LINGUIST List 2.707

Fri 25 Oct 1991

Disc: R-Linking

Editor for this issue: <>


  1. Joe Stemberger, Re: 2.692 Phonology, Pauses, R-linking
  2. David Stampe, 2.681 R-linking
  3. Jim Scobbie, R-linking

Message 1: Re: 2.692 Phonology, Pauses, R-linking

Date: Tue, 22 Oct 91 18:20 CDT
From: Joe Stemberger <>
Subject: Re: 2.692 Phonology, Pauses, R-linking
In a comment on r-linking, Geoffrey Russom states that:
>My understanding is that you get r-linking primarily to avoid a hiatus
>between vowels. With tense "u," there's an off-glide "w" to avoid the
>hiatus, so it's not surprising to find that Ellen doesn't get r-linking
>in that environment.
This is assuming that [uw] occurs in English. In almost all dialects of
English, whether British or North American, the vowel is actually
pronounced [u:]. This has been recognized for more than 30 years by
phoneticians studying English. There are SOME dialects that have [uw], but
they are relatively uncommon and "nonstandard" (on both sides of the
The reason that transcriptions with [uw] are so common is that it was the
preferred transcription of the American Structuralists. They pointed out
that English mid-vowels are diphthongized, so that treating the high
vowels as diphthongized fit the pattern of English better than treating
the high vowels as long. Such patterning is no longer of great concern to
most phonologists, but [uw] and [iy] have been retained in transcriptions
out of tradition.
As far as r-linking goes, though, it would be useful to know whether the
dialect in question is one of those relatively uncommon dialects that does
in fact have a diphthong. If so, then Geoffrey Russom's comments would
provide a possible explanation. But if not ...
---joe stemberger
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: 2.681 R-linking

Date: Wed, 23 Oct 91 05:43:09 -1000
From: David Stampe <stampeuhccux.uhcc.Hawaii.Edu>
Subject: 2.681 R-linking
Isn't "intrusive r" just an underlying /r/? That's the analysis that
Patricia Donegan describes in a forthc. book on historical phonology
ed. by Ch. Jones. Intrusive r occurs only in dialects of the UK, US,
Oz, NZ, that that de-rhoticize /r/ to [schwa] (syllabic or not) in
syllable-rimes, so that the distinction between /r/ and /schwa/ is
manifested only finally before a vowel, e.g. `copper is' vs `Cuba is'.
If learners miss such rare examples as these, the distribution of [r]
and [schwa] will seem complementary, and it is fully accounted for by
analyzing all [schwa]s as de-rhoticized allophones of /r/. On this
re-analysis, the reason for "intrusive r" being heard in `Cuba[r] is',
as well as in `coppe[r] is', is simply because Cuba also ends in /r/.
The same analysis extends to [schwa]s that arise in some of these
dialects by the centering diphthongization of low or lax vowels as in
saw, pa, baa, yeah. If `saw' [s.O.schwa] is analyzed as de-rhoticized
/s.O.r/, then `sawing', where de-rhoticization is blocked by the
following vowel, naturally will be pronounced [s.O.r.I.eng].
In some dialects and speech styles, nonsyllabic [schwa]s are fully
assimilated to certain preceding vowels, i.e. deleted. This is true
both of the [schwa] from historical /r/ as in `soar' and of the
[schwa] from diphthongization as in `saw'. If both are analyzed as
/s.O.r/, the complete phonetic convergence of `soar' and `saw' as
[s.O(:)], and `soaring' and `sawing` as [s.O.r.I.eng] is explained.
(A slightly different analysis, better for speakers who pronounce
etymological and intrusive r as [schwa.r] before vowels, instead of
just [r], would be that [schwa] is analyzed as a diphthongal phoneme
/schwa.r/, paralleling /a.r/ in `car', /a.y/ in `eye', etc. The point
remains that the "intrusive r" is synchronically basic, not inserted.)
The hypothesis that intrusive r is actually an underlying r seems to
fit the facts better than the alternative hypotheses I know of, viz
(1) Analysis of intrusive [r] as a "linking" sound (a consonant
 inserted to avoid vocalic hiatus). Besides not explaining why a
 rhotic should be used when the adjacent sounds aren't rhotic, this
 also does not explain why intrusive r occurs only in dialects with
 de-rhoticization: this is not normally a condition on "linking"
 sounds: for example, glottal stop is used in hiatus in English and
 other languages with no corresponding glottal-deletion process.
(2) Analysis (e.g. by Bill Labov) of intrusive r as due to false
 analogy or hypercorrection:
		soa' : soaring :: saw : X
 This fails to account for the fact that intrusive r is regular,
 and not just randomly distributed among individual lexical items.
(3) Analysis (by Theo Vennemann) of intrusive r as "rule reversal":
 Beside the phonetically motivated process
		r --> schwa / __Vowel
 speakers posit a complementary, distributionally motivated rule
		schwa --> r / __nonVowel
 This avoids the problems of analyses 1 and 2, but it introduces
 another: if the r is inserted by a phonetically unmotivated rule,
 why do intrusive-r speakers say that it is as difficult (some say
 more difficult) to avoid [r] in `I saw it' as it is to pronounce
 [r] in `I hear them'?
The hypothesis that /r/ is already there in underlying representation
would explain this: it's hard not to say /r/ in `I saw it' if /r/ is
in what you're trying to say in the first place!
More technically, it's well known that glottal insertion blocks the
"insertion" of r:
		[ay sO(.schwa) ?Ed] `I saw Ed'
But phonetic processes don't block rules! The hypothesis that /r/
is there to begin with avoids such problems: glottal insertion (a
fortition) would apply before de-rhoticization (a lenition), and
therefore it could only feed de-rhoticization:
	/ay sOr Ed/	[ay sO(.schwa) ?Ed] `I saw Ed'
	/ay hIr Ed/	[ay hI.schwa ?Ed] `I hear Ed'
The /r/-as-basic hypothesis is also supported by spellings like Eeyore
for (H)ee(h)aw, or Marmie for Mommie, and also by the "hypercorrect"
pronunciations, in rhotic dialects under the influence of prestigious
non-rhotic dialects, of idear, vehercle, chester drawers, etc.
David Stampe <>, <stampeuhccux.bitnet>
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: R-linking

