This completes Network Grammar’s review of particle/preposition ambiguity. One conclusion is that, in some circumstances, NG must deliver alternative propositions and leave cognition to determine which is the right one.
Nero is giving up
Without Vesuvius, sentence (104) becomes:
(18a) Nero is giving up
Nero is giving up Vesuvius
With Vesuvius the sentence can be comprehended two ways – with the junction giving__Vesuvius or the junction up__Vesuvius. There is no mystery: the distinction is prosodic. With a slight pause and then slight emphasis, up becomes a different phonological word and the junction UP__Vesuvius (i.e. a preposition phrase) is formed, there being no junction giving__UP.
(104a) Nero is giving UP Vesuvius
With no pause or emphasis, the particle is comprehended because there is a junction giving__up. Vesuvius then gives an argument rather than an adjunct.
(104b) Nero is giving up Vesuvius
The ideas presented above for disambiguating sentence (104) before delivery to cognition cannot be the whole story. Consider sentence (106) which does have particle-or-preposition ambiguity.
(106) John is looking up the street
If (106) has the prosodically distinct UP, a preposition phrase is recognised but it attaches to John, not to giving. It indicates the location of John while he is looking: a slightly odd reading but not impossible.
Ignoring the prosodic distinction and concentrating on (106) with plain up, a preposition phrase is still possible: it indicates where John is fixing his gaze. But a particle reading is also possible, indicating that John is using a street-map.
The particle reading can be achieved unambiguously by delaying up.
(107) John is looking the street up
But the possibility of (107) does not give (106) the preposition reading unambiguously. And (106) may be disambiguated either way by a continuation.
(108) John is looking up the street for a parking space
(109) John is looking up the street in Google Maps
My guess is that achieving unambiguous delivery to cognition from a sentence like (106) cannot be achieved. Using an incremental, left-to-right method as developed here in LanguidSlog, disambiguation would be implausibly complicated because language could have evolved to achieve the same effect by simpler means – for example, particle always following direct object as in (107).
The last word on the subject is therefore that disambiguation reliant on semantics or context must happen in cognition rather than in sentence processing. That being so, sentence processing delivers to cognition alternative propositions when multiple junctions are syntactically possible. One of these analyses would make better sense with the immediately preceding and following propositions in the ‘bundle’ and the other would simply fade away.
Nero is giving Olivia up Vesuvius
But would sentence (103), contrary to the NG analysis in LS30, deliver both OLIVIA / UP / VESUVIUS as well as GIVE / UP / SACRIFICE?
(103) Nero is giving Olivia up Vesuvius
Both Olivia__up and giving__up must be junctions in the lexicon. But no amount of prosodic tinkering can change native-speaker intuition – that up is the dependent of giving, not Olivia. It’s even difficult to contrive a preceding question that would prepare the ground for comprehending the Olivia__up junction. Who where is Nero giving? would surely prompt an answer something like Olivia up Vesuvius is Nero’s gift, not sentence (103).
However if the discourse had established that Nero owns two slave-girls, one located up Vesuvius and the other elsewhere, then Olivia__up would be comprehended.
My guess is that the two propositions are indeed delivered for (103). It’s easy to see how one of them ‘fades away’. That would be generally true for such sentences, even where the semantics of one reading are impossible:
(110) John is eating scrambled egg on toast
This is a consequence of NG processing one junction at a time. Of course, if scrambled egg on toast is a single lexical item, the issue doesn’t arise.
That’s all on particle vs. preposition. Have I been too superficial? If you want more on this – or on any other topic – please comment and I’ll discuss the points you raise.
Next week I have a bit of a whinge about the lack of any reaction to my ‘challenge’ way back in NG6. Really, if linguists can’t counter that, their ‘explanatory’ theories will lose any value.