In an earlier posting I passed on a little puzzle from Neal Whitman, about coordinate sentences with quotations in the first conjunct. Neal had asked about a proscription in Bill Walsh’s Lapsing Into a Comma (2000) against things like
(1) “I’m finished”, Kim said, and left the room.
Neal hadn’t found an earlier explicit statement of such a proscription, but he was pretty sure it had been in the air, since he’d uncovered some indirect evidence of it — in the children’s books written by Beverly Cleary; latest posting from Neal here. (These are charming books, so doing research on them is scarcely onerous.)
The point is that for man’y years Cleary almost always failed to repeat the subject in the second conjunct of examples like (1) — writing
QUOTATION, SUBJECT said, and VP
QUOTATION, SUBJECT said, and SUBJECT VP
(so blithely violating Walsh’s proscription) — but that in her 1990 book Muggie Maggie (well before Walsh’s book was published) she switched to 100% subject repetition.
Then Neal got a chance to look at the 1991 Cleary book Strider and discovered that it went back to nonrepetition. So in 1990 someone (an editor, a stand-in for Cleary, or, least likely, Cleary herself) cleaved to no-subject-repetition. Investigations continue.
I’m going to put that issue aside here, and look at the way the parts of these speech reports are put together. The reports have three parts:
A: a subject, referring to the person speaking (SUBJECT);
B: a verb (most commonly, say) of report (SAY); and
C: a representation of what was said (QUOTE).
A and B always stay together, in either order, so there are four ordering possibilities, along two dimensions:
Quotation Fronting (QF): whether C appears in its default place, following B in a predicate that combines with A (-QF), or whether C appears in clause-initial position (+QF).
Quotative Inversion (QI): whether B appears in its default place, following A (-QI), or whether the subject and main verb are inverted (+QI).
[Neal uses "quotative inversion" as a label for QF, though this isn't the standard usage in the literature on types of inversions of subject and main verb.]
The alternatives are then:
1 (-QF -QI): A + B + C (SUBJECT-SAY-QUOTE) Kim said, “I will go”
2 (+QF -QI): C + A + B (QUOTE-SUBJECT-SAY) “I will go”, Kim said
3 (-QF + QI): C + B + A (QUOTE-SAY-SUBJECT) “I will go”, said Kim
4 (+QF +QI): B + A + C (SAY-SUBJECT-QUOTE) Said Kim, “I will go”
As far as I can tell, these are truth-functionally equivalent. But they’re not equivalent with respect to the way information is organized in them or with respect to style. And they are somewhat different syntactically (beyond the ordering of their parts).
1 and 2 are neutral stylistically, but differ in what gets highlighted, A or C.
3 sounds journalistic, a version that’s unlikely to occur in speech, except in circumstances where A (the SUBJECT) is long or complex and is better postponed until the end of the clause. [There's a considerable literature on Complex NP Shift, under various names (in things like "I saw yesterday a dog that was larger than any I'd ever seen before", in contrast to the very dubious "I saw yesterday a big dog"). But what's interesting here is that a construction that normally has stylistic associations can be recruited for other ends -- in this case, to put off material that's difficult to process or produce.]
Finally, 4 is very mannered, very “literary” in character (Quoth the raven, “Nevermore”.)
Now, the syntax. QI fractures a verb (B) from its subject (A) and, apparently, in so doing makes B inhospitable to premodification:
1 Kim rapidly said, “I will go”
2 “I will go”, Kim rapidly said
3 ?? “I will go”, rapidly said Kim
4 ?? Rapidly said Kim, “I will go”
Postmodification is ok in the 3 and 4 cases (as in the other two):
3 “I will go”, said Kim rapidly
4 Said Kim rapidly, “I will go”
This means that (integrated) adverbials that pretty much must be premodifiers are out in 3 and 4:
3 * “I will go”, always/never said Kim.
4 * Always/Never said Kim, “I will go”
In addition, QI is a “main-clause phenomenon”; contrast
2 “I will go”, we believe Kim said.
3 * “I will go”, we believe said Kim
4 * We believe said Kim, “I will go”
No doubt there are more syntactic differences here between +QI and -QI.
In any case, treating -QF and +QF as separate (and incompatible) constructions, and similarly for -QI and +QI, while allowing the compatible constructions to combine freely, gets the syntactic forms right and gets their association with truth-functional semantics right (I think), but it provides no place to “hook” stylistic differences to — and types 3 and 4, in particular, are quite different in this respect.
What I suggested in a paper some time ago is that sometimes combinations of constructions can have lives of their own: they are composites, with discourse-structural, stylistic, or sociolinguistic values of their own, sometimes with semantics not entirely predictable from that associated with the combining constructions, and often with syntactic peculiarities of their own. (Similar moves are made in other flavors of “construction grammar”.) So, on this view, type 3 is a composite of -QF (the default ordering of A, B, and C — a special case of the default ordering of subject, verb, and direct object) and +QI, the combination being associated with one set of values, and type 4 is a composite of +QF and +QI, with the combination associated with a different set of values.
Such a view allows us to say that “the same” construction can have a variety uses. For instance, Subject-Auxiliary Inversion (exhibiting the order of auxiliary verb, subject, and complement of the auxiliary) is “used” in a large number of other constructions; the semantics is that of Subject-Predicate Clause, plus whatever semantics is contributed by each of the constructions that use SAI: several types of main-clause questions, several types of frontings (obligatorily with fronted negatives, so/such-phrases, and some others, optionally with certain fronted comparatives, etc.). There are extensive inventories of these SAI uses, along with accounts of the specific semantics associated with the uses and of the way they signal information structure and values of various sorts.