How to think about MRS structures that use a non-verb as the MRS Top like: "Where are you?"

I’ve had a working assumption that the Top variable of an MRS will always be a variable introduced by a verb in the sentence.

"Where are you?" ends up treating loc_nonsp as the predicate which introduces the Top variable for the MRS structure and doesn’t include a variant of _be_v at all:

                      ┌pron:x3
  pronoun_q:x3,h11,h12┤
                      │                ┌place_n:x4
                      └which_q:x4,h7,h8┤
                                       └loc_nonsp:e2,x3,x4

Logic: pronoun_q(x3, pron(x3), which_q(x4, place_n(x4), loc_nonsp(e2, x3, x4)))

In fact, it looks like loc_nonsp gets treated like a verb in other ways too. For example, "Where are you on the table?" ends up generating an MRS which has a stative preposition which is modifying the event variable that loc_nonsp introduces:

                      ┌pron:x3
  pronoun_q:x3,h11,h12┤
                      │                  ┌_table_n_1:x14
                      └_the_q:x14,h16,h17┤
                                         │                ┌place_n:x4
                                         └which_q:x4,h7,h8┤
                                                          │   ┌_on_p_state:e13,e2,x14
                                                          └and┤
                                                              └loc_nonsp:e2,x3,x4
Logic: pronoun_q(x3, pron(x3), _the_q(x14, _table_n_1(x14), which_q(x4, place_n(x4), and(_on_p_state(e13, e2, x14), loc_nonsp(e2, x3, x4)))))

Is there a principle involved here that determines when a non-verb will end up being the Top of an MRS structure and treated as if it were “verb-like” in this way? Any tips on how should i think about this?

Your working assumption is almost always right, but the verb “be” is the most frequent exception. When “be” combines with anything other than a noun phrase complement, it does not add a semantic predicate of its own, instead only constraining the tense and aspect for the clause. So in “the child is sleeping” there is a “_sleep_v_1” predicate but no additional predicate from “is”, and thus the difference between the MRSs of “the child is sleeping” and “the child slept” is only in the tense (present vs. past) and aspect (progressive vs non-progressive). We do the same for “the child is happy” and “the child is in the park”, where again “is” does not add a predicate of its own, and thus the main predicate for the clause comes from the complement phrase “happy” or “in the park”. This treatment of “be” simplifies the semantics of “The child is laughing and happy”, and minimizes the differences in semantics for “the child who is afraid of spiders” and “the child afraid of spiders”. However, when “be” combines with a noun phrase complement, as in “the lead actor is Pacino”, the main predicate supplied by “Pacino” doesn’t have any argument slot for “the actor”, nor any place to reflect the tense and aspect from “is”, so we treat this so-called “identity be” as a separate verb from the copula “be” of the earlier examples, and this identity “be” does supply a semantic predicate of its own. Other verbs that do not contribute their own predicate include the “have” of “we have arrived”, “do” as in “do we need to leave”, and “will” of “we will leave”.

1 Like

Very clear, thanks @Dan.