I am proud to announce that my student Sigwan Thivierge (co-advised with Maria Polinsky) has defended her PhD thesis!
The thesis, Phasehood and phi-intervention, presents a novel syntactic analysis of the Georgian agreement system. While this system has been the topic of a great deal of research in recent decades, what sets Sigwan’s analysis apart is the way the “basic” and “inverse” agreement paradigms receive a truly unified analysis, one that does not require reference to separate sets of exponents for each paradigm, does not reify the notion of “Screeve” or “paradigm” in the syntactic representation, and derives both the similarities and differences between the two paradigms solely from differences in the relative height of v and the subject. (On the issue of these differences in hierarchy, see also: Bejar 2003, Lomashvili and Harley 2011, Marantz 1989, and McGinnis 1995, 1997.)
In so doing, the thesis provides novel evidence for phasehood as an epiphenomenon (cf. Abels 2003, Rackowski & Richards 2005, Halpert 2019, i.a.). Specifically, particular projections behave as phases with respect to certain features (e.g. phi-features) because their heads are themselves probes for the relevant features; and the presence of those features on the head renders the entire projection a viable target for any higher probe seeking those same features. Since the allegedly ‘phasal’ category constitutes a closer target, relative to any higher probe, than anything contained within that category, minimality dictates that elements contained therein cannot be targeted under these circumstances, thus deriving the ‘phasal’ behavior from independently-motivated locality considerations.
Not only are phases epiphenomenal on this view, but they are also not static: unlike in Chomsky’s (2001) conception of the phase, the behavior of a given projection can seem ‘phasal’ at one stage of the derivation but not another. In particular, once an intervener has been targeted by a syntactic operation (e.g. Agree), it can be subsequently ignored for the purposes of minimality, and a more deeply-embedded goal can be targeted (an instance of the more general Principle of Minimal Compliance; Richards 1998, 2001, i.a.). This means that a given XP can act as a ‘phase’ at one point in the derivation, but as a non-phase at a subsequent point. As Sigwan shows, this is indeed attested in the Georgian agreement system, in particular when it comes to number-agreement in the inverse paradigm.
(UPDATE: The thesis is now available on lingbuzz.)