Ambiguity in spelling: singer (was: dot-calm)

[Mark A. Mandel]

David Bowie <db.list at PMPKN.NET> writes:

>>>

The "singer" complaint seems unfair to me--confusion through spelling
occurs (cf. <read>), but in spoken language the difference between [rid]
and [rEd], or [sINr] and [sIndZr] is completely transparent.

<<<

I was using the word to construct an argument by analogy, which I didn't put
clearly. Let me have another go at it, at risk of stating the obvious at
some points:

1. The rare word <singe-r> should not be criterial in constructing a
description of English orthography, despite the word's (/'sIndZ at r/) regular
derivation from "singe" via the productive agent/implement suffix /@r/.

2. Likewise, the rare word /'meS at r/ should not be criterial in constructing
a description of English phonology, despite its regular derivation from
"mesh" via the same suffix.

3. And in general rare words are poor candidates for minimal-pair
construction, because a pair doesn't exist for anyone whose vocabulary lacks
*either* member.

So much for the analogy. Let's get to the meat of the question. Look at
Ben's list of minimal pairs for /S/ vs /Z/ (quoted with reformatting here;
"etc." as in original):

        "mesher" vs. "measure"
        "Asher" vs. "azure"
        "Aleutian" vs. "allusion"
        "cash" vs. "cazh" (short for "casual")
        "shush" vs. "zhuzh" (a Queer Eye-ism)
        etc.

I doubt that those pairs actually exist for most speakers.

To check my intuitions, I looked at the British National Corpus of just over
100 million tokens. Take out the last two pairs, whose second members are so
recent and subculture-bound that Ben rightly felt that he had to gloss one
and source the other, implicitly (imho) conceding their irrelevance to the
great bulk of the English-speaking population. For the rest, the BNC lists
the following frequencies per million*
(http://www.comp.lancs.ac.uk/ucrel/bncfreq/lists/1_1_all_alpha.txt)

*Rounded frequency per million word tokens (down to a minimum of 10
occurrences of a lemma per million). I've added to get the values for
"measure" and "asher").

Word            PoS             Freq    forms

  measure       NoC + Verb      177     all forms as common noun or verb
  allusion      NoC             3       allusion, allusions
  azure         NoC             1
  Aleutian      NoP             0       Aleutian, Aleutians
  asher         NoC + NoP       0       asher, Asher
  Mesher        NoP             0

I may have missed a form or two, but the relative orders of magnitude are
clear.

Now here's the basic argument:

1. None of those minimal pairs has *both members* present at even 1ppm in
the BNC.

2. And therefore, I'd say, they shouldn't count for the phonemic distinction
of [Z] vs. [S].

3. But those two phones are distinct in English. (I hope *that* won't be
disputed here!)

4. And therefore minimal pairs are not necessary for a determination of
phonemic distinctiveness.

Whew! G'night, y'all.

m a m

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org