Relationship of voiceless–voiced and VOT

[freknu]

What is the relation between the voiceless–voiced contrast and VOT? I recall having seem in somewhere in a discussion before, but I can't remember exactly where.

Further studying the behaviour of stops in the spirant rule — and why I perceive plain voiceless and devoiced voiced as different — I'm inclined to analyse the devoiced stops as aspirated, which seems to be a simple solution.

ps → fs~ps
bs → pʰθ̠
ts → sː~ts
ds → tʰθ̠
ks → xs~ks
gs → kʰθ̠

The spirant rule necessarily applies to all environments of stop before voiceless, but it's usually before s or t. The s-pattern is the one I've focused on, since t might behave differently (it has no spirant allophone) — at least dt [tʰː] (vs. tt [tː]) seems to follow the pattern, but bt and gt are uncertain.

[Daniel]

In feature theory, there are two features:
[±voiced]
[±aspirated]
(sometimes by other names depending on which implementation of the theory you're looking at)

This gives us:
ph: [-voiced,+aspirated]
p: [-voiced,-aspirated]
b: [+voiced,-aspirated]
bh: [+voiced,+aspirated]

However, it turns out that [bh] is really something else, a special kind of voicing (breathy voice) associated with the release of the stop. So while that's an interesting topic, it isn't relevant to VOT really.

For VOT, then, the question is simply when the normal ("modal") voicing begins-- before, at or after the release of the stop? VOT of course can be measured to the millisecond, but in general I think those three distinctions are all that really matters in languages (taken as approximations).

b: negative VOT; voicing occurs during the closure; "prevoiced"
p: 0 VOT; voicing begins immediately
ph: positive VOT; voicing begins after the release, with a pause of aspiration

Different languages divide this space differently:
--English has <b> [p] and <p> [ph] technically (both are unvoiced, one is aspirated), where b has a VOT of 0, and p has a positive VOT.
--Spanish has <b>* [b] and <p> [p] (both unaspirated, one is voiced throughout), where b has a negative VOT and p has a VOT of 0.
(Spanish p = English b)
Conclusion: "voiced" and "voiceless" are often applied loosely to different languages, where there is just a binary contrast.
--Thai has a three-way contrast: b, p, ph.
--Hindi has a four-way contrast: p, ph, b, bh

[*I'm ignoring spirantization for Spanish where <b> isn't always a stop.]


Note: all numbers for VOT in this post are approximated/idealized. I don't know that a VOT of exactly 0 actually exists, for example. I think something like ±20ms is about normal for "0", and maybe ±80ms for aspiration. But it varies by language, and I might be remembering that incorrectly.

[freknu]

Hmm... so would you agree that p–pʰ would be a simpler explanation than p–b̥?

Normally aspiration wouldn't affect the perception of plosives, but in spirantisation it could be the reason — either an inherent feature or something innovated due to orthographic representation — why e.g. voiceless ks and devoiced gs seem to be different.

[Daniel]

I'm inclined to analyse the devoiced stops as aspirated, which seems to be a simple solution.

ps → fs~ps
bs → pʰθ̠
ts → sː~ts
ds → tʰθ̠
ks → xs~ks
gs → kʰθ̠


...

Hmm... so would you agree that p–pʰ would be a simpler explanation than p–b̥?

I'm not sure. It's very hard to know whether a stop preceding a fricative is aspirated or not!
Perhaps that could even explain what's going on, that it's assimilating. Or, more than that, perhaps these are no longer separate segments at all but rather affricates.

I wouldn't go with the "simpler" explanation, but whatever is better.

In the end, remember that "P" and "B" are used for a lot of languages, but as I noted above actually relate to different phonetic properties. Those phonetic properties can be determined by analyzing recordings. That "analysis" isn't even really an analysis, certainly not arbitrary. But then how we analyze it systemtically (such as with feature theory) may be.

Normally aspiration wouldn't affect the perception of plosives

It depends on the language. In a language like English voicing and aspiration both are equally predictive, so we don't know whether speakers are learning it based on "voicing" or based on "aspiration" or perhaps more directly on "VOT". The results are the same. In a language like Thai or Hindi, it is in fact very important.

but in spirantisation it could be the reason

either an inherent feature or something innovated due to orthographic representation

I don't follow. Why would orthography change aspiration? You mean because speakers want sounds that are written differently to also be pronounced differently? That's not a common kind of effect, though I suppose anything is possible (if unlikely!).

why e.g. voiceless ks and devoiced gs seem to be different.

I'd look into the research on this for German, Catalan, Slavic, etc. There must be some information out there about whether these sounds are neutralized or still phonetically contrastive.

[freknu]

I'm not sure. It's very hard to know whether a stop preceding a fricative is aspirated or not!

Perhaps that could even explain what's going on, that it's assimilating. Or, more than that, perhaps these are no longer separate segments at all but rather affricates.

That's an interesting idea, I didn't think of affricatives.

I wouldn't go with the "simpler" explanation, but whatever is better.

What I meant was not introducing unnecessary complexity. Using p b̥ b would necessarily require an explanation for a three-way voicing contrast, while p pʰ b only requires a two-way contrast, with aspiration possibly being due to assimilation as you mentioned.

