LINGUIST List 6.530

Sun 09 Apr 1995

Disc: German Affricates

Editor for this issue: <>

Directory

Desrochers Richard, Re: 6.472 German Affricates

Alexis Manaster Ramer, Re: 6.524 German Affricates

Message 1: Re: 6.472 German Affricates

Date: Fri, 7 Apr 1995 17:54:57 -Re: 6.472 German Affricates
From: Desrochers Richard <desrochrERE.UMontreal.CA>
Subject: Re: 6.472 German Affricates

I agree with A. Manaster-Ramer that perceptual considerations should overcome distributiuonal or etymological considerations (in my opinion, etymological ones are the least important, if they have any). But as just pointed out one native-speaker of German, perceptual cues are not quite obvious and may sometimes contradict themselves. Even syllable "boudaries" are quite variable for a same speaker. But we must always try to tell apart phonetic considerations from phonological statements, which do not respond to the same criteria. In Quebec French, there is a conditioned variant [ts] for /t/ before high front vowels and glides. There is absolutely no constant difference instrumentally detectable between this [ts] and the sequence /t+s/ in other words like "tsar" or "tzigane", and this [ts] may be heterosyllabic as well as tautosyllabic between vowels if we rely on phonetic facts. We can use phonetic arguments only if phonetic "facts" are constant enough. English /tS/ is sometimes analysed as one phoneme, sometimes two phonemes, since phonotactic considerations contradict themeselves, and phonological statements are more (ideally) submitted to criteria like "generality of statements", "economy of description or of entities" in the absence of clear phonetic criteria. The ideal situation is when all arguments converge; as to German tsch, I think the best way to settle for a "decision" (always dependant on theorical assumptions) is by a thorough survey of occurences, status of words, speaker's intuition, preferred syllabification. I would guess, for the time being, that even if it occurs in word initial position,the number of examples is quite limited, and means only that there is no constraint on this pronunciation word-initially. What strikes me hte most is how difficult it seems to find intervocalic tS with no (putative) morpheme-boundary (in which I do not quite believe, but this is not the point) on either side. (The fact that there is a apocope in entschuldigung has only to do with what I would call "lexical" reduction at the phonetic level; French does not have the sequence /ms/ word-initially, and for a very frequent word like "monsieur" you have [msjo"] or, to stay with the same kind of expression, a word like "excusez" becomes [sku"ze], although nobody has ever demonstrated that graphic "x" is a sequence and not a monophonematic /ks/. Anything can happen in allegro - or, say, hypoarticulated speech). Maybe one other argument possible would be to find german dialects where tS has become t, whereas other post-consonantal S are intact. In the worst case, we only end up with undecidable analyses like for English tS. R. Desrochers

Message 2: Re: 6.524 German Affricates

Date: Sat, 8 Apr 1995 08:59:09 -Re: 6.524 German Affricates
From: Alexis Manaster Ramer <amrCS.Wayne.EDU>
Subject: Re: 6.524 German Affricates

Re: Ellen Contini-Morava's posting, the point is an affricate and a sequence of stop + fricative are NOT "audibly indistinguishable". The fact that they contrast in Polish (a fact which was known to Bloomfield and which I just see mentioned in Sihler's book on Greek and Latin) proves that beyond any doubt. The only question is which of the two phonetic types is used in those situations in those languages where there IS no contrast. Just in case, the Polish forms that contrast are written 'cz' vs. 'trz' and examples of the contrast include 'trzech' ('three'- gen. pl.) vs. 'Czech' ('Czech').

Alexis MR

Message 3: Re: 6.524 German Affricates

Date: Fri, 07 Apr 95 22:01 PDT
From: benji wald <IBENAWJMVS.OAC.UCLA.EDU>
Subject: Re: 6.524 German Affricates

I didn't at first pay attention to the "one-phoneme-or-two" discussion emanating from the analysis of German affricates". The problem is classically familiar to me, but at first I considered it a pointless game of pin the label on the phoneme, using the game rules of a persistent but out-moded phonemic theory -- the "bi-unique" phoneme with its Neogrammarian/behaviorist ideological trappings . However, a quote from Alexis caught my attention when Leo Connolly cited it with a retort. It goes:

Alexis Manaster Ramer (amrCS.Wayne.EDU) writes:

)Richard Desrochers and David Powers seem to me to miss the point )about the distinction between one-segment affricates and two )segment sequences of stop+fricative. Distributional and )etymological arguments are irrelevant if the distinction is )audible (whether there is a contrast in the language in question )or not).

More recently Alexis added:

"If you believe in both phonemes and) underlying representations, for example..."

