The Mixed LBE Generalisation, Branan’s Generalisation and implications for the theory of left-branch extraction
James Morley
November 2024 

This thesis explores two putative generalisations over left-branch extraction (LBE) – one new, one from recent literature – and argues that they have serious implications for standard theories of LBE; before sketching a new proposal which captures them. First, the thesis focuses on motivating the novel generalisation, which I call the Mixed LBE Generalisation (MiLBEG): that dependencies which can extract left-branch elements in a given language always show a ‘mix’ of canonical A- and A’-properties within that language. I provide evidence from a genetically diverse sample of 35 languages, demonstrating that their LBE-permitting dependencies consistently exhibit one of three ‘mixed’ profiles of A/A’-properties: in addition to a non-zero set of A’-properties, they show the A-property of either (i) A-minimality, (ii) lacking weak crossover effects or (iii) inducing case-assignment to, or control of predicate-agreement by, the moved left-branch element. When these mixed properties are present, LBE is possible; when they are absent, it is not. The second section of the thesis addresses the second generalisation, first proposed in Branan (2018) and here labelled Branan’s Generalisation: that non-argument extraction (of which I take LBE to be a subset) – is only possible from nominals which can control φ-Agree. I demonstrate that contemporary theories of LBE cannot account for this generalisation; this includes both the standard phase-based theory of Bošković (2005, et seq) and more recent proposals which do engage (in some form) with Branan’s Generalisation, like Davis and Branan (2019) and Branan (2022). In the final section of the thesis, I argue that this problem can be resolved by exploiting the insights of MiLBEG. Specifically, since all LBE-permitting dependencies show a mix of A/A’-properties – as per MiLBEG – I argue that all LBE-permitting dependencies must be performed by composite probes containing [φ] and [A’] features (e.g. Coon and Bale 2014; van Urk 2015; Lohninger 2023). I then propose, moreover, that a recent theory of composite probing - Branan’s (2022) - is able to model the range of ’mixed’ dependencies described in the first half of the thesis, with one important additional consequence: the relevant probes are guaranteed to obey Branan’s Generalisation. Insofar as we assume that LBE can only be performed by probes of this type, then both generalisations are accounted for. The final section then provides more speculative discussion of how, in turn, to derive the assumption that LBE is always performed via composite probes, based on Thivierge’s (2021) phasehood-as-φ-intervention model.

Format:	[ pdf ]
Reference:	lingbuzz/008603 (please use that when you cite this article)
Published in:	University of Cambridge (MPhil thesis)
keywords:	syntax, left branch extraction (lbe), composite probe, weak crossover, a/a', possessor extraction, external possession, phase theory, phase unlocking, phi-agreement, phasehood-as-phi-intervention, phasehood, locality
Downloaded:	278 times

 

[ edit this article | back to article list ]