6

Some languages such as Georgian for example have final-obstruent devoicing and voiced /b,d,ɡ/ become devoiced to [pʰ,tʰ,kʰ] before voiceless consonants and before a pause, for example the word "ზოგადად" ("zogadad" "generally") has a word-final /d/ which is devoiced to [tʰ].

does any language have the opposite process?

1
  • 2
    Some people think Hittite did, but the evidence that this change actually happened word-finally, and that the result were actually voiced consonants, are both questionable.
    – Draconis
    Sep 17, 2022 at 18:43

3 Answers 3

7

There are two known cases in synchronic phonology. The first is Lezgian, where plain voiceless stops become voiced word-finally, resulting in (sing~pl) alternations like k’arab ~ k’arabar 'bone(s)', fend ~ fender 'trick', qejd ~ qejder 'remark', where the stems ends in a voiced obstruent, vs. rib ~ riper 'awl', qʷed ~ qʷeter 'partridge', meg ~ meker 'hair'. Aspirates are not affected (markʰʷ ~ markʰʷar 'stack), and ejectives are aspirated (wirtʰ ~ wirt'er 'honey'). You can retrieve the final consonant's laryngeal specification by looking at its realization before a vowel.

The second case, Somali, is less clear. While voicing is (technically) contrastive in stops, there is no root-final contrast, and you only find voiced stops. There is an alternation ilig 'tooth' ~ ilko 'teeth' indicating that at least in that case /k/ becomes [g] finally. However, there do not seem to be any lexical contrasts. This may or may not "prove", in a given theory, that there is final voicing – such a rule may be necessary to express the generalization that all final stops are voiced.

There are ample examples of sentential sandhi where underlyingly voiceless consonants assimilate to a following segment, which can result in voiceless consonants becoming voiced before a voiced segment. The above cases are different in that there is no following segment, voiced or otherwise.

2
  • 2
    Is it not rather unconventional to mark aspiration before labialisation in aspirated labiovelars? Or is it conventional for Lezgian? I notice Wikipedia writes it the normal way, with labialisation first. Sep 17, 2022 at 22:40
  • Isn't it a morphonological, not phonetic ('synchronic phonological') rule in Lezgian? As far as I know, this 'voicing pattern' is diachronically explained through intervocal devoicing, so at the end of the day, it isn't a true final voicing (see Yu 2004).
    – Aer
    Oct 3, 2022 at 14:24
5

On an empirical level, phonemic voicing of word-final obstruent consonants appears to exist in some languages (in Sanskrit and in varieties of Polish spoken in some regions such as Poznań-Kraków; for sibilants but not other obstruents in a variety of Catalan); despite this, there has been theoretical controversy about whether phonological processes of final consonant voicing should actually exist.

Hock 1975 says:

"In generative phonology, it is a generally accepted doctrine that, since word-final devoicing (WFD) is a very common and natural phenomenon, the obverse phenomenon, namely word-final voicing, should not be found in natural language.

(page 219)

Hock then discusses some data that is problematic for that doctrine, and proposes some analyses.

Voicing neutralization in general: an overview

There are a few kinds of voicing neutralization processes that we can see in languages. (Or rather than voicing, we can use the term "laryngeal contrast"). Wetzels and Mascaró provide an overview. I'm going to try to give a brief summary to provide context before addressing your question.

It's common for languages to neutralize voicing contrasts in sequences of obstruents followed by other obstruents. In many languages, such as Dutch, Yiddish, and I think the majority of Slavic languages, we see regressive assimilation in obstruent clusters that results in voicing before a voiced obstruent and devoicing before a voiceless obstruent.

Depending on the language, voicing neutralization of obstruents may also occur in some contexts other than clusters of multiple obstruents. Many languages neutralize voicing contrasts for word-final obstruents—even when these are followed without any intervening pause by a word starting with a vowel or sonorant consonant.

Russian provides an example of a devoicing process that affects word-final consonants: Padgett states that "final devoicing occurs at the end of a phonological word", specifying that "one or more prepositions plus the following major category item constitutes a phonological word". That word-final obstruents in Russian devoice even when followed across a word boundary by a sonorant is illustrated by the example "/otkaz lenɨ/ [otkas] [lenɨ] 'Lena's refusal'" (page 3).

