INDEX variable not present elsewhere

[goodmami]

In the ERG trunk (r26802) item 974 of the CSLI test suite ("It was in January that Browne was hired.") has a broken MRS:

[ TOP: h0
  INDEX: e2 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ]
  RELS: < [ _in_p_temp<7:9> LBL: h1 ARG0: e4 [ e SF: prop TENSE: untensed MOOD: indicative ] ARG1: u5 ARG2: x6 [ x PERS: 3 NUM: sg IND: + ] ]
          [ proper_q<10:17> LBL: h7 ARG0: x6 RSTR: h8 BODY: h9 ]
          [ mofy<10:17> LBL: h10 ARG0: x6 ARG1: i12 CARG: "Jan" ]
          [ loc_nonsp<18:22> LBL: h1 ARG0: i13 ARG1: e14 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] ARG2: e4 ]
          [ proper_q<23:29> LBL: h15 ARG0: x16 [ x PERS: 3 NUM: sg IND: + ] RSTR: h17 BODY: h18 ]
          [ named<23:29> LBL: h19 ARG0: x16 CARG: "Browne" ]
          [ _hire_v_1<34:40> LBL: h1 ARG0: e14 ARG1: i21 ARG2: x16 ] >
  HCONS: < h0 qeq h1 h8 qeq h10 h17 qeq h19 >
  ICONS: < e14 focus e4 e14 topic x16 > ]

Note how the INDEX e2 is not the ARG0 of any EP. In 1214, it would have been the ARG0 of the _be_v_itcleft, but now ICONS has taken over its duties of showing focus.

[goodmami]

@danflick but should _in_p_temp be the INDEX? or _hire_v_1?

Also, this happens in simpler sentences with it-clefts, like "It was Browne who was hired"

[danflick]

With the adoption of ICONS constraints, it seemed good to get rid of the awkward `be_v_itcleft_rel" for these it-cleft constructions, and instead make use of the topic/focus ICONS elements. While this is cleaner for simple examples like "It was Abrams who won", the loss of an independent predication for the "be" creates an interesting problem for examples where the tense and/or aspect differ. How do we encode the contrast between "It was Abrams who won" and "It is Abrams who won"? And even worse, where do we record the perfect aspect of "It had been Abrams who was winning" when we only have one eventuality in the MRS, namely the ARG0 of _win_v_1_rel which is busy being progressive but not perfective. Or "It will be Abrams who won"?At present, the 2018 version identifies the INDEX of the clause with that embedded verb's ARG0, so it is no longer left unbound, but this analysis ignores the tense/aspect contributions of the "be", and is clearly inadequate. So we leave this issue open, and hope someone will propose a better account.

[guyemerson]

Is there a semantic difference between "It was Abrams who won" and "It is Abrams who won"? Genuine question -- I'm struggling to think of a situation where one would be acceptable and the other wouldn't.

[goodmami]

Looking at this again...

While _be_v_itcleft is ugly and introduces syntactic terminology into the semantics, I do miss having something around to hang the TAM properties on. Here are some other options (not necessarily better or well-motivated or well-thought-out):

• Introduce an eventualization predicate analogous to nominalization which wraps MRS fragments and makes an event (in this case makes an event into an event).

• Reanalyze it-clefts as be-identification. E.g., "It was Abrams who won" is like "It was Abrams" instead of "Abrams won" (where "who won" is a relative clause). The sentence is, after all, an answer to a question like "Who won?" and not "What happened?". But I don't think this helps with non-nominal clefts ("It was in January...").

[guyemerson]

If the TAM properties persist in the semantics, that should be because there's a clear difference in meaning. As I commented above, I don't know when "It was Abrams who won" would be acceptable but "It is Abrams who won" wouldn't be.

Maybe the semantic difference is clearer with a habitual:

• It is Abrams who waters the plants.
• It was Abrams who waters the plants. (But now it's Kim who waters the plants.)

But even here, I would rather say "It was Abrams who watered the plants" or "It was Abrams who used to water the plants". This might suggest that the syntactic TAM properties are constrained but don't contribute to the semantics, much like tag questions. I could be wrong, of course -- maybe there are some better example sentences I haven't thought of.

Unless there's a clear semantic motivation for keeping the TAM properties, I think it's better to leave them out.

[goodmami]

Ok, it's a fair point for discussion.

So if it's syntactically constrained then we would expect to proscribe:

• It is Abrams who won.
• It was Abrams who wins.

But I don't have much problem with those, though the second may need more context.

But what about Dan's example using non-default aspect:

• It had been Abrams who was winning.

Does the perfective of the it-cleft somehow unify with the progressive _win_v_1? E.g., is it just the same (except maybe for info structure) as:

• Abrams had been winning.
• It was Abrams who had been winning.
• It had been Abrams who had been winning.

Or if it's just syntactic constraints that don't contribute to the semantics, then maybe it's the same as:

• Abrams was winning.

I can see making a case for unifying, but dropping them looks like a mistake. But if they unify then maybe the "It is Abrams who won" examples get ruled out as it-clefts.

