LINGUIST List 2.722

Mon 28 Oct 1991

Disc: r-linking

Editor for this issue: <>

Directory

1. , Re: 2.707 R-Linking
2. Nancy L. Dray, p.s. re r-linking

Message 1: Re: 2.707 R-Linking

David Stampe argues that intrusive-r is best treated as underlying /r/ which
gets deleted in non-linking environments. I have always favoured a rule-
reversal analysis, but I'm not sure there's a lot of evidence either way.
The rule reversal analysis certainly fits the historical facts equally well.
Once the rule r --> 0 / C or pause has been reversed to
0 ---> /[V,-hi,+back]__(#)V the extension to words with no historical /r/ is
explained as a natural dropping of lexical restrictions on a rule, there is no
need to assume, as Stampe suggests for the underlying /r/ analysis that
speakers think /Vr/ is in complementation with V# because they don't often hear
examples of historically V final words like banana, saw, paw, pa, in hiatus
with a following vowel. (I'm slightly misrepresenting his argument - he takes
[a:] and [O:] as /a/ and /O/, which doesn't seem at all motivated for
most non-American non-rhotic dialects.)
Against the rule-reversal analysis, Stampe points out that some non-rhotic
speakers find it as hard to avoid intrusive-r in "I saw it" as to pronounce [r]
in "I hear them". I don't know of any hard evidence on how easy speakers find
it to not implement obligatory sandhi rules, whether they are phonetically
motivated or not, so this seems a difficult line of argument to pursue. I
don't suppose there are that many phonetically unmotivated sandhi rules in the
world's languages anyway, so it would be hard to get a real grip on this
anyway.
Stampe points out that glottal insertion (as in "I saw [?]Ed") blocks
r-insertion. [I have actually heard instances in which both intrusive-r and
glottal insertion took place, but I expect they were slips of the tongue -
they were on the BBC, and may have been attempts to avoid intrusive-r in words
with non-historical /r/.] As Stampe points out "Phonetic processes don't block
rules". I wonder if ?-insertion couldn't also be a "rule" in non-rhotic
varieties. It seems to me that ?-insertion is used as an alternative to
intrusive r [particularly as a way of avoiding it, as I suspect was the case in
the BBC examples]. I would probably have to accept that ?-insertion is also a
phonetic process in non-hiatus environments in these dialects, which will not
appeal to some.
There is some phonological evidence for r-insertion, however:  derived from
V-reduction can also trigger r-insertion in many non-rhotic varieties. As in
"The wind[] [r] isn't broken" [hypothetical] or "See ya [r] Ian" [attested in
natural speech]. We don't want to claim window and you have underlying /r/,
nor to we want to have vowel reduction produce [r], so here there seems no
viable alternative to r-insertion.
Stampe points to spellings such as EEyore and marmie as support for the
underlying /r/ analysis. I don't think we can assume spellings necessarily
directly reflect underlying representations. It would be sufficient for
speakers to observe that <ore> may represent [O:] and that <ar> may represent
[a:] for them to use such spellings creatively. [In fact, in non-rhotic
varieties <ar> is the only unambiguous way of representing [a:] in non-
prevocalic environments; so it is very useful. - I'm discounting <ah> which
is pretty rare in the orthographic system, and not used at all preconsonantally
.] To prove the point, we can also observe that in non-rhotic varieties
final <a> (bigga, supa etc.) can be used in the orthographic representation of
words in final // instead of <er>. (I have observed a lot of this in Oz,
where it seems to be particularly common in advertising.)
Ian Smith
York U

A postscript to my note re "the issue-r-is":
It dawned on me that this may not actually have been a case of
r-linking after "u" after all. Perhaps for speakers of this
dialect "issue" has a schwa or other vowel at the end,
at least in rapid speech. The spelling of "issue" and my own
citation form may well have altered my perceptions (or at least
my recollections) here. Does anyone know more about what the
final vowel for "issue" would be for dialects that have this
kind of r-linking? Thanks.
NLD