A discussion of word-formation in Old Qári, along with a discussion of the phoneme inventory.
Phoneme inventory
The Old Qári consonant inventory consists of the five fortis stop phonemes denoted p t k kʷ ʔ; four lenis stop phonemes denoted b d g gʷ and nine continuant phonemes, denoted w y x s r l m n ŋ. It should be noted that most reconstructions are agnostic about the precise articulation and phonation of the stop phonemes – fortis and lenis are used here as labels without prejudice as to whether the the distinction was one of voicing, aspiration, glottalisation etc.
The vowel inventory is remarkably small, consisting only of two phonemic vowels. Rather than being distinguished along the axes of height and backness, the vowels appear to have been underspecified for height, the front vowel we denote with e and the back with a. In addition, there is a subphonemic anaptyctic vowel which we denote ɨ.
The continuant phonemes have vocalic allophones (for the distribution of which vide infra), denoted u i ə ṣ ṛ ḷ ṇ. Note that the three nasal consonants merge in their vocalic allophone to ṇ.
Adjacent stops, when differing in phonation, undergo anticipatory assimilation. Thus, a sequence of underlying pd is realised as bd, while bt surfaces as pt.
Stress in Old Qári is variable, yet predictable. Where present, a grade vowel (see below) is always stressed, otherwise the first syllable after the initial consonant of a word’s root is stressed.
Roots
Roots can consist of up to four consonants. Typically, uniconsonantal and biconsonantal roots are found in grammatical words and derivational morphology, although biconsonantal roots are not unusual in lexical words from the earliest strata of the language (for example, √xs ‘heart’). Lexical words are most frequently derived from triconsonantal roots, with quadriconsonantal roots being rare. Quadriconsonantal roots, such as √slst ‘sparkle, glitter’, are generally transparently derived as reduplicated forms of bi- or triconsonantal roots (in this case √slt ‘shine faintly’).
Roots have the following selectional restrictions:
- if a root contains two adjacent stops, they must differ in place of articulation;
- if a root contains two adjacent continuants, they cannot be identical; and
- triliteral roots cannot consist solely of stops or solely of continuants.
Roots also have three grades, denoted by the presence or absence of a phonemic vowel:
- the zero-grade lacks a phonemic vowel
- the front-grade adds the front vowel e: generally adds a dynamic meaning
- the back-grade adds the back vowel a: generally adds a stative meaning
From roots to words
To form a lexical word from a root, the root and any suffixes need to be vocalised. There are four principles which determine what the surface instantiation of what the resultant word looks like:
- suffixes cannot occur immediately following a grade vowel;
- the maximal consonant cluster is CC;
- clusters are disallowed in anlaut and auslaut; and
- consonant clusters are avoided wherever possible.
As an example, let us examine a few derivations of the root √gxr ‘ooze’. The back-grade formations of this root produce words related to blood, ‘that which oozes’.
The most “basic” derivative here is the word for blood itself, which is a root noun (i.e. has a suffix in -Ø). Preferentially in the derived grades the grade vowel occurs as far to the right as possible, so in this case we have the form gxrá. Noting that the maximal consonant cluster is CC, then either x or r is going to have to vocalise. Given that the vocalic allophone of a continuant is not permitted adjacent to a vowel, we arrive at the form gərá ‘blood’.
The suffix -n produces adjectives related to the root, so in this case something like ‘bloody’. Noting principle one, a form like gxrán is impermissible because a suffix cannot immediately follow a grade vowel. A form like gxárn also doesn’t work, because clusters are disallowed in anlaut, which leaves us with the form gáxrn. Again, we arrive at the limitation on maximum two adjacent consonants, so one of those three final continuants will vocalise. We’ve already seen that we can’t have a vocalic allophone next to a full vowel, it’s going to be either r or n. According to principle four, consonant clusters are to be avoided, so we arrive at gáxṛn ‘bloody’.
The suffix -nt produces words indicating something like possession of the root, so in this case ‘having blood’ (Old Qári has the generative metaphor BLOOD IS BRAVERY, so ‘having blood’ here indicates ‘brave’). Following the same path as we did above, we get the underlying form gáxrnt. If we vocalise the r again, we have an auslaut consonant cluster. Therefore the only acceptable outcome is gáxrṇt.
Zero-grade formations of the same root produce words related to pus and, by extension, disease. Let us look at the root noun derivation: in the form gxr one of the continuants must vocalise and again to avoid the creation of a consonant cluster we see the vocalisation of x, giving gər. Old Qári content words obligatorily have at least two syllables, so in lieu of an overt suffix the anaptyctic vowel –ɨ is suffixed to give gərɨ. By the same rules as Words without a grade vowel always stress the first syllable after the initial consonant of the root, so the final form produced is gə́rɨ ‘pus’.
Roots which begin with two stops, such as √pkn ‘run’ operate along much the same lines as above, but will make use of the anaptyctic vowel ɨ to resolve impermissible consonant clusters. Thus, deriving the front-grade root verb (i.e. has a suffix in -Ø), we arrive at pɨkné ‘to run’. The zero-grade derivative pɨ́kṇ ‘gazelle’ satisfies the constraint against lexical words consisting of a single syllable, as well as that against vocalic hiatus.
The Qári reflexes of these words are gyá [ˈɡʲɑ] ‘stain’, gári [ˈɡɑɾi] ‘red’, gárat [ɡɑˈɾæʔ] ‘male adolescent initiate into the warrior cult’, hor [ˈxɔɾ] ‘semen’, pengké [ˈpɛŋke] ‘to flee’ and pikipiki [pikiˈpiki] ‘id’ for what it’s worth. Note the innovative reduplication, potentially of a diminutive origin, in the Qári reflex of pɨ́kṇ.