In this paper, we applied Bresnan et al.ís (2007) Generalized Linear Model approach to model the English dative alternation to a corpus that shows more variation in text genre and discourse type: the ICE-GB corpus.
In a direct comparison, using only variables currently available for both data sets, we are able to explain 90.8% of the variability in the ICE-GB data as compared to 94.5% in the Switchboard corpus, possibly showing that the variation in genre in ICE-GB decreases the predictive power of the model. As expected, both models showed that the theme is especially pronominal and the recipient is often indefinite and non-local (third person) in the NP-PP construction, while in the NP-NP construction, the theme is longer than the recipient, the recipient is pronominal and the theme is indefinite.
Next, we extended the model for ICE-GB by including a number of syntactic variables. Only word order had a significant effect. This observation and investigation of errors has led us to wonder whether the dative alternation should be modelled in the traditional fashion or perhaps (also) according to the order of the recipient and the theme.
Reference: Daphne Theijssen (2008). Using the ICE-GB Corpus to model the English dative alternation. Online Proceedings of the Aston Postgraduate Conference on Corpus Linguistics.
Paper (pdf; 75kB) ; BibTeX
back to publications