Stops in English

[HoskhMatriarch]

OK, does anyone know what the difference between the two sets of stops in English is? It's not voicing, because I'm pretty sure I say dogs as [dɑks], but it's still clearly distinct from docks. I think docks might have a preglottalized k or something, I'm not sure.

[Zaarin]

From what I understand, the difference in English is voice onset timing. Though for me, even devoiced dogs and docks would be distinct, because I have /ɔ/ in the first and /ɑ/ in the latter...

[Richard W]

The difference varies across dialects.

For post-vocalic obstruents, possibly the key recognition cue is vowel length. Vowels are longer before 'voiced' stops. I'm English, and for me the traditional short vowels are shorter than the long vowels. I came to the conclusion that beat and bid had about the same vowel length about 40 years ago, and it still seems to hold true for me. (I have had vowel changes since then.)

The final stops are still contrasted by voice in my speech, and I have only one pair of vowels that come close to being distinguished only by length, those of led and laird. Unfortunately, the long vowel is fairly rare (possibly non-existent in my vocabulary) before voiceless stops, and fairly rare before fricatives - scarce is the only example that comes to mind. A possibly source of confusion for introspection is that in some accents the juncture before flexions -s and -ed manifests itself as extra length on the vowel. I think of this as Scottish, and I believe I don't have it in my speech.

There are other differences that normally go with voicing. Peter Ladefoged reported that in Californian English, voiced stops had all the attributes of voicing except voicing itself! What is noticeable in other accents that the onset of voicing in word initial 'voiced' stops is later than in many other languages. For example, to acquire an English accent, Poles have to be taught to start the voicing later than in Polish. (I still wonder whether my informant was training spies! It may just be that his university has very high standards.) The aspiration of word-initial voiceless stops in English is fairly well known. It's not impossible that, like Danish, English is belatedly going down the path of the High German consonant shift.

[Matrix]

For me, dogs and docks are [dɑgz] and [dɑks].

[MadBrain]

Phonemically, the difference between p/t/k and b/d/g is voicing.

Phonologically, this is realized as:
- Voiceless stops (+aspirated accented syllable initially)
- Partially voiced stops.

Unlike fully voiced stops (French etc) but similar to slack voiced stops (Javanese, Shanghaiese), English voiced stops don't require lowering the glottis, so voicing naturally stops halfway through. The result is that in something like 'aba', the 'ab' transition is voiced, 'b' itself is devoiced, and then the 'ba' transition is voiced. You also tend to still have the other side effects of voiced stops, such as lowered pitch.

English b/d/g have the intention of being voiced stops, even when they're not that much voiced in actual realization.

[Travis B.]

This is how thing work in my dialect, and which applies in many other English dialects, but cannot be stated to be a rule across the board.

Fortis plosives in initial position and at the start of stressed syllables are voiceless and aspirated; all other plosives are not aspirated.

Fortis plosives at the start of unstressed syllables and fortis plosives in onsets after sibilants are voiceless and unspirated.

Lenis plosives in syllable onsets are not aspirated, and are in free variation between being voiced and voiceless, particularly in initial position in everyday speech. (In careful speech they are more likely to actually be voiced; particularly on the phone I also catch myself also prenasalizing or realizing as implosive initial /b/.)

Intervocalic fortis plosives are voiceless and unspirated, except for intervocalic /t/, which is usually but not always voiced.

Intervocalic lenis plosives are in free variation between being voiced and voiceless unspirated, except for intervocalic /d/, which is always voiced. (In careful speech they are more likely to actually be voiced.)

Fortis plosives in postvocalic position or after a nasal, /r/, or /l/ are normally preglottalized, with /t/ in particular more frequently than not in everyday speech becoming [ʔ].

Vowels are short (regardless of whether they are traditionally considered long or short) if there is a fortis obstruent not preceded in the same cluster by a lenis obstruent before the next vowel, across word boundaries, and all other vowels are long. This includes diphthongs, which do not have any difference in their length from monophthongs of the same formal length, except for diphthongs and triphthongs formed by l-vocalization, which are longer than monophthongs and diphthongs without l-vocalization of the same formal length.

For me personally, all final consonant clusters not followed in turn by vowels in the following word, and many consonant clusters, are devoiced, but this does not apply to all people.

[finlay]

The idea of a phoneme, while still useful for various contexts, apparently starts to break down when you look at stuff like this - essentially you're right, the distinctions at the beginning of the syllable and the end are different, as it's purely VOT/aspiration at the beginning and there's an element of glottalization at the end. My phonetics teacher basically used to be a non-believer in phonemes and cited the overall prevalence of syllabic writing systems as a kind of anecdotal evidence of that.

And yeah there are further complications too - I'm pretty sure I lengthen the vowel before past tense -ed and not a regular -d, as Richard said, which is one of the holders-on among the scottish features of my accent (one of the ones that people wouldn't notice easily). The classic example is brood / brewed, which I think are clearly different.

[Travis B.]

Trying to eliminate phonemic fortisness versus lenisness would make things much more complex in my variety at least.

For starters, it would require the introduction of vowel length and glottalization in the underlying form. But at the same time it would require extra rules to explain variation in vowel length at morpheme and word boundaries (in that vowel length there depends upon what follows).

Furthermore, eliminating "silent" phonemes (i.e. elided intervocalic /t d n nt nd b v ð/), at least in my dialect, would require introducing not two but three degrees of vowel length as well as pitch accent, but this has the problem that the elided phonemes are in free variation with their realized counterparts (and the frequency of elision varies wildly from individual to individual), meaning that attempting to do so would require introducing extensive allomorphy and indicating that this is really phonological in nature rather than being part of the underlying forms.

At the same time, it would not eliminate a VOT distinction in the underlying form, but make it more complicated in that it would require the introduction of three overlapping levels of VOT, voiceless aspirated, voiceless unaspirated, and sometimes-voiceless-unaspirated-and-some-times-voiced.

In essence, one can only make things significantly more complicated by trying to move rules from phonology into underlying forms, only to likely find that such is effectively impossible, with one having to introduce new phonological rules to compensate.