From mcginnis at ucalgary.ca Fri Jun 1 20:00:20 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Fri, 1 Jun 2001 14:00:20 -0600 Subject: A follow-up Message-ID: Dear DM-List, I just wanted to follow-up a bit on the reply by Martha to my query, and indeed it was very helpful! I found it interesting that you should mention Levelt and his massive 1989 monograph. Although not stated so, my question was an attempt to clarify the DM-position on lexical selection and its relation to what seems, to be its opposite, Levelt's 1989 "Lexical Hypothesis", which requires all structure to have a Lexeme as its source or instigator; that is, an Unaccusative structure could only be instigated by a prior lexical selection of an Unaccusative. The reason for Levelt and DM to be associated recently in my head, was the fact that an MIT publication co-edited by Marantz on Neuro-Linguistics contains an article by Levelt in which he reiterates his 1989 hypothesis. I was wondering if DM was entertaining any implementations of ideas in line with Levelt's, which would remedy the massive over-generation which seems unavoidable in a model with Post-syntactic selection of open-classes? In my acquaintance with the DM literature, although its position is very clear on late insertion of Vocab Items, early selection of Open-Classes has not been explicitly ruled out, at least to my knowledge. Is it the case that DM is willing to tolerate massive overgeneration or does it have a means to rule it out? I do realize that DM need not be a production model of language. Thanks in advance, Mark __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/ From mcginnis at ucalgary.ca Mon Jun 4 17:01:45 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Mon, 4 Jun 2001 11:01:45 -0600 Subject: Martha McGinnis: A follow-up (reply to Mark Volpe) In-Reply-To: Message-ID: Dear Mark, Again, I'm not sure that 'early insertion' of roots, as in GB, Minimalism, LFG, etc., addresses the issue you've raised. As far as I can see, the issue concerns the interface between the interpretive system and the grammar. You seem to be asking how a theory with post-syntactic interpretation (like GB) ensures that appropriate lexical items are chosen. In my understanding, DM doesn't insist that interpretation is purely post-syntactic. One possibility is that the entire derivation is accessible to the interpretive system -- from lexical selection and syntactic combination to Vocabulary insertion. I recall Alec Marantz suggesting in class lectures that "all choices are interpreted": the derivation is interpreted to the extent that it makes choices among possible options. Such choices occur at every stage of a derivation. Chomsky's recent work takes a similar line: he argues that the derivation is sent off to PF and LF in a series of subchunks ('phases'). I believe an 'interpret-all-choices' approach also bears some resemblance to Levelt's model, in which (as I recall) the conceptual/intentional system drives the selection of 'lemmas' from the lexicon, but is also informed (via feedback) of later stages of the syntactic derivation, including the selection of phonological 'lexemes'. Perhaps you can find an empirical argument that the under/overgeneration issue you raise is connected to the issue of early vs. late insertion of roots, but otherwise the two seem to be independent. One issue that does arise under late insertion is how to deal with the apparent suppletion of roots. For example, is 'people' a suppletive plural of 'person'? Does 'people' spell out a root or a functional node? If all roots compete for insertion, as DM assumes, then 'people' can't be a root specified as plural (nor 'person' a root specified as singular) because such roots would block the insertion of roots not specified as singular or plural, like 'human' or 'cat'. So under DM, either this is not true suppletion (it's more like the relation of 'cow' to 'cattle') or else 'person' can spell out a functional (non-root) node, which lacks Encyclopedic content. How could we distinguish the two possibilities empirically? This would be an interesting issue to explore -- I'm not sure how much has been written about it. Regards, Martha mcginnis at ucalgary.ca From mcginnis at ucalgary.ca Tue Jun 5 22:25:15 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 5 Jun 2001 16:25:15 -0600 Subject: Mark Volpe: Let me try your patience, one last time. (reply to Martha McGinnis) Message-ID: Dear DM-Listers, Forgive me, especially Martha, and thanks for your responses. I'm not at all advocating early insertion. I'm a strong supporter of DMs position that phonologic realization of morpho-syntactic features should be post-syntactic, moreover, that syntax should make use of only abstract features (w/1 possible exception). The premise I'm trying to explore is whether such a post-syntactic phonological realization of abstract features is compatible with an early *selection* of Open-Class items, i.e., Roots. Perhaps the early selection (but not insertion) of underived Roots would be able to insure that Unaccusative structures are built for Unaccusatives, etc. In terms of Levelt's 1989 model, it is as you've described, Martha,("intentional drives Lemma selection"), in his words: "a preverbal message is the first step...initiated by the conception of some communication intention"(107). This is followed by "formulation processes [which are] lexically-driven" (181). Assuming post-syntactic Vocabulary Insertion with a Syntax carried out using abstract morpho-syntactic features, would (does) DM advocate selection of Open-Classes, i.e., Roots to "mediate grammatical encoding" (1989:181) as advocated by Levelt, given a properly formulated mechanism, ? Also, in your previous responses, you've given several examples which raise the crucial issue of whether Paninian Principles might not be applicable to Root selection. This is something I would like to hear much more about, but for the mean time, I'm not arguing that early *insertion* is preferable to late insertion, but that early *selection* might be preferable late selection. The idea that late selection might lead to over-generation was suggested to me by something you wrote in a previous message: "any root Vocabulary item can be inserted into any root node, and if the result makes sense to the conceptual system, it converges; otherwise, it crashes. So 'I arrived the plane' is no good because the Encyclopedic content of 'arrive' is incompatible with a transitive structure". This, for me, indicates that an inordinate number of crashes may occur before a convergence. Let me reiterate my thanks for your patience in answering my questions (and for letting me take up a lot of time). Cheers,---Mark V., Stony Brook __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/ From mcginnis at ucalgary.ca Mon Jun 11 17:35:03 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Mon, 11 Jun 2001 11:35:03 -0600 Subject: Martha McGinnis: Early selection of roots (reply to Mark Volpe) In-Reply-To: Message-ID: I'll respond again to Mark's posting, but others should of course feel free to put in their two cents on these issues as well. >The premise I'm trying to explore is whether such a >post-syntactic phonological realization of abstract >features is compatible with an early *selection* of >Open-Class items, i.e., Roots. Perhaps the early >selection (but not insertion) of underived Roots would >be able to insure that Unaccusative structures are >built for Unaccusatives, etc. This seems like a logical possibility, but as far as I know it's not a route that has been taken so far by anyone adopting DM assumptions. As I think I said in earlier postings, the role of LF and the competition among roots for insertion seem to be a couple of the issues at stake, with empirical consequences you could explore. It seems to me that even if roots are only 'selected' early, Vocabulary insertion would work the same way for roots as it does for non-roots, i.e., competition among items specified for certain features that play a role in the syntax. This is a different approach to root insertion than the one taken in the DM literature so far, so I do think the same empirical issues arise. Or maybe you have some other ideas about how to test your hypothesis. Remember that DM is just a framework of assumptions for hypothesis-testing. Within this framework, it's been proposed that roots aren't distinguished from one another until Vocabulary insertion takes place, and that all roots compete with each other for insertion (see Marantz's 1997 PennWPL paper and '"Cat" as a phrasal idiom' ms., and the Harley & Noyer paper cited in my first posting). As far as I can see, there's no contradiction between adopting the assumptions of DM and discovering that this proposal is wrong. Regarding overgeneration, see my last posting about the possibility of (quasi-)continuous interpretation of grammatical choices. This might block the kind of overgeneration you describe; it all depends on the details. As ever, it seems reasonable to ignore the issue until someone finds evidence that bears on it. I'm not sure I've added anything to my earlier postings, but I hope this helps. Regards, Martha mcginnis at ucalgary.ca From mcginnis at ucalgary.ca Fri Jun 15 16:39:58 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Fri, 15 Jun 2001 10:39:58 -0600 Subject: Martha McGinnis: Number marking in Bayso Message-ID: A non-text attachment was scrubbed... Name: not available Type: text/enriched Size: 2323 bytes Desc: not available URL: From mcginnis at ucalgary.ca Thu Jun 21 17:12:41 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Thu, 21 Jun 2001 11:12:41 -0600 Subject: Martha McGinnis: Squib Message-ID: Dear DM-list, I've written a squib based on the interchange Heidi Harley and I had on this list back in February, with an additional section on aspectual compositionality in idioms. The upshot is that the facts support DM, rather than Jackendoff's or Nunberg et al.'s theories of idioms. If anyone would like to see the squib, it can be downloaded from my website: http://www.ucalgary.ca/~mcginnis/papers.html "On the systematic aspect of idioms" Regards, Martha mcginnis at ucalgary.ca From