Authors: Suet Ying Lam and Heeju Hwang
Contact Email(s): lsy317@connect.hku.hk, heeju@hku.hk
7 thoughts on “Interpretation of null pronouns in Mandarin Chinese does not follow a Bayesian model”
Leave a Comment or Question Below
You must be logged in to post a comment.
Very interesting work, thank you so much!
I am curious about the strong subject bias of null pronouns in the experiments. I wonder if this is also because of the verbs wants to/think?
As a speaker of a null pronoun language, I am much more likely to interpret the null pronoun in these constructions as referring to the subject, because of wants to/think, than I would in a sentence like ‘Megan laughed at Ethan because __ is funny/mean/etc. ‘
In a nutshell: I am wondering whether the production bias of null pronouns measured in the free prompts is a reliable indicator of null pronoun N1 orientation in the ” __ wants to / ___ thinks” paradigm.
Thanks again!
Thank you for your question!
A brief answer: My personal opinion is that the free prompt is a reliable indicator of null pronoun N1 orientation in the ” __ wants to / ___ thinks” paradigm. Of course, giving a verb here does limit people’s continuations to some extent. But I checked the data in free prompts: People seem to continue more with a verb phrase when using a null pronoun. The structure you mentioned is rare (i.e., using a copula verb), so I do not think it will affect a lot. But we will seek a method to try to minimize this effect.
From a native speaker’s perspective, using a null pronoun to refer to an object (regardless of the syntactic structure, using this verb or not) makes me feel weird, especially in a written condition. For sure, this needs further study to prove. But that’s why I am not surprised about the extreme bias I have found.
Again, thanks for your question! That’s a really good point!
Session Question from Petra Schumacher: How did you assure that the null pronoun was produced? What other referential forms were generated?
Thanks for posting questions, Monica. And thanks for the question from Petra Schumacher!
I am a bit confused with the first question: do you mean how can I assure that null pronoun was produced in free prompts? It would be a bit difficult to explain because it is instinctive for me to identify it: basically, the null subject of a predicate will be interpreted as a null pronoun. Same in the pronoun prompt, the null subject of the verb “wants to/think” will automatically become a null pronoun for the participants. It is impossible for them to understand it as something else – must be one of the referents are “thinking/wanting to”. Across the 100 responses I collected, I did not see anyone misunderstood the null pronoun here.
For the second question: Overt pronouns and names were the majority.
A typo on the last sentence: across the 2000 (null pronoun prompt) responses…
Interesting study, great to see more data on this! I also wanted to know about the verbs wants to/think. I wondered if having these verbs in the (null) pronoun prompt condition, but not the free prompt condition, may make the model’s predictions less precise, because the interpretation data is biased by the verb.
I also wondered how many null pronouns were elicited?
And a final comment: it’s a real challenge to elicit completions for the null pronoun, because you need to indicate to participants that there is something there (hence the verb). I wonder if you could try doing this with priming instead. In our study we have a slightly different problem in the dieser prompt condition, which is that dieser can have an adnominal use e.g. dieser Mann “this man”, and this is probably more frequent than the pronominal use. We didn’t want to have to throw away a lot of data where dieser is used like this, so we primed participants by having several fillers at the start of the experiment which were unambiguously dieser used as a pronoun. It worked really well, I wonder if you could try something similar for null pronouns?
Thanks for the comment! The suggestion for priming is insightful.
Regarding your first question about the null pronoun prompt, as I have replied to Rodica above: I think there might be some effects, but not large enough to affect the prediction. But thanks for your comment, we will think about it.
For the null pronoun: The null pronoun occurs really rare, especially when using IC verbs: in experiment 1 where only overt pronoun prompt were used, no more than 10% of null pronouns were elicited (less than 5% in IC), and almost all of them were at the subject position. In a separate production study, we also found that null pronouns elicit very infrequently. In experiment 2, actually the null pronoun prompt we used is already a prime for the production of null pronoun: much more null pronouns were produced in experiment 2. In TOP verbs, it is around 20%.
And thanks again for your suggestion on using priming, that’s a pretty good idea, but it seems to be quite different in Mandarin. In Mandarin, null pronouns will not be used in forms like “this man”. If there is a null subject, it will be a null pronoun, especially there is a verb attached. But we will take a look at how priming might work on Mandarin null pronoun.