ABSTRACT
Models of language production must account for how speakers select referential expressions, a process that is complicated by the range of options available to speakers (e.g. he, Pim, Dr. Levelt). This paper reviews models of reference production, and suggests that they fall into two classes: (1) Pragmatic selection models, and (2) Rational models. We review evidence for and against each class of model, focusing on the use of pronouns. We argue that pronoun selection is not primarily driven by considerations of efficiency, and current models of accessibility do not explain the range of variation in production choices. We identify several challenges for future work in this area.
Acknowledgments
Thank you to the organisers of the International Workshop on Language Production (Antje Meyer, Ardi Roelofs, and Laurel Brehm) for the venue to discuss the ideas in this paper, and to the people who attended the conference for feedback.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 This paradigm has been used for multiple experiments. Two of the characters were named Lady Mannerly and Sir Barnes for Rosa and Arnold (Citation2017) and Zerkle et al. (Citation2015), but were renamed The Duchess and The Duke for other experiments (Langlois et al., Citation2018; Zerkle & Arnold, Citation2016; Zerkle et al., Citation2017; Zerkle & Arnold, Citation2019). All other characters were labelled with definite descriptions in all experiments.
2 There was an additional manipulation of audience knowledge, where the addressee could not hear the second sentence in Exp. 1, and could in Exp. 2. This did not affect the cognitive load effect.