East ǃXoon (Traill)

Traill describes the phonations of the East ǃXoon dialect as plain ⟨a⟩, murmured ⟨ah⟩, or glottalized ⟨aʼ⟩. [a o u] may also be both glottalized and murmured ⟨aʼh⟩, as well as pharyngealized ⟨a̰⟩/⟨aq⟩ or strident ('sphincteric') ⟨a̰h⟩/⟨aqh⟩. [a u] may be both pharyngealized and glottalized ⟨a̰ʼ⟩, for 26 vowels not counting nasalization or length.

Murmured vowels after plain consonants contrast with plain vowels after aspirated consonants, and likewise glottalized vowels with ejective consonants, so these are phonations of the vowels and not assimilation with consonant phonation.

Vowels may be long or short, but long vowels may be sequences rather than distinct phonemes. The other vowel quality sequences—better known as diphthongs—disregarding the added complexity of phonation, are [ai, ae, ao, au, oi, oe, oa, ou, ui, ue, ua].

All plain vowels may be nasalized. No other phonation may be nasalized, but nasalization occurs in combination with other phonations as the second vowel of a sequence ("long vowel" or "diphthong"). These sequences alternate dialectally with vowel plus velar nasal. That is, the name ǃXóõ may be dialectally [kǃxóŋ], and this in turn may be phonemically /kǃxóɲ/, since [ɲ] does not occur word-finally. However, this cannot explain the short nasal vowels, so Taa has at least 31 vowels.

A long, glottalized, murmured, nasalized o with falling tone is written ⟨ôʼhõ⟩. A long, strident nasalized o with low tone is written ⟨òqhõ⟩, since Traill analyzes stridency as phonemically pharyngealized murmur. (Note that phonetically these are distinct phonations.)

West ǃXoon (DoBeS)

DoBeS describes the phonations of the West ǃXoon dialect as plain, a e i o u; nasalized, an en in on un; epiglottalized or pharyngealized, aq eq iq oq uq; strident, aqh eqh iqh oqh uqh; and glottalized or 'tense', aʼ eʼ iʼ oʼ uʼ.

East ǃXoon (Traill)

Consonants in parentheses are rare.

The nasal [ɲ] only occurs between vowels, and [ŋ] only word finally (and then only in some dialects, for what are nasal vowels elsewhere), so these may be allophones. [β], [l], [j] also only occur in medial position, except that the last is an allophone of rare initial [ɟ]. [dʒ] and [w] (not in the table) occur in loans, mostly English.

Taa is typologically unusual in having mixed-voice ejectives. Juǀʼhoansi, which is part of the same sprachbund as Taa, has mixed voicing in [d͡tʰ, d͡tʃʰ, d͡tsʼ].^[14]

Taa may have as few as 83 click sounds, if the more complex clicks are analyzed as clusters. Given the intricate clusters posited seen in the non-click consonants, it is not surprising that many of the Taa clicks should be analyzed as clusters. However, there is some debate whether these are actually clusters; all non-Khoisan languages in the world that have clusters allow clusters with sonorants like r, l, w, j (as in English tree, sleep, quick, cue), and this does not occur in Taa.

There are five click articulations: bilabial, dental, lateral, alveolar, and palatal. There are nineteen series, differing in phonation, manner, and complexity (see airstream contour). These are perfectly normal consonants in Taa, and indeed are preferred over non-clicks in word-initial position.

The DoBeS project takes Traill's cluster analysis to mean that only the twenty tenuis, voiced, nasal, and voiceless nasal clicks are basic, with the rest being clusters of the tenuis and voiced clicks with x, kxʼ, q, ɢ, qʰ, ɢqʰ, qʼ, ʔ, h and either mˀ or nˀ. Work on Taa's sister language Nǁng suggests that all clicks in both languages have a uvular or rear articulation, and that the clicks considered to be uvular here are actually lingual–pulmonic and lingual–glottalic airstream contours. It may be that the 'prevoiced' consonants of Taa, including prevoiced clicks, can also be analyzed as contour consonants, in this case with voicing contours.