Date: Thu, 24 Oct 91 12:12:05 PDT
From: Jim Scobbie <scobbieCsli.Stanford.EDU>
Subject: R-linking
>From: Geoffrey Russom <>
>Subject: Re: 2.681 R-linking
>My understanding is that you get r-linking primarily to avoid a hiatus
>between vowels. With tense "u," there's an off-glide "w" to avoid the
>hiatus, so it's not surprising to find that Ellen doesn't get r-linking
>in that environment. Linking "-r" is also sometimes present word-finally
>before a pause (my British grandmother said "akapulker" for
>"Acapulco" even at the end of a sentence). But a word-final "-r" after
>tense "u" amounts to a final consonant cluster [wr], which is so awkward
>for English speakers that epenthesis or vowel shortening would be used to
>correct it.
My understanding of /r/ suggests the above is correct. I wonder though if
the grandmother in question was in fact rhotic. As a rhotic speaker, I lack
certain diphthongs associated with glide /r/ in non-rhotic dialects. For
example I have no [i] diphthong, so 'idea' has a final [r]. 'Theatre' also
has an extra [r]. We hear non-rhotic television, and have to construct the
appropriate forms mentally, so for years I thought Gibraltar had *no* final
[r] but that Chicago *did* have a medial [r]. Silly me. Wells in his 'Accents
of English' talks about this stuff at interesting length.
QUERY: for those speakers who have merged /r/ and /w/... what happens about
/w/-/r/ insertion. I'd love to read a description of this. Some speakers I
know have the very rounded /r/, a labio-dental approximant, and it's not
merged with /w/, alwight? The case I'm interested in is a real /w/ /r/ merger.
James M. Scobbie: Dept of Linguistics, Stanford University, CA 94305-2150
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue