The interplay of particle up with verb give and its arguments and possibly an adjunct present a big challenge. NG’s solution must explain all the syntax. Ideally it would also show that delivery of the PARTICLE proposition is timed such that LS25’s idea about semantic processing is possible.
This piece proposes an extension to the model established way back in LS7 and LS8. The extended model is powerful. As well as answering the questions posed by particles, it might provide alternative analyses for some of the trickier sentences seen earlier. LanguidSlog will not cover those possibilities but it will routinely use the extended model wherever appropriate from here on.
Dependent up may occur after a parent give form with no other dependent intervening or, for active sentences, a dependent noun optionally intervening. There must be a reason why the give__up junction is:
- allowed if the give form has not yet acted as parent to a following dependent
- allowed if the give form is active and has already acted once in a give__(noun) junction
- disallowed if the give form has already acted as parent to any other following dependent – to, by, a further bare noun, an adjunct.
There must also be a reason why a sentence with give__up becomes awkward if an adjunct is added to the right of and close to the verb. This effect can occur even if up is absent.
NG capabilities seen already
Earlier analyses in LanguidSlog show give forms creating incomplete propositions to accommodate arguments of the predicate. Active forms create three propositions, passive forms two. These propositions allow temporary uncertainty about the thematic roles of bare nouns. Resolving the uncertainty sometimes depends on a to– or by-phrase making the role of a noun unmistakable.
The examples in LS24 might suggest that the GIVE / GOAL proposition explains the behaviour of up. While it remains incomplete, give__up can be accommodated. But give__up is ungrammatical if encountered after GIVE / GOAL is complete and delivered. Does GIVE / PARTICLE / SACRIFICE somehow attach to GIVE / GOAL?
No, the two propositions only have give in common and must be separate. Furthermore a sentence such as (9a) delivers GIVE / GOAL / POPPAEA and then happily accommodates give__up.
(9a) To Poppaea Nero is giving up Olivia
We need a means by which, in the processing of a sentence, a junction may limit possible junctions later on. For example, a typical sentence starting Nero is giving can have giving__Olivia, giving__to and giving__up in various orders. Any of these is possible until giving__to occurs and makes giving__up unavailable.
The key to this is the C (category) concept in the propositions defining words and rules. In the example, C1 for giving is in the rules for giving__Olivia, giving__to and giving__up. When giving__to is processed, C1 is replaced by C2. C2 is associated with the same P (phonological word) and M (meaning) concepts for giving. However C2 is only in the rules for giving__Olivia and giving__to. Therefore up cannot occur in the sentence after to Poppaea.
Readers are now respectfully asked to (re)visit LS7 and LS8 to ensure a good grasp of the fundamentals of NG words and rules. The following diagram is meant to be logically identical to the first one in LS8.
The diagram has been distorted because we’ll shortly add another word and it’s not easy to fit in.
Also the annotation is slightly modified. Phonological word PI is is, PG is giving and PT is to. The other concepts are subscripted correspondingly. The C for giving is CG1 to distinguish it from CG2 which will appear shortly.
Alert readers may object that only a single M / R / M proposition is shown for is__giving but the NG analyses have three (R = AGENT, THEME and GOAL respectively). I can only say that the diagram would become hopelessly complicated if I tried to show all that.
We now add giving__Olivia (skipping over to__Poppaea which doesn’t advance the discussion).
Subscript O indicates Olivia.
The diagram looks OK but it’s not, because we could instead add giving__up, the rule for which also uses CG1.
If there is a rule that allows CG2 as well as CG1 for giving__Olivia but no rule that uses CG2 for giving__up, the following diagram gives the right result, allowing …giving to Poppaea Olivia but not …giving to Poppaea up.
The function of the proposition CG1 / X / CG2 is to transfer activation from the QUO concept to the SIC. But what is X, the REL concept? Neither of the Cs has any content so it feels odd to say that CG1 is instantiated by CG2. But let’s settle for INST and move on.
Shades of grey
This approach may help us understand why grammaticality is not always black-and-white. Consider sentence (100).
(100) Nero gave sadly Olivia to Poppaea
LS24 called this ‘clunky’. I don’t think it’s too bad. It’s not the sort of sentence that can’t be processed automatically and gets passed to cognition to sort out (like a learner deals with a sentence in their target language).
After Nero__gave, both gave__sadly and gave__Olivia are available. The occurrence of gave__sadly activates a C that disallows gave__Olivia. If gave__Olivia does occur, it can be processed because the original C is still there.
But gave__sadly also disallows gave__up. In (97), two successive junctions needing recourse to the original C turn ‘clunky’ into ‘bad’.
(97) * Nero gave sadly Olivia up to Poppaea
In (96), ‘bad’ is caused differently – by up__Olivia being treated as a preposition phrase.
(96) * Nero gave sadly up Olivia to Poppaea
I have no idea how the C / X / C propositions are acquired. But that doesn’t prevent us from looking next time at some tabular analyses of give…up sentences.
Here’s something to think about. The emphasis in this piece has been on narrowing the range of junctions available later in the processing of a sentence. The method could also broaden the range. Any ideas about possible uses for that capability?