Ellipsis alternation: Analyzing syntactic variation
Joanna Nykiel, University of Silesia in Katowice
At the most general level, this presentation engages the question of which areas of syntax reflect not grammatical principles but other factors such as processing preferences (see Miller & Chomsky1963, Hofmeister & Sag 2010, Sprouse et al. 2012, Chaves 2013, Culicover 2013, Hofmeister et al. 2015). Culicover (2013), for instance, argues that if a structure that is licensed by the grammar is degraded or infrequent in certain contexts, then these facts should be explained by extragrammatical factors. Syntactic variation thus provides a likely locus of extra grammatical factors.
I address one instance of syntactic variation, an alternation characteristic of elliptical constructions, where Ps appearing in antecedent clauses may (1a) or may not be repeated (1b) in stranded phrases (remnants).
(1) a. A: Go change it.
B: To what?
A: To one of your old costumes.
b. A: Go change it.
B: To what?
A: One of your old costumes.
While available in a wide range of languages, this alternation is crosslinguistically unusual in English in that it shows a higher frequency of remnants without Ps than remnants with Ps (see Rodrigues et al. 2009, Merchant et al. 2013, Nykiel 2013, 2015, 2016, Abels 2016). I offer evidence from corpora of spoken US English that the frequency of remnants with Ps compared to the frequency of remnants without Ps is determined by processing preferences. I use a mixed-effects logistic regression model and conditional inference trees to analyze the data. The results speak in favor of analyzing this alternation as always grammatical (see Sag & Nykiel 2011, Boškovic 2013; for an alternative perspective, see Merchant 2001, 2004) but subject to processing constraints that affect the frequency of remnants, both in English and crosslinguistically.