A Phonology Sketch

I wanted to share a rough conlang phonology sketch I’ve been working on, inspired as a kind of mix between Chukchi, Nahuatl, Mongolian, Cherokee, PIE, and Muscogee amongst others. The goal was to try to have something slightly minimalist in terms of phonemes but still “interesting”. I’ve provisionally given it the label Konqara, but not sure if any existing conlang or natlang already uses that name. I’ve left some comments and todos inline in red.


The 15 consonants of Konqara are shown in the table below in IPA, and in the native orthography between brackets <> where there is a difference.

Stopptkkʷ <q>
Fricativeɬ <l>ʃ <s>x <h>xʷ <w>
Affricatetɬ <tl>tʃ <ts>
Liquidrjw <u>
Consonant Phonemes

I wanted a written form that only uses the core letters of the alphabet without too many digraphs (there are only 15 consonants afterall), but it’s hard to know how to represent xʷ especially. No single letter seems to fit, and writing coda xʷ as ‘hw’ seems extremely ugly. W seemed at least in the right ballpark, but that means I need another convention to write the semivowel. X would be another possibility, as it is sometimes used for the unlabialised velar fricative, so xʷ wouldn’t be such a stretch.


There are only four vowel qualities, shown in the diagram below. The peripheral vowels are written i, a, and o, and the reduced vowel (schwa) is written e.

The vowel /ə/ shows a wide range of realisations depending on surrounding vowels and consonants, although this variation is not contrastive. It is always short and centralised compared to the other vowels, but it approaches [ʊ̆] when it follows a labiovelar and [ɪ̆] following a post-alveolar. In other positions, there is a weaker influence from preceding vowels and coda consonants.

I should describe allophones of the other vowels, which are likely to be varied in a three vowel system.

The Syllable

Syllables have the structure (C)(C)V(C)(C), although onsetless syllables are limited to word-initial position. An onset may be made up of any consonant or a consonant followed by a glide. When the onset contains two consonants, the following additional restrictions apply:

  • Sequences of two glides are not allowed
  • The consonants t, ʃ, and tʃ may not be followed by j
  • Velar and labiovelar consonants may not be followed by w

Decide: does it fit to allow labiovelars to be followed by j?

In coda position, clusters may contain at most two consonants which must strictly decrease in sonority. The glides cannot occur as part of a coda. The remaining consonants are divided into three basis classes which must occur in order of decreasing sonority.

Sonorants m, n, r

Fricatives ɬ, ʃ, x, xʷ

Occlusives p, tɬ, tʃ, t, k, kʷ

These coda clusters are subject to the following additional restrictions:

  • p does not occur as a coda directly after a vowel
  • The sequences np, ɬtɬ, and mixed clusters of velars and labialised velars xkʷ and xʷk are not allowed

Syllabification and Word Shapes

A Note on Vowel Sequences

It is not clear whether the vowels i and o and the glides j and w are separate phonemes or allophones of unified underlying phonemes. The argument in favour of allophones is the restricted and predictable distribution of vowels and glides, and surface alternations between vowels and glides at morpheme boundaries. Additionally, epenthetic schwas never surface next to glides. An alternative analysis with separate underlying glides and vowels would require rules converting schwa+glide or glide+schwa into the equivalent surface vowels, or forbidding any consonant sequences which would require an epenthetic schwa next to a glide.

On the other hand, if it is assumed that vowels and glides are unified phonemes, then additional restrictions are still required. Since hiatus is not allowed and /a/ has no glide equivalent, then any sequence which would produce hiatus or gliding of /a/ is forbidden. This suggests that glides and vowels may be different phonemes.

In either case, only surface sequences with the form (G)VGVGV…GV occur. If the glides are underlyingly vowels, then the rule is to make the final vowel a surface vowel, and then convert all even numbered vowels from the end into glides and all odd numbered underlying vowels into surface vowels. For example, /ioa/ would be [iwa], and /iaia/ would be [jaja].

Not sure whether I need this section, but it’s a bit frustrating that different parts of the description of syllabification are easier to describe with different assumptions about whether glides and vowels are the same or not. This seems pretty common in natural languages too. E.g. in Latin glides and vowels were not distinguished in the orthography and seem to mostly be predictably syllabic or non-syllabic, but there are corner cases which might not be predictable.

Syllabification Within a Root or Lexical Morpheme

The Algorithm

The position of ə is predictable from the structure of the word, and it is assumed that in the underlying representation, ə is not present. The placement of epenthetic vowels is chosen to break up consonant clusters which are too long and produce legal words. It is assumed here that the glides j and w are allophones of the vowels i and o.

Within an individual morpheme, the rule is to scan from right the left, constructing maximal codas of descending sonority with at most two consonants. If this maximal coda is preceded by a consonant, then ə should be inserted, otherwise the preceding vowel is taken as the syllable nucleus. At this stage, no restrictions apart from sonority are taken into account: if the syllabification produces an illegal coda cluster, then it is resolved by the predictable changes described in the following section.

