Speaker: Paul Kiparsky (Stanford University)
Date/Time: Friday 9/21 3:30-5 pm
Location: Singleton Auditorium BCS (46-3002) [Note unusual venue]
Title: UG-driven syntactic change: the word-order cycle
It has been claimed that head-final syntax changes to head-initial syntax, but not conversely. Indeed, major families like Niger-Congo, Afro-Asiatic, Sino-Tibetan, Indo-European, and Uralic, all with a significant proportion of SVO languages, are reconstructed on good evidence as proto-SOV. Wherever such changes can be tracked in the historical record, they can be seen to have the characteristic “drift” properties of proceeding in small but discrete steps in a constant direction over long periods of time, which are puzzling in many ways. Why not in shifting directions, or in a single big step? Why do different languages drift in the same direction? And how can the unidirectionality be reconciled with the persistence of typological diversity?
Arguing against accounts based on processing efficiency or representational economy, and against the historical hypothesis of Gell-Mann and Ruhlen 2011, I adopt the view that drift reflects biases in acquisition. I derive the headedness shift from a grounded and empirically improved version of the Final-Over-Final Constraint (Biberauer et al.), crucially in conjunction with the assumption that linearization is done in the syntax, and is sensitive to word structure, in particular to whether functional categories are expressed as syntactic heads or as morphological affixes. (Evidence for this comes from binding theory and from the lexicalist literature.) These principles generate biases which predict exactly the attested diachronic trajectories. I formalize this idea with a language acquisition model where learners favor the most probable language consistent with previously encountered data, definable in OT as the language with the greatest ranking volume (Riggle 2010).
My proposal makes a number of typological predictions, such as that languages with clause-final complementizers, and languages with no syntactically visible functional heads (such as Japanese) are uniformly V-final. A historical prediction of my proposal is that the word-order shift is in fact not unidirectional but cyclic. That is, it should also be the case that OV languages were once VO. For some head-final languages, there are morphological indications that this is correct (Trask 1977).