David Adger's (2000) paper "VSO clause structure and morphological feature checking" contains some surprising data concerning the use of prepositions with the definite article.
The basic data shows that prepositions come in two forms, one of which is used when its complement NP is introduced by an overt definite article, and the other used in all other cases:
le peann - with a pen leis a' pheann - with the pen *leis peann *le a' pheann dha thaigh - to a house dhan an taigh - to the house *dhan taigh *dhan thaigh *dha an taigh ri Dhàibhidh - to David *ris Dhàibhidh ri seo - to this *ris seo ri mo bhràthair - to my brother *ris mo bhràthair ri àite mo ghaoil - to the place of my love *ris àite mo ghaoil
Thus, it appears on first analysis that the preposition is sensitive to the presence of an overt definite article, rather to any kind of semantic definiteness, which is curious but handlable, assuming the following categories:
le :- PP/NP-def,-len dha/ri :- PP/NP-def,+len leis/dhan/ris :- PP/NP+def
The problem arises with data like the following, where a preposition takes a VP argument with a preposed object:
Dh'fheuch mi ri [teine a chur air dòigh] - I tried to [set a fire up] *Dh'fheuch mi ris [teine a chur air dòigh] Dh'fheuch mi ris [an teine a chur air dòigh] - I tried to [set the fire up] *Dh'fheuch mi ri [an teine a chur air dòigh]
This it seems that the dependency is purely one of adjacency rather than anything syntactic. Could we handle this using a postprocessing transducer? Is it possible to simulate this in the syntax using a special feature for all categories "starts with a definite article"?
No comments:
Post a Comment