Semantics (p. 41) [via PaperHive@docloop]

[docloop[bot]]

Annotation imported from PaperHive.

Regarding this part:

The example in (48c) shows that the passive participle cannot be formed with unergative intransitive verbs.

Remi van Trijp wrote:

For me, example 48c is unacceptable because it doesn't make sense semantically... whereas 48b is coercion of the verb in a caused-motion argument structure pattern

Link to original comment. About docloop.

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

Yes, but "die getanzte Frau" is interpretable but still ungrammatical. There has to be a direct object in the active. This can then function as the noun that is modified. And of course the semantic role of the adjectival participle has to be compatible with the noun. As for the coercion: This is the point I argue for. In HPSG and Standard LFG such coercions are lexical rules. If you believe in lexical integrety (as people in CxG, HPSG and LFG uually do) than there is no verb to apply the construction to. "getanzten" is an adjective. If you apply your coercion construction to the verb stem, you basically apply a lexical rule. This is what I argue for.

Link to original comment. About docloop.

[docloop[bot]]

Annotation imported from PaperHive.

Remi van Trijp wrote:

The answer in Cognitive/Goldbergian CxG is that such overgeneralizations/overgenerations are avoided through statistical preemption (Goldberg), entrenchment (Ambridge), or some other probabilistic preference. So they would rather use statistical preferences than positing more constraints on the grammar. As for coercion: this might indeed be just a matter of notational variation (though see Goldberg 2013). But "getanzten" is not necessarily an adjective, it can also be considered as a verb that merely functions as an adjective, in which case you would not violate lexical integrity.

Link to original comment. About docloop.

[docloop[bot]]

Annotation imported from PaperHive.

Stefan Müller wrote:

> The answer in Cognitive/Goldbergian CxG is that such overgeneralizations/overgenerations are avoided through statistical preemption (Goldberg), entrenchment (Ambridge), or some other probabilistic preference. So they would rather use statistical preferences than positing more constraints on the grammar. But the statistical preemption has to work on something. Usually the weights are associated with constraints. I would like to see the constraints and then we can talk about weights and statistics. > As for coercion: this might indeed be just a matter of notational variation (though see Goldberg 2013). But "getanzten" is not necessarily an adjective, it can also be considered as a verb that merely functions as an adjective, in which case you would not violate lexical integrity. Yes, the verb is inside the adjective and lots of its properties are still active. Adverbs can combine with it and so on. Maybe something with mixed categories could work for this. But note that a ASC that has the form [SUBJ Verb OBJ OBL] accounts for actives. We need an allostruction for the passive case [SUBJ Verb OBL] and what is needed for the adjectival cases is totally different: [OBL Adj/Verb]. So you cannot say that it is the same coercion as for verbs. In my approach you need one construction adding the object and the oblique. Passive is independent of this and adjective formation too.

Link to original comment. About docloop.

[docloop[bot]]

Annotation imported from PaperHive.

Remi van Trijp wrote:

Indeed, in the usage-based construction grammars, researchers argue for many (allo)constructions, because psychological evidence seems to point to redundancy in the inventory. The fact that you do not need allo-constructions won't strike a usage-based linguist as a problem (on the contrary). Nevertheless, I agree with you that as a linguist, we should be interested in the obvious relations between the active and passive constructions. In my opinion, these relations should be explained through how these patterns have emerged in the language in the first place. I would hypothesize that these relations are the side-effect of cultural evolution (including learning and usage processes).

Link to original comment. About docloop.