Linguist Forum
Specializations => Phonetics and Phonology => Topic started by: Venom on July 08, 2015, 11:21:09 AM
-
Hi! I have the following utterance here:
[. . .] to advertise, market and distribute tobacco products to our kids...
/ˈdɪstrɪbjuːt təˈbækəʊ/
Is there any connected speech phenomena applying here regarding the /t/ sound? Because I don't pronounce both /t/, just one of them, but how do i transcribe it then? /ˈdɪstrɪbjuː təˈbækəʊ/ or /ˈdɪstrɪbjuːt əˈbækəʊ/? I know I should maybe write it in one word, but my lecturer don't want us to.
I'm thankful for every help that I get. I'm sorry if my English is not good.
-
In a broad transcription, you would write the phonemes, not the pronunciation. So you would write one t at the end of one word, and another at the beginning of the next: /...t t.../
But in a narrow transcription, you would write the pronunciation. It makes no sense (technically) to write separate words in a pronunciation, especially when there is a phonological merger between the words. So you would write: [...t...] with no spaces between the words.
If your instructor doesn't agree with that, then ask them what you should do. I don't understand their reasoning, EXCEPT that it is a confusing exercise if you write EVERYTHING as one word. So maybe it doesn't matter.
Another option is to write the word-final t as an unreleased stop [t ̚], implying that it blends into the next one. But writing two segments here in general seems inaccurate to me in a narrow transcription.
The simplest answer may be the write it twice. If your instructor does not agree, then you need to ask them what to do, because this isn't a general "answer", just a preference :)
-
Hey djr33, thanks for your answer!
Maybe I haven't expressed myself clearly. I'm writing an assignment for a seminar. So we have to transcribe an excerpt and then apply some of the covered topics to it, such as allophones, -morphs, stress patterns and connected speech phenomena. And I was wondering if I can apply any phenomenon to the above-mentioned issue, such as assimilation. I hope it's clear now :)
Did I get you right that I have to enclose square brackets if I'm referring to pronunciation?
-
// = phonemes, broad transcription, lexical information
[] = phonetics, narrow transcription, specific pronunciation
As for the assignment, we won't do homework for you, but so far your questions are specific so that's ok.
In terms of what process is going on here, there isn't anything specific. You could call it assimilation and merger or something like that.
At a much more technical level, beyond what I assume would be expected (or accepted?) on your assignment, what is going on here is a ban on gemination in English. If you say "hot dog" slowly you hear t then d, but if you say it faster, you just hear one sound. In some languages like Italian, Japanese and many others, you can have sequences of identical sounds next to each other. But in English, you can't, and to avoid that, one disappears. From a historical perspective we'd call that something like "assimilatory loss", but that's not quite right for what you're doing, just looking at interactions within a pronunciation today.
My guess about what you're missing is what your instructor told you to do regarding how narrow the transcription should be. If you are supposed to maintain separate words, then I would guess it's close to a broad phonemic transcription. But again, that just depends on your instructor, so ask them :)
-
It is manifestly unfair to require students to narrowly transcribe an excerpt of English, unless they have been working extensively with a corpus of speech from the same individual. It would be legitimate to ask students to measure durations of objectively-identifiable acoustic events, such as determining the time at which voicing ceases going into the t, but it's impossible to convert that into a meaningful transcription without a corpus of data that establishes the pattern of post-vocalic voicing durations in t versus d versus ɾ. However, life is not always fair.
In the sequence that you are focused on, it would matter what dialect the speaker speaks. In US English, there is a flapping rule that would turn /t/ into [ɾ] between vowels (e.g. "compute a sum", "distribute alimony"), if the stop is (derivationally) syllable-final. In the case of /t#t/, you might think that there is a special rule degeminating double t, but maybe not, since we don't have contrastive lexical geminate consonants. The problem is that there is no process that would definitively create geminates (which then do or do not simplify). We don't have inflectional prefixes or suffixes that potentially result in geminates (like in Arabic, /sit-t/ → [sit:]) -- basically, the question doesn't arise. There are competing phonological ideologies about that situation. One holds that the grammar has to explicitly include statements to describe what doesn't exist in the language, and the other holds that the grammar only includes statements that are justified by contextual changes in pronunciation (such as voicing assimilation and epenthesis in the formation of plurals and past tenses).
Still, you can combine identical consonants between words, so it's a legitimate question whether there is also degemination in that contexts. Speaker intuitions are pretty much useless for resolving this point. There is clearly a mental difference between see Candy and seek Candy (and for a control and calibration, I suggest adding sick Candy and seek Andy), but to know for sure whether in any actual speech rate there is neutralization, you would have to (a) set up an experiment to measure the acoustic properties of these utterances and (b) set up a perception test to see if speakers can distinguish the two utterances. I'm willing to bet money that the utterances remain distinct, thus no degemination.
That all said, in the Estonian dialect of English, you get whopping long geminates in phrasal utterances like "distribute tobacco", so it depends on what dialect you're dealing with.
This BTW is way above your present pay grade.
-
In the case of /t#t/, you might think that there is a special rule degeminating double t, but maybe not, since we don't have contrastive lexical geminate consonants. The problem is that there is no process that would definitively create geminates (which then do or do not simplify). We don't have inflectional prefixes or suffixes that potentially result in geminates (like in Arabic, /sit-t/ → [sit:]) -- basically, the question doesn't arise.
That's incorrect. The situation arises specifically in this situation: where we find t#t at a word boundary.
