), where the round brackets around C and second V mean that the syllable-initial consonant is optional and the syllable may have a long vowel or diphthong.
Elbert & Pukui have pointed out that "Certain combinations of sounds are absent or rare." For example, no content word has the form CVV?V, and the form CVVCV, is also not common. They also noted that monovocalic content words are always long.
Word stress is predictable in Hawaiian for words with three or fewer moras (that is, three or fewer vowels, with diphthongs and long vowels counting as two vowels). In such cases, stress is always on the penultimate mora. Longer words will also follow this pattern, but may in addition have a second stressed syllable which is not predictable. In Hawaiian, a stressed syllable is louder in volume, longer in duration and higher in pitch.
For other Hawaiian words longer than three moras, stress is not predictable. However, every word can be analyzed as consisting of a sequence of these stress units:
Etymology is not a reliable guide to stress. For example, the following proper names are both composed of three words, of 1, 2, and 2 moras, but their stress patterns differ:
Phonological processes at work in Hawaiian include palatalization of consonants, deletion of consonants, raising and diphthongization of vowels, deletion of unstressed syllables, and compensatory lengthening of vowels. Elbert & Pukui have cited Kinney (1956) regarding "natural fast speech" (vowel raising, deletion of unstressed syllables), and Newbrand (1951) regarding Niʻihau dialect (free variation of [t] and [k], deletion of consonants, allophone of /a/, vowel raising).
Kinney (1956) has studied tape recordings of 13 or 14 native speakers of Hawaiian. She noted assimilatory raising of vowels in vowel sequences. For example, /ai/ was very frequently pronounced [ei], /au/ was often [ou], and /io/ was often [iu]. She cited specific words, such as /mai/ (directional adverb) as [mei], /mau/ (plural morpheme) as [mou], and /lio/ ('horse') as [liu]. The pronunciation of the island name Maui, Maui, /maui/, was ['m?u.i], with the quality of [?] compared to that of u in English cut. She observed deletion of unstressed syllables, such as /ke akua/ ('God') pronounced [ke 'ku?], and /hele akula/ ('go') pronounced ['h?l? 'kul?]. She also documented pronunciations of /loa?a/ ('gotten') as ['lo], and /pua?a/ ('pig') as ['pu].
Newbrand (1951) found that a Niihauan wrote ⟨t⟩ and ⟨k⟩ interchangeably, and freely varied the pronunciation of both ⟨t⟩ and ⟨k⟩ as [t] or [k]. She found /?a?ohe/ ('no') pronounced [?a'?ohi], showing vowel raising of /e/ to [i]. She documented /noho ?ana/ ('staying') pronounced [noo '?n?], showing deletion of the glottal consonants /h/ and /?/. The vowel quality of stressed short /a/ was noted as [?].