I'm not a fan of "believing in" as opposed to trying to make sense out of FACTS, but I want to join the issue on the key concepts of "distribution", "etymology" and "audible" mentioned in the first quote. To begin with, I can't see why any of the above has any more bearing on the classic issue of "one segment or two" than on "one phoneme or two". So my comments below don't make that distinction.

With respect to "distribution", I can't imagine how that can be irrelevant to any phonemic analysis, whatever belief system the analyst operates with. However, I'm willing to concede that there was a larger context to the discussion which I missed, but which makes sense out of what Alexis is saying about distribution. Therefore, I'll drop any further comment on that aspect of the quote.

With regard to "etymology" being irrelevant, I could only interpret that as the archaic position that grammatical information such as morpheme (and word????) boundaries are irrelevant -- and I was surprised that Alexis would make that claim in view of his previous interest in the "rider" vs. "writer" contrast in the vast majority of North American (t-flapping) English dialects. Connolly's example with the lack of effect on syllable-final German phonetic processes of the diminutive suffix (clitic or affective suffix?) -chen was similar in point. Does Alexis maintain that before the flap the contrast between the two diphthongs is "phonemic"?

If so, as was argued pointlessly in the 50s, does that mean that there is a phonemic contrast just before flapped /t/, or that in all contexts (once a phoneme always a phoneme) there is a phonemic contrast, so that write and ride differ both in a vowel phoneme and (perhaps) a final stop phoneme?

Next the issue of audibility -- and that might even affect the answer to the question above about writer/rider. Audibility is a vague notion, despite Alexis's pride in his German case. Like Alexis, I can teach people to hear the difference in English, if they can't without help, at least for my dialect and many others. But that does not mean that I assume they use it in decoding speech -- or would ordinarily be able to consistently distinguish the words in isolation based on the length and/or vowel quality cues, EVEN THOUGH THEY MAY CONSISTENTLY MAKE THE DISTINCTION IN SPEECH THEMSELVES. Note that here I am being more relevantly specific than Alexis's example of Cantonese tones. I may not be able to distinguish all the Cantonese tones, but is Alexis considering that maybe some Cantonese speakers make all the tone distinctions but can't "hear" all the differences? That's the issue I'm raising with respect to the "audibility" of the writer/rider distinction.

Meanwhile, I don't in fact know what a decontextualised test of writer/ rider for speakers who make the distinction would produce. A more interesting example to me is the difference between PRINCE and PRINTS (if there is one). (This is another "old chestnut", to be sure.) The reason is as follows. In most dialects of English, including mine, final -t before n-, as in PRINT, TENT etc is reduced to a glottal stop with "inaudible" apical occlusion. There is no release of the final "t". (Similar things happen, by the way, in many German dialects). This notwithstanding, I cannot, for the life of me, figure out whether I glottalise PRINT when I pluralise it to PRINTS, no matter how much I think about it or say it to myself. Sometimes I think I do or can. Sometimes I think I don't -- because if I try to emphasise it to get a CLEARLY audible contrast with PRINCE, it feels exaggerated. In any case, I think I do make a contrast, but even I, for all my auditory acuity, do not think I would be able to hear my contrast if I heard it played back to me. I would, of course, be interested in what other English speakers think, and also whether there are comparable problems in German.

In the meantime, my point is that "audibility" is not a sufficiently clear criterion for establishing PHONETIC contrasts (as input for "phonemic" contrasts?) If, as I suspect, there is a widely used distinction between PRINTS and PRINCE, it is a result of a learned phonetic process by English speakers such that syllable coda /t/ is glottalised following /n/ (and in certain other contexts in various dialects -- unconditionally in some British dialects). Whatever cross-glide "t" which may be produced in some of these dialects between /n/ and /s/ is a distinct phonetic process (possibly not even learned) which does not have the same phonetic effect of glottalisation.

We should see if this process exists, instead of just babbling about it. I hope it does, because I am quite sure it is totally useless to speakers for distinguishing words, or whatever phonemes are supposed to do. Then we can rethink what phonemic analysis is or should be, as if there weren't enough examples of the phenomenon in the literature (e.g., Labov's studies of "near mergers" -- maybe our first paper on them in 1969 had something to do with my agitation over the simplistic notion of "audibility").

The moral may be that such "biunique" concepts of the phoneme as Alexis seems to endorse in his above statement may have some force in the economy of any particular language, and therefore that form of analysis works UP TO A POINT when an analyst is exploring the phonological structure of a newly encountered language, BUT that it will not ALWAYS work (in ANY language), and more complex processes will have to be recognised (as most theories do -- but many then go too far, of course, and start building much of the history of the language into their analyses from morphophonemic residues). The more complex processes then challenge the "relevance" of such notions as "audibility", blindness to "etymology", i.e., grammatical structure (and, in fact, lexical structure) and, inevitably, that speakers of a language are in all ways ignorant of the (recent) history of the language they speak. Benji