Whether or not there is a "specific rule" is a matter of a theory. Many theories, however, would provide for either a rule (or equivalent) or a default of some sort. Theories that would handle this include Optimality Theory (where some ordering of constraints would apply to this situation, as they do to all others), or a theory more focused on articulation itself (that, arguably, if English has no geminates, one's tongue cannot move to produce them at a word boundary either). In a more formal rule-based theory, we could say there is a rule or not. But certainly there are rules for things like epenthesis, which arise when there are difficult pronunciations, as this would be, so there is no clear argument that there could not be such a rule, though there might not be, if we take it to be undefined, in which case some sort of default would have to apply, and I don't know what that would be.
That all said, in the Estonian dialect of English, you get whopping long geminates in phrasal utterances like "distribute tobacco", so it depends on what dialect you're dealing with.
Estonian dialect of English? You mean L2 English influenced under the phonology of Estonian? Of course that would be different. I don't really follow the point, more than just giving examples of other languages (like Estonian, or Arabic).
But in general, this just goes to my earlier point: it's fine for an instructor to have you do this kind of exercise as practice, but if there are going to be graded "right answers" the details for this sort of thing should be specific somewhere or not required. I would personally argue the grade if that was not discussed, or make a note/footnote on the assignment. I would suggest focusing on the bigger picture rather than these details for now. Ask your instructor regarding preferences for grades, or ask us or your instructor for deeper questions about the real "right answer" here (although it depends on the expectations of the assignment and particular theory).
-
The point is that there is no degemination in English, although one might be misled into thinking that there is because (a) there aren't any geminates within morphemes and (b) the spelling present t/tt, p/pp which one might misinterpret as indicating that there is degemination. So they only arise in phrasal concatenations, and phonetically they are not as dramatic as in Estonian or Italian.
-
No, I disagree. Whether or not there is a rule of degemination is unknown for the same reasons you suggest that there is not such a rule. There may very well be a rule against having identical consonants in a row. Additionally, ruling out the possibility of lingering historical effects is uncertain to me: why can't it be the case that "illiterate" are "irreplicable" actually have underlying geminates today? Or at least in some common compounds.
Look at the common misinterpretation of "duct tape" as "duck tape". It seems to me that SOMETHING is going on there, whether you want to call it a "rule" or not-- and such "rules" are just descriptions of processes that occur in a language, so there is an argument to be made.
Regardless, again, this is all abstract theorizing and depends on whether our theory has rules or not, which is not a very interesting question either purely abstractly or in this very specific context. If something along the lines of Optimality Theory is right, though, then very clearly there would be "degemination" in English, for obvious reasons. If not, I guess we're left with "it just happens!" ;)
-
Let's first focus on phrasal /Vt#t/, where there is no degemination. This isn't a matter of interpretation, it's a matter of direct observation. If you disagree, then you must have in mind some counter-evidence that shows that /Vt#t/ and /V#t/ neutralize; otherwise, I don't see what your basis for claiming the existence of actual degemination in English is.
In OT, one cannot say that "there is always degemination", since degemination is epiphenomenal. By stipulation in orthodox OT, every language is claimed to have every constraint so whatever constraints "drive" the phenomenon of degemination are always there. But in rule-based theory, the formal objects (features, a syntax of rules) are also "always there", but again degemination is from the theoretical POV epiphenomenal -- degemination is something that can result from assembling a rule of a particular character, so it is potentially there, but it is only actually there if there are facts that force the assembly of such a rule. Just as in OT, where a constraint ranking that results in degemination is there only if there are facts that force such a ranking.
"Why can't" is not a valid argument for a linguistic representation -- you can also ask "why can't the underlying form of 'boy' be /bœ̄/" or "why can [t] derive from /ħ/?". Since there are geminates in English (distribute tobacco) and there is no sufficient reason to claim that there exists a degemination process (however analyzed), we have a reason to reject the claim that there is a geminate in illegal etc. In a compound like "hat tip", the geminate survives; so claiming that there is degemination in English is like claiming that there is a process turning /ħ/ into [t] (or [h], it doesn't matter which).
(As an aside, duck tape was first invested using cotton duck, where "duck" is a kind of canvas. People started to think it was called "duct tape", thinking that it's useful for heating ducts, but it isn't. Duck Tape is probably then a re-creation of the original term in the mid-70's)
-
Let's first focus on phrasal /Vt#t/, where there is no degemination. This isn't a matter of interpretation, it's a matter of direct observation. If you disagree, then you must have in mind some counter-evidence that shows that /Vt#t/ and /V#t/ neutralize; otherwise, I don't see what your basis for claiming the existence of actual degemination in English is.
It depends on speech rate. Certainly I can say "it takes" as two separate words or faster as [Iteks]. I don't see the problem here. The point is that in a language like Italian, Japanese or Finnish, fast speech would not cause this to reduce to a non-geminate, at least not without being very fast speech, beyond what is required in English. If your theory of phonology doesn't explain that, it isn't adequate for the data. Whether there is a 'rule' is just a matter of terminology, but there's something going on, some difference between English and Finnish. That's all I'm saying.
In OT, one cannot say that "there is always degemination", since degemination is epiphenomenal.
Yes, and phrasing aside, OT accounts for degemination when it occurs. I haven't intended to claim anything more specific than that.
Since there are geminates in English (distribute tobacco)
There are lots of dialects of English. In mine (basically General American English), I probably would not produce two stops there at my normal speech rate. It's possible, if I articulate it especially clearly.
The point is that the data in the first place apparently does not have a geminate there. The simplest answer, of course, is that Venom is wrong and there actually is a geminate. But assuming that's not the case (and it sounds entirely plausible to me that there is no geminate there), then the question is how it is explained, nothing more, nothing less.
In a compound like "hat tip", the geminate survives;
Same here.
So I think the issue is simply empirical and related to different dialects. There is nothing to argue about, and nothing wrong with either dialect. But maybe that's why we haven't been on the same page with this?
Regarding duct/k tape, that's very interesting. What a weird mixed etymology!