However, there are other languages that do not neutralize voicing contrasts on word-final consonants in this context: for example, Hungarian and Yiddish retain a voicing contrast in word-final position except for when the following word starts with an obstruent consonant, in which case neutralization occurs with regressive assimilation in the resulting cluster of obstruents.

There are Polish varieties with word-final voicing

While standard Polish is described as having word-final devoicing, as in Russian, there are apparently varieties of Polish where neutralized word-final obstruents are instead voiced before vowels or sonorants. (As far as I know, all versions of Polish show regressive assimilation of voicedness and voicelessness in obstruent clusters, whether word-internal or across word boundaries.) Unfortunately, I forget in which paper I first encountered this fact, but it seems to be described by Wojtkowiak and Schwartz, who give the following example:

while the Standard Polish realisation of sok morelowy ‘apricot juice’ would be [sɔk mɔrɛlɔvɨ], Poznań-Kraków Polish would opt for [sɔɡ mɔrɛlɔvɨ].

(page 125)

Sanskrit sandhi voices stops word-finally

In Sanskrit, underlying voiceless stops are replaced word-finally by their voiced equivalents. This is seen before words starting with vowels. (Before a word starting with a voiceless obstruent, a word-final stop may be devoiced by regressive assimilation of voicelessness; before a word starting with a sonorant consonant, further assimilations may apply).

Some Romance varieties voice word-final sibilant consonants

Bermúdez-Otero (2001) reports on some complex data regarding word-final voicing neutralization in some variety of Catalan (I know Catalan has different varieties that can have different processes of consonant assimilation, but I don't know the details). Bermúdez-Otero gives the example of "llop amic [ʎo.pə.mik] 'friendly wolf'" with devoicing; vs. the underlying voiced consonant [β] seen in "lloba [ʎo.βə] 'she-wolf'" where the consonant is followed by a vowel within the same word (page 18).

This neutralization also applies in Catalan to consonants at the end of prefixes; like the word llop, the prefix sub- has [p] before a vowel, and [b] only before a voiced consonant; Bermúdez-Otero cites "subalpí [su.pəl.pi] ‘subalpine’" (page 23).

However, Catalan sibilant consonants show a different result of neutralization: "In the coda, underlying /ʒ/ surfaces as an affricate whose voicing is determined by the environment: [dʒ] before voiced sounds and [tʃ] elsewhere" (page 12).

Voicing of word-final sibilants at word boundaries in fact seems to be found in some other Romance varieties. French shows it as a historical process: historical word-final s was generally lost in French (some words retain [s]), but in certain contexts where a word historically ending in s is followed by a vowel-initial word, we find as a "liaison" consonant the voiced fricative [z]. A common example is between plural articles and a following noun starting with a vowel: les amis [le.za.mi] or [lɛ.za.mi] "the friends". In contrast, non-sibilant obstruents resulted in voiceless liaison consonants: un grand ami [œ̃.ɡʁɑ̃.ta.mi] "a great friend".

Some varieties of highland Ecuadorian Spanish are reported to have voiceless [s] both word-initially and word-medially, but voiced [z] before a following vowel-initial word and as a possible prepausal realization, resulting in a contrast between [az iðo] has ido “you have gone” and [a siðo] ha sido “s/he, it has been” (examples taken from Bradley and Delforge §2.3, (8)).

Works cited

Bibliography:

  • Wetzels, W. Leo and Mascaró, Joan. "The typology of voicing and devoicing". Language - Volume 77, Number 2, June 2001, pp. 207-244
2

Final obstruent voicing is thought to have happened in early Latin (or its ancestor Proto-Italic). There are inscriptional forms like FECED "made", where the final -D continues the Proto-Indo-European 3sg. ending *-t. There are no certain examples for stops other than the dentals, though, and obviously since we don't have recordings of early Latin or PIE, we can't be totally sure of the phonetics.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.