Linking: Sibilant with Other Sibilants (Is + she)

[zoltankr]

Hi, I read in the Beverly Lujan's American Accent Guide that there is no break between sibilants adjoining each other between words.

What is she doing?

I think "is she" sounds like /ɪzʃi/ in the question above but I want to know, you as a native American pronounce "is she" as /ɪzʃi/ or more like: /ɪʒʃi/ (with a /ʒ/ sound as in rouge) ?

Any suggestion would be appreciated

[Daniel]

It depends on the rate of speech.

Slowest /ɪzʃi/ > /ɪʒʃi/ > /ɪʃi/ fastest

I don't think anyone would ever say /ɪʒi/ though. The voicing of the second phoneme seems to be more important, not sure if that's a general phenomenon.

Not so often you might hear /ɪsʃi/, but I think the most natural/normal pronunciation is /ɪzʃi/.

[panini]

I think there are two phonological outputs, [ʍʌts ʃi duɪŋ] and [ʍʌɾ ɪz ʃi duɪŋ]. Then you can compress these outputs to get innumerable body outputs where the lingual and laryngeal gestures of [z] and [ʃ] overlap to some extent, which you could transcribe as [ʍʌɾ ɪz ʃi duɪŋ], [ʍʌɾ ɪz ʃi duɪŋ], [ʍʌɾ ɪzʒ ʃi duɪŋ] and so on (where point size represents variation in duration or voicing and place). In the ideal careful pronunciation [ʍʌɾ ɪz ʃi duɪŋ], there is a tight alignment of cessation of voicing and retraction of the tongue. You can slide those two gestures around considerably, which yields quite a number of percepts such as [ʍʌɾ ɪʃi duɪŋ]. I am not persuaded that the /z/ ever deletes without an acoustic trace, but perhaps someone can cook up an experiment to establish that.

[Copernicus]

Wow, panini.  This takes me back to my very first refereed publication--a squib in Linguistic Inquiry entitled Wuzzywant (LI vol 6 no 4, 1975).  The question I addressed was the cause of two different casual articulations of the phrase "What does he want?"  These can be written in eye dialect as "wuzzywant" and "wussywant"--a voiced vs. voiceless articulation of the intervocalic sibilant.  I argued that the former was a phonological reduction of the uncontracted expression "What does he want", but the latter a reduction of the contracted expression "What's he want".  An earlier analysis by Richard Sproat had treated both as phonological reductions of the same phonemic string, but my position was that contraction was not really a phonological reduction.  It was a morphosyntactic operation that created a different underlying phonological representation.

You claim that there are "two phonological outputs" that can be "compressed" differently, and that is not all that different from my 40-year-old original analysis of "Wuzzyant" vs "Wussywant".  The main difference, of course, is that I would call those "two phonological inputs" that have separate phonological derivations.  Theoretical differences aside, we have the same basic intuitive analysis--that the voiceless articulation comes from the contracted form "What's she doing" and the voiced articulation from the noncontracted form "What is she doing".  We can argue over whether contraction ought to be considered a matter of phonology or morphosyntax, but I think you have a good intuitive analysis here.

ETA:  I'm not sure that I can agree with djr's intuition that you would never hear /ɪʒi/ as a variant pronunciation, but that is a matter for empirical investigation.

[panini]

By Sproat, I assume you mean Laferriere. That's a bit before his time, like, he would have been in high school.

[Copernicus]

I think you are right that it was Laferriere.  I wrote that article so long ago that I forgot the details, and I did remember challenging Sproat on another issue--perhaps Welsh syntax a few years later.  So I misremembered.  Thanks for the correction.

[panini]

The main difference, of course, is that I would call those "two phonological inputs" that have separate phonological derivations.

From what I can tell, that results from different premises regarding the distinction between phonetics and phonology. In my theory, there is a distinct linguistic entity "a phonetics", which gives continuous physical interpretations to discrete symbolic representations (and vice versa). Thus I deny the SPE position that phonology and phonetics are the same thing (alternatively said, that phonetics is only that which is universally necessitated by the laws of physics and the nature of human anatomy). Referring to the above-mentioned continuum of body outputs, SPE has the option of assigning integer values to features in order to get a continuum of durations or amounts of voicing, but the fundamental premise behind giving that variation a phonological treatment is that the output is still composed of a series of discrete segments. This is contrary to the general view of phoneticians, who recognize that segments are at most the abstract cognitive basis for instructing the various articulators to move, with little respect for the deeper categorial origin of those movements. So, the various outputs represented mnemonically as [ʍʌɾ ɪz ʃi duɪŋ],  [ʍʌɾ ɪz ʃi duɪŋ], [ʍʌɾ ɪz ʃi duɪŋ] are not the result of changing a segment of one category to a segment of another category, they result from different continuous amplitude values being fed into the articulatory system.

If you treat all gradient differences in body output as resulting from phonological rules, then of course you might term my two phonological outputs inputs. My claim is that what happens to "what is she doing" is almost entirely in the phonetics. The devoicing of the cliticized variant of the copula in "it's" is part of phonology, as is flapping (this is controversial and non-obvious -- some people insist that you can't have flap as a phoneme in English, but that's another thread). Otherwise, there isn't any phonology going on in these variant utterances.

Not your job, but if you'd care to explicate the NP theory of phonetics vs. phonology, that could be illuminating.

[Copernicus]

I think that we actually have some things in common, terminological issues aside.  I don't have a problem with your "continuum" point when it comes to criticizing SPE.  IIRC, Kisseberth did some nice studies on so-called "low level" phonological rules such as final devoicing that showed a continuum of voicing features (phonation, consonant length, length of preceding vowel).  That is, there were a lot of "trace" features in devoiced consonants that differentiated them from voiceless consonants, although the data was inconsistent about the presence of trace features.

Regarding the labels "phonology" and "morphophonology", I have a somewhat longer historical perspective on their use.  Originally, Baudouin de Courtenay settled on the terms "physiophonetics" and "psychophonetics" as respective labels for alternations that belonged to one category or the other.  Trubetzkoy was never Baudouin's student, but he and Jakobson were Moscow-trained linguists who were admonished by their teachers not to pay attention to Baudouin's radical ideas, which, of course, drove them to study Baudouin's theory in great detail.  They settled on the terms "phonology" and "morphophonology" for what Baudouin was calling "physiophonetic" (single-phoneme) and "psychophonetic" (two-phoneme) alternations.  Physiophonetic alternations were not purely "allophonic", because they could represent phonemic neutralization, and Trubetzkoy handled phonemic neutralization under the rubric of "archiphonemics" within his phonological theory.  However, in an Appendix to his book, he called "morphophonology" a subject that did not belong to phonology.  It belonged to a completely separate area of grammatical description, and that is consistent with Baudouin's fundamental dichotomy between physiophonetics (your "phonetics") and psychophonetics (which you include under the rubric of "phonology").  I could go into much more detail on Russian schools of phonology, whence came modern phonological theories, and how phonological theory evolved in the West, but that would not be appropriate here.  Suffice it to say that Natural Phonology maintains Baudouin's original alternational dichotomy in terms of its "Process" vs "Rule" dichotomy.  Processes are responsible for physiophonetic alternations and Rules for psychophonetic alternations.  (In the context of the thread topic, English contraction would be a morphonological/psychophonetic/Rule-based operation, not a phonological one.)

There is a lot more I could say about the Rule/Process dichotomy here, but I fear that I have already gone well beyond the boundary of adhering to the thread topic.  Sorry for the digression.