Dialects of English exhibit microvariation in a surprisingly niche corner of the grammar: the objects of certain prepositions may be optionally realized as a gap for speakers of British English (BrE) -- but not for speakers of e.g. American English (AmE) -- just as long the entire PP is embedded under a have- or with-possessive (cf. (3)).
1) a. Kim usually buys breadi with olives {in/on/…} iti. (BrE: ok; AmE: ok)
b. That breadi has olives {in/on/…} iti. (BrE: ok; AmE: ok)
adi with olives {in/on/…} __i. (BrE: ok; AmE: *)
b. That breadi has olives {in/on/…} __i. (BrE: ok; AmE: *)
di—there are olives {in/on/…} __i. (BrE: *; AmE: *)
b. *I won’t eat that breadi if you’ve put olives {in/on/…} __i. (BrE: *; AmE: *)
Previous work on these Prepositional-Object Gaps (POGs) has shown them to be further constrained in various ways that are poorly understood (Griffiths & Sailor 2015, Sailor & Griffiths 2017, Stockwell & Schütze 2019). For example, while it is clear that POGs are only licensed under have/with possessives (3), it is less clear exactly which prepositions tolerate POGs in their complements (and whether they form a natural class), given the confound that ground omission is possible in certain types of PPs across all dialects of English (4) (Svenonius 2010):
4) a. We stood on a bridgei. Below __i we could see barges laden with wine.
(BrE: ok; AmE: ok)
b. There’s a boxi on the table—go see what’s inside __i.
in and on (Stockwell & Schütze 2019). Following this and other empirical refinements revealed by the experiment, we revisit our prior analysis of POGs (Sailor & Griffiths 2017): namely, that they involve a kind of PP-restructuring licensed by the possessive superstructure, leading to A-movement of the ground into the possessor position.
