Hi,
I'm really enjoying this. You're making me think.
I'm going to take your points out of sequence. I think I'm still replying to your post; if I misrepresent what you're saying, please correct me.
First, the summary of what I'm going to say: A lot depends on theory, and how you frame your terms. To me, ergativity in English is primarily a side topic to voice, and the only "marked" voice in English is the passive. All others rely on semantics and indirect evidence (such as your very detailled and useful post about the transitivity system in English). BUT: how do you frame the evidence there is systematically? In syntax? Make it part of the lexicon? In other words, what exactly is it that the term "ergative" adds to a combination of transitivity and lexical tagging? I'm still thinking about your suggestion to speak of "ergative structures" rather than "ergative verbs". This is an interesting approach, de-emphasising the lexicon in that respect; but I'm trying to ignore it for this post, mostly because I'm not done thinking it through.
Second, I think I've used the term "semantic" very loosely in my other post. There's reference, and then there's cognitive framing. (Or content and point of view.) The cognitive framing is harder to get at and interpret, mostly because these things aren't always immediately visible. We're talking about "ergative structures" in English, or the "middle voice", because we've noticed these constructions in other languages (Basque for ergativity; Ancient Greek for Middle voice; etc.). That is we have to strip away the structure and get down to the point-of-view meaning that the structures imply. And then we have to go back to English and look for expressions of said point-of-view meaning in this language. (Something similar is going on when linguists are probing "shall/will" along the lines of futurity/modality, within the discussion whether English has a future tense or not. The consensus is it doesn't, but the discussion - assuming "will/shall" as tense-modals - has been productive, if not conclusive.) But the thing is this: if you're bringing concepts to a language from outside (which is usual in comparative linguistics) you need an anchor; conventional structural methods - such as your "what syntactic operations yield well-formed usage?" approach - have their limitations. So do semantic (referential or framing). This makes ergativity/unaccusativity hard to think about, before you choose your approach.
Examples follow:
This is about the sentence, "He died a cruel death."
MrPedanticThe object here is a cognate object (it is implied in the verb
itself) and thus belongs to a slightly different model. (I would say
that it only exists to provide an adverbial opportunity: "he died a
cruel death" = "he died in a cruel way".)
I agree that, framing-wise, the object functions much like an adverbial. But it's an "object" in syntax, which has implications that are incompatible with adverbials. Most relevant, here, "die" is now prone to passivisation.
"A cruel death was died," does sound odd (I'll get to it in a minute), but I wouldn't bat an eyelid at "Many deaths were died that night." Interestingly, it's hard to put this into the active voice, mostly because no subject seems appropriate. (?"The Soldiers died many deaths that night."; ?"The army died many deaths that night."...). To me, all the examples I can think of (plural nouns, collective nouns...) don't express the passive meaning. The closest I come is "Many people died that night." Anything else I can think of is of questionable grammaticality.
However, "A cruel death was died," although it sounds odd, doesn't sound ungrammatical in the least (at least not to me). It's also not a semantic problem; I understand the sentence perfectly well, both reference- and framingwise.
The reason, I think, this sounds odd is a pragmatic one. I think this one sounds odd because it's hard to find a context for this utterance that justifies the passive, which is a "marked construction". You generally expect "marked" constructions to be there for a reason. I suspect in the right context the above sentence would be perfectly fine. (It's a matter of Grice's conversational maxims, the maxim of manner, in particular.)
This is where the "frame-semantics" of syntactic constructions become complicated, I think. How do language structures tie in with cognitive structures (e.g. To what extent do we buy the Sapir-Worf hypothesis?)
So, from this I go to self-observation:
MrPedantic and precisely because of that distinction, I would call "break" here ergative (ex. 5) , and "die" unaccusative (ex. 2).
See, I had the hardest time even to grasp what that meant, not now in this thread, but when I first discovered the distinction. That's because, learning English, I didn't train to see the difference. It wasn't necessary, as ergativity/unaccusativity isn't expressed through syntactic structures, but only indirectly through what operations are possible on the verb; this I pretty much took care off either through lexical list-tagging, or through collocation. If there is a hidden logic to it that I applied in learning, it never became conscious. (It's quite possible that I had a practical knowledge, but no discoursive one of this subject; but why, then, is it so hard to grasp?)
If we go back to the list and sift through the operations there, we'll find that "die" behaves different from "break" in the way we specified. But here's the catch: to apply that structural method, we have to assume that "break" in 5.a = "break" in 5.b = "break" in 5.c etc.; i.e. that "break" is the same lexical item in all these instances. That's because syntax has a hard time to differentiate between "signifier" and "signified", or "sign" and "concept".
Notice, for example, how your 5.a is already the transitive, while systematically it should be the intransitive agentive: 5.a *He broke. (i.e. "He caused/performed the action of breaking." as opposed to "He underwent the process of breaking," which is 5.b, now, and would be 5.c)
So I'd amend this, to:
5a. *He broke.
5b. He broke the plate
5c. The plate broke.
5d. The plate was broken.
5e. The broken plate
5f. The plate broke easily
And the comparison with "die" would be two-fold:
1. = sign; 2. = concept
5.1a He died.
5.2a He killed.
5.1b *He died the man.
5.2b He killed the man
5.1c The man died.
5.2c *The man killed.
5.1d *The man was died.
5.2d The man was killed.
5.1e *The died man [cf. The dead man.]
5.2e The killed man.
5.1f The man died easily. (<-- What's the difference to 5.1a? Should I add an * before it, as this is out of place, here?)
5.2f *The man killed easily. (<-- Is this not available, because 5.1f is available?)
5.1a, 5.1c, and 5.1f seem to be much the same. And this is the problem I have systematising a structural comparison. One possibility, I see is to re-cast 5a as reflexive 1. *He died himself./2. He killed himself. I might try to justify this through dying being a process you undergo, thus if you add an agentive/causative to core meaning (which is not in slot a, but in slot c) the verb becomes by necessity reflexive ("He caused himself to die.")
But these things are all a bit... tentative. I fear it's more rationalised than rational, if you get my drift.
[Interesting aside: you used the term "anticausative" alongside "ergative" and "unaccusative" for break in your thread. Bears repeating, as it's something I'm also still thinking about; a very interesting concept I haven't come across yet.]
MrPedantic But I find a semantic difference too: the first presents the sign from
the point of view of the reader, and the second, from the point of view
of the writer.
Now, that's an interesting observation. I'd argue that the semantic difference is not referential (it refers to the same state of affairs), but it's a framing difference. If we view the sign as a proxy for the agent, we're importing the difference of active vs. voice into a construction that's free of the syntactic properties that normally accompany this framing device in English. "Reads," then, is ergative, while "says is a straightforward accusative verb (one that takes the accusative (which isn't marked in English - except, perhaps, for pronouns, where it's indistinguishable - morphologically - from the dative; the conventional term would be "direct object").
MrPedantic Also, although the same few verbs tend to recur as examples in these discussions, actual usage is more imaginative.
That's what makes language so fascinating, isn't it? Nice example, there, too.