Why subject relatives prevail: Constraints versus constructional licensing
Language and Cognition
Relative clauses containing subject relative-pronouns (e.g. that go to Utah all the time) are the prevalent type both across languages (Keenan and Comrie 1977) and in conversation, accounting for 65% of relative clauses in the American National Corpus (Reali and Christiansen 2007) and 67% of relative clauses in the corpus examined for this study, the Switchboard corpus. This fact appears attributable to parsing preferences, as per Hawkins (1999 Hawkins ( , 2004 , Gibson (1998) and Gibson et al.
... ) and Gibson et al. (2005) : subject extractions are the most local filler-gap dependency and therefore impose the lowest burden on short-term memory. This explanation, however, not only lacks strong psycholinguistic support but also fails to explain a major pattern in Switchboard: subject relatives are not preferred across the board but only as modifiers of postverbal (object and oblique) nominals. We propose that the preference for subject relatives is an effect not of general-purpose interpretive or encoding constraints but rather of constructional licensing: the subject relative belongs to an entrenched syntactic routine, the Presentational Relative construction, e.g. I have friends that clip articles (McCawley 1981; Lambrecht 1987 Lambrecht , 1988 Lambrecht , 2002 . We investigate this hypothesis by examining the formal, semantic and pragmatic