* DoBeS only matches 17 series to Traill, as the Ʞkʰ–Ʞqʰ and Ʞkʼ–Ʞqʼ distinctions he discovered had not yet been published. DoBeS ⟨Ʞh⟩ and ⟨Ʞqh⟩, respectively, correspond to the former pair, while ⟨Ʞʼ⟩ and ⟨Ʞʼʼ⟩ (presumably in that order, as uvular clicks tend to have a delayed release) correspond to the latter pair.

Traill's account of East ǃXoon leaves for voiceless series of clicks without equivalents with a voiced lead. The DoBeS account of West ǃXoon, which uses voicing for morphological derivation to a greater extent than East ǃXoon does, has four additional series, written nꞰʼʼ, gꞰʼ, gꞰqʼ and nꞰhh in their practical orthography. The first three match the unpaired glottalized series of Traill, Ʞˀ (= ᵑꞰˀ), Ʞkʼ, Ʞqʼ. If Traill's ɡꞰh series is the voiced equivalent of plain aspirated Ʞʰ, rather than delayed aspirated, that would leave the DobeS nꞰhh series as voiced delayed aspiration.

All nasal clicks have twin airstreams, since the air passing through the nose bypasses the tongue. Usually this is pulmonic egressive. However, the ↓ŋ̊Ʞh series in Taa is characterized by pulmonic ingressive nasal airflow. Ladefoged & Maddieson (1996:268) state that "This ǃXóõ click is probably unique among the sounds of the world's languages that, even in the middle of a sentence, it may have ingressive pulmonic airflow." Taa is the only language known to contrast voiceless nasal and voiceless nasal aspirated (i.e. delayed aspirated) clicks (Miller 2011).

West ǃXoon (DoBeS)

West ǃXoon has 164 consonants in a strict unit analysis, including 111 clicks in 23 series, which under a cluster analysis reduce to 87 consonants, including 43 clicks.

These are written in the practical orthography (Naumann 2008).^[17] Marginal consonants are not marked as such.

Vowel nasalization is only phonemic on the second mora (in CCVV etc. syllables), as it is a phonetic effect of the ⟨nꞰhh⟩ clicks on the first mora. The ⟨nꞰhh⟩ clicks do not make the following vowel breathy, maintaining a contrast between ⟨nǂhha⟩ and ⟨nǂhhah⟩. Likewise, while ⟨gꞰʼ⟩ clicks do make the following vowel creaky, there is a delayed onset to the vowel and the amplitude of the glottalization of ⟨gǂʼaʼ⟩ is less than that of ⟨gǂaʼ⟩ with a phonemically creaky vowel.

In an attempt to keep the phonemic inventory as symmetric as possible, the DoBeS team analyzed as segments two of the click types that Traill analyzed as clusters. These are the pre-glottalized nasal clicks, ʼnꞰ, which Traill had analyzed as /Ʞ/ + /ʼn/, and the voiced aspirated clicks, gꞰh, which Traill had analyzed as /ɡꞰ/ + /qʰ/.

The expectation, from the morphology of ǃXoon, for voiceless-voiced pairs of click clusters led to the discovery of several series not distinguished by Traill. (This morphology appears to be more pervasive in West ǃXoon than in the East ǃXoon dialect that Traill worked on.) These are voiced click types which may not exist in East ǃXoon at all, namely nꞰʼʼ, nꞰhh, gꞰʼ, and gꞰqʼ. It also lead to the rediscovery of two series that Traill had not been able to publish before his death. Thus the DoBeS team distinguishes two series, Ʞqh and Ʞh, for Trail's Ʞqh and Ʞkh, as well as Ʞʼʼ and Ʞʼ for Traill's Ʞqʼ and Ʞkʼ (or perhaps vice versa). If Traill's Ʞkh series is to be analyzed as kꞰ + h, then that would require a different assessment of Traill's delayed-aspiration series.

Under the contour analysis of Miller (2009), the distinction between simple and contour clicks largely parallels the DoBeS identification of clusters, apart from the last four rows (Ʞʼʼ, nꞰʼʼ, Ʞhh, nꞰhh), which are considered to be simple clicks.