More Saltations

A supplemental page for the article

Bruce Hayes and James White (2015) Saltation and the P-map.  Phonology 32:267-302.

Caution: these are just casual notes, meant for possible future follow-up. If you want to really know if these are saltations, or how they work, you'll have to do some research yourself.


Austronesian, Taiwan.

Source:  UCLA M.A. thesis by Jennifer Kuo

In atonic position (stress is penultimate), /e/ regularly reduces to [u], even though [i] is closer to [e] than [u] is and is legal in stressless syllables. For example, from /pemex/ ‘hold’ and /qunedis/ ‘lengthen’ we have these derivations:

/pemex/    /pemex-an/            /qunedis/         /qunedis-an/              UR
ˈpemex    peˈmexan                quˈnedis           quneˈdisan                Penultimate Stress
ˈpemux]   [puˈmexan]            [quˈnedis]        [qunuˈdisan]             Vowel Reduction

As in several examples in the Hayes/White paper, these saltations have a diachronic story to tell:  modern [e] descends from schwa. Schwa is no closer to [i] than it is to [u], so the ancestor stage was arguably non-saltatory. But later, schwa fronted to [e], creating the saltation. In the taxonomy of Hayes and White (2015), this is a case of "Flanking".

Child phonology

Source:  Stemberger, J.P. (1993). Rule ordering in child phonology. In M. Eid & G. Iverson (Eds.), Principles and prediction: The analysis of natural language (pp. 305-326). Amsterdam: John Benjamins

Final  /g/, /ng/ (angma) are realized as t, n.  The [t] is due to regular Final Devoicing.

But final  /k/ is realized faithfully as [k].  Hence /g/ saltates over /k/ to arrive as [t].

Obviously not diachrony; though in child phonology the existence of rampant near-neutralization makes any claim made without measurement vulnerable to doubt.

Catalan final devoicing + affrication

Catalan (Romance, Catalonia) is a language with a quite general process of Final Devoicing, applying to all final underlying obstruents. However, Catalan Final Devoicing has an interesting wrinkle: the voiced palato-alveolar fricative [ʒ] of course “ought” to become [ʃ] in final position, but instead it turns into [tʃ]. This is despite the fact that [ʃ] is perfectly well-formed in final position. The crucial comparison can be seen  below:

[bɔʒ-ə]      ‘crazy-fem.’    /bɔʒ/ -> [bɔtʃ]   ‘crazy-masc.’
[fluʃ-ə]     ‘loose-fem.’    /fluʃ/ -> [fluʃ]    ‘loose-masc.’

A very helpful source on this pattern is Eulàlia Bonet and Maria-Rosa Lloret’s paper “Fricative–affricate alternations in Catalan”, Probus 2018; 30(2): 215–249. They give the history behind this saltation: the reason [ʒ] in [bɔʒ-ə] presently alternates with [ʃ] word-finally is that historically, it was [bɔdʒ-ə]. Thus, long ago, this alternation was simply Final Devoicing, as with all the other Catalan obstruent pairs. But later on, a process of Intervocalic Lenition converted [bɔdʒ-ə] to [bɔʒ-ə], making the [ʒ] ~ [tʃ] alternation saltatory. Later on, additional changes reintroduced intervocalic [dʒ] into Catalan, making it hard to set up a synchronic analysis that recapitulates history. (Note, however, that this is exactly what Bonet and Lloret do; they resort to abstract UR’s, /d+ʒ/ vs. /dʒ/, to make it work.).

Overall, the Catalan case would be classified as an instance of “Flanking” in the Hayes and White (2015) paper: historical [dʒ], alternating with [tʃ], “moved over” to [ʒ], thus flanking the phoneme [ʃ]. That is, Catalan supports the view taken by Hayes and White that saltations have unnatural, diachronic origins.

Is this pattern productive?  Or is there any evidence that it is being “repaired” in Catalan? A intriguing observation that bears on this point is made by Bonet and Lloret in a footnote:  “Although in native words intervocalic [ʒ] alternates with word-final [t͡ʃ], a few examples from loans can be found with intervocalic [ʒ] alternating with word-final [ʃ], as in beix [bɛ́ʃ] ‘beige.SG’, [bɛ́ʒus] ‘beige.PL’.” It seems plausible to suggest that such cases are indeed grammar-repair. While native forms preserve the old, unnatural alternation ([ʒ] ~ [tʒ]), the novel forms that are assigned a novel alternation pattern ([ʒ] ~ [ʃ]) which avoids saltation.

Curious readers are also referred to Max Wheeler's book (2005) The Phonology of Catalan.

Ilokano Reduplication

Source:  Bruce Hayes and May Abad (1989) "Reduplication and syllabification in Ilokano" (1989) Lingua 77, 331-374.

In reduplication, glides can reduplicate as long vowels to fill a heavy-syllable template, as in /RED-bwaja/ --> [bu:bwaja].  But simple long vowels cannot thus lengthen:  /RED-pusa/ --> [puspusa], not *[pu:pusa]. Thus /w/ can become [u:], but intermediate [u] cannot.

Like a number of the examples in the Hayes/White paper, this has a probable historical source:  [bwaja] used to be [buhaja], where the phonotactic impossibility of *[buh-buhaja] could have given rise to [bu:buhaja], just as in Modern Ilokano /RED-da?it/ becomes [da:da?it] (*[da?da?it]).  This is a workable scenario because onset clusters like [bw] did not occur in earlier Ilokano; they generally from historical sources like [buhaja].

Ewe Tone Raising

Source:  George N. Clements (1978) Tone and syntax in Ewe.  In Donna Jo Napoli (ed.) Elements of Tone, Stress, and Intonation. Washington, D.C.: Georgetown University Press.

Tone sequences of the form H M H  (High - Mid - High) surface (non-finally) as R R R (R is "Raised", a super-high tone).  But we are given to believe that H H H is stable.

Spirantization in Kuot

An email kindly sent to me by Connor Mayer:

Hey Bruce,

I was looking at this language isolate Kuot spoken in Papua New Guinea for the Uyghur intonation project I'm working on ... and I noticed that it looks to have some saltation.


According to Wikipedia:

When vowel-initial suffixes are added to stems that end in voiceless consonants, those consonants become voiced. For example:

/obareit-oŋ/ [oba'reidoŋ] he splits it

/taɸ-o/ [ta'ßo] he drinks

/marik-oŋ/ [ma'rigo ŋ] he prays

The phoneme /p/ becomes [ß], not [b].

/sip-oŋ/ ['sißɔŋ] it comes out

/irap-a/ [i'raßa] her eyes

They also have intervocalic [b] (although there's no data showing stem-final /b/ in the Wikipedia article):

nebam-tuaŋ my feather

It seems pretty similar to the Campidanian examples, but I didn't notice it in any of your or James White's stuff so I thought it might be another interesting case to add to the repertoire.