As for my pronunciation, I am willing to accept that it isn't strictly speaking one segment. It may be an unreleased t at the end of a word, followed by a release at the beginning of the next word. This is phonetically identical (or very close??) to a medial t which is constricted then released as 'one segment'. (Of course one distinction here would be flaps, which I don't believe would occur in my speech across word boundaries except in lexicalized compounds etc. Can't think of a clear word to test that in at the moment, but I think it's true.)
-
I'm amazed linking hasn't been mentioned: /t_t/
-
"Linking"?
-
Em...? It's not like linking is something I made up on my own, as far as I can tell a standard suprasegmental IPA feature, usually written /s‿a/ but also with an underscore for simplicity /s_a/.
Or if you prefer: liaison or sandhi.
-
There are lots of dialects of English. In mine (basically General American English), I probably would not produce two stops there at my normal speech rate. It's possible, if I articulate it especially clearly.
Well, you are in a position to provide the relevant quantitative evidence for distinctness, if it's there. This has been a long-standing problem in phonology, I would say originating with David Stampe, who was a wizard at introspecting minutia of phonetic transcriptions that simply do not bear up under close scientific scrutiny. As for dialects of English, I'm willing to take a position that in General American English, there is no speech-rate dependent degemination i.e. neutralization of VC#CV and V#CV in comparable prosodic contexts. The proper scientific method for addressing the claim is experimentally. You could surprise me and prove that there exists a dialect where there is phrasal degemination, but introspection about neutralization is not a valid experimental method.
-
Freknu, interesting. "Liason" I'm used to in the context of French specifically. But I haven't come across the type of analysis that would apply to English-- that's because I don't tend to read things about phonology, not because it doesn't exist. That may be the best answer to the original question. (However, having taken a few introductory courses for phonology, I don't believe any of them mentioned that, so it might be beyond what is expected for the assignment. I'm still a little puzzled about what exactly is expected, though.)
You could surprise me and prove that there exists a dialect where there is phrasal degemination, but introspection about neutralization is not a valid experimental method.
Introspection is about all that is reasonable to expect for an informal discussion on a forum. And whether or not there may be subtle acoustic cues, my introspection is fairly accurate: there is a minimal distinction, if any, when spoken at a natural or fast pace. I can certainly articulate both clearly, but I often would not. Whether it is identical/neutralized at whatever level of experimental measurement you'd like to use is another question. But even if not, and perhaps then the term "degemination" is technically wrong, it seems to be missing the point of the original question: there's definitely some kind of liason/linking going on there, so what is it?
As for whether my evidence stands up, it's no better or worse than yours. In both cases it's none, except for our observations. And since we can't observe each other, there's little point to arguing over the data. I could record myself, but given my bias that wouldn't prove much. But if you'd like, I can, and maybe it will come across to you as natural.
-
Wow, thanks for the answers. I mentioned the "flap t" at other parts in the text. Our definition was " [. . .] occurs when an intervocalic /t/ is following an accent". I mentioned words like "availability". But "distribute tabacco"? I'm not sure.
I also mentioned linking and intrusive r, but we never spoke about liaison with other letters (or should I say sounds?). So I guess it's too specific for my shortened assignment about 8 pages.
thank you for your help :)
-
I have no problem with the existence of phonetic geminate consonants in English across morpheme boundaries, as in words like unnecessary and misstatement. What we ought to recognize is that there are various tempos and styles that affect pronunciation. In fast and casual speech (hypoarticulation (https://en.wiktionary.org/wiki/hypoarticulation)), degemination is a fairly common phenomenon in English. Hence, I can easily pronounce unnecessary and misstatement with geminate or nongeminate pronunciations, depending on the tempo of my pronunciation. However, that is based on my intuitions. I believe that my claim would be borne out in instrumental acoustic analyses of people pronouncing those words, so that might make an interesting experiment for folks to try. It is fairly easy these days to record pronunciations and generate spectrographic analyses with appropriate software, so I invite others to try it. Just ask folks to read sentences that contain those words and don't mention that you are interested in the pronunciations of those words.
Generally speaking, what causes degmination is the collapse of syllable structure, and that seems to be a general trend with all lenition (weakening) processes that affect articulation. For example, flapping in American English is not just something that happens to intervocalic alveolar consonants. It is also the case that the following syllable be relatively unstressed. The stronger articulation on the preceding consonant tends to attract consonants, and flaps tend to occur in syllable-final positions.
There is a contrary tempo of articulation--so-called "hyperarticulation (https://en.wiktionary.org/wiki/hyperarticulation)"--that favors strengthening processes such as gemination and lengthening.
-
It is fairly easy these days to record pronunciations and generate spectrographic analyses with appropriate software, so I invite others to try it. Just ask folks to read sentences that contain those words and don't mention that you are interested in the pronunciations of those words.
This, alone, would not be sufficient (I'm not picking on the fact that you didn't mention measuring durations). The key is that you have to collect a corpus with potentially minimally-contrasting pairs, such as see Candy and seek Candy. There is no magical duration number that corresponds to "degeminated", so if you find that the stop closure in token #42 of seek Candy is 100 ms, that doesn't show that there is degemination.
Since it is clear and agreed that degemination, if it exists, is speech-rate dependent, that means that speech rate has to be considered, meaning controlled. There is no independent litmus test for "speech rate", and it isn't even established what speech rate is -- it may be a high-level cognitive register choice, or it may be a lower-level continuous articulatory setting – or both. In which case, you need to both generate multiple "speech rates" for each speaker, and encode for said "speech rate".
Focusing on consonant closure durations might seem like an obvious way to test whether /k#k/ and /k#V/ or /V#k/ become acoustically identical, but since no two utterances are ever literally identical, you really also need to include more nuanced information on the acoustic structure of the consonant (burst amplitudes, aspiration durations, so on) and also information on the preceding vowel (and the following, just to not miss anything).
