It’s just a Phase, it’ll pass: Measuring Unlocking in Malagasy vs. Dinka

This talk has a double aim: to demonstrate that micro-variation between languages is a conspiracy of the derivation; and to draw theoretical implications on the status of Phase Unlocking (Rackowski & Richards 2005; Branan 2018; Halpert 2019, a.o.). The first task will involve comparing Malayo-Polynesian Malagasy and Nilotic Dinka Bor (van Urk 2015), which show striking similarities: they share an identical version of the Austronesian voice system, whereby one (any) argument per clause moves at the left periphery and is designated as the pivot, in-between a V2-topic and a derived subject. They also share an identical distribution of A/A’-properties in clausebound pivot movement. Where they (only) differ is in long-distance pivot movement: while in Dinka it retains all A-properties of its clausebound counterpart, in Malagasy it shifts to a full A’-profile; yet they both proceed via “Phase Unlocking”, as embedded-pivot extraction requires the matrix verb to Agree with the containing CP, so as to render it transparent. In fact, both languages require a prior step of (covert) A-movement of the CP complement, before subextraction of the embedded pivot to an outer specifier of the CP’s landing site. But while in Dinka the CP’s landing site goes as far as matrix SpecvP (van Urk & Richards 2015), I will argue that in Malagasy the CP lands even higher, at the matrix left-periphery pivot projection. I will propose that the two languages differ in the structural height of Phase Unlocking: it happens by matrix v° in Dinka, but by (roughly) matrix C° in Malagasy. This height difference is epiphenomenal on their distinct way of (i) deriving their VO order and verbal complex (V- vs. roll-up VP-movement), and of (ii) bringing the pivot theme to Specv/VoiceP (V2 Agree vs. smuggling); their difference in A/A’-properties of cross-clausal movement then stems from an interplay of Cyclic Agree (Béjar & Rezac 2009) and the Edge Condition (Chomsky 2008). The second task is to address the following implications: unlocking piggy-backs on independently established relations between a higher probe and a locality domain; it is therefore constrained by derivational timing and locality under a weak version of the PIC (Chomsky 2001). Moreover, it requires not mere in-situ Agree with the phase, but an additional step of (covert) movement of the phase to the specifier of the unlocking head. Finally, I will illustrate how and why unlocking is in complementary distribution with successive-cyclic movement.