In the world’s languages, speech is often organised into sequences of speech sounds called syllables. For instance, in English, “golf” has one syllable, while “defenestration” has five syllables. However, in British Columbia, Canada, there exists a language which appears (emphasis on the appears here) to do things a little differently. In fact, it has been proposed that this language has no syllables, or at least challenges the notion that speech sounds are pretty much always organised into syllables. Today we will take a look at Nuxalk, or Itnuxalkmc, a Salishan language spoken mainly in Bella Coola, BC.

Also referred to as Bella Coola in linguistics literature, the Nuxalk language is spoken by around a dozen native speakers today. While classified as a critically endangered language, Nuxalk is currently taught in the schools in the province of British Columbia and the Nuxalk Nation. Within the Salishan language family, Nuxalk forms its own branch, with the other major branches being the Coastal Salish branch and the Interior Salish branch, the latter of which contains the Shuswap language or Secwepemctsín which we have covered previously.

Nuxalk has a total of 3 vowels, /i a o/, all of which exhibit allophonic variation depending on where in the word or syllable they occur. There are also 29 consonants, with stop consonants either being aspirated or ejective. Velar and uvular consonants (like /k/ and /q/ respectively) may also be labialised. But it is the consonants which we are particularly interested in today.

The Nuxalk language, alongside some Salishan languages, is particularly notable for its unusual word formation, as entire words can contain long sequences of consonants, or just exist as a consonant sequence with no apparent vowel. This includes words like p’s (bend) and sps (northeast wind), and even words that translate to entire sentences like clhp’xwlhtlhplhhskwts (then he had had in his possession a bunchberry plant).

When presented with a long sequence of consonants like this, one would expect that there has to be some kind of epenthesis occurring, where a schwa could be inserted to break this sequence up. But having listened to some spoken Nuxalk, it seems that these long sequences of consonants generally do not have such epenthesis occurring. With no vowels nor other sonorant sounds that can function as vowels, Nuxalk poses a challenge to linguists in defining how many syllables such obstruent-only words contain. And so, to start off, what makes a syllable?

According to the 2013 Cambridge Dictionary of Linguistics, a ‘syllable’ is defined as a unit of organising speech sounds, and syllables consist of smaller parts, which are:

• An optional onset
• An obligatory rhyme or rime, which is split into
  • An obligatory nucleus
  • An optional coda

The onset may consist of a consonant or a consonant cluster, while the nucleus may contain a vowel or a syllabic consonant like the Czech and Slovak r and l. Lastly, there is the optional coda, which occurs as a consonant or a consonant cluster. Thus, in the word ‘dog’, d would be the onset, o would be the nucleus, while g would be the coda. With this in mind, let us take a look at an example of a Nuxalk word, scs or transcribed as [sxs], which means ‘seal fat’. How would we define which is what? Is there a nucleus, and how many syllables are there?

The first claim that Nuxalk ‘does not contain syllables’ could be traced back to the analyses conducted by Stanley S. Newman in 1947, in his publication in the International Journal of American Linguistics titled “Bella Coola I: Phonology”. Here, Newman observed that there were no restrictions on the sequencing of consonants, and a lack of contrastive stress or tone which could be used to help determine where one syllable begins and the other ends.

Following this initial proposal, several linguists have proposed their own models to argue against Newman’s claims that there were no syllables in Nuxalk. The objectives mainly pertained to two key areas, (1) in obstruent-only words, where are syllable boundaries defined, (2) which sounds would function as a ‘nucleus’ in such words. This leads to the 1955 publication by Charles F. Hockett, titled A Manual of Phonology, where he argues that each obstruent in such Nuxalk words would constitute their own syllable, with the obstruent playing the role of the onset. Such a concept was also supported by E. C. Fudge in 1976, where he also posits a similar syllable structure for obstruent-only Nuxalk words, but the obstruent here may play the role of the onset and the coda, with the nucleus being a “null syllable peak” transcribed as *.

But is there a general pattern in which segment types could function as nuclei in such Nuxalk words? One of the first studies which aimed to elucidate this was conducted by Greenberg, and published in 1962. Using a rather limited set of data, Greenberg thought that every Nuxalk word should contain a sonorant segment (defined as m, n, l, i, a, u), a non-lateral fricative (s, x, xʷ, χ, χʷ), or an alveolar plosive (c, t), or a combination of these.

The notion of syllabic consonants is not new. The Slavic languages are particularly well known for its use of [r] and [l] sounds to function as syllable nuclei, and can be seen in words like Vltava, the name of the river that flows through Prague, Czechia, and Macedonian прв (first). In Hokkien, there are syllabic nasal consonants ‘m’ and ‘ng’, the latter of which is also a family name, usually written with the character 吳 or 吴. Similarly, /m/ sound in Bantu languages like Swahili can also be syllabic when followed by a consonant, like the word mtu (person).

