LINGUIST List 2.674

Thu 17 Oct 1991

Disc: Phonological Issues

Editor for this issue: <>

Directory

  • Richard Ogden, RE: 2.669 Phonology and Sound Change
  • Michel Jackson, R-Insertion
  • Jim Scobbie, (Re: 2.669) Autosegmental representations
  • John Phillips, Re: Query intrusive r-insertion

    Message 1: RE: 2.669 Phonology and Sound Change

    Date: Thu, 17 Oct 91 9:18 BST
    From: Richard Ogden <RAO1vaxa.york.ac.uk>
    Subject: RE: 2.669 Phonology and Sound Change
    Thank you MIchael Jackson for your comments on [r] and vowels. But I speak British English, and although I said my voiced alveolar approximants are often rounded as well, I also said they're velarised and the vowels after them are retracted; not the same as rounded. On the other hand the other liquid in my system of liquids at the syllable onset is always 'clear' by comparison; I think phonology ought to say *something* about that, because there seems to be a liquid system with two members, one of which is clear and the other dark. OK so you might describe the retracted vowel afterwards as 'spreading', coarticulation or whatever -- but that says nothing interesting about the liquid system in my dialect of English. (Other dialects of English have it the other way round: dark [l], clear [r]). It seems to me that a feature that means /dark/ has as its phonetic interpretation both [velarised] and [rounded], and I don't see why I'd want two phonological features to do this, since the contrast is /dark/ vs /clear/ in my liquid system. You might be interested in: John Kelly & John Local: Long-domain resonance patterns in English in 'Speech input.Output: techniques and applications', IEE Conference Publication No 258, 1986 Richard Ogden

    Message 2: R-Insertion

     Date: Thu, 17 Oct 91 12:36:31 EDT
    From: Michel Jackson <jacksonshs.ohio-state.edu>
    Subject: R-Insertion
    This posting is LONG. >In J.C. Wells' "Accents of English"(1982), volume 1, p. 226, he gives the >rule for r-insertion in RP English (both linking /r/ and intrusive /r/) >as: > 0 -> r / [-high V]__ #0 V > > ... stuff deleted ... > >To my humble non-native ears, r-insertion after /u:/ appears to be >possible, even though it is a close vowel, as in 'you and me' /ju:rnmi:/ >and 'hue and cry' /hju:rnkraI/. Am I right in assuming this? > > ... more stuff deleted ... > >Does anyone know how common intrusive r-insertion is in RP? Is it the >predominant phenomenon in environments defined by the rule or not? > >Richard Piepenbrock Gimson's _An Introduction to the Pronunciation of English_ (one of the standard references on the topic) describes the following situation: i) historical syllable-final /r/, called 'linking' ii) inserted intervocalic /r/, called 'intrusive' Gimson says (p. 258 of the 1989 edition) In the case of words which end with orthographic r, and /r/-link is regularly inserted between the final vowel of the stem /schwa, long accented mid central (backwards epsilon), long low back unrounded (script a), and long lower mid back rounded (open o)/ and any initial vowel of the suffix. ... This process applies to derivatioanl as well as to inflectional suffixes. Gimson's examples included things like "blur" (no /r/ in isolation) / "blurring" (with /r/) and "familiar" (no /r/ in isolation) / familiarize (with /r/). These are simple cases that appear to be correctly described by the rule you cite above. In liason contexts between words, Gimson describes a more complex situation. His discussion (p. 302 ff.) is as follows: RP retains word-final post-vocalic /r/ as a linking form when the following word begins with a vowel, i.e. in those cases where an [r] soudn existed in earlier forms of RP ... The vowel endings to which and /r/ _link_ may ... be added are /long low back unrounded (script a), long lower mid back rounded (open o)/ and those single or complex vowels containing final [schwa] (/schwa, long accented mid central (backwards epsilon) lax high front-schwa centering diphthong, upper mid front-schwa diphthong, lax high back-schwa centering diphthong/), e.g., in _far off_, _four aces_, _answer it_, _fur inside_, _near it_, _wear out_, _secure everything_. By analogy, this /r/ linking usage is extended to all /long low unrounded, long lower mid rounded, schwa/ endings, even when there is no historical (spelling) justification. Such _intrusive_ /r/s are to be heard particularly in the case of /schwa/ endings. ... Less frequently, analogous links (unjustified by the spelling) are made with final /long low unrounded, long lower mid rounded/ ... Gimson continues It shold be noted that, in synchronic terms, the same process is in operation whether the /r/ link inserted is historically justified (linking) or not (intrusive). However, he observes (p. 303) There appears however, to be some gradation in the likelihood of occurrent as follows: (1) Where a word ends in a non-high vowel, the insertion of /r/ is _obligatory_ before a _suffix_ beginning with a vowel ... (2) Before another word, the insertion of such an /r/ is _optional_. However, it occurs in the vast _majority_ of cases where a historically justified _linking_ of /r/ is possible. (3) Where the /r/ link is _intrusive_, speakers tend to use it after /schwa/ ... more readily than (4) after /long low unrounded/ or /long lower mid rounded. There is considerable resistance to (5) the insertion of intrusive /r/ before a suffix (e.g. _strawy_ ...) Thus, I would say that Gimson's account is very consistent with Wells'. Gimson adduces no examples of linking or intrusive /r/ after high vowels. It should be mentioned that this may be a result of the fact that Gimson analyzes what might be considered underlyingly high vowels before historical /r/ as phonemic diphthongs /I, U, e/ in words like _near_, _secure_, and _wear_. This analysis allows him to preserve the "non-high" generalization at the expense of a more "surfacy" phonemic analysis. Which is, of course, exactly what you had suspected. So much for RP. There are, however, similar dialects in the US, & I have a few black relatives who speak such dialects in the Southern Conn. area (although their speech is also influenced by Virginian remnants and New-Yorkisms) who have wildly generalized intrusive /r/s. In unguarded moments, i've heard a cousin of mine say 'I see[r]'im about once a week' and 'I saw[r]'em yesterday'. This is of course anecdotal. But if you have had some exposure to r-drop / r-intrusion dialects on this side of the Atlantic, it might well account for your intuition that /r/s can be intruded after high vowels.

