I wrote:
> I just found myself having to choose between the following
> sentences:
> 
> 1.  Astrophysics' gain is linguistics' loss.
> 2.  Astrophysics' gain is linguistics's loss.
> 3.  Astrophysics's gain is linguistics' loss.
> 4.  Astrophysics's gain is linguistics's loss.
> 
> [NB these are all phonologically different]
> 
> Would we agree that (1) is what we would expect?
> 
> Which one sounds right to you? One definitely felt righter
> to me than the other three, but I am so surprised that I
> don't want to risk biasing your judgements by telling you 
> which.

Summary of actual preferences. "+" = OK. "(+)" = sort of OKish.
"-" = bad. These values are relative to each other.


           1   2   3   4
Joe        +   -   +   -
Rob        +   -   +   +
And       (+)  -   +  (+)
Chet      (+)  -   +  (+)
Nik                +
Jasp       -   -   -   +
Ros        -   -   -   +
Dylan      -   -   -   +
Dick       -   -   -   +


The first thing that strikes me about this data is Dylan's
deviancy. By the regular rules I induce from the data, he
should be "Dyl".

More (pukka) comments below.


Which would we expect it to be?
(1) or (4). Me & Nik thought it would be (1), because of the 
consciously learnt rule that 's is suppressed when the
host already contains plural -s, and we take the -s on the
end of astrophysics and linguistics to be morphologically
the plural -s.

Ros: "4 for me, too. working on the practice that all 
singular nouns have their genitive formed by adding 's."
That's a different practice from the one I consciously
use. But the difference is only apparent when you have
a singular noun that contains the plural suffix -s.
I'm not sure whether Dick & Jasp think that "astrophysics" 
and "linguistics" don't contain the plural suffix -s,
or whether they actually apply a different rule, a la
Ros. Dick says these words aren't plurals, which is
true, but I don't know whether he's saying 
  (a) the presence of a plural -s morpheme in a singular
      noun is irrelevant to the expression of 'S [= Ros's
      position]; or
  (b) the -s in "linguistics/astrophysics" is not
      a plural -s morpheme
      [either because (i) it's some other kind of
       soundalike morpheme, or (ii) because there is no
       plural morpheme, since -s is just a bit of 
       phonology at the end of the shape of most
       syntactically plural nouns]

Rob says that only the (4) pattern is possible for
"Asterix'(s)", "Vercingetorix'(s)". I imagine that we
all agree with this, and that consequently the
judgements shown above are not primarily affected
simply by the fact that the host ends in /s/ (the 
phoneme).


Looking at the judgements, Ros+Jasp+Dylan+Dick are
all clear in applying a rule for 'S expression that
because of either (a) or (b) above treats "linguistics"
& "astrophysics" like all other singular nouns.

For the rest, what stands out is firstly that (3)
is either okay or the preferred form, even though
it appears to be inconsistent, and secondly that
(2), the alternative inconsistent version, is 
unanimously rejected.

It seems to me that any judgement that (1) or (4) is
OK is easily accounted for -- it's easy to tweak the
rule accordingly. But the judgement that (3) is
good or even the best, while (2) is bad, is hugely
perplexing.

Painfully, I have not even an inkling of a possible
explanation. Unless there are some weird euphony
constraints operating over the entire sentence?

--And.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%