Hi Benoit,

Well, this should be becouse semantics seems underlying... the everithing. We all agree that syntax is important, but ... What comes to my mind concerning the argument structure and these genetive-dative-like constructions is a bit seen in languages that lost their cases (such as French), but mostly in these that lost their cases recently (such as Bulgarian). We did an experiment (using French too) which demonstrated  that the mental calculus of who does what to whom, of the very same syntactic form, is very different depending exlusively on semantics (of Entities). So,,, Patherns alone do not solve the problem of OBJ2.  Yes, the trial to decode what is said by humans usind a machine can not succeed based on syntax alone.The puzzle will go on.

Here the experiment:

https://www.researchgate.net/publication/255575344_EXPERIMENTAL_SUPPORT_OF_ARGUMENT-BASED_SYNTACTIC_COMPUTATION

I hope this helps to understand that the problem is deep..

Cordialement,

Velina

Article EXPERIMENTAL SUPPORT OF ARGUMENT-BASED SYNTACTIC COMPUTATION

Hi Benoît,

Yes, it's true that (2), SUBJ V OBJ OBJ2, looks less formal and more functional than (1), NP V NP NP. However, just because 'Subject', 'Direct Object', etc. are often referred to as grammatical 'functions', this doesn't mean that we can truly see them as (purely) 'functional' or semantic. Rather, as I know you know, the semantic aspect of the ditransitive construction is formulated in terms of, in the prototypical case, thematic roles such as 'agent', 'recipient' and 'patient'/'theme', and Jackendovian primitives such as CAUSE and RECEIVE. 

Besides, 'SUBJ' etc. may capture more cases than just 'NP', the latter being only a typical but not the exclusive way in which subjects can be instantiated. The subject in the ditransitive construction could also be a clause: 

(3) [That Trump is now the President of the US] gives [many people] [a headache].

Likewise, OBJ2 could also be a clause:

(4) [The teacher] told [the pupils] [that they would go on a picnic].

That is why it's preferable, perhaps, to formulate the syntactic side of argument structure constructions in terms of grammatical functions rather than phrase-structure categories (maximal projections). Another advantage is the one you mention, namely that there might be semantically different constructions, notably the one instantiated by e.g.They elected him president, share the same (preferential) phrase structure categories as the ditransitive.

Finally, and related to this, by labelling the elements as 'SUBJ', '(Indirect) OBJ', etc., we have a more precise of how these constituents behave syntactically. For instance, we know that the element labelled as a (Direct) Object (i.e. OBJ2) might quite likely be used as the SUBJ in a corresponding passive sentence, something which is not the case for the second post-verbal NP in the construction we find in They elected him president:

(5) That they would go on a picnic was told to the pupils.

(6) *President was elected to him.

So, labels such as 'SUBJ', 'OBJ', etc. are quite formal things after all. They just name different aspects of 'form' or 'syntax' than labels such as 'NP' do.

Thank you for your answers, Velina and Bert!

Indeed, Bert, there is in fact something very formal about the ‘semantic’ criteria that I mentioned. The two examples in (5) and (6) are particularly helpful in that regard. So, my example (2) might not really be an issue to the way constructions are defined within CxG, as opposed to what I originally thought. 

But does that mean that – in spite of the terms used – I should look at my representation in (2) strictly in terms of a (complex) ‘formal’ description? From the perspective of CxG, that should be the case. But is it really? I am not sure.

Hi Benoit,

Concerning (2), SUBJ V OBJ OBJ2, lI would do this exactly - take it as s more complex formal description which provides the arguments' structure.

Bert will tell better.

(By the way Bert's reasoning arround the passive form has been experimantally supported in onother study related to the mental calculus of the three arguments and the grammatically correct free word order used in the expeirment.)

