Yet another challenge for Network Grammar:
(179) The claim was withdrawn
(180) The claim that the moon is a pancake was withdrawn
(181) The claim that the claim that the moon is a pancake was withdrawn is inaccurate
(182) * The claim that the claim that the moon is a pancake was withdrawn
(183) * The claim that the claim that the moon is a pancake was withdrawn is inaccurate is preposterous
Thanks to whoever thought up these sentences. Sorry but I can’t trace who that was.
Obviously we have to look at how relative clauses can be nested and at why there’s ungrammaticality if the numbers of subject noun phrases and verb phrases are unequal. But first we should look at the multiple occurrences of the claim in (181) and how they are interpreted as referring to different claims, not the same one.
Repeated noun phrases
In a single clause, multiple occurrences of phonologically identical noun phrases cannot refer to the same thing:
(184) A dog1 bit a dog2
This is OK although it more usually be expressed as:
(185) A dog1 bit another dog2
If there’s only one dog, a reflexive is used:
(186) A dog bit itself
In a second clause the choice of determiners affects meaning:
(187) A dog1 bit a man1 and a man2 shot a dog2
(188) A dog1 bit a man1 and the man1 shot the dog1
The choice of determiners in the first clause is less significant:
(189) The dog1 bit the man1 and the man1 shot the dog1
In (181), the claim occurs in two clauses, the second subordinate to the first and that causes two claims to be understood rather than one.
The reduced form works for a variant of (180):
(190) The claim the moon is a pancake was withdrawn
But it doesn’t really work for a variant of (181):
(191) ? The claim the claim the moon is a pancake was withdrawn is inaccurate
The reason for this could be that repetition of the claim is more likely to be a piece of disfluent speech and is treated as such – in upstream phonological processing perhaps.
In (181) we need to distinguish between two CLAIM concepts, the first qualified by the second and the second qualified by MOON etc.
Isn’t it also the case that the specific CLAIM in (180) must be distinguished from the generalised CLAIM that is permanently in the network? The particular CLAIM is qualified by MOON but the generalised one is not. Because of the possibility of another occurrence of the claim as in (181) it’s not possible to qualify the generalised CLAIM even temporarily.
The implication is that a noun generates a temporary concept to serve the purposes of the sentence in which it occurs. For (180) and (181) this could be CLAIM′ which instantiates the permanent CLAIM. So at The__claim:
The node that becomes CLAIM′ is in the lexical entry for the but it has no other properties. It can be assumed that a (determiner)__(noun) junction also creates a proposition reflecting the semantics of the particular determiner – the, a, this etc.
This process applies to any noun occurrence except perhaps a mass noun without a determiner – Porridge is nice as distinct from This porridge is nice. The process must also apply to nouns with no determiner. For example, indefinite count-noun plurals and proper nouns:
(192) We supply balls for Wimbledon
(1) John kissed Lucy
There are many types of ball and many Johns:
The INSTANTIATED BY proposition must actually be formed as soon as the noun is encountered, i.e. before looking for junctions between the noun and preceding words. What happens when a junction with a determiner is found is creation of the ‘determiner semantics’ proposition mentioned above. However, for simplicity, the diagrams only show the additional CLAIM concept being created at the__claim.
The latest instantiation
The process links CLAIM′ as shown with full activation. When claim occurs again, this is in effect the M node in the P / C / M word.
So, at claim__that:
The resulting proposition cannot be delivered yet because it is missing a concept.
In (181) the next junction is the second occurrence of the__claim. This is treated the same as the first occurrence except that what is instantiated is CLAIM′, not CLAIM.
I hope this is not confusing. In this diagram and the previous one, propositions from earlier junctions are included. In this one, only CLAIM′ / INST / CLAIM″ is newly created.
The next junction in (181) is that__claim. This also uses the latest instantiation of CLAIM to create:
It’s difficult to draw this together with the already-created propositions. However the ‘complementary nulls’ (see LS12) principle causes it to combine immediately with CLAIM′ / QUALIF / (null):
Note that the first of the QUALIF propositions means ‘claim1 is qualified by something’ while the second means ‘something is qualified by claim2’.
A second claim__that junction comes next and the resulting proposition is easily added:
Thus a sequence The claim that the claim that… can continue unconstrained although it becomes difficult to draw with single symbols for INST and QUALIF.
Actually sentence (181) continues …that the moon… If that__moon is processed in the same way, a proposition CLAIM″ / QUALIF / MOON is created.
Is MOON the most approriate SIC concept? The subordinate clause will create PANCAKE / INST / MOON so the proposition looks fine since the claim is more about the moon than about pancakes.
But no, this is irrelevant. The junction between the subject noun in a subordinate clause and a preceding that is not possible. The noun would be the dependent in the junction but it must also be dependent in a junction with the clause’s verb. A word can’t be dependent more than once. The junction must be (relative pronoun)__(verb).
Sorry for cheating
It’s been useful to make the wrong assumption in this piece because the main message has been about the particularising of nouns. Next time we’ll look properly at how the hierarchy of clauses is built.