I understand that all this attention to detail is a nuisance, but it's necessary, in order to make a valid scientific claim.
-
It is fairly easy these days to record pronunciations and generate spectrographic analyses with appropriate software, so I invite others to try it. Just ask folks to read sentences that contain those words and don't mention that you are interested in the pronunciations of those words.
This, alone, would not be sufficient (I'm not picking on the fact that you didn't mention measuring durations). The key is that you have to collect a corpus with potentially minimally-contrasting pairs, such as see Candy and seek Candy. There is no magical duration number that corresponds to "degeminated", so if you find that the stop closure in token #42 of seek Candy is 100 ms, that doesn't show that there is degemination.
There are a lot of things I failed to mention, and I have no problem with you designing an experiment in any way you wish. It is obvious that one would need to measure durations, so I didn't feel a need to mention it. What is worth mentioning is that asking people to read lists of minimal pairs would not be the best approach, since that method might obscure the effects of varying speech tempos. Instead, one would want to elicit a corpus of reasonably natural pronunciation. Having people read sentences that contain the interesting phonetic targets is just one method that could be used.
Since it is clear and agreed that degemination, if it exists, is speech-rate dependent, that means that speech rate has to be considered, meaning controlled. There is no independent litmus test for "speech rate", and it isn't even established what speech rate is -- it may be a high-level cognitive register choice, or it may be a lower-level continuous articulatory setting – or both. In which case, you need to both generate multiple "speech rates" for each speaker, and encode for said "speech rate".
Actually, just having people read sentences works just fine. For example, Galina Barinova conducted a series of experiments in Russia in the 1960s and 1970s by just randomly picking people off the street, shoving a microphone in their faces, and asking them to read sentences. She elicited a very rich corpus of data that included many careful and casual articulations of the same words.
Focusing on consonant closure durations might seem like an obvious way to test whether /k#k/ and /k#V/ or /V#k/ become acoustically identical, but since no two utterances are ever literally identical, you really also need to include more nuanced information on the acoustic structure of the consonant (burst amplitudes, aspiration durations, so on) and also information on the preceding vowel (and the following, just to not miss anything).
Geminate pronunciations and degemination abound in English both at word boundaries and internal morpheme boundaries, not to mention with all consonants that can begin and end morphemes. So you can make the experiment as large and detailed as you want, but I suggest keeping it reasonably simple at first. Aspiration doesn't have to be a factor, if one chooses to limit the data to voiced consonants: "real love" vs "reel of", "bad ear" vs. "bad deer", "the song" vs "this song", etc. I think you'll find that lax vowels are particularly good at sucking up geminate pronunciations across word boundaries in running speech.
-
Venom, what dialect are you working with? There's quite a bit of variation, so we should put this discussion in some context. I'm referring (almost entirely) to general American English, just because that's what I speak, but that might be misleading for another dialect.
...But "distribute tabacco"? I'm not sure.
Hmm... I would produce that as one [t] at my normal speech rate. If I was speaking VERY quickly I might flap it, but that might be not because of a flapping rule but just because at that very fast speech rate it would be a short stop anyway, so maybe indistinguishable from a flap. I'm not convinced I'd ever flap across word boundaries except where those words are phonologically merged (as in a case like going to > gonna, but with [t]).
Copernicus, I agree with your post, very good points. I didn't mean to suggest at all (if I did, not sure) that degeminiation is a requirement in English. But it's more likely than in a language like Japanese, Italian or Finnish, so something is going on there [or not going on], but not in the other languages.
I have no problem with the existence of phonetic geminate consonants in English across morpheme boundaries, as in words like unnecessary and misstatement.
Same. Especially "un-necessary" seems ok, and in very careful pronunciation, "mis-statement" although that's a little forced for me. (Dictionary pronunciations would certainly do that.)
As for the last two replies, Panini, I think you're making this too complicated. There is no such thing as a perfect experiment and I have seen plenty of real research with experiments less perfect than the ones you describe. So while all of these are legitimate factors, the modern method to account for all of that is to simply try to balance for all factors you can and then gather enough data that statistics will handle the rest. As you describe, we would have contrasting pairs (or near pairs) and work from there. Still, I don't know that we need to resort to a formal experiment to make this point. Let's see if you agree:
Background: Languages like Japanese, Italian and Finnish have contrastive geminates lexically. English does not.
Hypothesis 1: In English vs. the other languages there would be a difference in roughly length for word-boundary identical-phoneme sequences.
Hypothesis 2: All of these languages would treat these word-boundary-geminate sequences identically, with no effect for length.
(Assume the normal balancing for things like default length of articulation, speech rate, etc.)
I'm pretty confident that no one would disagree with that background. (If they did, it might relate to some of these rare compounds with potential geminates in English, but I don't think anyone actively claims English has geminates in the way the other languages do.)
After that, I also think that most people would agree with Hypothesis 1 rather than Hypothesis 2. There are lots of details to work out, yes, but it's not all that complicated to see that there is a difference, where English is more likely to degeminate, or whatever you want to call it. Do you disagree?
-
I'm basically reminding people that sloppy experimental design yields non-results: i.e. GIGO. There is a basic minimum standard for an acceptable experiment, and I have not suggested anything that goes over that minimum. The background dichotomy between languages with lexical geminates (Finnish) vs. ones without (English) is a false dichotomy -- there is a at the phonological level a quadrichotomy between existence vs. non-existence of lexical geminates, crossed with existence vs. non-existence of phrasal geminates. Some language actively and phonologically purge all geminates (regardless of speech rate).
