Joanna is, as ever, quite right about the fascination these sentences 
have had for linguists.  I remember reading Paul Postal's stuff on 
"the cross-over phenomenon" (early term, since superceded, I guess) 
as a grad student and thinking that no more impenetrable bit of prose 
had ever come my way.  (I am now older and my standards are lower.)

Anyway, one extremely readable and clear exposition of this is in R. 
L. Trask's book _Language: The Basics_.  It's in the chapter in 
syntax, and I believe he calls it "Lanacker's Rule" for convenience.

Kathleen Ward

>Please note that coreference between the noun and pronoun is not
>possible in the starred sentences (that is, in those sentences, the
>pronoun cannot be interpreted as referring to the same individual as the
>name in the sentence):
>
>The sentences below are the samples:
>
>(1-a) When Tom was young, he was rich.
>(1-b) When he was young, Tom was rich.
>(1-c) He was rich when Tom was young.  *** he does not equal 'Tom'
>(1-d) Tom was rich when he was young.
>
>(2-a) I like Jack because he is handsome.
>(2-b) I like him because Jack is handsome. *** 'him' is not 'Jack'
>(2-c) Because he is handsome, I like jack.
>(2-d) Because Jack is handsome, I like him.
>
>(3-a) It is not until you lose health that you realize its importance.
>(3-b) It is not until you lose it that you realize the importance of
>health.
>
>(4-a) After you lose your membership in it, the younger generation seems
>pretty bad.
>(4-b) After you lose your membership in the younger generation, it seems
>pretty bad.
>
>(5-a) Tom said that he had not murdered his wife.
>(5-b) He said that Tom had not murdered his wife. *** 'he' is not 'Tom'
>
>These cases are examples of a phenomenon quite famous in syntactic
>theory; people have been trying to develop an adequate account of why
>coreference is blocked in such cases for a long time now; many people
>believe that there are answers. I don't know what the current set of
>rules is. But it's important to note that coreference is blocked in
>these cases;  you can't always alternate the pronoun/noun order and
>preserve the meaning of the sentence. Keywords in the syntax literature
>would be 'C-command', 'coreference', 'constraints on coreference', and
>'binding'. Be prepared for full-bore syntactic theory; it might be
>helpful to consult an introduction-to-generative-syntax text for a
>discussion accessible to someone with little/no linguistics training. I
>believe a lot of intro books would discuss this issue, since it is so famous.
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>Johanna Rubba   Assistant Professor, Linguistics
>English Department, California Polytechnic State University
>One Grand Avenue  • San Luis Obispo, CA 93407
>Tel. (805)-756-2184  •  Fax: (805)-756-6374 • Dept. Phone.  756-259
>• E-mail: [log in to unmask] •  Home page: http://www.calpoly.edu/~jrubba
>                                        **
>"Understanding is a lot like sex; it's got a practical purpose,
>but that's not why people do it normally"  -            Frank  Oppenheimer
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~