Despite this, there are numerous counterexamples that refute Greenberg’s proposal, likely due to the small dataset Greenberg worked with. These include words like p’qʷ (fine, loose, powdery) and t’kʷ (to bleed), both of which feature none of the segments Greenberg set out.

This is where we bring up the topic of using fricative consonants as syllabic nuclei in languages like Nuxalk. Hoard’s 1978 publication mentioned how syllable boundaries may be defined once a nucleus is identified. Here, Hoard set out six primary segmental generalisations identified in Nuxalk words that only contain obstruents:

• Syllables with obstruent nuclei may take on a single segment for an onset.
• Only plain stops and affricates can occur in the onset.
• Only fricatives can then occur in the nucleus.
• Ejective stops and affricates are always syllabic.
• Syllabic stops and affricates do not take an onset.
• Syllables with obstruent nuclei do not take a coda.

By Hoard’s generalisations, a word like q’s would have two syllables, while sq’ʷ would also have two syllables. The word qstχ may be parsed as having two syllables, [qs.tχ].

These models may be organised into a few hypotheses, from Newman’s No Syllable Hypothesis to the Simple Syllable Hypothesis referred to by Bagemihl. In the early 1990s, the moraic licensing theory was proposed by Bagemihl, and that its maximum syllable structure is CCVVC, but this was refuted by Cook in 1994. This theory entailed that Nuxalk does not delete segments that cannot be syllabified, but instead, deletes segments that cannot become morae, that is, the smallest unit of timing in spoken language. According to Bagemihl, not every mora would belong to a syllable in Nuxalk, which is refuted by Cook as misleading. Cook asserts that the unsyllabified obstruents in Nuxalk persist in the language because Stray Erasure is not allowed, that is, the process of deleting segments which cannot be incorporated into a well-formed syllable.

This brings us to the Good Enough Nucleus Hypothesis proposed by Mellesmoen in 2021. In this hypothesis, fricatives may function as syllable nuclei if there are no more sonorant sounds available, and when epenthesis is a ‘more marked option’. It must be noted that fricatives are not the ideal sounds to be used as syllable nuclei, just that they are good enough to act as one when no other strategies that result in a prosodic word are favourable. This makes syllable structures such as CF, where F is a fricative consonant, possible, and words like t’xt (stone) would have one syllable, in contrast to the zero or three in predecessor analytic methods. This in a way is similar to Hoard’s 1978 publication, but also accounts for certain patterns that arise in Nuxalk, with one salient one being reduplication patterns.

Nuxalk is not particularly unique in terms of its usage of such bizarre syllabic nuclei, as similar usage has been documented in the Salishan, Wakashan, and the Chemakuan languages in the region, and also in some Berber languages over in North Africa. Some varieties of the Miyako language spoken in Okinawa, Japan also have some similar syllable structures, such as the words kff (to make), psks (to pull), and ff (comb).

Even with this model of Nuxalk syllable structures, there are still some unanswered questions. What happens when a Nuxalk word consists of only stop consonants, like q’t (go to shore)? Nevertheless, the Good Enough Nucleus Hypothesis presents a development in the linguistic understanding of how Nuxalk syllables work. Rather than being interpreted as having “no syllables at all”, analyses such as Optimality Theory and those of reduplication patterns in Nuxalk have led to a more comprehensive understanding of Nuxalk syllable structures.

Further Reading

Bagemihl, B. (1991) ‘Syllable structure in Bella Coola’, Linguistic Inquiry, 22(4), pp. 589-646.

Cook, E. (1994) ‘Against moraic licensing in Bella Coola’, Linguistic Inquiry, 25(2), pp. 309-326.

Fudge, E. (1976) ‘Phonotactics and the syllable’, Linguistic Studies Offered to Joseph Greenberg, Anma Libri, Saratoga, California.

Greenberg, J. (1962) ‘Is the vowel-consonant dichotomy universal?’, Word, 18, pp. 73-81.

Hoard, J. E. (1978) ‘Syllabication in Northwest Indian languages, with remarks on the nature of syllabic stops and affricates’, Syllables and Segments, North Holland, Amsterdam.

Hockett, C. F. (1955) ‘A manual of phonology’, Indiana University Press, Bloomington.

Levin, J. (1985) ‘A metrical theory of syllabicity’, PhD dissertation, Massachusetts Institute of Technology, Massachusetts.

Mellesmoen, G. (2021) ‘Syllables and reduplication in Bella Coola (Nuxalk)’, Papers for the International Conference on Salish and Neighbouring Languages, 56, pp. 226-243.

Newman, S. (1947) ‘Bella Coola I: Phonology’, IJAL, 13, pp. 129-134.