ABSTRACT
In formulating a referring expression, speakers may choose between an explicit expression (such as a proper name or a noun phrase) or a reduced form such as a pronoun. We investigated whether speakers are influenced by their conversation partners to produce full noun phrases instead of pronouns and whether this differs depending on whether their partner was a native or a nonnative English speaker. Participants took turns describing and matching cartoons with a (confederate) partner, who used either full noun phrases or pronouns when referring to discourse entities. We found that participants were more explicit or less explicit in their own referring expressions depending on their partner’s behavior on the turn before and adapted to the same extent with native and nonnative partners. We conclude that speakers adapt their production of referring expressions based on what their partners say but do not make strategic adjustments based on who their partner is.
Acknowledgments
The authors would like to thank Steve Pitcher-Cumming for designing the stimuli used in the experiment.
Data availability
The data sets analyzed for the study along with the analysis script can be found in the Open Science Framework repository (https://osf.io/ekm8a/).
Notes
1. Participants occasionally dropped subject pronouns as an alternative to using a pronoun or full NP; hence, coding took into account all three categories of reference.
2. Participants’ preference for full NPs over pronouns may seem surprising given that they are somewhat unnatural for given/repeated entities, which feature heavily in the materials. One possibility is that the participant cartoon set, which always displayed the character’s name under each character’s appearance, may have encouraged participants to use proper names. Importantly, however, this property of the stimuli was consistent across conditions and, hence, would not have influenced participants’ pattern of adaptation in the experiment.
3. As the experiment utilized multiple confederates, we first ran a preliminary analysis to check whether participants’ behavior differed significantly across confederates. We modeled the outcome variable on the sole predictor of confederate ID (sum-coded) separately for the native-partner and the nonnative-partner confederate conditions. Models included by-participant random intercepts but no random slopes since confederate was manipulated between subjects. Neither model showed an effect of confederate ID on participants’ reference productions (all ), confirming that participants behaved similarly across the two confederates in both partner conditions.
4. Our analysis collapses pronouns and omissions into a single category; however, we note that an ordinal regression taking into account all three referring expression types as separate categories produces the same pattern of results.
5. Based on a reviewer’s suggestion that the influence of priming and/or audience design may have varied over the course of the interaction (e.g., strategic audience design effects may have been more prevalent at the start of the interaction), we conducted an additional analysis including experimental phase (first half of trials versus second half of trials; sum-coded) as a predictor. This model indicated no effect of phase nor its interaction with either prime type or confederate (all ), suggesting that any effects of priming and audience design remained stable over time.
6. We thank two anonymous reviewers for their comments on this point.