Authors: Hyosik Kim, Ming Xiang and Masaya Yoshida
Contact Email(s): hyosik87@gmail.com, mxming@gmail.com, m-yoshida@northwestern.edu
3 thoughts on “The Structure of Antecedent Influences Processing of Ellipsis”
Leave a Comment or Question Below
You must be logged in to post a comment.
Really interesting work! Thanks for sharing. I’m really excited by the question of how complexity effects differentially impact referential versus syntactic dependencies.
I wondered how you thought about the CP > NP intervenor complexity effect in the context of sluices, though. If the CP > NP difference is because of decay of the filler over a longer distance in the NP condition (since it doesn’t have the intermediate trace), then I wondered how we should think about this penalty in the sluice context? Do you think that the relevant set of processes ( encode – maintain – retrieve filler) are engaged in the same way at the point of processing the sluice?
Thank you so much for your interest and your question! I think that is exactly what causes the penalty of the ellipsis site. It looks like a complex set of processes for a single word, but I think that the processing of the ellipsis site should be very much like that of antecedent clause. I personally think this is why the backward sluicing experiment will be very interesting. I did not talk about it in detail during my presentation, but I think the backward sluicing experiment is exactly tackles that question. In the backward sluicing, if the parser builds a structure at the ellipsis site, the structure will contain the follow up input incrementally as the source of the ellipsis site and in that case the sluicing remnant will serve as a filler. Then the processing will be easier if there is a CP boundary. In other words, we would expect to see the complexity effect (or the decay effect) in both (i) and (ii) because (ii) will be the representation of the ellipsis site of (i). If this is confirmed, then I think it strongly show the decay of the filler effect in sluicing.
(i) I don’t know which manager, but [the consultant claimed that/the consultant’s claim about] the new proposal had pleased one of the managers…
(ii) I don’t know which manager [the consultant claimed that/the consultant’s claim about] the new proposal had pleased.
Recently, me and my colleagues ran a backward sluicing experiment with semantic congruency manipulation. We found that the reading time was slower at the spillover region (after ‘expand’) in both (iii) and (iv) and I think it is very suggestive that the sluicing remnant served as a filler and retrieved at the point of the verb.
So, I highly expect to see the complexity effect in backward sluicing as well.
(iii) The mayor couldn’t remember [which donor/which department], but the proposal would [expand] the town hall for a rich donor.
(iv) The mayor couldn’t remember [which donor/which department] the proposal would [expand] the town hall for a rich donor.
*(iv) The mayor couldn’t remember [which donor/which department] the proposal would [expand] the town hall for.