NG35 proposed that, after delivery of propositions from the first clause, the set of propositions for the verb is carried forward but with activation and SIC concepts cleared. In this post the idea is examined more closely. This leads to discussion of sentences where a clause is formed from the brought-forward verb plus three fresh nouns. The grammaticality of these is dubious and therefore a full NG account of their operation is judged inappropriate.
Carrying forward the argument structure
In (122) and (126) to (128), processing of the first clause and clearing of the SIC concepts would be:
(122) etc John gave Mary apples and xxx…
This is valid even if xxx is coordinated with apples (both being unmarked), generating a complete XXX / COORD / APPLE proposition. Eventually a sequence of nouns coordinated with apples must end. Then either the sentence also ends or else the sentence continues with xxx for which there is no apples__xxx junction and xxx must therefore be part of a following clause.
The analysis for the rest of (122) is quite straightforward:
(122) John gave Mary apples and Nancy bananas
The analysis for (126) must accommodate the new verb:
(126) John gave Mary apples and sold Nancy bananas
Three fresh nouns
An analysis for (127) would have to handle three successive gave__(noun) junctions with concepts for the nouns eventually linked to AGENT, GOAL and THEME.
(127) John gave Mary apples and Ken Nancy bananas
This is difficult. Sentence (122) shows how two successive gave__(noun) junctions give propositions for GOAL and THEME. When the third such junction is encountered in (127) the process would have to backtrack to ensure GIVE / AGENT / KEN. A process that backtracks when details emerge from later in the sentence is not allowed in NG. The ban is one of the things that distinguishes NG from most other grammars. Fortunately there are at least three ways that (127) can deliver the intended cognitive effect without backtracking.
First, handling of the sentence is not fully automatic but downstream cognition has to do some ad hoc processing – in much the same way that the ungrammaticality of non-native speakers or infants is dealt with.
Second, there is ‘priming’ that allows the correct propositions to be formed. For example, utterance of (127) might come immediately after Which boys gave which girls which fruit? Thus the hearer already has GIVE / AGENT / BOY, GIVE / GOAL / GIRL, GIVE / THEME / FRUIT, BOY / INST / KEN , GIRL / INST / NANCY and FRUIT / INST / BANANA . So activation by the sentence of KEN, NANCY and BANANA has exactly the required effect.
Third, there is some prosodic variation on Ken ensuring that the word is treated as clause subject. To be unmistakable, John needs to have had the same variation: JOHN gave Mary apples and KEN Nancy bananas. Emphasis (or whatever) on JOHN changes the C concept for gave such that the only junction KEN can form is with this C, generating GIVE / AGENT / KEN. All the other nouns are processed as in (122) regardless of which C is in play.
Could NG apply?
It’s tempting to think that instead of the above possibilities there’s a fully analytic explanation. This would have each gave__(noun) junction deliver a concept to complete one of the brought-forward propositions. A following gave__(noun) junction would somehow displace the earlier noun-concepts in a sort of circular fashion – proposition 2 to proposition 3 to proposition 1:
Intriguingly this could work for the equivalent of (127) with indirect-object fronting:
(132) Mary John gave apples and Nancy Ken bananas
It works if you believe that (132) delivers the meaning for Nancy gave Ken bananas, not Ken gave Nancy bananas. That does seem to be the case and the meaning is not definitely switched by emphasising MARY and NANCY.
However the idea is rejected here for two reasons. One is that (132) and (127) seem not to be processed effortlessly: they feel garden path-ish . The other is that it’s not easy to see how the ‘circular displacement’ would work. The 4:2 split of activation would have to be involved to allow (122) to work; and perhaps everything remains undelivered until clause-end. Suggestions would be welcome.
Only two successive (verb)__(noun) junctions occur in the to-dative equivalent to (128) and the analysis seems straightforward.
(133) John gave apples to Mary and Ken bananas to Nancy
But there is a problem. The proposition GIVE / GOAL / KEN should be delivered at bananas to be consistent with all previous NG analyses. KEN needs to be retained so that it can be displaced on to the AGENT proposition at to.
An explanation of that is needed. It’s tempting to glibly say that complete, fully-active propositions are not actually delivered until clause-end. Perhaps that is so, but a comprehensive review of LanguidSlog is needed before that can be said with real confidence.
La garde recule?
That last point is important but not threatening to the whole project. In contrast the question of backtracking needs to be handled with great care. The story for sentences (127) and (132) is sound enough but there are more problematic cases to be explained. Any weakness in the argumentation could be seized upon by diehard generativists to rubbish NG. For example, see recent comments on NG4 by ‘KA’.