"Consider" as an object copular verb

Hi! I’m puzzled by the ERG’s analysis of these two sentences:

(1) She considers him a fool.

dmrs (19)

As far as I could understand it, this analysis with consider as a two-place predicate seems spurious:

dmrs (20)

(2) She considers him to be a fool.

dmrs (15)

See, respectively, analysis of (1) and analysis of (2).
In both cases, consider is treated as a three-place predicate (abastracting away from parses that don’t seem relevant in the case at hand). By contrast, believe in (3) is analyzed as a two-lace predicate, see here.

(3) She believes him to be a fool.

dmrs (18)

Is there any reason for this discrepancy? Isn’t consider an object-raising verb like believe in (3)? Cf. Abeillé (2021:517-519):

Copular verbs such as be or consider are analyzed as subtypes of subject-raising verbs and object-raising verbs respectively […]. They share their subject (or object) with the unexpressed subject of their predicative complement. […] A copular verb like be or seem does not assign any semantic role to its subject, while verbs like consider or expect do not assign any semantic role to their object. […] Consider selects a subject and two complements, but
only takes two semantic arguments: one corresponding to its subject, and one corresponding to its predicative complement.

Any ideas on this topic would be highly appreciated, since I would like to implement analogous constructions in PorGram, a new HPSG grammar of Portuguese I’m developing in collaboration with @arademaker. Using the the abstract types of the matrix.tdl core component of the Grammar Matrix customization system, I already have manually implemented some object raising verbs.

1 Like

The analysis with two-place “consider” has different semantics. In this case, “a fool” is a modifier, meaning that the subject is the fool, e.g. compare:

(4) She arrived a fool.

As for raising verbs, the ERG has a much more fine-grained inventory than often seen in the literature. No doubt @Dan would have more to say about this particular contrast, but my first thought is that asserting (2) not only asserts something about her beliefs but also asserts something about her actions towards him (in the sense that “considering” is a performative action like “naming” or “claiming”), while (3) only asserts something about her beliefs (in the sense that we are only talking about a mental state, and nothing performative).

(Of course, I can also imagine that there is some variation between English speakers in how they use the word “consider”!)

For example, I would find it natural to say “she has considered him to be a fool” to mean that she has given a judgement about him (and perhaps this has created a precedent that we’re supposed to follow). In contrast, I would only take “she has believed him to be a fool” to mean that she has had this belief at some point (without creating a precedent for us).

1 Like

In all analysis mentioned by @leonel the same sense was used _consider_v_1. This is strange for me since propbank seems to provide two different senses for the two cases above:

  1. She consider him to be a fool
  2. She as a fool consider him (for something)


Surely WN contains more senses (OpenWordnet-PT Browser consider) but we know that only senses with different syntactic behavior are relevant to ERG.

@leonel analysis 1.a and 2 have arg1-3 but analysis 1.b only arg1-2 since mod points to consider and not the other way around right? So for me 1.a and 2 are semantically equivalent right?

Thanks for your comments, @guyemerson. For me, the most import question can be spelled out this way: is consider in the relevant reading of (1), i.e., the one (roughly) paraphraseable with (2), a persuade-type verb, i.e., an object control verb, or a believe-type verb, i.e. an object raising verb? Object control verbs impose selectional restrictions on their objects, since they are semantic arguments of the corresponding predicates, while raising verbs don’t, because their object is not one of its semantic arguments.
In the usual analysis of consider in (1) and (2), this verb is an object raising verb. It expresses a two-place relation between an experiencer (EXP) and a state-of-affairs (SOA), just like believe in (3), expect in (4) etc. The predicate expressing the SOA, on its turn, imposes selectional restricts on its subject (the object of the matrix verb), see (4). By contrast, the ERG analysis (1) (in the reading of (2)) is a relation between three participants.

(4) a. She considers inflation a risk.
b. ??? She considers inflation a fool.
(5) We expect it to rain tomorrow. (Abeillé, 2021:491)

@arademaker It looks like the ERG has 6 lexical entries which share the _consider_v_1 predicate, and one other lexical entry with _consider_v_as: http://svn.delph-in.net/erg/tags/2020/lexicon.tdl

@leonel By “usual analysis”, do you mean Abeillé’s analysis? There might be a disagreement here about what “consider” actually means. Compare (1) with:

(6) She declared him a fool.
(7) She labelled him a fool.

If “label” is analysed as a control verb, then an awful lot of ditransitives would have to be analysed that way too.

I agree that (4b) is implausible, but I think this is a weak argument for a particular semantic analysis. Maybe I’m missing something from the argument, but otherwise we could apply the same argument to the following, to conclude that “give” is an object control verb:

(8) a. She gave democracy a boost.
b. ??? She gave democracy a hug.

Abeillé’s analysis looks reasonable, but it would be difficult to incorporate into the ERG without radically changing the analysis of copular constructions in general. For “he is a fool”, the ERG introduces a _be_v_id predicate. Maybe we could introduce a corresponding id predicate for a copula without the verb “to be”, but this would require some care. If we can manage to get rid of _be_v_id entirely, then Abeillé’s analysis becomes more attractive. But there would still be the question of where to draw the line between control verbs and ditransitives.

Sorry if I’ve raised more questions than answers :sweat_smile: