21 thoughts on “Laura Michaelis (University of Colorado Boulder): Null Instantiation in Sign-Based Construction Grammar (Rui Chaves, Paul Kay & Laura Michaelis)”
Nice talk! Thank you, Laura. (and Paul and Rui) Looking forward to the Q&A.
?Thank you, Stefan. I don’t know how important this is but I messed up the timezone in my slide at the end announcing the Q&A. I thought it was 1800h UTC but it’s 1800 CEST (UTC +2). I hope I didn’t confuse anyone. The calendar time adjust works great by the way.
Thank you for the talk!
I am puzzled why the context requirements are placed into the description of the daughter rather than at the mother. I assume they will be inherited by the mother, so maybe this is not really an issue. I just wonder what your motivation is for putting them into the daughter.
Manfred, as you know PK and RC replied by email;( I’m copying their responses here in case anyone wants to see. Thank you for the question.
Rui: Yours is a very good question. Either the features in question (CNTX) is underspecified lexically and we assume that certain constructions would instantiate it as indicated, or they are not underspecified, in which case they would have to be instantiated in the MTR instead, as you suggest. I’m not sure what is the most pragmatically appropriate decision here, but either one sounds plausible to me (as you note). If the latter, we could assume that general CNTX inheritance constraints would ensure that the DTRSs information is registered in the MTR. Thanks for spotting this! We need to fix it.
Paul: As you point out, the CNTXT value in the case at hand is the same in DTR and MTR, so where it’s spelled out versus where it’s just a tag doesn’t matter. I fear my answer, therefore, can be nothing more
than autobiographical, and so of no general interest. I guess I think of a lexical rule [MTR A, DTR B] as ‘converting’ B to A. Thus, in this inapt,
but hopefully innocent, image, B exists prior to A and so, if there is a constraint to be encountered in the context, B will be the first to
encounter it. Of course this is foolish talk in a constraint-based context, but *mea culpa.*
Thank you, Stefan and Bob. We look forward as well to the Q&A. Paul just pointed something interesting out to me: the example ‘I should not have contributed’, which was intended to show that special properties of INI arguments prevents them from having wide scope over negation or a modal, doesn’t illustrate that point very well. The point was supposed to be that this sentence couldn’t mean ‘There’s some amount of money that I should not have contributed’, but the reason might be that the ‘contribution’ argument of contribute can’t have wide scope because its referent is not ‘given’ enough to receive the contrastive reading it would have in the wide scope existential reading. I might have to use a DNI verb, like win, to make this point.
Thank you for the very interesting talk! I am very interested in null realization of subjects and complements. I like the idea of lexical and context co-lisencing. Of course, I worked more in the frame/valency perspective using the interaction between ARG-ST and VAL (also taking into account Pustejovsky’s argument typology).
Did you consider the idea of having different lemmas for different realizations instead of one lemma with different realizations? At least for some lemmas.
This is related to the question how complex does the VAR feature become? At what cost?
Thanks, Petya. Please give me a reference to your work. I would say our account too is centrally based on the interaction between ARG-ST and VAL, but the picture gets clouded somewhat because we don’t prevent every sign of the type covert-sign from appearing in VAL. It’s a very interesting question about different lemmas. In some cases we’re having a hard time deciding whether the lexical relations that the constructions are licensing are word-word, lexeme-lexeme or lexeme-word relations. For example, is the generic-habitual version of ‘arrest’ in ‘The police seldom arrest ø without probable cause’ a present-tense word or a habitual-generic lexeme that happens to be inflected for present tense? Regarding the VAR feature, I think it’s the INDEX feature that becomes more complex on our account, since we propose a taxonomy of indices. But maybe your could elaborate regarding VAR? Thank you for your comments!
Hi Laura! Thanks a lot for the reply! Very useful to me (as well as the presentation – I should read more on this in your and your co-authors’ publications).
I like the idea of having the covert-sign in surface realization, because it is still meaningful.
Recently I have been working on subject realization (including null and weird subjects) and optionality of arguments(complements) in Bulgarian. This was part of a Bulgarian-Russian project, so unfortunately my publications are in Bulgarian or Russian (http://62.44.116.41:8080/exist/apps/evogramm/pages/publ.html#PO)
Yes, I meant exactly this example with ‘arrest’. This affects also the typology of subject-less sentences that are of interest to me like general ones, indefinite ones, impersonal ones, etc. Where is the boundary between a lexeme and its variant…
In Bulgarian I assume that VAR will be even more dense because of the fact that it is a null-subject language. Your explanation is enough – thanks!
