We move away from give and analyse sentences with other ditransitives, ones that allow either the to-dative or the double-object form but not both.
Beth Levin has documented hundreds of different verb classes according to their syntactic behaviour. One class includes give along with sell, feed etc.
For NG to be valid it must work for all verb classes. Obviously it’s impossible here to show that to be so, but encouragement comes from looking at just a few other verbs.
There are other classes of ditransitive and some of them can form semantically coherent sentences with our Romans. We’ll try deny and restore in active and passive forms.
A verb that allows the double-object but not the to-dative is deny.
(46) Nero is denying Poppaea Olivia
(47) * Nero is denying Olivia to Poppaea
Absence from language knowledge of a denying__to junction makes (47) ungrammatical. (Other prepositions may be possible but a resulting preposition phrase would be an adverbial adjunct, not an argument of deny.)
In contrast, restore allows a to-dative but not a double-object.
(48) Nero is restoring Olivia to Poppaea
(49) * Nero is restoring Poppaea Olivia
(49) cannot be prevented by omitting restoring__(noun) from language knowledge as that junction is needed for (48). Instead the is__restoring junction must create a set of incomplete propositions different from that for is__giving. This prevents goal from being derived from a bare noun but still allows it from a preposition phrase.
(48) Nero is restoring Olivia to Poppaea
In (49) a GOAL proposition can’t be derived because that REL concept doesn’t occur anywhere. The last junction creates the proposition RESTORE / THEME / OLIVIA with six units. It’s complete but contradicts the RESTORE / THEME / POPPAEA that has already been delivered. (More typically, ungrammaticality is sensed where a proposition remains incomplete at sentence-end.)
Two bare nouns before the verb
(50) Olivia Nero is restoring to Poppaea
The analysis of (50) still relies on the assumption that the propositions from the first two junctions have different priority when combining with those from the third. The attraction of (null) / INST / NERO to one but not the other is even more enigmatic than in LS14 because the two propositions RESTORE / AGENT / (null) and RESTORE / THEME / (null) seem essentially identical.
There must be asymmetry somewhere. Note that there is asymmetry in sentence (43) as analysed in LS17; this is similar to (50) in that it too has two bare nouns before the verb. So, despite what was suggested in LS13, could the asymmetry be activation – say, activation on one of the propositions from is__restoring but not the other?
We’ll return to this issue when a more complete set of strange behaviours in NG has been collected.
With deny, fronting doesn’t pose the same questions.
(51) Poppaea Nero is denying Olivia
For passive restore, as expected, the equivalent of (11) works:
(51) Olivia is restored to Poppaea by Nero
The equivalent of (12) does not:
(52) * Poppaea is restored Olivia by Nero
Language knowledge must lack a restored__Olivia junction.
For passive deny, the equivalent of (12) works:
(53) Poppaea is denied Olivia by Nero
The expectation is that the equivalent of (11) should not work:
(54) ? Olivia is denied to Poppaea by Nero
To me, (54) is just about acceptable, contradicting Levin (1993).
NG explains how idiolects can vary
A virtue of the NG analysis is that it can easily account for differences of opinion on (54) by claiming that the junction denied__to may be omitted from an idiolect or may be included even if denying__to is omitted. And any other variation of perceived grammaticality in marginal areas may similarly relate to presence/absence of a single junction.
To minimise tedium, I’ll not include any analysis of sentences with monotransitive verbs. If you’re interested, look at the sentences listed in LS15 that omit (to) Poppaea, substitute a verb like love, and do the NG analysis. Very simple.
More and more
What about interrogatives? Yes, the easier ones next week; harder ones to follow.