Qaqet language

The voiceless stops /p t k/ are often aspirated word-initially and unreleased word-finally. They have two other predictable allophones: voiced /b d g/ following a nasal, and lenited /v r q/ intervocalically. These last three phonemes probably arose historically out of their stop counterparts and are in the process of being phonemicized. While the alternation is still for the most part predictable, it has become contrastive for some aspectual verb stems. Also, loanwords don't undergo the lenition process: akar 'cars' (from Tok Pisin kar).[2]

The velar stop is almost always palatalized in the vicinity of the vowel /i/. This holds for its voiced and fricativized allophones as well. Hence: a=vadem-ki [aβaⁿdəmɟi] 'trap', vrli-ki [βɽiʝi] 'cousin'. However, there are a couple of counterexamples, where palatalization doesn't occur even though it would be expected. For instance, in the word [aɣiapki] 'chicken', the velar fricative is not palatalized, even though it is followed by /i/. This might be explained by an earlier form of the word containing the sequence /ai/. In fact, the alternative realization [aɣaiapki] is attested.[2]

The voiced stops appear as allophones of voiceless stops (as explained above), but also as phonemes in their own right. They can occur initially, intervocalically as well as after consonants, but not at the end of a syllable. They are usually prenasalized.

The voiceless fricative /s/ alternates with [h] in free variation. Speakers tend to associate [s] with the southern and [h] with the northern dialect, respectively. In practice, however, both variants seem to be used by speakers of all dialects, with no change in meaning.[2]

Qaqet has four short vowel phonemes:

The close vowels are usually realized as lax [ɪ ʊ]. /i u/ can have allophones of semivowels [j w] when preceding other vowels. The open vowel /a/ undergoes assimilation to adjacent sounds: it is realized as back [ɑ] before [ɣ], but as front [æ] before palatal consonants, as well as front vowels in the next syllable. There is also variation across speakers between [a] and [ɐ]. The phoneme /ə/ is much shorter than the other vowels, especially in the vicinity of sonorants, where it is frequently elided, particularly in rapid speech. The vowel [o] only occurs in recent loanwords, such as botol 'bottle' (from Tok Pisin).[2]

Vowel quantity is generally not contrastive in Qaqet, but long vowels do occur as the result of vowel sequences across morpheme boundaries. Sequences of /a/ and a high vowel become long monophthongs: a=ilany [eːlæɲ] 'leg/foot', a=ulan [oːlan] 'eel'. The vowel /i/ generally does not have a long counterpart, but is realized as [eː]: [meː] ~ [miː] 'most'. The short vowel /ə/ does not have a long counterpart. There are a small number of roots with long vowels which cannot be explained as the result of underlying vowel sequences, e.g. laan 'type of bamboo' vs. lan 'bones'.[2]

There are diphthongs /ia/, /iu/, /ui/, /ua/, which occur even within roots. The diphthongs /ai/ and /au/ do occur, but are more commonly realized as long vowels [eː] and [oː].[2]

Qaqet's syllable structure can be summarized as (C)(C)V(V)(C). All consonants can occur at the onset of a syllable. Consonant clusters at the onset of a syllable usually consist of an obstruent followed by a sonorant, e.g. slep 'intensely', brasuqa 'eagle'; but there are also clusters of a nasal and a liquid, as in mrarlik 'cross'. In coda position, the plain stops /p t k/, the nasals /m n ɲ ŋ/, the fricative /s/ and the liquids /ɽ l/ can occur, but not the voiced stops /b d g/ or the lenited versions of the stops /β r ɣ/.[2]

Within the noun phrase, determiners (including possessor indexes, articles, indefinite pronouns and demonstratives) usually precede the head noun, while modifiers (adjectives, numerals, quantifiers, prepositional phrases, directionals) tend to follow it, although there are exceptions. In the following example, the noun is preceded by the demonstrative luqa and the article ama-, while the adjective follows it:[2]

As can be seen from the example, adjectives agree with the noun they modify in noun class and number.

Nouns are almost always preceded by some determining element: an article, a possessor index, a demonstrative or an indefinite pronoun, or combinations of these. The generic noun marker a contrasts with the articles ama and ma. The former can be described as a Stage III article following Greenberg (1978):[6] it probably evolved from a definite article, but has now lost all of those functions and merely marks nouns. It cannot occur with most proper names, nor in combination with possessor indexes. The articles ama and ma can occur even with proper names, and they have a more specific meaning than the general noun marker a. Their main function is to mark a noun for referentiality. Moreover, ama can connect a qualifying adjective to a preceding noun, as can be seen in the example above. The difference between ama and ma is that the latter marks a noun as inherently identifiable. Hence, it frequently occurs with proper nouns.[2]

Below is a list of Qaqet numerals (see Hellwig 2019: 90-91, Tab. 31).[2]

Qaqet's number system is basically decimal, with base words for 10, 100, and 1000. However, 5 is used as an auxiliary base, such that the words for 6-9 are formed by adding the numerals for 1-4 to the number 5. The word for 1 consists of the root qunas and the noun class suffix -ka, while the word for 2 uses the same root with the dual suffix -iam. The same logic is applied to form the numerals for 10 and 20. Addition is expressed through the element ngena- 'plus, together with'.

There are two structures for forming higher multiples of ten. One puts malev 'ten' first, then adds the factor with the article ama-. The other puts the factor first and inserts the preposition ne- 'from, with' (nama- is a contraction of that preposition and the article ama-). Both of these structures were obtained through elicitation. In practice, Qaqet speakers generally switch to Tok Pisin for numerals above 10.[2] It is not known whether any of the higher numerals are used in natural speech.