The question is fairly simple when framed as pertaining to phonological systems, and in English (a) there is nearly no evidence one way or the other whether anything in the grammar eliminates geminates lexically but there do not seem to exist any geminates (owing to historical accident) but (b) there are clearly phonological geminates created at the phrasal level, by concatenation of identical consonants. The complication comes when one wants to make phonetic claims about pronunciations of phonological geminates, as a function of speech rate or whatever. A propos (a), unnecessary and various compounds containing C1+C1 either show that in fact there is no phonological degemination in English, or that putative degemination is limited to certain lower levels (in the Lexical Phonology sense). I'm trying to be open-minded about the claim that there is lexical phonological degemination in English, so I'm not pointing to this as an counterexample to that claim, but such words only make the case for degemination in English harder to sustain.
I don't understand the supposedly competing hypotheses that "English differs in length for C1#C1 from other languages" vs. "English and languages with lexical geminates treat C1#C1 identically, with no effect for length." What does "effect for length" mean? Are you proposing that the only alternative is that English C1#C1 is durationally identical to what is found in languages with lexical geminates? That implies that there is a fundamental identity in the duration of geminates in languages with lexical geminates, which we know is not true.
Since it's not controversial that a phonetic contrast V#CV vs VC1#C1V exists in English at normal speech rates, the only issue remaining is whether under some circumstance there is an actual phonetic neutralization of V#CV vs VC1#C1V.
There is a simple reason why it is less likely that there is phrasal degemination in English compared to languages with lexical geminates. There is a greater potential for neutralization as a result of degemination in English, compared to those other languages. There are many minimal pairs like "see/seek", "see/seal", "see/seat" and so on, where concatenation of one verb vs. the other with an object noun beginning with k, l, t yields distinct utterances. Degemination would render that distinction null, and that is perceptually a bad situation, which is likely to block degemination (obscuring distinct utterances is sub-optimal). The other languages do not face that quandry (or only minimally, in Finnish).
So as you can see, we disagree on the intuition that English is more likely to degeminate. I think maybe Modern Hebrew is more likely to degeminate, but we'd have to determine this empirically.
-
So as you can see, we disagree on the intuition that English is more likely to degeminate.
Hm, ok. You're welcome to disagree. The points made in your post are interesting, but it doesn't have much relevance to my strong intuition that I would pronounce-a-two-consonant-sequence-like-it-was-one (~degeminiate).
Perhaps you have a more technical definition of "degeminate", as a theoretical concept, in mind, as opposed to merely something that has a similar phonetic effect. I have no such concept in mind, just that phonetic effect.
Regarding which languages are likely, as I've said, we'd need to pick a dialect of English first. I can't speak for all of them, but my American English is certainly likely to.
I think maybe Modern Hebrew is more likely to degeminate, but we'd have to determine this empirically.
Why would Modern Hebrew in particular have a very different phonological structure from English? I know that's a weird question, but looking at it at face value, Modern Hebrew is the combination of Ancient Hebrew plus several European languages (some or all of which do not have contrastive gemination). So while historically there are certainly more instances of Hebrew words losing geminates, that in particular isn't relevant as a synchronic process. But you think it is more likely to, at word boundaries, etc., not produce two consonants? I can't comment on Hebrew, although I would expect most research on it to be regarding the loss of gemination historically. So, yes, an experiment would be needed.
Or is this, as I suggested, a more technical issue: because this is more relevant for Hebrew (historically, and in the potential daily speech of its speakers), there is actually a "rule" in Hebrew, while the situation is unknown in English, so there is no "rule"? I would be open to that idea, but I don't think it really changes the data, and it may be conflating diachronic change with synchronic grammar to take that argument too seriously, though I certainly agree it more readily provides data.
There are many minimal pairs like "see/seek", "see/seal", "see/seat" and so on, where concatenation of one verb vs. the other with an object noun beginning with k, l, t yields distinct utterances. Degemination would render that distinction null, and that is perceptually a bad situation, which is likely to block degemination (obscuring distinct utterances is sub-optimal).
That argument does not hold up. Peripheral/sporadic semantic mergers are normal in historical sound change (I can cite a textbook for this, one I've been teaching from, if you wish). One easy example is the pin~pen merger in Southern American English. Usually it doesn't cause too many problems but for a few words it does. It doesn't stop the merger. The change is "blind" to such situations, which I imagine would be especially true when that happens across word boundaries since those words would still, in the vast majority of their usage, be distinct.
So in the end, rather than debating this too much more, let me just ask a question: Where do you think we disagree? I'm not so sure anymore.
-
I am trying to figure out what your belief is. Earlier I thought you were assuming that, as a self-evident matter of fact, English has degemination at the phrasal level (self-evident in the same way that the existence of flapping is self-evident -- to any trained linguist who has familiarity with American English). Now I think that your position is that you personally have such degemination as a fast-speech process (and there may be others like you out there). By "degemination", I mean some process (phonetic or phonological) which "shortens" identical consonant sequences to the point that they become the same as single consonants. The litmus test for degemination as opposed to simple "reduction in duration" would be actual neutralization. I think we've covered the ground regarding how to actually perform the test: I hope that you agree that it would be a rather challenging experiment, but it could be done if you constructed the right kind of stimuli, and plan to make appropriate measures. I fervently hope that you agree that it would be kind of a waste of your time to do this, and that it would be actually harmful to conduct a crappy experiment. (This is the converse of the "teaspoon of sewage" rule, that a teaspoon of sewage in a barrel of wine yields a barrel of sewage. But in linguistics, a teaspoon of experimental dressing-up tends to turn a barrel of sewage into credible results).
