A while ago, we covered the possibility of a language that distinguishes between fewer than two vowels, and reached a rather ambiguous answer. However, there have been arguments and proposals made by some linguists that there can be languages that do not distinguish between vowels at all. And so today, we will take a look at one example, and dissect how linguists reach such an outlandish proposal.

When we refer to a ‘language without vowels’, we do not mean that a language is only ever stocked with consonants. Instead, we are talking about the absence of the vowel’s role in distinguishing between lexical terms (like ‘bug’ and ‘bag’). This is because many phonological inventories one would see when looking up the features of a language showcase phonemic vowels, and that these vowels are used to distinguish between different words.

Firstly, a quick overview of the Kabardian language. The Kabardian language is a Northwest Caucasian language under the Circassian branch predominantly spoken in Russia’s North Caucasus republics, Kabardino-Balkaria and Karachay-Cherkessia. It is commonly thought of as a dialect of the Adyghe language, though historically, during the Soviet era, it was treated as a separate language from Adyghe. Spoken by almost 2 million native speakers today, Kabardian shares many grammatical and phonological features with the other Circassian language, such as a large number of consonants, juxtaposed against a very small number of vowels.

The Kabardian language seems to distinguish between at least seven surface vowels, [i], [e], [a], [u], [o], [ɨ], and [ə]. These surface vowels are a result of some form of modification from the nature of the consonant preceding the vowel. For example, palatalised consonants or sounds with this /j/ sound would produce the [i] or [e] sounds, while labialisation, or consonants that carry the /w/ sound, would result in the [u] and [o] sounds. Thus, the first step used to boil down these surface vowels would be to identify these common patterns of sound modifications, and boil them down to the central vowels.

This is where we get the commonly-accepted position that Kabardian has a vertical vowel system. This is corroborated by acoustic analyses like those conducted by John D. Choi, which suggested that Kabardian has three vowels, distinguished by vowel height (and hence a vertical vowel system). These three vowels would be, in decreasing vowel height, [ɨ], [ə], and [a]. The surface vowels mentioned above would be deemed as ‘contextual variants’ of these central vowels, since their variation depends on the nature of the preceding consonant. Other analyses proposed the three-vowel system that includes the vowels [ə], [a], and the long vowel [a:] instead. However, most conservative interpretations suggested that there are two contrastive vowels instead, which are [ə] and [a], much like the vowel systems of languages like Upper Arrernte.

But where do we go from here, and boil these three vowels down to zero?

The main proposals of the lack of phonemic vowels in Kabardian stem from the publications made by Kuipers, with the latter garnering more academic attention. The first vowel we will focus on is the long vowel [a:]. Here, Kuipers suggested analysing this as a sequence instead, as /ħa/. This sequence only occurs in the plural suffix in Kabardian grammar. Additionally, [a:] is the only syllable-initial vowel. As such, by interpreting this long vowel as a ħa sequence instead, it would entail that Kabardian syllables always have a consonant initial. Through the proposition of some metathesis in the ħa sequence, this would become ah, and like the modification of phonemes by frontness or backness mentioned earlier, this modification would become that of apparent length. This reduces the vowel inventory down to two vowels, that of the conservative interpretation.

Next, Kuipers had to show that the vowel [ə] is predictable in relation to stress patterns in Kabardian. To him, Kabardian appeared to be essentially a string of consonants, with some vowels inserted in a predictable pattern. He noticed that the [ə] only appeared after consonants, and not [a]. Additionally, [ə] never appeared before a consonant that is followed by [a], with some exceptions being after certain consonants that precede a sequence -ra. Where this [ə] can appear is dictated by certain rules concerning the type of consonants that surround a position of interest, which may factor in the sonority of these other sounds in the sequence. In a stressed syllable, where [a] does not appear, [ə] would appear instead.

