On the origin of spines: Implications for an (emerging) Theory of Features

In the past few decades, numerous studies have explored the architecture of functional projections dominating the vP domain. A common view is that languages may share a universal inventory of features (Rizzi 1997, Cinque 1999), but which features are represented by separate projections and which are unified by one is a matter of parametric variation (Giorgi & Pianesi 1997). For example, famously, some languages are Voice-splitting where VoiceP and vP are two separate projections whereas others are Voice-bundling where the features and functions of Voice and v are unified under one projection (Pylkkänen 2008, Harley 2014).

In this talk, I provide a comprehensive account of how the Voice-bundling/splitting parameter works in complex event nominalizations (CENs), which has never been articulated before. A common view is that CENs across languages have a defective passive-like Voice which is a separate projection from an nP (Alexiadou 2017). Using evidence from Lithuanian, an understudied Baltic language, I show that this is not a linguistic universal. Lithuanian CENs can have an active thematic Voice, which has a projected agent and assigns structural object case to the theme. Most importantly, CENs exhibit a remarkable Voice-bundling: the features of this active Voice and the nominalizing n head are bundled and occur on a single head. I provide evidence that this type of bundling is part of the presyntactic lexicon (List 1 in DM).

This study has important implications for syntactic theory. First, it shows that syntactic atoms can be complex i.e., come in bundles, which is problematic for cartographic approaches (Cinque 1999) including theories like Nanosyntax (Caha 2006, Starke 2009). Second, the bundles of features can come from the presyntactic lexicon rather than being assembled in syntax via head movement or at PF via fusion. Lastly, bundling can be cross-categorical. While previous studies have suggested that bundled-up features are domain homogeneous (H.Á Sigurðsson 2010) e.g., verbal domain features bundle exclusively with the features from that domain, I demonstrate that this may not be the case. Specifically, I show that the features of the nominal domain and the features of the verbal domain can occur on one head.