Now, you've also advanced the position that most people would agree with Hypothesis 1, and this concerns me somewhat because the discussion has thereby shifted from questions of objective truth to social perception of scientific hypotheses. I presume that "most people would agree" isn't actually relevant to, or evidence for, your claim, which I now think is that "English is more likely to degeminate" (than e.g. Japanese, Finnish, Italian). This is not a scientifically testable claim, it's a pre-theoretical claim about the probability of a particular outcome, were the experiment to be conducted. In the current context, it's an arbitrary claim, because we don't have a logical basis for preferring one outcome vs. the other. So the procedure would be to say why you expect a particular experimental outcome. I've offered a functional explanation for a factor that would tend to make neutralization the unlikely outcome (not impossible, just not likely). I think we have intuition stalemate (intuitions are sacred and not subject to empirical disconfirmation).
Having said that doing the experiment would be a waste of your time, it might actually be of some value to conduct a pilot experiment, by cooking up a script with appropriate minimal contrasts which you read and record, and perform basic duration measurements on (preceding V duration, consonant closure, and release). Let the target be in a long-enough frame that it would be difficult to subconsciously manipulate production, include multiple tokens of a given string, and target two speech rates (normal and "as fast as you can while actually saying the word"). Do the stats. Surely you have nothing better to do this summer (hah!). I've had numerous intuitions about phonetic properties of my speech and have been surprised to look at the measured phonetic reality, so you may develop new intuitions.
Although I didn't lay out the reasoning, my claim that Modern Hebrew is more likely than English to have actual neutralization is based on the (uncontroversial) fact that MH has active synchronic phonological processes avoiding geminates. It is possible that said phonotactic tendency is restricted to word-internal contexts, but since there is no evidence that it is so restricted, it follows that MH is likely (not guaranteed, likely) to have phrasal degemination. I don't know of any actual experiment that has been performed to test this. In contrast, there is no active evidence for active degemination in English, and a certain amount of evidence (in the form of morpheme-concatenations like "unnatural, unnecessary" as well as phrasal concatenations) that there is no degemination (thus, degemination is unlikely).
My argument does crucially rely on the distinction between actively eliminated strings, which directly motivate a thing in the grammar (rule, or package of constraint rankings, depending on your meta-theory), versus an observation that imaginable strings don't exist, such as *blɔrt or *binni. My distinction is motivated, at the theoretical level, by the premise that historical change in a broad sense "explains" all pattern-facts of synchronic language -- basically, regularities exist because historically-speaking, something happened to give rise to these regularities; but not every historical fact has a corresponding object in the grammar of a language, for instance, the phonological relationship between "father" and "paternal" is not part of the grammar of English.
-
Let's forget about morpheme and word boundaries for the moment, because they just confuse the issue. A geminate consonant is just a string of two identical consonants, and degemination is the collapse of that string into a single consonant. The facts in English are that geminate consonants can only exist when there is an intervening word or morpheme boundary, but my (non-generative theoretical) position is that such boundaries have no ontological status in phonology. They merely establish the existence of a syllable boundary, and phonological rules (processes) only apply within prosodic domains, not lexical ones. So there is no such thing as "phrasal degemination" from my perspective. Rather degemination applies within syllable boundaries, so it occurs in running speech when syllable boundaries shift.
So a word like "nonnatural" can be pronounced either with a geminate or nongeminate nasal consonant, depending on speech tempo, which affects syllable structure. In rapid and casual speech, stress will attract the first segment of the geminate, which is the actual trigger condition for degemination. In careful speech, the syllable boundary remains stable, giving rise to a geminate pronunciation. Obligatory degemination is responsible for the fact that simplex English morphemes do not contain geminate consonants, but it is also responsible for the observable phenomenon of degemination in casually articulated English. In other words, degemination is a natural articulatory constraint in English that applies when fed appropriate inputs. Otherwise it has no observable effect on articulation--except, of course, when English speakers try to learn a language like Finnish or Estonian, which has morpheme-internal geminate consonants. Then degemination manifests as part of a typical English accent in those languages.
-
Let's forget about morpheme and word boundaries for the moment, because they just confuse the issue.
Hmph! That's like me saying "Let's forget syllable boundaries because they just confuse the issue". But I'm willing to play along to a limited extent.Rather degemination applies within syllable boundaries, so it occurs in running speech when syllable boundaries shift.
So a word like "nonnatural" can be pronounced either with a geminate or nongeminate nasal consonant, depending on speech tempo, which affects syllable structure.
Under those assumptions, what syllable and segmental structure do you have without the assumed tempo-governed effect on syllable structure, and what syllable structure change is associated with tempo; what is the principle that refers to syllable structure and results in degemination? I want to make sure I understand your analysis.In rapid and casual speech, stress will attract the first segment of the geminate, which is the actual trigger condition for degemination.
Lemme see if I get this right. When a consonant is next to a stressed syllable, and that consonant is the first of a sequence of identical consonants, then that first consonant is resyllabified to the stressed syllable. So in seek candy, you would ordinarily get sijk.ˈkændij, but in casual speech this would resyllabify to sij.ˈkkændij, except that no onset (or coda) can have a geminate, by your hypothesis, therefore kk becomes k. Is that how it works in your theory?
-
I think that you've got the basic idea, panini. Phonological processes play two essential roles in coordinating speech production--enhancing auditory clarity and reducing articulatory complexity. Degemination reduces the effort needed to pronounce a prosodic unit. Stress attracts segmental units from neighboring syllables, and that resyllabification feeds degemination. So you actually observe its effects in casual articulation. Word and morpheme boundaries tend to have the effect of preserving syllable boundaries, especially in normal or careful articulation. So no degemination there. However, if you try to learn a language such as Finnish (or Estonian, which has trigeminate phonemes), natural English degemination interferes with articulation, not to mention perception. You can try to pronounce foreign geminate consonants, but you have to make an effort to suppress degemination in order to succeed. That is, phonology plays a direct functional role in the coordination of the speech tract during speech production.
