Re: Ergativity

Thomas Wier wrote:
<<<
Semantics have no (direct) effect on transitivity.  Take the English
triplet "dine", "eat", and "devour".  In each case, there is some notional
entity being eaten, but each verb has different syntax from the other
two.  "Dine" in always intransitive: *"I dined the food". "Eat" is
optionally intransitive: "I ate the food" ~ "I ate". "Devour" is
always transitive:  *"I devoured".  The test of transitivity is a word's
behavior in syntax (adjusting for the possibility of elision); there
is no Platonic "transitivity" floating in grammatical space here.
>>>
Indeed. If you think:
(i) I --->eat--->something
(ii) I --->eat [--->something] (= I inflict an action on something)

then your logical path is exotropic active and infers: "Since in the active voice
(i) "something" is a direct object, then in (ii) it is an implied direct
object."

Now, if you step aside from your mother tongue, then you might as well consider the following:
(i) I <---[am] rich
(ii) I <---eat (= I am engaged in an activity sustaining my life)
(iii) I <---[am] rich ---[with] gold
(iv) I <---eat ---[with] rice

In this system based on endotropic attributive voice, you don't have any direct
object anymore--or maybe both "something" and "gold" are direct objects? :-)
Some languages tag the switch from the first system to the second. For instance Indonesian:

(i) Saya kaya [emas].
I <---rich ---[gold]
I am rich [with gold].

(ii) Saya makan [nasi].
I <---eat ---[rice].

(ii) Saya memakan nasi.
I --->eat --->rice.

To sum it up, it might be difficult to decide whether a system is "ergative",
"active", "accusative" or "split-" on the basis of how "to eat" works, or all
the more whether it has an "implied direct object". In this regard, "to eat" is
a confusing example because it precisely lies at the junction of active state
and action, which are expressed in natural languages with either an attributive
voice ("active state") or with an active voice ("transitive action"). And this
is quite apart from the fact that, like you write it, these "syntactic" systems
really depend on the underlying "semantic" treatment of the actors and their
predicate (do they act or undergo action, do they have will, is the state or
activity time-consuming, social, repetitive, natural, etc.)