Robert Levine (levine@ling.ohio-state.edu)
Mon, 22 Jun 98 20:35:36 EDT
While I agree with Carl that there may be reasons to make ARG- ST a feature bearable by phrases, the case of purposives isn't necessarily clear enough to base any strong claims on, one way or the other. You should be aware that Michael Johnson, in a collection of papers on HPSG that I'm coediting with Georgia and that has just gone to press at Cambridge, argues that purposive modification of just the type you illustrate requires the syntactically overt presence of what he calls an affected object in the sign of the modified VP, an effect he gets without direct reference to ARG-ST, and that whether a verb's object counts as affected or not is reflected as a partition of verbs in the type hierarchy. Thus, _do so_ replacement and ellipsis and the like will not satisfy the requirements for what he calls `recipient' purposive modification, since this requirement is not semantic in nature, but requires the overt presence of a daughter sign which reflects the affected object of the verb in question. The mechanisms he uses to access the presence of an affected object make somewhat different predictions than what you propose, since on his account, purposive modification may itself supply an affected object for a higher (or further right--righmore?) recipient purposive, while (at least on various people's story) adjuncts don't show up on ARG-ST lists, so that any account which required a more external recipient purposive to access information about an affected object supplied by an internal purposive would't be able to do so, so that what are indeed well-formed multiple purposives of a certain type would be predicted to be bad, something which his account avoids. You might contact him and get a copy of his paper; he's at contact him and get a copy of his paper; he's at johnston@cse.ogi.edu. Then again, maybe purposives of the sort you're looking at are optional complements and standard tests for VPhood like ellipsis or _do so_ are telling you this[??] Bob Levine
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:38:20 PST