During the years I worked mostly on the nominal group in Bulgarian in relation to HPSG. However, since I am in a Slavic Department, so I had to write my books in Bulgarian: on the NP phrase where I am also interested in the valency of the deverbal and relational nouns plus a small part on binding (which I intendedto translate in english and get feedback from the community); and on Grammatical modeling of Bulgarian, where I share my experience on developing a Bulgarian Resource Grammar with a valency lexicon, and on Universal Dependencies experience on Bulgarian language model.
I really enjoyed your talk. Thanks!
There’s one context which you didn’t elaborate on, but which, for some odd reason, I found myself working on, and that is sportscasting. What examples do you have in mind? Any references?
Thanks, Nurit. Unsurprisingly, perhaps, Josef is the match report maven. We wrote about them in our 2010 C&F paper on genre-based NI. The implicit argument is the ball, which is DNI even though not topical (the players are topical).
He hammered Ø wide of Gary Walsh’s exposed net.
Paramatti put the ball back into the box and Panadic headed Ø into the net.
Hi Guy, yes it is salient and that is confusing. The point was only that the accessibility DNI affordance is not freely available but sanctioned by a fairly contextually constrained construction. We just don’t yet understand the use conditions on the accessibility construction! Following on something Oana David said in her thesis, I believe the use of accessibility-based DNI is constrained aspectually and by genre. In the narrative fiction genre, and in particular in event-sequence clauses (I grabbed and pulled ø with all my might) accessibility DNI is quite felicitous. Perhaps not otherwise. There are also idioms like Gimme! and the kid formula ‘Can I see?’. But the plot thickens: I note I just used SEE yesterday with DNI in this comment to Manfred:
Manfred, as you know PK and RC replied by email;( I’m copying their responses here in case anyone wants to see
SEE is not ordinarily a DNI verb, at least not outside the playground context I mentioned (Can I see?).
Thanks, interactions with aspect and genre would certainly make it more complicated!
Perhaps the question under discussion is relevant? In the conversation on slide 18, the question is whether the child pulled the fire-alarm box. But if it’s accepted that the child touched the fire-alarm box and the question is specifically whether the child *pulled* it, I think it sounds more natural:
A: Had your little girl pulled this fire-alarm box that you know of?
B: No, sir. I admit she had pushed it, but pushing it would not set off the alarm, and nobody had see her pull it.
A: And they suspected she pulled it?
B: She was touching the box, but she was pushing, not pulling.
A: And the child could have pulled, not pushed?
B: I was watching the whole time, and I only saw her push.
I would be interested to know your thoughts on the value of ACCESS (Slide 20). How could this “x” look like in order to indicate that the threshold is crossed? Or at least: In what direction should we look for in order to find an answer to this question?
Thank you, Elodie! We do not yet understand the interacting contextual conditions on accessibility-based DNI. It appears that having an overt antecedent in an immediately preceding clause improves felicity, as does visual presence of the entity referred to. In addition, in narrative event sequences it’s highly felicitous, but perhaps only with verbs of use and object manipulation (Their list was in front of me, and I scanned ø quickly hoping something would jump out and grab me). But what we had in mind with the accessibility threshold is Gundel Givenness. If a referent is not merely mutually identifiable but also highly salient or ‘in focus’ it seems to improve felicity of DNI omission, e.g., [looking at fire alarm box] I just pulled ø and nothing happened.
I’m wondering about the scope of null instantiation. I understand that unbounded dependency gaps are not an example of null instantiation. What about the null subjects in null subject languages and the nominal arguments realised as clitics? Are they a case of null instantiation? I ask mainly because I have argued that null arguments in Welsh (associated with either agreement morphology or a clitic) need to be realised as empty elements in the syntax. (I have argued the same for gaps.) So it would be nice from my point of view if these things fell outside the scope of null instantiation. (I assume Welsh has other kinds of null instantiation but it is not something I have looked into.)
Nice talk! Thank you, Laura. (and Paul and Rui) Looking forward to the Q&A.
