Skip to content

ErgSemantics_OneAnaphora

Jump to bottom
EmilyBender edited this page May 13, 2014 · 5 revisions

ESD Test Suite Examples

Browne's barks.
The dog chased that one.
Three bark.
That's enough.

Linguistic Characterization

This phenomenon involves anaphoric noun phrases which either don't include a pro-form at all (Browne's, Three) or have the pro-form one, which is unusual among English pronouns in that it can take dependents. All of these expressions take their interpretation from the (linguistic or non-linguistic) context. Though one anaphora in particular are famous in the syntactic literature because Baker (1978) claimed that the antecedent could only be an N' (not a simple N), in fact Payne et al (2013) convincingly argue that the constraints on the possible antecedent of one in this use are semantic rather than syntactic. In particular, they show that the antecedent can be any pragmatically plausible expression of (in their terms) type <e,t>. In MRS terms, this means that the antecedent should be a subgraph rooted in a predication denoting an entity. Where Payne et al (2013) focus specifically on expressions involving the form one (e.g. The dog chased that one.) Nerbonne et al (1989) group these together with ‘null-head anaphora’ (e.g. Browne's barked) under the term ‘N-bar anaphora’. The ERG's semantic analysis further groups N-bar anaphora with various NP anaphora (Three bark., That is Kim.).

Motivating Examples

  • Sandy has a blue glass, but we couldn't find another one for Kim.

  • Sandy always likes the pictures of Kim better than the ones of Pat.

  • [Looking at a display of chocolates:] I want one!

MRS Fingerprints

generic_entity(ARG0 x1)

Interactions

  • Some, but not all, partitive expressions also involve generic_entity.

Reflections

  • Payne et al (2013, 804) point out that phrases like Mary's former mansion present an interesting puzzle for semantic compositionality in that there are two interpretations of former: the phrase can either refer to something of Mary's that used to be a mansion or (more commonly) to a mansion that used to be Mary's. In the latter interpretation, former has to be predicated of the relation holding between Mary and mansion, which in the ERG isn't introduced by Mary. [FIXME: This note doesn't belong here; it's relevant to Payne et al's analysis because they are trying to establish that no nouns are inherently relational, and any noun can be promoted via a type-shifting rule to a relational interpretation. I think we want to hold onto this observation, though, and just figure out where to put it.]

Open Questions

  • The dog chased that one./The dog chased those ones. both have a card in them corresponding to one. The card for ones at least has CARG 2+, but I wonder what information this is adding that the NUM attribute isn't already capturing. In other words, why not have one introduce generic_entity only?

  • Payne et al (2013) distinguish between one_d and one_ct, where the former is a ‘determinative’ that stands in for an NP and the latter is a count noun. Might it make sense to have one_d introduce card_rel (and generic_entity), but one_ct only generic_entity?

  • The example That's enough has two generic_entity EPs in it, one corresponding to that and one to enough. If generic_entity is supposed to be an instruction to find a relevant (linguistic?) antecedent, we are claiming that these instructions are the same for demonstrative pronouns as for one anaphora and partitives, modulo what we can get from the quantifier EP. On the surface, this claim is a bit surprising to me, but I'm not sure what kind of evidence could be used to support or refute it.

  • Looking at the testsuite examples for partitives, I see generic_entity in one (Who is the oldest of all?), but not the others. What is the pattern here?

  • ‘subgraph rooted in a predication denoting an entity’ --- do we not have a more streamlined term for that yet? Also, I couldn't find ‘entity’ anywhere on the Basics page. Is it meant to be there?

  • ‘subgraph rooted in a predication denoting an entity’ --- do we expect these to always correspond to <e,t>?

Expert External Commentary

Grammar Version

  • Note the grammar version (trunk + svn revision) consulted as this was written or updated.

References

  • Baker, C. L. (1978). Introduction to generative-transformational syntax. Englewood Cliffs, NJ: Prentice-Hall.

  • Payne, J. & Pullum, G. K. & Scholz, B. C. & Berlage, E.(2013). Anaphoric one and its implications. Language 89(4), 794-829. Linguistic Society of America. Retrieved May 12, 2014, from Project MUSE database.

  • Nerbonne, J., Iida, M., & Ladusaw, W. (1989). Running on empty: Null heads in head-driven grammar. In Proceedings of the Eight West Coast Conference on Formal Linguistics. Stanford: CSLI.

More Information

Home | Forum | Discussions | Events

Clone this wiki locally