mcginnis at ucalgary.ca Fri Jun 1 20:00:20 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Fri, 1 Jun 2001 14:00:20 -0600 Subject: A follow-up Message-ID: Dear DM-List, I just wanted to follow-up a bit on the reply by Martha to my query, and indeed it was very helpful! I found it interesting that you should mention Levelt and his massive 1989 monograph. Although not stated so, my question was an attempt to clarify the DM-position on lexical selection and its relation to what seems, to be its opposite, Levelt's 1989 "Lexical Hypothesis", which requires all structure to have a Lexeme as its source or instigator; that is, an Unaccusative structure could only be instigated by a prior lexical selection of an Unaccusative. The reason for Levelt and DM to be associated recently in my head, was the fact that an MIT publication co-edited by Marantz on Neuro-Linguistics contains an article by Levelt in which he reiterates his 1989 hypothesis. I was wondering if DM was entertaining any implementations of ideas in line with Levelt's, which would remedy the massive over-generation which seems unavoidable in a model with Post-syntactic selection of open-classes? In my acquaintance with the DM literature, although its position is very clear on late insertion of Vocab Items, early selection of Open-Classes has not been explicitly ruled out, at least to my knowledge. Is it the case that DM is willing to tolerate massive overgeneration or does it have a means to rule it out? I do realize that DM need not be a production model of language. Thanks in advance, Mark __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/ From mcginnis at ucalgary.ca Mon Jun 4 17:01:45 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Mon, 4 Jun 2001 11:01:45 -0600 Subject: Martha McGinnis: A follow-up (reply to Mark Volpe) In-Reply-To: Message-ID: Dear Mark, Again, I'm not sure that 'early insertion' of roots, as in GB, Minimalism, LFG, etc., addresses the issue you've raised. As far as I can see, the issue concerns the interface between the interpretive system and the grammar. You seem to be asking how a theory with post-syntactic interpretation (like GB) ensures that appropriate lexical items are chosen. In my understanding, DM doesn't insist that interpretation is purely post-syntactic. One possibility is that the entire derivation is accessible to the interpretive system -- from lexical selection and syntactic combination to Vocabulary insertion. I recall Alec Marantz suggesting in class lectures that "all choices are interpreted": the derivation is interpreted to the extent that it makes choices among possible options. Such choices occur at every stage of a derivation. Chomsky's recent work takes a similar line: he argues that the derivation is sent off to PF and LF in a series of subchunks ('phases'). I believe an 'interpret-all-choices' approach also bears some resemblance to Levelt's model, in which (as I recall) the conceptual/intentional system drives the selection of 'lemmas' from the lexicon, but is also informed (via feedback) of later stages of the syntactic derivation, including the selection of phonological 'lexemes'. Perhaps you can find an empirical argument that the under/overgeneration issue you raise is connected to the issue of early vs. late insertion of roots, but otherwise the two seem to be independent. One issue that does arise under late insertion is how to deal with the apparent suppletion of roots. For example, is 'people' a suppletive plural of 'person'? Does 'people' spell out a root or a functional node? If all roots compete for insertion, as DM assumes, then 'people' can't be a root specified as plural (nor 'person' a root specified as singular) because such roots would block the insertion of roots not specified as singular or plural, like 'human' or 'cat'. So under DM, either this is not true suppletion (it's more like the relation of 'cow' to 'cattle') or else 'person' can spell out a functional (non-root) node, which lacks Encyclopedic content. How could we distinguish the two possibilities empirically? This would be an interesting issue to explore -- I'm not sure how much has been written about it. Regards, Martha mcginnis at ucalgary.ca From mcginnis at ucalgary.ca Tue Jun 5 22:25:15 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Tue, 5 Jun 2001 16:25:15 -0600 Subject: Mark Volpe: Let me try your patience, one last time. (reply to Martha McGinnis) Message-ID: Dear DM-Listers, Forgive me, especially Martha, and thanks for your responses. I'm not at all advocating early insertion. I'm a strong supporter of DMs position that phonologic realization of morpho-syntactic features should be post-syntactic, moreover, that syntax should make use of only abstract features (w/1 possible exception). The premise I'm trying to explore is whether such a post-syntactic phonological realization of abstract features is compatible with an early *selection* of Open-Class items, i.e., Roots. Perhaps the early selection (but not insertion) of underived Roots would be able to insure that Unaccusative structures are built for Unaccusatives, etc. In terms of Levelt's 1989 model, it is as you've described, Martha,("intentional drives Lemma selection"), in his words: "a preverbal message is the first step...initiated by the conception of some communication intention"(107). This is followed by "formulation processes [which are] lexically-driven" (181). Assuming post-syntactic Vocabulary Insertion with a Syntax carried out using abstract morpho-syntactic features, would (does) DM advocate selection of Open-Classes, i.e., Roots to "mediate grammatical encoding" (1989:181) as advocated by Levelt, given a properly formulated mechanism, ? Also, in your previous responses, you've given several examples which raise the crucial issue of whether Paninian Principles might not be applicable to Root selection. This is something I would like to hear much more about, but for the mean time, I'm not arguing that early *insertion* is preferable to late insertion, but that early *selection* might be preferable late selection. The idea that late selection might lead to over-generation was suggested to me by something you wrote in a previous message: "any root Vocabulary item can be inserted into any root node, and if the result makes sense to the conceptual system, it converges; otherwise, it crashes. So 'I arrived the plane' is no good because the Encyclopedic content of 'arrive' is incompatible with a transitive structure". This, for me, indicates that an inordinate number of crashes may occur before a convergence. Let me reiterate my thanks for your patience in answering my questions (and for letting me take up a lot of time). Cheers,---Mark V., Stony Brook __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/ From mcginnis at ucalgary.ca Mon Jun 11 17:35:03 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Mon, 11 Jun 2001 11:35:03 -0600 Subject: Martha McGinnis: Early selection of roots (reply to Mark Volpe) In-Reply-To: Message-ID: I'll respond again to Mark's posting, but others should of course feel free to put in their two cents on these issues as well. >The premise I'm trying to explore is whether such a >post-syntactic phonological realization of abstract >features is compatible with an early *selection* of >Open-Class items, i.e., Roots. Perhaps the early >selection (but not insertion) of underived Roots would >be able to insure that Unaccusative structures are >built for Unaccusatives, etc. This seems like a logical possibility, but as far as I know it's not a route that has been taken so far by anyone adopting DM assumptions. As I think I said in earlier postings, the role of LF and the competition among roots for insertion seem to be a couple of the issues at stake, with empirical consequences you could explore. It seems to me that even if roots are only 'selected' early, Vocabulary insertion would work the same way for roots as it does for non-roots, i.e., competition among items specified for certain features that play a role in the syntax. This is a different approach to root insertion than the one taken in the DM literature so far, so I do think the same empirical issues arise. Or maybe you have some other ideas about how to test your hypothesis. Remember that DM is just a framework of assumptions for hypothesis-testing. Within this framework, it's been proposed that roots aren't distinguished from one another until Vocabulary insertion takes place, and that all roots compete with each other for insertion (see Marantz's 1997 PennWPL paper and '"Cat" as a phrasal idiom' ms., and the Harley & Noyer paper cited in my first posting). As far as I can see, there's no contradiction between adopting the assumptions of DM and discovering that this proposal is wrong. Regarding overgeneration, see my last posting about the possibility of (quasi-)continuous interpretation of grammatical choices. This might block the kind of overgeneration you describe; it all depends on the details. As ever, it seems reasonable to ignore the issue until someone finds evidence that bears on it. I'm not sure I've added anything to my earlier postings, but I hope this helps. Regards, Martha mcginnis at ucalgary.ca From mcginnis at ucalgary.ca Fri Jun 15 16:39:58 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Fri, 15 Jun 2001 10:39:58 -0600 Subject: Martha McGinnis: Number marking in Bayso Message-ID: A non-text attachment was scrubbed... Name: not available Type: text/enriched Size: 2323 bytes Desc: not available URL: From mcginnis at ucalgary.ca Thu Jun 21 17:12:41 2001 From: mcginnis at ucalgary.ca (Martha McGinnis) Date: Thu, 21 Jun 2001 11:12:41 -0600 Subject: Martha McGinnis: Squib Message-ID: Dear DM-list, I've written a squib based on the interchange Heidi Harley and I had on this list back in February, with an additional section on aspectual compositionality in idioms. The upshot is that the facts support DM, rather than Jackendoff's or Nunberg et al.'s theories of idioms. If anyone would like to see the squib, it can be downloaded from my website: http://www.ucalgary.ca/~mcginnis/papers.html "On the systematic aspect of idioms" Regards, Martha mcginnis at ucalgary.ca