Gold phonology
The phonology was similar to its parent language, Tapilula, but unlike Tapilula, the Gold language has evolved closed syllables. Nevertheless, many syllables are CV and it is rare to have more than one closed syllable in a word.
The sound of the language is fairly guttural, like that of Tapilula. It would never be mistaken for any relative of the Babakiam branch when spoken aloud, nor would it be mistaken for Thaoa, which lost its tones early on.
Consonants
There were labialized consonants in Gold, but they are not considered phonemic because unlike in Khulls and Poswa, they can only occur before a vowel. Thus it is better to consider this as simply a /w/ inserted between a syllable onset and its nucleus. This also means /w/ itself is phonemic rather than being considered, as in Khulls, just an allophone of /ʕʷ/. THus, with labialized consonants ignored, the setup is:
- /p b m w t d n s z l č ǯ j k ġ ŋ h g ḳ ʕ/
The velar ejective ḳ is the only ejective in the language, although the clusters /pḳ/ and /tḳ/ could occur, even word-initially, leading to marginal phonemes /ṗ/ and /ṭ./ More commonly, however, /ṗ/ and /ṭ/ also appear as allophones of /ḳ/ after the syllabic nasals /ṁ/ and /ṅ/.
The only voiced stop that occurs with a frequency on par with the voiceless stops is /d/, which becomes a fricative [ð] between vowels. The stops /b ġ/ and the voiced affricate /ǯ/ are rare, occurring only in positions where a recent sound change caused the voicing of their respective voiceless counterparts.
Final consonants
The final consonants are /k ḳ l n s ʕ/. Syllabic consonants /ṁ ṅ ŋ̇/ do, however, exist.
Ejective allophones of voiceless stops
The voiceless stops p t k are aspirated in all positions. However, when preceded by the voiceless ejective ḳ, which occurs at the end of some stressed syllables, the ejective articulation bleeds across the syllable boundary and results in allophonic ejectives ṗ ṭ at the beginning of the next syllable. These ejectives also lose their aspiration, since Gold and related languages do not allow any consonant to be simultaneously aspirated and ejective.
Phonemicity of /w/ and labial clusters
The phoneme /w/ can be analyzed as /ʕw/, in line with Khulls where [ʕʷ] and [w] are merely allophones of each other, and with similar situations in the early histories of the branches to the east and west of Khulls. If this is done, /w/ becomes a glide which can only occur after another consonant, and therefore can be considered a modification of that consonant. This creates symmetry between the plain /w/ and the very common sequence /hw/, whereas all of the other consonants take a following /w/ much more rarely.
Labialization was originally not contrastive before /u/, since in most cases it had been triggered by a following /u/. However, as the phoneme /hʷ/ evolved steadily towards a rounded bilabial fricative, a new contrast opened up when a much later sound change created new plain /h/ in various positions, including before /u/. This /h/ sound was somewhat rounded but remained acoustically distinct from the proper /hʷ/.
In an unrelated sound change, other consonants which had previously been allophonically labialized before any following /u/ came to develop a contrast as well between the weak allophonic labialization that had been present for thousands of years and the new stronger type which was seen as analogous to a consonant being followed by /w/ (if voiced) or /hʷ/ (if voiceless). However, this contrast was only present before a following /u/, since in all other environments the new consonants merged with the preexisting labials. In the script, this contrast was marked by writing the weak labialization with the symbols for the plain consonants and the strong labialization with the symbols for the consonant + /w/, except in the case of /hʷ/, which had its own symbol.
Though this sound change may seem to have been a disruption of a previously well-balanced system, it actually made the consonant system more symmetric, since there was now a four-way contrast between ka kʷa ku kʷu, whereas previously there had only been ka kʷa k(ʷ)u (here, k and a are used as examples; the same change happened for all other phonemes as well).
Treatment of coronals followed by /w/
The parent language, Tapilula, had had an unusual series of labialized alveolar consonants tʷ dʷ nʷ. These were the only remnants of an older inventory in which almost all consonants had had unique labialized forms. Most of the labialized consonants in Tapilula had produced bilabial consonants in Gold, but the alveolars resisted the change. Even by the time the Gold language began splitting apart into its many daughter languages, these consonants still remained, but they no longer contrasted with a sequence of a coronal consonant followed by /w/ (written /hʷ/ when voiceless).
However, the pronunciation of these consonant clusters was different than that of other labial clusters. In word-initial position, tʷ dʷ nʷ were generally pronounced /tl dl nl/, and had begun to tend towards a simple /l l l/ in western dialects but /t d n/ in eastern dialects. In any other position, the consonants were generally retained as full clusters /tl dl nl/. But no dialect pronounced these clusters with the strong labialization characteristic of the other labial clusters.
Consonant sandhi and marginal phonemes
Bleeding of syllable-final /s/
A syllable-final /s/ before another consonant often metathesizes across the syllable boundary, meaning that its own syllable becomes open and the next syllable comes to begin with a cluster. In most clusters, the /s/ also changes to /h/, which reflects its original pronunciation in the Tapilula language. Thus, there is not actually a sound change of /s/ > /h/, but rather a lack of the otherwise common sound change of /h/ > /s/ in syllable-final position.
The common sequence /sd/ was pronounced and is generally Romanized as /dh/. For example, səs "sleep" + dŭdi "flower" produce sədhŭdi "sleep flower", which is considered to be a pure CVCVCV words. However, although /dh/ patterns as a single consonant rather than a cluster, this was not usually considered a phoneme, because it could only occur between two vowels, unlike the traditional voiceless aspirates /pʰ tʰ kʰ/ which are spelled with the simple letters p t k. In the daughter languages, /dh/ evolved to /t/ (in Khulls and Thaoa) or to /s/ (in Babakiam in initial position; elsewhere it disappeared). The changes in Thaoa and Khulls happened independently. An earlier pronunciation in all three branches was IPA /θ/.
Because the voiced bilabial stop /b/ arose only after long vowels and word-initial vowels (generally from disappearing noun classifier prefixes), the sequence /sb/ was far, far rarer than /sd/. Nevertheless, it followed a similar path, becoming /bh/ by analogy, and then evolving to /p/ in all three languages (it did not become /f/ in Babakiam because it was never a fricative, unlike /d/).
Vowels
Pure vowels
The vowel inventory was /a i u ə/, where the schwa vowel /ə/ was much rarer than the other three. There are also syllabic nasals /ṁ ṅ ŋ̇/.
Syllabic nasals
The syllabic nasals cannot have tones, but they can occur in closed syllables (e.g. /pṁp/), which is analogous to having a high tone. There were no sound changes that created new syllabic nasals in Gold; therefore, all syllabic nasals trace back to syllabic nasals in the parent language as well. They could contrast with sequences of a plain nasal plus a schwa in either direction, although the only non-syllabic nasal that could occur in syllable-final position was /n/ (although this had place allophones matching following consonants).
Tones
Tones were not well developed in Gold. Syllables could be high or low, and when a high tone occurred immediately before a low tone of the same vowel, this resulted in a falling tone which was considered a long vowel and is Romanized with a macron. However, there is no long form of the schwa; there are only ā ī ū. Note that high tone is Romanized with a grave accent, as in à, to keep in line with its descendants where this tone develops a final glottal stop.
Although there were only two tones, vowel sequences like àa and aà were becoming more common, and this is what led to the long tones of Khulls and its descendants, which are spelled ā and á respectively. Long tones also existed in Thaoa and Poswa but died out. The àa ~ ā type is much more common than aà ~ á. These could also occur with diphthongs, but only on the ā tone. That is, ài was common but aì was entirely absent, even over morpheme boundaries.