He also went further into talking about post- and pre-accentual stress, and how [ə] does not necessarily play a distinctive role here. However, what I find weird is that Kuipers introduced an additional notation denoting stress, and how minimal pairs would differ by the presence or absence of such notations, drawing a few examples to illustrate his point. From this analysis focusing on how [ə] is related to stress, he suggested that the phonemicity of the [ə] vowel would cause complications in how morpheme boundaries are drawn, and that its elimination is perhaps the most parsimonious way of interpreting the vowel system of the Kabardian language.

To get rid of the final outstanding vowel, Kuipers proposed that [ə] and [a] are simply differing features of openness. Note that when we talk about vertical vowel systems like this one in Kabardian, the vertical axis denotes the openness of the vowel, with the upper end being close, and the lower end being open. Essentially, [ə] and [a] only differ by how open they are, with [a] being the more open one. And so, Kuipers would have thought, why not make this feature of openness as a consonantal feature instead? Thus, he proposed that a consonant can carry a feature where the resulting epenthetic vowel would be more open, and hence producing the [a] sound instead of [ə]. With this argument, Kuipers had effectively wiped out the vowels in Kabardian.

Anderson suggested that Kuipers’ analysis could lend some plausibility, by applying a framework where certain segments may be ‘contrastively unspecified’ and other phonological features may not be contrastive. Meanwhile, Wood’s 1990 review suggested that how these vowels are assimilated into the surrounding consonants are more complex in nature than previously thought, resulting in some ambiguity over the phonemicity over the /a/ and /a:/ vowels.

Halle’s 1970 publication is perhaps one of the more significant rebuttals to Kuipers’ vowel-less argument, and asserted that Kuipers’ arguments would eventually lead towards the two-vowel hypothesis. Halle also remarked that Kuipers’ method of reducing vowels essentially redefined vowels as a certain phonological feature, such as a feature of openness. Thus, when extended to every language in the world, it could technically be argued that all of them are ‘vowel-less’. In Halle’s publication, he pointed out how Kuipers had to add new phonological notations to reduce the two vowels, such as the ‘stress phoneme’, marked by an apostrophe. Additionally, Kuipers was criticised as failing to establish the predictability of Kabardian vowels, and thus failing to demonstrate that Kabardian has fewer than two phonemic vowels, let alone zero. This publication is quite interesting, and I recommend checking it out.

As exemplified by Halle, Kuiper’s vowel-less proposition is generally not entirely accepted by the linguistic community, although the door is still open on the verity of this claim. Some of the more conservative analyses published in the early 1990s would argue for a two-phoneme analysis instead. The majority of linguistic opinion seems to be that while Kuipers’ vowel-less proposition is not entirely attested, we should not rule out the possibility of the existence of vowel-less languages, despite the lack of such an attested example. As Université de Montréal linguist Kevin Tuite put it, the existence of object-first languages, that is, object-subject-verb and object-verb-subject word orders was initially snubbed from ‘universal grammars’, but attested counterexamples were raised, with these languages predominantly spoken in the Amazon. As such, as with the example of Kabardian and previously, Wandala, while not being examples of vowel-less languages, perhaps such a language may exist, or have been in existence at some point in history.

Further reading

Anderson, J. (1991) ‘Kabardian disemvowelled, again’, Studia Linguistica, 45(1/2), pp. 18-48.

Choi, J.D. (1991) ‘An acoustic study of Kabardian vowels’, Journal of the International Phonetic Association, 21(1), pp. 4-12.

Colarusso, J. (1992) A Grammar of the Kabardian Language, University of Calgary Press.

Halle, M. (1970) ‘Is Kabardian a vowel-less language?’, Foundations of Language, 6(1), pp. 95-103.

Kuipers, A.H. (1960) Phoneme and morpheme in Kabardian (Eastern Adyghe), The Hague: Mouton.

Wood, S.A.J. (1990) ‘Vertical, monovocalic and other “impossible” vowel systems: a review of the articulation of the Kabardian vowels’, Working Papers, 36, pp. 191-212.