I suppose that, in the spirit of full disclosure, I should mention that my undergraduate and graduate experiences were heavily influenced by the brilliant phonetician, Ilse Lehiste, who also tried to beat correct Estonian pronunciation into my skull. So I had the experience of trying to learn to pronounce and perceive geminate and trigeminate phonemes. Ilse also taught me the rudiments of instrumental acoustic phonetics, although I did not go on to specialize in it. We had a state-of-the-art phonetics lab at Ohio State back in the 1970s.
-
Copernicus's post 2 above is basically my position on this.
A couple points to address, then I'm happy to drop it:
I am trying to figure out what your belief is.
I was merely defending my original point that whatever phenomenon is going on is that those two phonemes are being produced as one sound, and it is perfectly reasonable to only represent one in a transcription of [my dialect of] English. It was not a "theoretical" claim in any deep sense. And I stand by that: there's just one sound there, at least under some pronunciations. A "dictionary" pronounciation could certainly have two, and some careful pronunciations may also. I'm not making any claims about frequency, but based on the original question, I find that analysis (2 sounds becoming 1) plausible. If not "degemination", then make up whatever word you'd like. But I am at least suspicious that something is going on here because the two consonants are identical. Maybe it's more of a merger of two stop consonants (so that "kt" and "tk" would merge the same way). I might be wrong about that. But I'm confident in the data. I didn't intend a deep theoretical discussion, though I've read some interesting responses so I'm content with that too.
Earlier I thought you were assuming that, as a self-evident matter of fact, English has degemination at the phrasal level (self-evident in the same way that the existence of flapping is self-evident -- to any trained linguist who has familiarity with American English).
The obviousness of a phenomenon is irrelevant. However, yes, to me the elimination of one of those stops in the original phrase is pretty self evidence, just like flapping, EXCEPT that flapping is so hard-wired into my linguist brain now it's hard to think of it as anything but obvious.
Now I think that your position is that you personally have such degemination as a fast-speech process (and there may be others like you out there).
Yes. My intuitions can't speak for others, but I also don't have any intuition that I am unusual in this. It may be that I do it more, because I often speak relatively quickly, but it sounds very natural to me in what I think of as "English". The original post, as I said, seemed completely plausible. I didn't feel the need to question the observation made.
By "degemination", I mean some process (phonetic or phonological) which "shortens" identical consonant sequences to the point that they become the same as single consonants.
Sure. There's a problem with this in that it assumes a category of "consonants" [segments] within the speech representation. That seems a bit narrow to me, given that we're talking about phonetic effects, rather than basic phonology. So maybe flapping is a "real rule", while this degemination (or whatever you want to call it) is just a "phonetic effect". I don't have any claim regarding that distinction. I just think it happens. The problem regarding segmentation vs. articulation is that with articulation we can see whether it's a flap or stop based on physical properties. For "one consonant" I don't know how to test that as physical properties. Certainly the tongue would not touch twice. But that may not be enough to consider it degeminated. As I said, though, I expect English is more likely to shorten it than a geminating language. (No, I don't personally have conclusive experimental results that prove this. It's an internet forum for discussion, not a conference or journal where I'm attempting to publish my ideas and change the field.)
I think we've covered the ground regarding how to actually perform the test
Sure, we could do an experiment. But I don't think we're going to. Given that, it's irrelevant. It doesn't mean I'm wrong. It doesn't mean you're wrong. Maybe you aren't interested in discussing non-experimental, statistically significant aspects of phonology, and that's fine. But the lack of an experiment is entirely irrelevant to your argument because it means we're both equally wrong (or right)-- we don't have the data yet. So if you want to do an experiment, go ahead. As I said, I could record myself pronouncing these words. Given that positive evidence (even the very small sample size, etc.), at least it's something! Now, time to falsify it, if you wish.
I hope that you agree that it would be a rather challenging experiment, but it could be done if you constructed the right kind of stimuli, and plan to make appropriate measures
Aren't all experiments challenging? Again, this is irrelevant. I've seen a lot of clever experiments and I have no doubt this one could be done by someone who is motivated. (I'm not, mostly for reasons of time, and my main area is syntax, not phonology.) Plus, we've already seen some very plausible ideas in this thread. Doable.
I fervently hope that you agree that it would be kind of a waste of your time to do this, and that it would be actually harmful to conduct a crappy experiment.
I'm confused. What do you mean by "this"? An experiment? I thought you were suggesting one. Regardless, I'm not wasting my time with an experiment because I'm not planning to do one anyway.
Now, you've also advanced the position that most people would agree with Hypothesis 1, and this concerns me somewhat because the discussion has thereby shifted from questions of objective truth to social perception of scientific hypotheses.
Untrue. Scientific "truth" does not exist. There is only the sharing of knowledge (via publication, etc.) and the conventionalization of that knowledge (through, for example, teaching in classes and what is accepted in mainstream theories and presented in textbooks). While there may philosophically be a "real truth" out there, science, as a field, does not have a way of identifying it. All we can do is talk to one another and decide among the various theories regarding which, after all of the evidence is presented, is most acceptable. We might be right. We might be wrong. But it's eventually subjective and interpretive, not objective. Hypothesis testing is great, but still there are infinitely many possibilities by definition.
So the way the field works is exactly what I described, except that in a more formal situation (again, remember, this is an internet forum, and it would be absurd to have an expectation of cutting edge experimental results with every post, I'm sure you agree), then we would have some data to back it up and argue our point so others agree.
An example of this is the requirement for a 5% of "by chance" statistical results. (That is p<.05.) This is entirely conventional and has no objective truth to it. It's a good guess, and we are ok with it, but that's an opinion, not a fact.
Where we may disagree is that I would like taking a guess (in the end everything is a guess, just some better or with more evidence than others), rather than having no guess at all. You're welcome to falsify it. That's science.
