Devoicing of voiced consonants

Voice and voicelessness in English consonants can be a complicated topic. Describing the phonology of English has never been an objective of this blog, which is why in today's post I describe the basics which can be confusing to Polish students of English in particular and may also be similarly difficult to speakers of other languages where the devoicing of consonants works in a comparable way, e.g. other Slavic languages, German or Finnish.

English has a few pairs of voiced-voiceless consonants, namely /b/-/p/, /d/-/t/, /g/-/k/, /v/-/f/, /z/-/s/, /ð/-/θ/, /ʒ/-/ʃ/, /dʒ/-/tʃ/. All of these are separate phonemes, distinct categories in the English speaker's mind. In languages such as Polish in some environments (around voiceless consonants, at the end of a word) a voiced consonant clearly becomes a voiceless one: the word mewa 'seagull' has /v/ represented by the letter <w>, but the Genitive plural mew before a pause contains /f/, as does mewka, a diminutive form (a small or young seagull) with the sequence /fk/ represented by <wk>. The same could be said about /b/ in żaba 'frog' and /p/ in żab 'frog-Gen-pl' and żabka 'little frog'.

Now when we look at the English words bird and frog, we will see that they are not spelt phonemically the same as Bert and frock respectively. Even though the /d/ and /g/ are devoiced at the end of a word before a pause or a consonant (i.e. excluding the bird is, the frogs where /d/ and /g/ clearly remain), these are not the same as /t/ or /k/. What is the difference then and why do English speakers distinguish between these consonants?

The difference lies in two things: vowel length and fortis-lenis. Before a voiced consonant, even when devoiced, the vowel is a fraction of a second longer: a short vowel becomes relatively longer (but we do not use the length mark [ː] to indicate that) and a long vowel or diphthong becomes even slightly longer than it would be when followed by a voiceless sound. The fortis-lenis distinction is just a Latin way of naming the force or loudness with which consonants are produced. A voiceless consonant like /p/, /s/ or /f/ is pronounced somewhat more loudly than a devoiced version (allophone) of /b/, /z/ or /v/. 

In the 20th century linguists described devoicing in English as partial. These days, however, they tend to write about devoicing, though bearing in mind that a devoiced consonant does not sound identical to a voiceless one.

Devoicing matters in learning about the pronunciation of the Past Simple or past participle <-ed> as well as the Present Simple, plural or possessive (Saxon genitive) <-s>, <-es> and <-'s>. When a textbook mentions that /d/ or /z/ should be pronounced after a voiced consonant, semi-vowel or vowel, that does not mean no devoicing can occur - that depends on the environment, the surrounding sounds or pauses. The vowel length and fortis-lenis differences apply regardless.

As time goes by and pronunciation changes some new distinctions can appear and others can be lost. One good example is /ʍ/, the voiceless equivalent of /w/. Nowadays heard in the speech of some speakers of regional dialects in the UK and the USA, it was the recommended standard pronunciation of <wh> in words like what, when, where, whale, overwhelm and whence. These days the distinction is omitted e.g. in RP-based pronunciation textbooks and words like whales and Wales are treated as homophones.

The final remark is that what I know about the issue is based on studying British RP and occasionally learning about the American GA accent. Other rules may apply in Englishes spoken by native speakers, bilinguals or non-natives around the world.

I deliberately used no reference sources to create this post. Let me know in the comments if anything requires correction. I may be a linguist but I do not consider myself an expert on the theoretical side of phonology.