Noun class, gender, and the workings of Agree: evidence from agreement with conjoined subjects in Xhosa
Vicki Carstens
August 2019

This paper begins as a reply to Taraldsen et al (2018), wherein it is argued that shared gender features do not underlie singular/plural noun class pairs in Xhosa. Taraldsen et al's claim is based on the fact that when singulars of a given noun class are conjoined, there is typically default subject agreement with the conjunct rather than agreement in the relevant plural class. My paper shows that similar phenomena are found in Slavic languages which have more canonical gender systems (masculine, feminine, neuter), and that an existing analysis (Boskovič 2009) easily explains the Xhosa facts that Taraldsen et al report. My paper then explores conjoined plurals, showing how a pattern of LCA (last conjunct agreement) with preverbal subjects reported in Mitchley (2015) parallels Serbo-Croatian and is also amenable to an analysis in Boskovič (2009). However, I also show that supplementary research of my own reveals a complex pattern of preferences determining agreement with conjoined plurals, in the speakers in my study, which would not be expected under Boskovič's approach, or under recent approaches of Marušič et al (2015), Nevins & Marušič (to appear), and Murphy & Puškar (2017). I argue for a novel analysis under which multiple gender values are copied from all conjuncts to T's uPhi, if &P is raising to preverbal position, and then those values are reduced to one by impoverishment, informed by interacting hierarchies of preference favoring (i) FCA > LCA > default, and (ii) agreement with least marked > most marked noun classes.
Format: [ pdf ]
Reference: lingbuzz/004683
(please use that when you cite this article)
Published in:
keywords: conjuncts, agreement, agree, noun class, gender lca, fca, default agreement, morphology, syntax
previous versions: v3 [August 2019]
v2 [July 2019]
v1 [July 2019]
Downloaded:312 times


[ edit this article | back to article list ]