So in the end, your posts are going off topic rather than actually addressing the point. If all you wish to say is "we don't have experimental data that proves any of this", then I certainly agree.
Surely you have nothing better to do this summer (hah!)
I wish. Dissertation is taking all of my time :)
I've had numerous intuitions about phonetic properties of my speech and have been surprised to look at the measured phonetic reality, so you may develop new intuitions.
This is interesting! But wouldn't you say that most of the time you're more right than wrong? I have not even come close to suggestion intuitions should replace actual experiments. That would be absurd. But given that we don't have experimental results, my observations/intuitions aren't better than nothing?
Although I didn't lay out the reasoning, my claim that Modern Hebrew is more likely than English to have actual neutralization is based on the (uncontroversial) fact that MH has active synchronic phonological processes avoiding geminates.
Right, I think I hinted at that in my reply. It may be true. But I don't see how we can KNOW that English does not, when English does not have such cases to test (except across word boundaries, which we're apparently considering to be less than optimal test cases). In a theory like Optimality Theory, some effect (whatever it may be) would be inherent in the ordering of constraints. In some other theory, perhaps this would be undetermined. But merely speculating doesn't do much here, given your own reasoning.
(I am interested in that you brought up historical degemination in MH, though, so thanks for that. I'm fairly familiar with Arabic, less Hebrew, and I didn't know that MH was degeminating/ed, so I looked up a few things about that.)
One argument to be made here is poverty of the stimulus: so again, depending on the theory, children may simply have no such rules because the context never comes up.
As an odd example here, I've encountered a situation where Spanish speakers produce word-final voiceless stops. (This was in an artificial grammar experiment where, oddly enough, there was no such input, but by mixing up aspects of Spanish and aspects of the artificial grammar, they arrived at a midpoint with word-final voiceless stops once in a while.) They did this enough that I noticed a pattern: for some reason, they seemed to be aspirating word-final [t]. Of course it might be due to having heard me speak in English (but are they really that aware of aspiration? I don't know), or it might be a natural grammatical property of Spanish that (non-existent) word-final voiceless stops are aspirated. I wouldn't claim my data proves anything, but it does make me curious and very open to the idea that languages may have "rules" (or "emergent effects" or whatever you want to call them) for situations that actually never come up. Another argument in support of it is that not all new sounds are as difficult to learn. Our grammars are complicated, and I think it would be underestimating their complexity to assume they don't have any more properties than the absolute minimum.
My argument does crucially rely on the distinction between actively eliminated strings, which directly motivate a thing in the grammar (rule, or package of constraint rankings, depending on your meta-theory)
The rules themselves (and your argument) also depends on a meta-theory. I have a feeling you are working within one, implicitly, which is why we seem to disagree more than we need to, when I'm speaking (to the greatest extent possible) about data rather than theory, trying to remain theory neutral since I have no particular theoretical claims at this time.
but not every historical fact has a corresponding object in the grammar of a language, for instance, the phonological relationship between "father" and "paternal" is not part of the grammar of English.
Indeed not, and that's immaterial here. Diachrony may inform synchronic study, but it can't determine grammar rules.
Now this post has become longer than I intended, but, anyway, that's where I'm coming from with this.
-
I think that djr made some very good points about the nature of this kind of venue. I was the first to mention that one might do an experiment to prove a point, but it was just a passing remark. The value of informal discussion boards like this is to suggest directions for more serious endeavors. Similarly, I can't really give a full-blown description and defense of Natural Phonology here, because there are a lot of very complex issues and references to past research that would be needed here. So the more superficial discussions here are meant to pique interest in issues that people might want to think about more deeply as they advance their own thinking.
To me, degemination belongs to a rather large family of processes that simplify articulations. In Natural Phonology, we call these "lenitions", but that is an old term that linguists have applied to phonological and morphonological processes long before there ever was a Natural Phonology theory. Assimilation is also a leniting process. In fast and casual speech, lenitions tend to feed each other in a way that erodes syllable structure in dramatic ways, much to the dismay of those trying to learn a new language. So you might expect a relatively simple two-syllable word like "cotton" /ka.tən/ but hear [kʰaʔn] (not sure how to represent a syllabic [n] with the inline system). A phonologist explains this in terms of a simplifying process that takes place in English homorganic clusters, where /ən/ → syllabic [n], which sets up the homorganic cluster simplification [tn] → [ʔn]. The output is "simpler" in that a glottal stop is inherently easier to pronounce than a voiceless stop formed by a closure in the oral cavity. And that allows us to explain pronunciations like [kʰæʔm] for "captain" (sometimes spelled "cap'm" in eye dialect), but with the new wrinkle that the [pt] cluster simplifies to [kʰæppn] through forward assimilation and [kʰæpn] through degemination, which can feed another round of forward assimilation to [kʰæpm], which sets up the homorganic structure that feeds glottal stop substitution to [kʰæʔm].
So what you end up with in fast speech is sometimes a recursive cascade of simplification that radically reduces and simplifies articulation at the expense of auditory clarity. The interesting thing, though, is that you get these really long, complex phonological derivations underlying the production of a simplified articulation. Native speakers are able to handle the phonological distortion precisely because they can calculate a plausible path to the surface articulation in their own minds. (Well, that and the fact that the context primes them to listen for a narrow range of words.) Degemination in all of this is just the tip of the iceberg, but it plays an important role in fast and casual speech tempos.
-
As you suggest, degeminiation may not be an isolated phenomenon and other clusters of similar but not identical sounds may also simplify. So I may be wrong to be specifically thinking of this as "degeminiation". Regardless, in response to the original question, the loss of one of those segments is still very plausible/natural for me. So maybe just calling it "lenition" would be fine, but doesn't explain much.
Interesting points, Copernicus.