    Message 3: (Re: 2.669) Autosegmental representations

    Date: Thu, 17 Oct 91 09:45:49 PDT
    From: Jim Scobbie <scobbieCsli.Stanford.EDU>
    Subject: (Re: 2.669) Autosegmental representations
    > Ellen Kaisse <kaisseu.washington.edu> says: > Richard G, echoed by Richard O, wonders if autosegmental > representation can deal with the spreading of non-distinctive features > and the partial assimilation of one segment to a neighbor. These > questions are somewhat orthogonal to the question of how you represent > phonological processes. That is, you tell the feature geometry what > features are represented at some stage in the derivation and it will > spread them for you. You need another part of your theory to tell you > what features are available. It is correct that the number and types of constraints operative at any particular 'level' of a phonological description need not hold for all such levels. But in order that we have some kind of continuity that enables us to claim that all these levels are the same kind of thing, ie phonological, we will have to define some ground rules (constraints on phonological representations) which are good for all levels. For example, in a phonological feature bundle it might be a good idea to ban any distinction between multiple copies of identical features. So [+cont] = [+cont] = [+cont] (A) [+cont] [+cont] [+cont] And we might want to ban the presence of contradictorily-valued features: So [+cont] (B) [-cont] is ill-formed. I think both these formal properties of phonological representations would be pretty well-accepted by many people as fairly basic to Autosegmental Phonology (AP). The second has *not* been universally adopted, especially in the hierarchical representations of AP, because it is useful for describing contour segments. I think proponents of this position are now on the defensive and need to prove their case. This is because affricates, a typical contour segment, can be treated as [+stop] [+fricative] which involves no oppositely valued features (Lombardi, Steriade). So, when we begin to discuss a phenomenon that starts to violate (A) or (B), we might want to question whether we are doing something akin to introducing new non-distinctive features (you could say that we are just relaxing the constraints on the *combinations* of features that occur -- because we are moving to the post-lexical level say), or whether we are changing the ground rules for what constitutes phonological form. The treatment of intrusive stops, and the partial spreading of rounding are cases (no pun intended!) in point. Ellen Kaisse points out that in current AP: > As for spreading of a feature partway into its neighbor, that's easy > enough: > X Y > / \ | > [-F] [+F] > (again, the backslash should be dotted, indicating spread of +F from > Y back onto X). What I just wrote will give you a contour segment X > which starts out with one value of a feature and ends up with > another. You can find actual uses in recent literature - e.g. Nick > Clements' 1987 CLS Parasession paper on intrusive stops. Kiparsky 1985 > (_Phonology Yearbook_ 2) also talks about gradient phonological rules of > this and other sorts in the LP model. Now this look to me like a re-definition of feature geometry rather than the introduction of non-distinctive segments using a novel feature combination. Of course, really it is both, since every alteration to the basic rules of feature geometry would allow new=non-distinctive structures. But is it a good idea to allow such latitude? You pays your money... If no underlying representations require a geometrical form X (1) / \ [+f] [-f] but fr certain purposes such a representation appears useful, then you can say that at the post-lexical level the definition of feature geometry changes and use (1). My opinion is that this produces a bad theory. My disseratation goes into this some. Alternatively you could hold the feature-geometry constant and deal with the phenomena that appear to require (1) in other ways -- by altering the timing and overlap of the exponents of the features in question. This is to claim that there are sophisticated, systematic, non-universal rules of phonetic implementation. Is this contravertial? Empirical considerations ought to help determine which is better, along with the type of theoretical considerations I've hinted at briefly above. The reply to Richard Ogden's query (how does AP deal with such and such an assimilation) is perhaps an "it shouldn't". If AP were a predictive theory rather than, well, then it might say to us that it is incapable of handling such-and-such a phenomenon, AND BE PLEASED ABOUT THE FACT. -- James M. Scobbie: Dept of Linguistics, Stanford University, CA 94305-2150

    Message 4: Re: Query intrusive r-insertion

     Date: Thu, 17 Oct 91 15:37:09 BST
    From: John Phillips <johnlanguage-linguistics.umist.ac.uk>
    Subject: Re: Query intrusive r-insertion
    The rule for r-insertion for me is that it can occur after any vowel in theory. In practice you don't get high vowels on the ends of words in standard English. `You' is [juw], so `you and me' is [juwmmij]; `hue' is [hjuw] so `hue and cry' is [hjuwnkraj]. An intrusive r in the latter sounds wrong to me. The former is a bit different because `you' is often pronounced [j]. In non-RP English English you sometimes hear things like [jrad] for `you had'. But children are taught at school that it is wrong to put an r in where it is not spelt. English regional dialects use intrusive r more freely, including after high vowels, e.g. in Gizzer bi. A birrer wo? (Give me a bit. A bit of what?) ['gIzbI] [bIr'wA]