Authors: Zhenguang Cai, Nan Zhao and Martin Pickering
Contact Email(s): zhenguangcai@gmail.com, nanzhao@hkbu.edu.hk, martin.pickering@ed.ac.uk
6 thoughts on “How do people interpret implausible sentences?”
Leave a Comment or Question Below
You must be logged in to post a comment.
Hi, very interesting work! The Gibson-et-al PNAS paper also showed that the overall noise rate in the input influences the probability of misinterpretations, such as increasing the rate of ungrammatical sentences or implausible sentences in the fillers. Have you looked at such effects as well already? Perhaps adjusting the noise rate could also be a means to induce even stronger priming effects. Best, Michael
Thanks Michael – we have also read your related work.
No, we have not yet manipulated noise but it would be an interesting point to address using structural priming. Another point would be the difference between interpretation and plausibility-rating (as you have done).
Thank you for the interesting talk!
With the parsing example for what happens upon encountering an implausible argument, it seems like a similar process could happen with an unpredicted but plausible argument. For example, “give a dog ___” can predict “a bone”, but could plausibly be followed with “to the groomer”.
Participants could either fail to reanalyze the structure and could end up with an implausible DO structure (“give a dog the groomer”) or successfully reanalyze and end up with the PO structure, but either way, would the DO structure still be active and interfere with priming in a similar way as in the example for implausible arguments? If so, then should the priming be modulated by something like cloze probability of the second argument in addition to plausibility?
Ah, that’s a very interesting point! Thanks for that, Juliana.
From the top of my head, I think implausibility of the initial analysis triggers syntactic reanalysis. So in the case of “give the dog the groomer”, people may reanalyse it into “give the dog to the groom”. But how committed people are to these alternative analyses may depend on the semantic interpretations. So related to the later question, I think the probablity of a renanalysis would depend on both the cloze (or syntactic probability) and also the resulting plausibility of the reanlysed interpretation. Of course, these are empirical questions that can be tested. Thanks for these points.
This was super neat, I think there’s a lot of converging evidence with work by Bob Slevc as well. I’m wondering if the manipulation here is specific to English, or whether similar non-wellformed priming effects would be expected in languages where there is more extensive case marking. So, for example, in some languages words and articles are marked for dative vs. accusative case. Typically these languages also have flexible word order, which makes them interesting for testing how much this is driven by reinterpretation rather than assuming a word is simply missing. Testing in languages with redundant case marking and in languages with flexible word order might be a great way to see whether plausibility (e.g. animacy in this case which seems to be the main source of signal) is driving the effects you see or simply “filling in the gaps” with more plausible interpretations (e.g. Levy 2008 Proceedings of ACL).
Thanks Cassandra, for the comment. Yes, Bob Slevs has done some very related works. I think the case-marked language thing would be interesting. I’d say that case-markings will increase reliance on the original syntax (and hence the literal interpretation), but how reanalysis would occur remains to be tested – I am not familiar with case markings though. I’d suppose similar reanalysis would occur, though probably to a lesser degree than in English.