Consider the following sentences, which all have the pattern
X, that's what I say.
where X is a complete clause:
[1a] Nobody ever got something for nothing, that's what I say. (source)
[1b] The previous spring's where it's born, that's what I say. (source; scroll up)
[1c] She'd be a fool to turn ye down, that's what I say. (source)
[1d] Never trust an ethnic restaurant full of white people, that's what I say. (source)
[1e] Let them know that they've got to be afraid of you, that's what I say. (source)
They all seem to be related to a corresponding pseudo-cleft in which X becomes a content clause, namely,
What I say is that X.
So, for example, we have
[2a] What I say is that nobody ever got something for nothing.
and similarly for the other examples.
Now, this reminds me of left-dislocation, as discussed in another question on the English stackexchange:
[3] A good attitude, that's what counts.
This is, apparently, a straightforward case of a left-dislocation of
[4] What counts is a good attitude.
However, all examples of dislocation I've seen discussed so far involve a dislocation of a noun phrase (such as a good attitude). True, CGEL, for one, says only that dislocation usually involves an NP (p. 1408), so presumably other things could be dislocated as well. In full, CGEL says
A dislocated clause has a constituent, usually an NP, located to the left or right of the nucleus of the clause, with an anaphorically linked pronoun or comparable form within the nucleus itself.
It seems to me like the examples [1a]-[1e] could fit this definition...
One specific thing that worried me for a time about the left-dislocation theory in the case of the examples [1a]-[1e] is that the word that appears in both the non-dislocated and the (allegedly) dislocated sentences:
[1a'] Nobody ever got something for nothing, that is what I say.
[2a] What I say is that nobody ever got something for nothing.
In contrast, in the attested examples of left-dislocation, that only appears in the dislocated sentence:
[3] A good attitude, that is what counts.
[4] What counts is a good attitude.
However, I don't think that this superficial breaking of the [3] vs. [4] pattern matters. After all, that in [2a] is a subordinator introducing the content clause, whereas in [1a'] it is a pronoun, just like in [3].
My real worry is that I haven't personally seen any attested examples of left-dislocation other than those of an NP. In fact, the predecessor of CGEL, Quirk, Greenbaum, Leech, and Svartvik, says that only NPs feature in left-dislocation (p. 1310)... On the other hand, CGEL leaves the door open for other kinds, but it doesn't provide any examples. And as these two are my go-to sources at the moment, I don't know what to think.
So: are the examples [1a]-[1e] all instances of left-dislocation of a content clause? Or are they mere comma splices, so that one should really have, e.g.,
[5a] Nobody ever got something for nothing; that's what I say.
instead?
