Templates: lexical/constructional (p. 16) [via PaperHive@docloop]

[docloop[bot]]

Annotation imported from PaperHive.

Regarding this part:

The lexical approach assumes that there are several lexical items for verbs like run .

Ash Asudeh wrote:

In many respects, the main point of both Asudeh/Dalrymple/Toivonen (2013), which is based on an LFG proceedings paper from 2008, and Asudeh/Giorgolo/Toivonen (2014) is that templates in LFG allow one to be agnostic about this general question by capturing information that can then be associated either with lexical items or with phrase structure configurations (constructions) as the empirical facts dictate. Note that the templates don't really do anything except *label/name* a functional description. LFG theory already allows functional descriptions to be associate with both lexical items and phrase structure nodes. For example, the familiar \up = \down that is itself an f-description, and of course much richer specifications can be added, including ones that contain glue semantics meaning constructors. But these are facts about the formal framework. It would be perfectly possible for us to use the framework to make the *theoretical* claim that all valence operations are lexical -- i.e. the templates are always added to lexical entries. However, notice that the way to do this is to use optionality and disjunction in our approach. For example, look at the lexical entry for 'crushed' in AGT (2014: (57)). This single lexeme is either the past form (@PAST) or the passive form (@PASSIVE). It is fundamentally a regular transitive verb in terms of its argument structure (@AGENT-PATIENT). Notice that our PASSIVE template remaps the argument structure correspondingly. This leaves us with two main consequences with respect to your book. First, if we did pursue the purely lexicalist analysis, we would ideally only have as many distinct lexical entries as there are distinct morphological realizations. We would not generate several lexical entries for 'run'. Second, I believe you may have misunderstood some aspects of our analysis of passive and how it interacts with phrase structure (I don't think we would need a specific phrase structure rule for passive benefactives, as you claim later; more on this at the appropriate spot).

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

> I believe you may have misunderstood some aspects of our analysis of passive and how it interacts with phrase structure (I don't think we would need a specific phrase structure rule for passive benefactives, as you claim later; more on this at the appropriate spot) My claims come with a lot of "ifs". If you want to have fixed c-structure then you have to have special passive cases. Let's discuss this below.

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

> First, if we did pursue the purely lexicalist analysis, we would ideally only have as many distinct lexical entries as there are distinct morphological realizations. We would not generate several lexical entries for 'run'. But you would have a disjunction in the one lexical entry, wouldn't you? You would then not be able to explain why some verbs do funny things that they normally do not do (She smiled herself an upgrade, where laugh is used ditransitively). In Goldberg's view and in the Lexical Rule view this comes about since the verb enters a certain construction. In your view it seems that you have to say. This verb is now like another verb that happens to be disjunctively specified.

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._

[docloop[bot]]

Annotation imported from PaperHive.

Ash Asudeh wrote:

[I don't know how to use the quote facility to target specific things in your comment, so I'll just respond to the two parts respectively] 1. We would have disjunctions in the lexical entries, but in our approach I guess this would be tantamount to saying that the verb in question ('run', 'smile' or whatever) appears in these various environments, including the "constructional" ones captured by CxG or the LR approach. In other words, we directly represent the outputs of those approaches, rather than treating the relationship pseudo-derivationally as in the LR approach. I think this is more true to the ethos of constraint-based approaches. I have always been suspicious of lexical rules for this reason. Also, notice that this is an internal disjunction, so the verb is not treated as accidentally homonymous (as it would be if there were several separate lexical entries for e.g. 'run'). 2. Still not sure if I understand the point about passive --- I guess maybe because the antecedent does not apply (we don't want to have fixed c-structure, I think, but I'm not sure if I understand exactly what you mean in the comment).

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

> In other words, we directly represent the outputs of those approaches, rather than treating the relationship pseudo-derivationally as in the LR approach. I think this is more true to the ethos of constraint-based approaches. I have always been suspicious of lexical rules for this reason. Also, notice that this is an internal disjunction, so the verb is not treated as accidentally homonymous (as it would be if there were several separate lexical entries for e.g. 'run'). HPSG talks about lexical items. Lexical entries are the elements that are stored and lexical items include these stored ones and those licenced by unary branching rules. You are right, we could change the phonology of lexical items and we sometimes do, since morphology is also done via lexical rules. Note that in conversion (like play-V -> play-N) you have unary branching constructions in which the phonology does not change. The lexical rules that derive the lexical item used in resultative constructions are of this kind.

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

>Still not sure if I understand the point about passive --- I guess maybe because the antecedent does not apply (we don't want to have fixed c-structure, I think, but I'm not sure if I understand exactly what you mean in the comment). Ida's starting point was that she wanted to explain the non-possibility of passive with benefactives by the fact that there is a benefactive construction with a fixed form. A fixed c-structure. V-NP-NP. If this is the only way you can realize the benefactive construction it follows that the OBJ cannot be missing due to passivization. So independent of whether you treat the passive lexically or not, the impossibility of the passive in benefactive constructions was thought to follow from the fixed c-structure. Either you assume this and run into problems or you allow for optional constituents in the c-structure but then you need a different explanation of the impossibility (for some speakers) to form the passive and you do not need the c-structure anymore.

Link to <a href = "https://paperhive.org/documents/items/CDuBamUa0iWI/discussions/DL1q4xIttM9">original comment. About docloop._