?Thank you, Stefan. I don’t know how important this is but I messed up the timezone in my slide at the end announcing the Q&A. I thought it was 1800h UTC but it’s 1800 CEST (UTC +2). I hope I didn’t confuse anyone. The calendar time adjust works great by the way.
A very interesting talk
Thank you on behalf of the band
Thank you for the talk!
I am puzzled why the context requirements are placed into the description of the daughter rather than at the mother. I assume they will be inherited by the mother, so maybe this is not really an issue. I just wonder what your motivation is for putting them into the daughter.
Manfred, as you know PK and RC replied by email;( I’m copying their responses here in case anyone wants to see. Thank you for the question.
Rui: Yours is a very good question. Either the features in question (CNTX) is underspecified lexically and we assume that certain constructions would instantiate it as indicated, or they are not underspecified, in which case they would have to be instantiated in the MTR instead, as you suggest. I’m not sure what is the most pragmatically appropriate decision here, but either one sounds plausible to me (as you note). If the latter, we could assume that general CNTX inheritance constraints would ensure that the DTRSs information is registered in the MTR. Thanks for spotting this! We need to fix it.
Paul: As you point out, the CNTXT value in the case at hand is the same in DTR and MTR, so where it’s spelled out versus where it’s just a tag doesn’t matter. I fear my answer, therefore, can be nothing more
than autobiographical, and so of no general interest. I guess I think of a lexical rule [MTR A, DTR B] as ‘converting’ B to A. Thus, in this inapt,
but hopefully innocent, image, B exists prior to A and so, if there is a constraint to be encountered in the context, B will be the first to
encounter it. Of course this is foolish talk in a constraint-based context, but *mea culpa.*
Thank you, Stefan and Bob. We look forward as well to the Q&A. Paul just pointed something interesting out to me: the example ‘I should not have contributed’, which was intended to show that special properties of INI arguments prevents them from having wide scope over negation or a modal, doesn’t illustrate that point very well. The point was supposed to be that this sentence couldn’t mean ‘There’s some amount of money that I should not have contributed’, but the reason might be that the ‘contribution’ argument of contribute can’t have wide scope because its referent is not ‘given’ enough to receive the contrastive reading it would have in the wide scope existential reading. I might have to use a DNI verb, like win, to make this point.
???
Thank you, Laura!
Thank YOU, Olga!
Thank you for the very interesting talk! I am very interested in null realization of subjects and complements. I like the idea of lexical and context co-lisencing. Of course, I worked more in the frame/valency perspective using the interaction between ARG-ST and VAL (also taking into account Pustejovsky’s argument typology).
Did you consider the idea of having different lemmas for different realizations instead of one lemma with different realizations? At least for some lemmas.
This is related to the question how complex does the VAR feature become? At what cost?
Thanks, Petya. Please give me a reference to your work. I would say our account too is centrally based on the interaction between ARG-ST and VAL, but the picture gets clouded somewhat because we don’t prevent every sign of the type covert-sign from appearing in VAL. It’s a very interesting question about different lemmas. In some cases we’re having a hard time deciding whether the lexical relations that the constructions are licensing are word-word, lexeme-lexeme or lexeme-word relations. For example, is the generic-habitual version of ‘arrest’ in ‘The police seldom arrest ø without probable cause’ a present-tense word or a habitual-generic lexeme that happens to be inflected for present tense? Regarding the VAR feature, I think it’s the INDEX feature that becomes more complex on our account, since we propose a taxonomy of indices. But maybe your could elaborate regarding VAR? Thank you for your comments!
Hi Laura! Thanks a lot for the reply! Very useful to me (as well as the presentation – I should read more on this in your and your co-authors’ publications).
I like the idea of having the covert-sign in surface realization, because it is still meaningful.
Recently I have been working on subject realization (including null and weird subjects) and optionality of arguments(complements) in Bulgarian. This was part of a Bulgarian-Russian project, so unfortunately my publications are in Bulgarian or Russian (http://62.44.116.41:8080/exist/apps/evogramm/pages/publ.html#PO)
Yes, I meant exactly this example with ‘arrest’. This affects also the typology of subject-less sentences that are of interest to me like general ones, indefinite ones, impersonal ones, etc. Where is the boundary between a lexeme and its variant…
In Bulgarian I assume that VAR will be even more dense because of the fact that it is a null-subject language. Your explanation is enough – thanks!
