I have recently been reading a lot of literature on the syntax of proto-indo-european, specifically theories that it was an ergative/absolutive or active/stative language rather than the nominative/accusative language of its daughter languages (like Greek and Latin). From what I can tell, typological studies (see especially Michael Silverstein’s chapter in Grammatical Categories in Australian Languages “Hierarchy and Features of Ergativity”) the ergative hypothesis is probably wrong.
However, there is a great deal of convincing evidence (see in particular Brigette Bauer’s Archaic Syntax in Indo-European: The Spread of Transitivity in Latin and French) the PIE was an active/stative language.
What is interesting about this is that in active languages transitivity is a very minor feature, when it occurs at all. Now ancient greek grammarians and linguists have long known that the accusative in greek (and other ancient indo-european languages) has various semantic roles apart from being a purely grammatical case (i.e. direct object). The proto-typical semantic roal seems to be that of goal.
However, if PIE was an active language, then the accusative reconstructed in earlier stages would not have as a primary role the status of direct object. Transitivity would be a minor part of the language, if it occured at all.
Borrowing from localist theory and cognitive linguistics, the locative/directional roles of the accusative may have been primary, and as a result of or along with metaphorical mapping of the directional accusative onto the semantic role of goal, when PIE changed into a nominative/accusative language the accusative was became the direct object.
In other words, my question is it is possible that the original role of the accusative was not direct object at all, but that this became a secondary role in later stages of PIE, and was subsequently passed on to greek and latin?
It seems awfully strange to me that, if this were the case, somehow every daughter language picked the same case to represent the direct object. And not just the direct object, but the two quite different semantic roles that fall under the “direct object” role in IE languages. A direct object can represent either —
the patient of a verbal action that changes state (“I chopped the vegetables,” “the horse pushed me”)
the focus of a non-state-changing action (“I see the book,” “I consider his plan.”) The accusative of respect is basically focus for adjectives, and that turns up in several PIE daughter languages, too.
If the accusative wasn’t already doing both these jobs in PIE I’d expect a bit more variation in how the daughter languages dealt with this.
But the change from an active language to a nom/acc language occured prior to the split into daughter languages. It is an early state in PIE. The point is that remnants of the active syntax exists in the daughter languages, although the change from active to nom/acc syntax occcured prior to the split (again, see Beekes, Bauer, Shmalstieg, and others on this). So the fact that nom/acc syntax exists in all the attested indo-european languages doesn’t say much about whether the same syntax was present in PIE at some point.
The idea is that there are several aspects of PIE which do not resemble nom/acc synatx. Although an ergative synatx has been posited since the turn of the century, it appears that modern typology of ergative languages rules this out. However, all the data fit very well within an active syntax.
If the accusative was originally a locative/goal case, and (per cognitive linguistic metaphor) was transformed into the direct object in PIE prior the split, wouldn’t this explain the remnants of “active” syntax in ancient indo-european languages?