Effects of similarity on speakers’ production of referring expressions
Item statusRestricted Access
Embargo end date31/07/2022
When people talk, they often wish to refer to a discourse entity that has been previously mentioned. Nevertheless, they often have several choices about how to do so. This thesis aims to understand the choice of referring expressions in Mandarin, where people can repeat the noun phrase, use a pronoun or omit the referring expression, with a focus on the role of similarity between discourse entities. I first investigated the role of similarity by varying occupational similarity between the target referent and the competitor: similar (e.g., killer-murderer) versus dissimilar (e.g., killer-proofreader). In two experiments, I found that the effects of similarity were a function of production conditions. When speakers saw pictures and heard sentences mentioning both the target referent and a competitor, then described actions involving only the target referent (Experiment 3.1), they tended to produce more explicit referential expressions (repeated noun phrases) when the target referent and the competitor had a similar occupation than when they had a dissimilar occupation. In contrast, when speakers saw pictures and heard sentences mentioning only the target referent, then described actions involving both the target referent and the competitor (Experiment 3.2), they tended to produce less explicit referential expressions (omissions) when the two entities had a similar occupation than when they had a dissimilar occupation. I interpreted the contrasting effects in terms of different stages of the effects of similarity: discourse processing versus linguistic processing. I also provided additional analyses which examined the extent to which speakers’ choice of referring expressions in Experiments 3.1 and 3.2 was affected by phonological similarity between the target referent and the competitor. The results showed that phonological similarity was a factor influencing speakers’ reference production. I then looked at the effects of same versus different numerosity, namely the notional number of entities on speakers’ choice of referring expressions under the production condition as in Experiment 3.1. I found that again, speakers produced more explicit 3 referential expressions (repeated noun phrases) when the target referent and the competitor had the same number of entities (e.g., both were single) than when they had a different number of entities (e.g., one was single, but the other was multiple). These effects occurred when plurality was expressed via numerals (Experiment 4.1) or the suffix men attached to singular nouns (Experiment 4.2). The same pattern of effects was found even when I expressed all single/multiple entities using bare nouns without a numeral or a suffix men (Experiment 4.3), suggesting that the locus of the effects of similarity could not be at the phonological or morphological level where the entities’ phonological and morphological features are represented respectively. Instead, these results would be compatible with both a semantic and a syntactic locus of the effects of similarity. I then looked at the effects of same versus different gender on speakers' choice of referring expressions. Mandarin is a language of particular interest because pronouns distinguish gender only in writing, but not in speech. I first showed that gender was a reliable index of similarity in a rating pre-test (Experiment 5.1). I then found that the pattern of the effects of occupational similarity was replicated when the target referent and the competitor had a different gender (Experiment 5.2). In the combined analysis of Experiments 5.2 and 3.1, I found gender congruency between the target referent and the competitor affected the use of pronouns – a pattern that was different from what I have found before (Experiment 5.2). I attributed the findings to the communicative effort of avoiding referential ambiguity. This hypothesis was validated in Experiment 6.1, where a written production version of Experiment 5.2 was conducted. By combining Experiments 6.1 and 5.2, I found that the effects of similarity yielded different results for pronouns in spoken production (where using pronouns is ambiguous) and written production (where using pronouns is unambiguous), suggesting the effects of similarity are associated with ambiguity avoidance. Taken together, these studies have the potential to highlight the field of reference production. We conclude that (a) the effects of similarity on speakers' choice of referring expressions are robust, which are shown across different domains (occupation, numerosity, and gender), (b) the exact consequence is dependent on the production conditions the speaker is currently acting within and the similarity domains being manipulated, (c) the locus of the effects of similarity in reference production cannot be at the phonological or the morphological level - instead, they can arise at the level of semantic or syntactic processing, (d) the effects of similarity on speakers' choice of referring expressions reflect a communicative effort mechanism by which referential ambiguity tends to be avoided.