It’s not so easy to find good books about Amazonian languages, but there are a number of interesting language features that seem more common there than globally. Two which I find quite interesting are nasal harmony and frustrative marking. This post will describe nasal harmony, and part 2 will cover the frustrative.

Nasal Harmony

Everyone (well, almost all linguists) is familiar with vowel and consonant harmony. The most famous forms of vowel harmony are cases like front-back, rounding, height and tongue root harmony, occasionally with some interaction with surrounding consonants (e.g. backing of velars to uvulars in Mongolian). Pure consonant harmony is probably less common, and tends to be agreement in anteriority or retroflexion amongst coronals.

Nasal harmony is interesting because it often has very visible effects on both vowels and many consonants. There is a good overview here, but the basics are:

1. Nasalisation spreads over a multi-syllable domain, which is normally within a single phonological word (but may be less than the entire word)
2. In most (all?) languages, nasality seems to be the dominant feature, so domains are oral by default and nasalised if a trigger/source segment is present
3. Nasal spread may be progressive, regressive, or both
4. The domain is normally bounded by either a word boundary or a blocking element. Blocking elements can either be specific consonants, vowels, word boundaries, or prosodic features (e.g. primary stress)
5. In examples I’ve seen in detail, there are normally (almost) full corresponding sets of oral and nasal vowels, so nasal spread does not neutralise vowel contrasts
6. A subset of consonants within or adjoining the nasalisation domain also assimilate in nasality (sometimes with contrast neutralisation)

By far the most common blocking elements are consonants, and the word boundary if not. Consonants follow a hierarchy, with sonorants least likely to block and stops most likely to block nasal spread:

laryngeals > glides > liquids > fricatives > obstruent stops 

If a consonant does not block nasal spread, it usually but not always also nasalises, with the exception of the glottal stop for articulatory reasons (if airflow is stopped in the glottis then it cannot flow through the nose). Voicing seems relevant, since some languages show a voiced stop/obstruent ~ nasal alternation, but have unvoiced obstruents as blockers which do not assimilate.

Example: Aguaruna

The following description comes from Simon Orverall’s grammar. In Aguaruna, a language of the Peruvian Amazon, the domain of nasalisation is a sequence of contiguous vowels and glides within a single phonological word. Compare the following:

1. [sṹw̃ɨ̃] ‘neck’
2. [súwɨ] ‘dark’
3. [ỹã́ỹã] ‘rat’
4. [yáya] ‘star’

Bound morphemes may trigger alternations in stems. Compare:

1. [nṹw̃ĩ], morphologically nu=ĩ
2. [ⁿdúβi], morphologically nu=i

Nasalisation itself is a marker of possession:

1. [nã́w̃ɨ̃], ‘his/her foot’, ‘our feet’
2. [ⁿdã́w̃ɨ̃] ‘a foot’

Nasalisation is blocked by non-glides, as in [húkĩ]. Nasalisation can also be realised as an excrescent nasal consonant when the nasal domain is followed by one of /p, t, k, ts, tʃ/.

Nasal consonants show more complex allophony than glides. They are nasals in front of a nasal vowel, prenasalised voiced stops before an oral vowel, and can be plain voiced stops before an oral vowel word initially. The following examples show denasalisation before an oral vowel:

1. /míʃu/ [bíʃu] ~ [ᵐbíʃu]
2. /yamái/ [yaᵐbɛ́i]

Nasal consonants, even when they partially denasalise before an oral vowel, appear to induce nasality to their left. For example /mínau/ is given as [míⁿdou] and not [bíⁿdou], presumably because the nasal /n/ triggers nasalisation of the preceding syllable. The transcription in the grammar is somewhat confused on this point: nasalisation of the vowels is not written when a nasal domain ends in a phonemic or excrescent nasal consonant, but allophony elsewhere suggests that the preceding domain is nasalised even so.

Languages without Blocking Consonants

The typology here suggests a fundamental split between the Aguaruna type, where a consonant normally either blocks and does not assimilate or permits spread and assimilates, and a type where these two properties are separated so that non-assimilating consonants do not block nasal spread. Tucano is of this type, with individual morphemes typically being nasal or oral and not mixed, even if they contain what would be blocker consonants in other languages. Note the non-blocking behaviour of intervocalic unvoiced stops and unvoiced obstruents in the following examples, even though they do not assimilate:

1. nĩtĩ ‘charcoal’
2. mãsã ‘people’
3. sũkũã ‘small of back’
   

Conclusion

This was a simplified overview of the topic, but I think it is interesting because nasal harmony probably has a lower profile than other types of vowel harmony. This may be due to its uneven global distribution, with less documented areas being hot spots, and also because no major world language I can think of has productive nasal harmony.