Cordialement

Hi Benoît,

I think that, yes, you could call it a complex 'formal' description, when you provide a CxG representation. There's little harm in doing that. But you're right that labels such as 'SUBJ', 'OBJ', etc. look less purely form-related than, say, 'NP'. That's because you can't read these labels directly 'off them' but need to take into account word order and, crucially also, the thematic roles before you can be certain a given NP is a (Direct) OBJ and not, for instance, a predicative complement. 

All best,

Bert

The underlying problem is that construction grammar does not accept sentence syntax as structurally distinct from other forms of 'form', like phonetic form in words. The solution discussed here, going back to school grammar functions, instead of using phrase structure notations inherited from generative grammar, of course makes sense, but hides the theoretical problem of CG. We (linguists) need to develop a semantically informed syntax of sentences in general, underlying all constructions and determining how to describe them – instead of raising the same question construction after construction.

Thank you for you answer,

I also thought that there might be a theoretical problem within CxG, hence my question. However, to what extent are the labels SUBJ, OBJ (etc...) not already a 'semantically informed' type of syntax, as Bert Cappelle has shown in his answer, that can be used as a descriptive tool across sentences/constructions? Are these not adequate enough? (This is a real question, I don't know.)

Best,

Benoît

@Benoît. In a sense, you may say that intuitive school grammar is just fine, for example for language teaching. The linguistic problem, however, is that string notation is insufficient, because many constructions have variable linear manifestations. Then the string Subject-verb-etc. is incorrect, if the same construction can become: Adverb-verb-subject-etc. because of the underlying sentence syntax, which also includes the rest of what is going on, besides the construction, in a sentence. Intuitive school grammar refers to our implicit,i ntuitive semantic understandings but does not describe these, and it does not systematically describe the linear variations in manifestation either. So what I am saying is that CxG has a considerable problem with syntax, which is why it is a sloppy theory of language. – Sure, constructions can be fun...

Hi Benoit, "SUBJ, OBJ (etc...) not already a 'semantically informed' type of syntax, as Bert Cappelle has shown in his answer " this is a question, yes. I think that they SHOULD be used as a descriptive tool across sentences/constructions.They are the most adequate, according to my reasoning. The problem is HOW to do this, as S-V-O-R seem being determined following the semantics of the nouns used (and the context).

Here cross-linguistic and brain studies about the importance of the argument structure for the assembling the sentence meaning.

https://www.researchgate.net/publication/38134865_The_Argument_Dependency_Model_A_neurocognitive_approach_to_incremental_interpretation

https://www.researchgate.net/publication/6781419_The_extended_argument_dependency_model_A_neurocognitive_approach_to_sentence_comprehension_across_languages

https://www.researchgate.net/publication/220702564_Towards_a_Neurobiologically_Plausible_Model_of_Human_Sentence_Comprehension_Across_Languages

You may take a look at the project that the authors have started...

Best

Article The Extended Argument Dependency Model: A Neurocognitive App...

Article The Argument Dependency Model: A neurocognitive approach to ...

Conference Paper Towards a Neurobiologically Plausible Model of Human Sentenc...

I think at least for English, the SUBJ / OBJ distinction is easy to justify based on pronoun forms (I / me, they / them). At least that is a clear surface difference in form (notwithstanding the complex distribution of oblique pronouns in various English constructions).

Different types of OBJ are really only distinguishable across configurations (passive, alternations etc.). But as Beth Levin's verb classes show, it's not so hard to agree on those for a large bulk of cases based purely on distributional criteria, even without semantics. And in any case, if you're using a CxG variant that postulates a 1:1 form meaning correspondence, then the different meaning also supports assuming constructional homophones such as resultative and ditransitive OBJ clusters.

@Amir. I agree that it is not hard to agree on the school grammar description of complements if you already have a description of verb valencies. I am just saying that the LINGUISTIC interest in these intuitive descriptions is to find out what if at all they contribute to the understanding of what SENTENCES are made of. Different constructions share the fact of appearing in sentences, and the latter have other features than manifesting constructions.