Different questions warrant different answers. For example, "Who came to the party?" seems to want an exhaustive list of party attendees, while "How can I get to Central Park?" seems to want a single answer, despite there being many possible ways to get to Central Park. Why does the first question want an exhaustive answer, and the second want a non-exhaustive one? In this work, we present data from a corpus search of naturalistic speech examining the distribution and co-occurrence of cues to (non-)exhaustivity present in the linguistic form of a question. This work builds off of Moyer & Syrett, in press.
Corpus data is important for at least two reasons:
-
From an acquisition perspective, corpus data provides an understanding of what kind of linguistic structure is present in the input to the language learner (cf. Syrett 2007, Dudley 2017). By taking the snapshot of the natural speech that adults produce, we can understand what children who overhear them might be exposed to.
-
From the processing perspective, cues are probabilistically linked to interpretation. By quantifying the distribution of a cue and its co-occurrence with other cues, we can understand how robust certain interpretations of a speaker's meaning are. (Degen & Tanenhaus 2015, 2016, 2018; Degen 2015; Elman, Hare, McRae 2004; MacDonald, Pearlmutter & Seidenberg 1994; Seigenberg & MacDonald 1999; Tanenhaus & Truswell 1995; McRae & Matsuki 2004, a.o.)
Our work helps to identify cues in raw data by tagging for clauseType, questType, embedded verb, and matrix verb.