Annotated samples for: I don’t think or I think … not – what is the difference?

This dataset contains manually annotated random samples of the constructions I don’t think and I think … not, compiled for a study on neg-raising in English (Horn 1978; Nuyts 1990; Pichler 2013). The dataset was created to investigate the distribution and interpretation of neg-raising constructions and their relationship to epistemic stance marking in spoken English. The data were extracted from the British Spoken subcorpus of Collins WordbanksOnline, which consists of transcripts of spontaneous conversations and broadcast speech produced between 1972 and 2009 (Collins 2009). Queries were designed to retrieve tokens of I don’t think and I think, after which occurrences of not in the complement clause were manually identified to obtain tokens of I think … not. To ensure comparability between the constructions, no tense, modal or other variation of the matrix predicate was included in the search queries. Randomized samples of the retrieved tokens were manually examined and filtered to exclude irrelevant cases. The remaining tokens were manually annotated for the presence or absence of neg-raising and for relevant contextual properties in the surrounding discourse. This dataset may be used to replicate or extend research on neg-raising and the various semantic functions of CTP-clauses in spoken English.