During the years I worked mostly on the nominal group in Bulgarian in relation to HPSG. However, since I am in a Slavic Department, so I had to write my books in Bulgarian: on the NP phrase where I am also interested in the valency of the deverbal and relational nouns plus a small part on binding (which I intendedto translate in english and get feedback from the community); and on Grammatical modeling of Bulgarian, where I share my experience on developing a Bulgarian Resource Grammar with a valency lexicon, and on Universal Dependencies experience on Bulgarian language model.
As somewhat related but not pure HPSG references in English I can mention these:
about the Valency Lexicon (https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.681.736&rep=rep1&type=pdf); about NN groups in Bulgarian in HPSG: https://benjamins.com/catalog/slcs.158.06ose
Let us keep in touch! 🙂
Let’s keep in touch, thank you!
I really enjoyed your talk. Thanks!
There’s one context which you didn’t elaborate on, but which, for some odd reason, I found myself working on, and that is sportscasting. What examples do you have in mind? Any references?
Thanks, Nurit. Unsurprisingly, perhaps, Josef is the match report maven. We wrote about them in our 2010 C&F paper on genre-based NI. The implicit argument is the ball, which is DNI even though not topical (the players are topical).
He hammered Ø wide of Gary Walsh’s exposed net.
Paramatti put the ball back into the box and Panadic headed Ø into the net.
I’m a little confused between slides 18 and 19. Isn’t the fire-alarm box salient here?
Hi Guy, yes it is salient and that is confusing. The point was only that the accessibility DNI affordance is not freely available but sanctioned by a fairly contextually constrained construction. We just don’t yet understand the use conditions on the accessibility construction! Following on something Oana David said in her thesis, I believe the use of accessibility-based DNI is constrained aspectually and by genre. In the narrative fiction genre, and in particular in event-sequence clauses (I grabbed and pulled ø with all my might) accessibility DNI is quite felicitous. Perhaps not otherwise. There are also idioms like Gimme! and the kid formula ‘Can I see?’. But the plot thickens: I note I just used SEE yesterday with DNI in this comment to Manfred:
Manfred, as you know PK and RC replied by email;( I’m copying their responses here in case anyone wants to see
SEE is not ordinarily a DNI verb, at least not outside the playground context I mentioned (Can I see?).
Thanks, interactions with aspect and genre would certainly make it more complicated!
Perhaps the question under discussion is relevant? In the conversation on slide 18, the question is whether the child pulled the fire-alarm box. But if it’s accepted that the child touched the fire-alarm box and the question is specifically whether the child *pulled* it, I think it sounds more natural:
A: Had your little girl pulled this fire-alarm box that you know of?
B: No, sir. I admit she had pushed it, but pushing it would not set off the alarm, and nobody had see her pull it.
A: And they suspected she pulled it?
B: She was touching the box, but she was pushing, not pulling.
A: And the child could have pulled, not pushed?
B: I was watching the whole time, and I only saw her push.
Thank you Laura, this is a great talk !
I would be interested to know your thoughts on the value of ACCESS (Slide 20). How could this “x” look like in order to indicate that the threshold is crossed? Or at least: In what direction should we look for in order to find an answer to this question?
Thank you, Elodie! We do not yet understand the interacting contextual conditions on accessibility-based DNI. It appears that having an overt antecedent in an immediately preceding clause improves felicity, as does visual presence of the entity referred to. In addition, in narrative event sequences it’s highly felicitous, but perhaps only with verbs of use and object manipulation (Their list was in front of me, and I scanned ø quickly hoping something would jump out and grab me). But what we had in mind with the accessibility threshold is Gundel Givenness. If a referent is not merely mutually identifiable but also highly salient or ‘in focus’ it seems to improve felicity of DNI omission, e.g., [looking at fire alarm box] I just pulled ø and nothing happened.
Very interestinng talk.
I’m wondering about the scope of null instantiation. I understand that unbounded dependency gaps are not an example of null instantiation. What about the null subjects in null subject languages and the nominal arguments realised as clitics? Are they a case of null instantiation? I ask mainly because I have argued that null arguments in Welsh (associated with either agreement morphology or a clitic) need to be realised as empty elements in the syntax. (I have argued the same for gaps.) So it would be nice from my point of view if these things fell outside the scope of null instantiation. (I assume Welsh has other kinds of null instantiation but it is not something I have looked into.)