Amongst the languages of the world, some are known for having an unusually large number of phonemic vowels in their phonological inventories, with one of the most popular examples being the Danish language, if we are discounting additional phonation types like breathiness. Today, I want to take a look in the other extreme, and try to dissect if languages with fewer than 2 phonemic vowels exist.

By some analyses, there are already languages which are proposed to have exactly two phonemic vowels, usually /a/ and /ə/, often with a range of allophonic variations. Examples of these languages include the Upper Arrernte language, and the Northwest Caucasian languages like Abaza, Abkhaz, and Ubykh. This paucity of distinctive vowels is usually made up for with a rich consonant inventory, which typically involves some form of secondary articulation like labialisation [kʷ] and palatalisation [tʲ], that may contribute to allophonic variations of the two vowels like [o] and [e].

Today, we will examine the case of the Moloko language, an Afroasiatic language spoken by around 10000 native speakers predominantly living in the Far North Province of Cameroon. Classified as a Central Chadic language, the Moloko language is the language that has been proposed to contain just one phonemic vowel, /a/, with an epenthetic schwa that serves to break up consonant clusters. Whether or not this schwa constitutes a phonemic vowel is open for debate, as this schwa occurs rather predictably, such as in unstressed syllables or in epenthesis.

Resources surrounding the Moloko language are rather scarce, largely attributed to the relatively poor documentation of the language prior to the 1970s. Contemporary sources concerning the analysis of vowels in Moloko and the Central Chadic languages are largely attributed to linguists Ekkehard Wolff, Catherine Bow, and Dianne Friesen, who have either published grammars or conducted analyses studying the vocalogenesis (development of vowels) in the Central Chadic languages, or studied vowel prosody systems to tell apart the phonemic vowels from the surface vowels.

Bow’s 1997 analysis has presented us with a total of 9 surface vowel sounds for Moloko, [i ɪ ɛ œ ə a o ʊ u], though the author also noted other transcriptions resulting from slight differences in vowel heights and laxness. Rossing’s analysis was also cited, presenting us with a set of 7 vowels thought to be phonemically distinguished, /i e ø ə a o u/. It was also noted that [i] and [u] were allophones of [ɪ] and [ʊ], resulting from assimilation to the approximants [w] and [j].

Bow’s thesis also discussed the two-vowel hypothesis and one-vowel hypothesis for Moloko, both of which were reduced from the Rossing analysis using this principle called the principle of Lexical Minimality, in which the phonological information used to tell apart one lexical item from another is minimised through some certain processes.

The main similarity between these hypotheses is how allophones arise from prosody or word-level processes associated with palatalisation (the ‘j’ part in ‘tj’) and labialisation (the ‘w’ part in ‘kw’). For the two-vowel system, [i] and [u] are considered allophones of /ə/, and [e] and [o] are considered allophones of /a/. The /ə/ may be omitted in fast speech or in other contexts. The one-vowel hypothesis pins on the extent to which this reduction from Rossing’s analysis occurs. Friesen suggests support for the one-vowel system as the presence of /ə/ is predictable, especially in epenthesis, leaving /a/ as the only underlying vowel phoneme. As such, the debate stands, is /ə/ phonemic, or is it just there to break up consonant clusters but also subject to some processes that occur at a word level?

Nevertheless, we can piece which surface vowels may be formed from their respective vowels and word-level processes. From Friesen’s grammar of the Moloko language, for the /a/ vowel, we have:

• Plain, no process – [a]
• Labialisation – [ɔ]
• Palatalisation – [ɛ]
• [j] adjacent – [a]
• [w] adjacent – [a]
• /j/ or inherent labio-velar adjacent – [œ]

And for the epenthetic /ə/, we have:

• Plain, no process – [ə]
• Labialisation – [ʊ]
• Palatalisation – [ɪ]
• [j] adjacent – [i]
• [w] adjacent – [u]
• /j/ or inherent labio-velar adjacent – [ø]

This presents another question, is there another linguistic model one could apply to assess the phonemicity of these surface vowels?

