Coordination post-verb presents a different set of problems. Before starting to look at what is possible and what not, this post presents some final thoughts on coordination of fronted material.
It also shows that the semantic properties of nouns must sometimes play a part in sentence processing. From here the sentences therefore use gave with John, Ken, Luke… for AGENT; apples, bananas, cherries… for THEME; and Mary, Nancy, Odette… for GOAL.
With these players, sentence (118) from LS33 becomes:
(119) Mary and Nancy John and Ken gave apples
This seems acceptable and describes one act of giving by a pair of boys to a topicalised pair of girls. However it is not possible to use the fronted construction to describe two acts of giving, each by one boy to one girl:
(120) * Mary John and Nancy Ken gave apples
Sentence (120) is comprehended as one act of giving by Ken to the three others. Although a THEME + AGENT pairing might provide some semantic help, multiple pairings before the verb are dubious:
(121) ? Apples John and bananas Ken gave to Mary
After the verb
But coordination of pairings is fine after the verb:
(122) John gave Mary apples and Nancy bananas
Coordination in (122) is of GOAL + THEME pairings but it is still possible to have separate coordinated GOAL or THEME after the verb:
(123) John gave Mary and Nancy apples and bananas
This seems simple enough to be treated in a similar way to pre-verb coordination. But before looking at that in detail, the coexistence of multiple GOAL and THEME nouns must be properly understood. Consider:
(124) John gave apples bananas cherries damsons and…
(125) John gave Mary bananas Odette damsons and…
A native-speaker would have no difficulty understanding these sentences as they unfold. The work is done in sentence-processing, not in cognition. This assertion is made because no garden-path is experienced and structures like these are so ubiquitous that they must have evolved to suit automatic processing.
At bananas, (125) is analysed like sentence (20), but (124) is not. Two nouns immediately following the verb can be GOAL-then-THEME or coordinated THEME or coordinated GOAL. As for the nouns in these sentences, apples and bananas are plausible as THEME, implausible as GOAL; while Mary is plausible as GOAL, less plausible as THEME.
Is previous usage involved?
Assessing plausibility might be based on usage history but that would present difficulties.
In incremental processing as described in these essays, there is no distinction between gave__(noun) where the noun turns out to be THEME and where GOAL. Recording what has actually been delivered to cognition would require a complicated mechanism to feed back into the C / R / C rules.
Furthermore, usage of a junction is not a reliable measure of plausibility because it is distorted by the frequency of occurrence of the words in the junction. Sentences (124) and (125) would be no more ambiguous if, instead of all common forenames and fruits, they included items such as Dorian and durian – as long as they are recognised by the hearer.
A better account can be based on the notion of ‘plausible as goal’ for a noun. Beth Levin (1993) shows that verbs allowing the double-object construction only allow a narrow range of concepts to be represented by nouns immediately after the verb, i.e. as ‘indirect object’. Roughly this can be characterised as animate concepts. However there are grey areas and idioms, such as John gave his shoes a shine, that need to be explained separately.
It follows that, for a junction like gave__(noun), the lexical-entry for the noun can be of two types: either marked for nouns that can act as indirect object, or unmarked for other nouns. NG allows marked and unmarked lexical-entries to link via different C nodes to rules-of-combination with give-type verbs.
For a sentence with two nouns following give, noun1 determines whether the unmarked C (for apple etc) or marked C (for Mary etc) is in play. When noun2 is processed, the first junction that is sought is (noun1)__(noun2). These junctions are only available if noun1 and noun2 are of the same type.
If that is so, the applicable C / R / C rule may be (unmarked noun) / COORD / (unmarked noun) as in sentence (124). Or the C / R / C rule may be (marked noun) / COORD / (marked noun) as in John gave Mary Nancy Odette Patty and… .
But there is no C / R / C rule with COORD and both (marked noun) and (unmarked noun). Thus in (125) there is no Mary__bananas junction and gave__bananas is tried instead. With noun1 marked and noun2 unmarked, the sentence is analysed as double-object like (20).
(20) Nero is giving Poppaea Olivia
If noun1 is unmarked and noun2 marked, the sentence is ungrammatical. It might be analysed as in (20) and delivered to cognition to salvage something – perhaps by assuming preposition to has been missed in articulation or audition.
Not 100% but not a problem
A correct analysis in automatic sentence processing is not guaranteed. But errors should be infrequent, surfacing in cognition as garden-paths. This suggests the analysis for (20) in LS15 is false.
With its two marked nouns, (20) is a GP, sentence-processing delivering POPPAEA / COORD / OLIVIA and leaving OLIVIA / COORD / (null) awaiting completion with another marked noun. But cognition should easily interpret the delivered bundle because OLIVIA / HASP / SLAVE can be presumed for the discourse and then GIVE / THEME / SLAVE is unexceptionable.
Although signifying a chattel, Olivia unmarked is a less attractive hypothesis because it would mean Nero gave Olivia a dog is a garden-path whether dog is marked or unmarked.
But what about coordination?
Having established what can coordinate with what, it’s quite straightforward. In (125) there are junctions for Mary__Odette, bananas__damsons etc. Note that the leftmost item is the one in a junction with the verb. For material preceding the verb (as in NG33) it is the rightmost.
The coordination story continues next week – and there might even be some tabular analysis.