Before the nucleus, a maximal onset is constructed as follows: if the vowel is preceded by a high vowel then this vowel becomes a glide. If another consonant precedes the (glide+)vowel, then this is also assigned to the syllable onset. This process repeats until the entire morpheme is syllabified.

This can be illustrated with /tʃiʃnkiontmn/. The steps are as follows:

  1. tʃiʃnkiontmn
  2. tʃiʃnkiont.mən
  3. tʃiʃn.kjont.mən
  4. tʃi.ʃən.kjont.mən

Because the schwas and gliding are written in the standard orthography, this gives the final written form tsisenkjontmen.

Similar processes of epenthetic schwa insertion and coda maximisation are described for Chukchi and Halh Mongolian (see “The Phonology of Mongolian”). Chukchi is simpler (maximum syllable CVC), whereas Halh has a much bigger consonant inventory and codas of up to three segments.

Assimilatory Processes

Several rules apply after syllabification and schwa insertion. Except where noted, the results of all rules are reflected in the orthography. These are:

  1. The sequence tj becomes tʃ
  2. The glide j deletes after ʃ or tʃ
  3. Velars become labiovelars before w
  4. w deletes after labiovelars
  5. A sequence of velars and labiovelars assimilates in labialisation to the final member of the sequence
  6. p becomes x in coda position after a vowel
  7. The nasal n assimilates in place to a following consonant, but this is only written for the sequence np, which becomes mp
  8. The sequence ɬtɬ becomes xtɬ

These rules may feed each other. For example, xkw 🡲 xkʷ 🡲 xʷkʷ.

Is it just me who finds the sequence of ɬtɬ a bit tricky? Nahuatl has a change ltl 🡲 ll, which doesn’t seem to be part of a more general process of cluster simplification.

Also may need to consider result of identical segment clusters and affricate + fricative clusters. In Mongolian clusters created by morphology seem to result in secondary geminates, even though I haven’t see any mention of root / inherent geminate consonants?

At Morpheme Boundaries

When a suffix or prefix is added to a stem, the new material is normally treated as an unsyllabified addition to an already syllabified core. Syllables are then constructed using the same leftward scanning coda maximisation principle, but with additional restrictions. These are:

  • A schwa may be deleted from the stem, but no new schwa may be inserted into already syllabified material within it
  • A stem or prefix final high vowel may become a glide to resolve hiatus if it is not already preceded by a glide…
  • but no vowel which has already become a glide within the stem may be turned back into a syllable nucleus

This makes the morphological structure of a word relevant to its syllabification.

Schwa Deletion

Consider the underlying sequence /tarʃt/. As a single morpheme, this would be syllabified as tarest [ta.rəʃt]. If the word is composed of the morphemes tars-t then the outcome can only be tarset [tar.ʃət] because a schwa cannot be inserted into a new position in the root tars.

But because a stem can be modified by vowel deletion, if a stem ending in -əC gains a suffix beginning in a vowel, then the schwa deletes and the final consonant becomes the onset of a new syllable. This happens in taren-as [tarən-aʃ], which becomes tarnas [tarnaʃ]. Similarly, a vowel-final prefix may trigger schwa deletion if the stem begins with a single consonant followed by a schwa and the coda of the first syllable is at most one consonant of lower sonority than the onset. Compare the outcomes of i-metsan, which undergoes deletion to become imtsan, and i-tseqni, which does not lose its vowel because /tʃkʷ/ is not a legal coda. These rules apply even if the suffix or prefix vowel is itself an epenthetic vowel. If vowels are deleted from the stem, then the changes and assimilations listed in the previous section apply.

This is shamelessly stolen from “The Phonology of Mongolian”, which argues that a similar rule explains apparently unpredictable schwas in some words. Chukchi also deletes schwas when affixes are added, but I haven’t seen any similar issues of morphology-driven irregular epenthetic vowel placement described.

Hiatus Resolution

If two non-schwa vowels are adjacent at a morpheme boundary, then additional resolution rules apply. These are:

  1. If the vowels are identical then merge them
  2. If the first vowel is high and not preceded by a glide, then convert it into a glide
  3. If the first vowel is preceded by a equivalent glide (ji, wo) then delete it so only the glide remains
  4. Otherwise, insert an epenthetic glide based on the vowel qualities:
    1. j if the second vowel is i, or the sequence is i.a
    2. w if the second vowel is u, or the sequence is u.a

These rules feed the normal assimilation rules, especially the regressive assimilation rules between t and j, and between velar consonants and w. The table below summarises the most frequent outcomes of vowels in hiatus.

Most Frequent Resolutions of Vowel Hiatus

Irregular Alternations

Some roots and affixes have irregular forms depending on whether a preceding or following morpheme begins with a vowel or a consonant. The most common alternation is between the glides and the consonants s and xʷ. Roots that end in these consonants may show irregular lenition of them to the corresponding glides when a vowel initial suffix is added.

In some morphemes a historical schwa has become a copy of a previous vowel, but still deletes when combined with certain affixes.

Stress and Word Accent

Stress is not distinctive. Words begin with a higher pitch on the first syllable, which then drifts downwards. Sentences typically show a downdrift effect, which each initial high pitch being lower than the initial pitch of the previous word.