This is largely the theme of Bow’s 1999 thesis, in which the Optimality Theory was applied. Briefly put, this is a linguistic model in which the observations we make in a language are resultant from optimising satisfaction given the conflicting restrictions and constraints, and a certain input. These constraints are ranked, with high-ranked ones being prioritised over lower-ranked ones. Under Optimality Theory, there are two primary forces at play, which are markedness, which languages tend to bias against, and faithfulness, which languages tend to bias towards. As such, an increase in markedness, where certain structures or constructions are favoured over other ones, would lead to a decrease in faithfulness, where the output tends to be close to the input, preserving distinctions between words.

Optimality Theory has been built upon by other linguists, leading to Correspondence Theory, where under perfect faithfulness, the output would be identical to the input. This places an additional set of constraints on the faithfulness bit. Under Bow’s analysis, both hypotheses would each violate a constraint on faithfulness, with the two-vowel hypothesis violating the ‘maximise’ constraint, where in this context, phonological deletion must not occur. The one-vowel hypothesis would violate the ‘dependent’ constraint, which prevents phonological epenthesis in this context. These violations occur because the two-vowel hypothesis would entail some form of deletion of certain vowels, while the one-vowel hypothesis would entail the presence of epenthesis during syllabification.

So, what about prosodic effects? After all, these processes have led to the various surface vowels in the Moloko language mentioned earlier. Here, there is this principle called the Richness of the Base, which states that “no constraints hold at the level of the underlying form”. This implies that the input should not be bound by any constraints, so that the output are resultant from any constraints that are applied, with this one being prosodic effects, where a prosody is associated with a syllable. If this is not associated in the input, then by faithfulness, the feature will not be associated in the output. She found that palatalisation mainly targets vowels, alveolar sibilant and affricate consonants, while labialisation mainly targets vowels and velar consonants.

Bow’s analysis using Optimality Theory and Correspondence Theory supports both the one-vowel and two-vowel systems, and that they do not really affect the outcome, leaving this up to, once again, the level of abstraction one uses in their analysis. Nevertheless, there are several arguments that might push in favour of the two-vowel hypothesis.

Firstly, epenthesis. The one-vowel hypothesis makes the /ə/ vowel (or schwa) an epenthetic vowel, but this schwa may be affected by prosody, leading to the surface vowels mentioned earlier. This would entail the proposition of three epenthetic vowels, [ə], [i] and [u], which would in turn violate the principle of Lexical Minimality, which states that the amount of phonological information used to distinguish between lexical items would be minimised. This makes an argument for the two-vowel hypothesis.

Secondly, optimisation. This would favour a closer correspondence, or perhaps greater fidelity, between the input and output when applying the analyses used in Bow’s study. Here, the two-vowel hypothesis demonstrates a closer correspondence despite the analyses supporting both hypotheses, making it yet another point in favour of the two-vowel hypothesis. This leads to the issue of transparency, which is favoured when talking about fidelity, faithfulness, and correspondence. This biases towards the two-vowel hypothesis, which has demonstrated to be more transparent than the one-vowel one.

So is there really a language with fewer than two phonemic vowels? According to past research, it ultimately boils down to how one analyses it. Bow’s analysis shows support for both hypotheses, but biases towards the two-vowel hypothesis can still occur when applying Optimality Theory and Correspondence Theory. I think exploring topics like this underscores the difficulties in assessing phonemicity of certain sounds, especially when it is nearly inconceivable to think about a language that has an absurdly restricted number of a certain type of sounds. Academic coverage concerning the Moloko language still stands pretty scarce, so the question is still relatively open on whether or not Moloko really has one or two vowels.

But I have found an even more unbelievable claim, that there exists a language with supposedly ‘no phonemic vowels’. How would that be possible? Check back later for a look into the Wandala language, and the competing analyses studying its vowel system.

Further Reading

Bow, C. (1997) ‘Description of Moloko phonology’, Summer Institute of Linguistics.

Bow, C. (1999) ‘The vowel system of Moloko’, MA thesis, University of Melbourne.

Friesen, D., Isaac, M. D., Gaston, A. & Samuel, M. (2017) ‘A grammar of Moloko’, Language Science Press, Berlin, Germany.