Authors: Rebecca Tollan and Bilge Palaz
Contact Email(s): rtollan@udel.edu, bilgeplz@udel.edu
9 thoughts on “The dual nature of subjecthood: Unifying subject islands and that-trace effects”
Leave a Comment or Question Below
You must be logged in to post a comment.
Thanks for an interesting talk! Two questions:
(1) I know you talked about this in the talk, but could you clarify why “who likes the dog” is acceptable?
(2) Why is removing “that” in the object extraction sentences improve the acceptability only by a tiny amount, while removing “that” in the subject extraction sentences improves the acceptability dramatically?
Thanks!
Shota
Hi Shota, thanks for your questions!
For (1), a subject wh question like “who likes the dog?” does not involve extraction out of a discourse feature-marked constituent (i.e., it’s extracted out of a matrix clause, and not an NP or that-CP), so a combinatorial discourse role can straightforwardly be formed between the filler and gap site.
For (2), including “that” drops acceptability much more so for subject wh than for object wh in that-clauses, due to the degree of the clash between a focused filler and gap site: extracting out of a that-clause violates CDRC, so the degree of clash between filler and gap is now a critical factor. In an object that-trace qu, the “clash” is between focused filler and focus (i.e., object) gap site – so, no clash! Thus, the mild degradation of the object that-t construction is due only to violation of CDRC. But for a subject, on top of violation of CDRC, the there is an clash between focused filler and topicalized gap site, so the ratings for subject that-t decrease much more substantially.
Thanks for your reply, Rebecca! I guess I am struggling to understand how those two answers fit together. On one hand, you seem to be suggesting that extracting a subject is fine on its own in your answer to (1), while in your answer to (2), you seem to be suggesting that extracting a subject has an additional penalty over and above the penalty of extracting out of that-clause. Am I missing something?
Maybe what I am not clear on is why the discourse-role conflict effect occurs only when there is an extraction out of a discourse feature-marked constituent?
Hi Shota,
under our hypothesis, extraction of a subject is problematic only when the subject (gap) is situated within a constituent already bearing a discourse role (in English: a that-clause but not a null clause), because the dependency chain then cannot be assigned a single discourse function (if the constituent in which the gap is located has a discourse function of its own already, the chain would be associated with 2 discourse roles instead of 1). Once this fails, the degree of clash between filler and gap is then at issue, and wh Q filler + subject gap gives rise to the biggest clash.
“under our hypothesis, extraction of a subject is problematic only when the subject (gap) is situated within a constituent already bearing a discourse role”
I guess my question is where this hypothesis comes from?
I mean, I think I am not clear on why the clash is conditionalized on another clash?
This was a very interesting talk! I’m wondering about the sentences with a null comp in experiment 3. Why do you think the ratings are equally high in RCs and questions when the complementizer is null? We might have expected a difference between questions and RCs even for the null comp condition, since these sentences involve extraction of a subject. Is the topicality of the subject reduced when the comp is missing?
Thank you Elaine! We found a difference between RCs and wh Qs for the null comp condition too: null comp RCs were rated lower than null comp wh Qs in our experiment (plus in one of the 2 replication studies; in the other replication study, null comp wh Qs were rated lower than null comp RCs – we think that this was due to the fact that we embedded the wh Qs for this replication round, which caused the ratings for them to drop). Regarding the topicality of the matrix/embedded subject, there’s a corpus study by Underhill (1988) which concludes that the matrix subject is more topic-like in a that-comp clause, whereas the embedded subject is more topic-like in a null-comp clause.