Perhaps it's time to consult the literature on the subject?
edit: here's a pretty thorough account by Jong-Bok

[guyemerson]

• It is Abrams who won.
• It was Abrams who wins.

But I don't have much problem with those, though the second may need more context.

Could you give a context where you would use the second?

With the perfect, I think I can see a difference:

With one throw of the dice, it was now Abrams who was winning
?With one throw of the dice, it was now Abrams who had been winning
?With one throw of the dice, it had now been Abrams who was winning

here's a pretty thorough account by Jong-Bok

I might be missing something but I couldn't find any mention of tense or aspect.

[goodmami]

• It was Abrams who wins.
  Could you give a context where you would use the second?

Hmm, it's a bit hard without discussing time travel and alternative histories :)

"The script had undergone several revisions. It was Han who destroys the Death Star in the first draft, but by the final draft it is Luke." (or "Abrams who wins"..."is Browne" if you want the original example)

But "who would destroy" or "who destroyed" are also acceptable, or even preferred.

here's a pretty thorough account by Jong-Bok
I might be missing something but I couldn't find any mention of tense or aspect.

No, sorry, I linked it because it had plenty of positive and negative examples with more variety than I'd seen in the three or four other papers I'd looked over. None of these mentioned tense, or did only in passing.

I just found this one by Edward Fichtner which mentions tense (and modals and adverbs and other uncommon patterns for it-clefts):

From pp14--15:

In the examples of the clefting transformations which have been examined so far, it was found necessary to adjust the number of the verb be introduced by Cleftization so that it agrees in number with its surface subject. The tense of the very be, on the other hand, has been modified without comment to correspond to that of the verb in the underlying sentence. [...] In fact, there are quite specific constraints which govern the assignment of tense markers to be in the Cleftization process. [...]
[...]
The principle which governs the extension of the tense markers from the verb in the underlying sentence to the verb be inserted by Cleftization is very similar to the one which governs the adjustment of the form of the verb in the reported statement in indirect discourse.

So Fichtner claims explicitly that it is agreement with the "verb of the underlying sentence", or the subordinate clause. He doesn't get very deep into the semantics, but I haven't seen any papers explicitly say that the properties get unified such that "It had been Abrams who was winning" means "Abrams had been winning" and not just "Abrams was winning".

[guyemerson]

"The script had undergone several revisions. It was Han who destroys the Death Star in the first draft, but by the final draft it is Luke." (or "Abrams who wins"..."is Browne" if you want the original example)

Nice.

[goodmami]

In the 2018 version I see a similar problem with item 551 in the MRS test suite: "That the cat chases Browne is obvious.":

[ "That the cat chases Browne is obvious."
  TOP: h0
  INDEX: e2 [ e SF: prop TENSE: pres MOOD: indicative PROG: - PERF: - ]
  RELS: < [ loc_nonsp<0:4> LBL: h4 ARG0: i5 ARG1: e2 ARG2: u6 ]
          [ appos<5:26> LBL: h4 ARG0: e7 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: x8 [ x PERS: 3 NUM: sg IND: + ] ARG2: x9 [ x PERS: 3 NUM: sg IND: + ] ]
          [ _the_q<5:8> LBL: h10 ARG0: x8 RSTR: h11 BODY: h12 ]
          [ _cat_n_1<9:12> LBL: h13 ARG0: x8 ]
          [ compound<13:26> LBL: h14 ARG0: e15 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: x9 ARG2: x16 [ x PERS: 3 NUM: pl IND: + PT: notpro ] ]
          [ udef_q<13:19> LBL: h17 ARG0: x16 RSTR: h18 BODY: h19 ]
          [ _chase_n_1<13:19> LBL: h20 ARG0: x16 ]
          [ proper_q<20:26> LBL: h21 ARG0: x9 RSTR: h22 BODY: h23 ]
          [ named<20:26> LBL: h24 ARG0: x9 CARG: "Browne" ]
          [ subord<30:38> LBL: h1 ARG0: e26 [ e SF: prop TENSE: untensed MOOD: indicative PROG: - PERF: - ] ARG1: h27 ARG2: h28 ]
          [ _obvious_a_to<30:38> LBL: h29 ARG0: e30 [ e SF: prop TENSE: untensed MOOD: indicative ] ARG1: i31 ARG2: i32 ] >
  HCONS: < h0 qeq h1 h11 qeq h13 h18 qeq h20 h22 qeq h24 h27 qeq h4 h28 qeq h29 > ]

Note that the INDEX e2 is not the ARG0 of any EP.

It seems the issue is fixed for this example in trunk, but not for the original sentence ("It was in January that Browne was hired."). I'm just noting this in case whatever changes fixed the "obvious" example could also fix the "in January" example.

[guyemerson]

I don't think this example would solve the problem that Dan raised. For "That the cat chases Browne is obvious", the tense/aspect of "is" can be pushed to "obvious". But for "It is Abrams who won", it isn't clear where the tense/aspect of "is" should go, and Dan has said he's left this problem open.

[goodmami]

Good point, Guy. Thanks!