Reading Roberts (2010) for my seminar on head movement, I learn this!
“…The auxiliary associated with the restructuring verb can vary according to the transitivity of the lower verb (this applies to a subset of restructuring verbs, according to Cinque 2004:59, n 48):
L’ho voluto mangiare mangiare agentive/transitive
it’have.1sg wanted eat.inf
“I wanted to eat it.
Sono voluto partire. partire unaccusative/intransitive
be.1sg wanted leave.inf
“I wanted to leave.”
!!!Like the kind of impersonal Agree relation McCloskey has between T and those special embedded lexical verbs in Irish, or the kind of Agree relation across verbal affixes that I think I have in Hiaki.!!! How come I didn’t know about this before?
(Also cool further note from Cinque: If clitic climbing stops partway up in a multiple-restructuring verb construction, the auxiliary must be HAVE. It can’t be BE, even if the lexical verb at the bottom is unaccusative. So auxiliary selection is all about establishing an Agree relation between the matrix auxiliary and the lexical verb, and if there’s an intervening Agree relation with an intermediate verb, the auxiliary selection doesn’t happen.)
Another cool fact from this domain: when you do VP ellipsis (or “modal complement ellipsis”, in Aelbrecht’s term)–elide/omit the complement of the modal verb, like “volare”, the aux has to be “have” again, even if the elided verb is intransitive (because its antecedent is). These facts were discovered by Andrea Beltrama in a paper in 2012ish. What’s interesting is that Agree usually goes into VPE just fine (“There were supposed to be bananas on the table, and then there weren’t “). But in Italian, ellipsis blocks the Agree relation. Aelbrecht has a timing story that could be worked up to make this a point of variation.
That is awesome, thanks for the pointer, Jason! Is that ellipsis in combination with clitic climbing? I’ll go look and see. Thanks thanks! This website is already worth what I invested!