I don't follow. Why would orthography change aspiration? You mean because speakers want sounds that are written differently to also be pronounced differently? That's not a common kind of effect, though I suppose anything is possible (if unlikely!).

That was what I was thinking, but if it's that unlikely I'm probably going astray — perhaps even more so since the dialects haven't even had any standardised orthographies.

why e.g. voiceless ks and devoiced gs seem to be different.

I'd look into the research on this for German, Catalan, Slavic, etc. There must be some information out there about whether these sounds are neutralized or still phonetically contrastive.

German, Catalan, Slavic, gotcha.

[freknu]

What about English backs /bæks/ versus bags /bægz/? Would it be more orthodox — and at this point avoiding unnecessary controversy — to analyse it as s–z contrast, even though all other consonants usually devoice next to s?

[ʋɛcçːɛ] ("the fold") versus [ʋɛɟʝːɛ] ("the wall"), as well as [tɕɵʉr] ("bull") versus [dʑɵʉr] ("animal") does have s–z, so perhaps it would be more accurate to describe plosive + {s} as a s–z pattern as well (although it's actually a spirant and not a sibilant)? Even though I personally cannot hear it and find z very difficult, always pronouncing zoo as [sʊː] and so on...

leitst [lɜits(t)] "is searched (mid.)" — leidst [lɜidθ̠(t) -ð̠(t) -z(t)] "is lead (mid.)"
leikst [lɜiks(t)] "is played (mid.)" — leigst [lɜigθ̠(t) -ð̠(t) -z(t)] "is hired (mid.)"

I cannot come up with a verb pair for p–b, but I can use substantives instead. NOTE: the genitive forms is normally -as, but the -a- is usually syncopated in compounds.

stopps [stɔps] "a stop's" — stobbs [stɔbθ̠ -ð̠ -z] "a stump's"

[Daniel]

What I meant was not introducing unnecessary complexity. Using p b̥ b would necessarily require an explanation for a three-way voicing contrast, while p pʰ b only requires a two-way contrast, with aspiration possibly being due to assimilation as you mentioned.

I think you're mixing up phonetics and phonology there. Phonetically, you may very well have a three-way contrast (or, really, even infinite variation). But in terms of phonemes, you would just have the original binary contrast. It just happens that the same binary contrast is realized differently in certain environments. That doesn't introduce new contrasts to the system; it just represents them differently in certain environments.

What about English backs /bæks/ versus bags /bægz/? Would it be more orthodox — and at this point avoiding unnecessary controversy — to analyse it as s–z contrast, even though all other consonants usually devoice next to s?

What do you mean about devoicing in English?

[ʋɛcçːɛ] ("the fold") versus [ʋɛɟʝːɛ] ("the wall"), as well as [tɕɵʉr] ("bull") versus [dʑɵʉr] ("animal") does have s–z, so perhaps it would be more accurate to describe plosive + {s} as a s–z pattern as well (although it's actually a spirant and not a sibilant)? Even though I personally cannot hear it and find z very difficult, always pronouncing zoo as [sʊː] and so on...

Not sure. Still looks like affricates to me. Coincidentally you have fricatives with the same place of articulation as a preceding stop? They're always the same place of articulation?

[freknu]

The affricatives are (as far as I know, always the same place of articulation):

labial — spɸ pɸ bβ
coronal — stɕ tɕ dʑ
dorsal — scç cç ɟʝ

(EDIT)

I've made quite a few topics on this, and I keep coming back to it, so it should be evident that it's something that bugs me! I just can't put my finger on it ...

So near, yet so far away.

[MalFet]

Normally aspiration wouldn't affect the perception of plosives, but in spirantisation it could be the reason — either an inherent feature or something innovated due to orthographic representation — why e.g. voiceless ks and devoiced gs seem to be different.

I'm late to the party and missing some of the earlier pieces I think, but the direction the analysis is going here strikes me as a bit fishy. Just as one technical comment, aspiration can most definitely affect the perception of plosives (even and especially in languages like English where it doesn't directly map a phonological contrast), and the kinds of tight phonetic analysis that you're interested in doing usually need to hew quite closely to diachronic processes of change in addition to synchronic phonotactics.

In other words, what are you ultimately trying to explain here? VOT has somewhat fallen out of favor over the last few years, mostly because it often doesn't map active feature contrasts particularly well. That's unfortunate, since VOT is usually relatively easy to measure and observe. Most everything else requires a bit more instrumentation. If your goal is to move your analysis into the acoustic exponents of phonotactic processes (as it seems to be, correct me if I'm wrong!), you're well past the point where native speaker intuitions are going to help you. It might be time to bust out the high fidelity recordings and Praat, at minimum.

[Daniel]

If your goal is to move your analysis into the acoustic exponents of phonotactic processes (as it seems to be, correct me if I'm wrong!), you're well past the point where native speaker intuitions are going to help you. It might be time to bust out the high fidelity recordings and Praat, at minimum.

I agree here-- freknu, what is the point of this part of the analysis? Are you trying to figure out grammatical structure? (That might already be done.) Or are you trying to describe technical details of phonetic realizations? (If so, check this out in Praat.)