Authors: Niki Koesterich, Maayan Keshev, Daria Shamai and Aya Meltzer-Asscher
Contact Email(s): nikik@mail.tau.ac.il, miyan.keshev@gmail.com, dariayoko@gmail.com, ameltzer@tauex.tau.ac.il
6 thoughts on “Interference in the comprehension of filler-gap and filler-resumptive dependencies”
Leave a Comment or Question Below
You must be logged in to post a comment.
Very interesting talk, thank you! Do you think that retrieval interference and encoding interference are mechanistically separable, or are they mechanistically the same? For instance, would you expect unique individual differences to predict susceptibility to retrieval or encoding interference?
Secondary question which I believe was partially answered by your response to Maryellen: Are there preferential differences/bias for resumptive pronoun or gapped constructions in Hebrew? Could this preference, if it exists, underlie the differences that you see in retrieval and encoding interference?
Thanks for you’re interesting questions! On some frameworks encoding and retrieval interference arise from the same underlaying mechanisms (e.g. self-organizing sentence processing), and in others these are more distinct processes. We currently don’t have a string position as to how the interference arises and I’m not sure our current results present a case for either direction.
When dealing with individual differences, as your question suggests I think things are even more complicated. First, one would have to establish that interference varies reliably between individuals. Second, since even if you assume distinct parsing mechanisms I expect these rely on multiple overlapping capacities that differ between individuals. This could obscure individual differences as a measure. So overall, I think it’s a very interesting question we should pay attention to, personally I am not especially optimistic about tackling this with individual difference, but this could also be my lack of experience with such studies.
As for the preferences/biases for resumption/gap resolutions: Yes! Hebrew speakers usually prefer to have a gap whenever possible. This is why it was important to us to test obligatory resumption as well (Exp.1). So we interpret the striking alignment between interpretation patterns in Exp1-2 as evidence that the frequency of/bias against resumption in the given construction does not affect readers’ (in)ability to discriminate the filler from the distractor.
Too bad I did not think to mention this online! So thanks so much for asking that!
Terrific talk – thank you. It was great to see a non-difference between gap processing and RPs. It might especially make sense if you don’t think of RPs as retrieving anything … if the dependency is already in place before you get there.
Since you did an ROC analysis, have you considered modeling the variance? Do you think it could be revealing about whether people are failing to parse on some trials, or rather arriving at some sort of fuzzy representation?
And a materials question: were those Obj-Exp verbs? (Or was that just the English translation/example?) Interesting to ponder, since your distractor is also an attitude holder.
-matt
Thanks so much Matt!!
It might be interesting to note, in the context of your first comment that for the prepositional RPs (Exp. 1) the dependency might not be 100% set before you get to the RP, yet we still did not a difference in discriminability between these and the optional, direct object, RPs which were used in Exp.2
We have not considered modelling the variance, and are just starting our way with ROC analyses, but I’m very interested to hear what you have in mind!
Materials were not consistently Obj-Exp, this example is such in Hebrew too, but it is not a general property of the materials. Still, we will have another look at the materials to check what proportion of the items had these verbs in the RC and how many main verbs induces a perspective/attitude holder status for the distractor. Thanks!
This was so interesting! Are there other structures in Hebrew where we’d also expect encoding interference, which are not predicted by frequency, to tease these options apart?
We had a chat about this on gather.town but I’m putting a summary here in case other people will be interested in the answer:
Another environment which is usually used in interference studies is subject-verb agreement. It is probably possible to find environments in Hebrew with and without gender cues for that case too. Maybe there’s less frequency bias there.
But it is also important to note that even in the current study, it’s not clear that frequency should underlie the effect we observed. As far as I know, the production interference that has been suggested to cause people to switch to subject relatives is (a) associated with similarity between the filler and the embedded (RC) subject, while here we manipulate the similarity to the subject of the main clause. (b) encourages usage of passive SR in English, while these are not as productive in Hebrew (the corresponding strategy for Hebrew speakers might be production of object resumptive pronouns – Fadlon et al., 2019)