This article includes a list of general references, but it remains largely unverified because it lacks sufficient corresponding inline citations. (April 2009) (Learn how and when to remove this template message)
Final-obstruent devoicing or terminal devoicing is a systematic phonological process occurring in languages such as Catalan, German, Dutch, Breton, Russian, Polish, Turkish, and Wolof. In such languages, voiced obstruents become voiceless before voiceless consonants and in pausa. The process can be written as *C[+voice] > C[-voice]/__#.
Most modern continental West Germanic languages developed final devoicing, the earliest evidence appearing in Old Dutch around the 9th or 10th century. However, Yiddish notably does not alter final voiced sounds; this appears to be a later reversal.
Of the North Germanic languages, Norwegian, Swedish and Danish (the latter of which has no voiced obstruents) do not have final devoicing. As in Danish, Icelandic stops are voiceless, but it has voiced fricatives which may also occur word-finally.
Gothic (an East Germanic language) also developed final devoicing independently.
Romanian does not have it. Other Romance languages such as French and Italian rarely have words with final voiced consonants for different reasons in their phonological histories, but borrowings from English into French or Italian that have a voiced final consonant (such as weekend) are not devoiced. Portuguese merges [s] and [z] in word-final position (nós and noz are homophones) but has a few words ending with voiced stops like sob (although some dialects feature an epenthetic vowel after the final consonant).
Note: Hungarian, which lies geographically between Germanic- and Slavic- speaking areas, does not have it. Terminal devoicing is indicated in the orthography in Turkish, but it isn't in Azeri. For example, the personal name M?hm?d is pronounced [mæh'mæt] in Azeri, with a final [t], even though it is spelled with a final ⟨d⟩. Meanwhile, the Turkish version of this name is also pronounced with a final [t], but is spelled with a more phonetically accurate Mehmet.
In Dutch and Afrikaans, terminal devoicing results in homophones such as hard 'hard' and hart 'heart' as well as differences in consonant sounds between the singular and plural forms of nouns, for example golf-golven (Dutch) and golf-golwe (Afrikaans) for 'wave-waves'.
The history of the devoicing phenomenon within the West Germanic languages is not entirely clear, but the discovery of a runic inscription from the early fifth century suggests that this terminal devoicing originated in Frankish. Of the old West Germanic languages, Old Dutch, a descendant of Frankish, is the earliest to show any kind of devoicing, and final devoicing also occurred in Frankish-influenced Old French.
English does not have phonological final-obstruent devoicing of the type that neutralizes phonemic contrasts; thus pairs like bad and bat are distinct in all major accents of English. Nevertheless, voiced obstruents are devoiced to some extent in final position in English, especially when phrase-final or when followed by a voiceless consonant (for example, bad cat [bæd? k?æt]).
Old English had final devoicing of /v/, although the spelling did not distinguish [f] and [v]. It can be inferred from the modern pronunciation of half with a voiceless /f/, from an originally voiced fricative [?] in Proto-Germanic *halbaz (preserved in German halb and Gothic halba). There was also final devoicing of [?] to [x] finally, evidenced by spellings like burh alongside burg.
Final-obstruents devoicing occurs in the varieties from Northern Germany. The German contrast between homorganic obstruents is more properly described as a fortis and lenis opposition than an opposition of voiceless and voiced sounds. Therefore, the term devoicing may be misleading, since voice is only an optional feature of German lenis obstruents. By contrast, the German term for the phenomenon, Auslautverhärtung ("final-sound hardening"), refers to fortition rather than devoicing. However, the German phenomenon is similar to the final devoicing in other languages in that the opposition between two different kinds of obstruents disappears at the ends of words, making homophones of such pairs as Rad ("wheel") and Rat ("council, counsel"), both pronounced [?a:t]. The German varieties of the north, and many pronunciations of Standard German, optionally involve voice in the distinction between fortis and lenis obstruents however.
Some examples from Northern German include:
|Bad [ba:t]||bath||Bäder ['b?:d?]||red! [?e:t]||talk!||reden ['?e:dn?]|
|Maus [mas]||mouse||Mäuse ['mz?]||lies! [li:s]||read!||lesen ['le:zn?]|
|Raub [?ap]||robbery||Raube ['?ab?]||reib! [?ap]||rub!||reiben ['?abn?]|
|Zug [t?su:k]||train||Züge ['t?sy:]||sag! [za:k]||say!||sagen ['za:?n?]|
|Fünf [ff]||five||Fünfen ['fvn?]|
Final-obstruent devoicing can lead to the neutralization of phonemic contrasts in certain environments. For example, Russian ('demon', phonemically /b?es/) and ('without', phonemically /b?ez/) are pronounced identically in isolation as [b?es].
The presence of this process in Russian is also the source of the seemingly variant transliterations of Russian names into -off (Russian: -), especially by the French, as well as older English transcriptions.
In compounds, the behaviour varies between languages: