LS24 summarised the empirical behaviour of give up. This time we look for an explanation of the varied semantic effects. Lexicon-driven sentence processing could account for a surprisingly large proportion of the data.
One possibility is that the QUO shared by the AGENT, THEME and GOAL propositions can have a different concept substituted for the one initially linked (the plain GIVE seen in earlier analyses). The replacement concept would be selected according to the presence/absence of null for SIC in those propositions (see LS24).
This idea is rejected because it would conflict with the assumption that a proposition is delivered to cognition as soon as it is complete. Compare:
(5a) Nero is giving up Olivia to Poppaea
(9b) To Poppaea Nero is giving Olivia up
When giving__up is encountered, in (5a) none of the propositions is complete; in (9b) all are complete. The semantic effect of up is the same in both sentences. It’s possible that the substitution of a different concept as QUO could work in (5a). But it can’t work in (9b) because all the propositions have already been delivered.
A better explanation of the effect of giving__up is that it adds another proposition to the bundle for the sentence. Where in sequence the proposition is added doesn’t matter. The proposition is GIVE / PARTICLE / SACRIFICE.
The proposition acts as a ‘manner’ adverbial and the REL concept could be MANNER. But it has been generalised to PARTICLE since particles may have other sorts of adverbial effect. Conceptual content can be concentrated on SIC, there being no obvious need to share it between SIC and REL
The SIC concept SACRIFICE is adopted so as to be compatible with but distinct from the eventual GRUDGE, QUIT or YIELD. This may be important because, when giving__up is encountered in (5a), all those possibilities are still open.
Selection between GRUDGE, QUIT and YIELD depends on the combination of AGENT, THEME and GOAL in the propositions delivered. ‘No ghost in the machine’ means the semantic effect accrues simply from the combination of propositions in the bundle:
The available senses for give up are:
Ignoring the AGENT column above, it’s possible to discern a pattern:
+THEME +GOAL → GRUDGE
+THEME –GOAL → QUIT
–THEME → YIELD
But these senses can’t result from the sort of ‘no ghost’ mechanism used in NG analyses. If GIVE / PARTICLE / (null) were created at up, how would the null then be replaced with GRUDGE or QUIT or YIELD? The particle can’t form a junction with the noun for THEME or GOAL: one or other would need to be the dependent but both are dependents of the verb.
Is GIVE / PARTICLE / SACRIFICE delivered as soon as up is encountered? That’s possible but the semantic effects would have to be achieved in the wider network. Here’s a possible model:
Labelled circles represent concepts but the graphics are otherwise quite different from the ‘triangle’ diagrams that appear from time to time in LanguidSlog. The dotted arrows represent progression through the network. SACRIFICE is activated by the sentence and THEME and/or GOAL may also be. Somehow one of YIELD, QUIT and GRUDGE is activated by progression from two other places – from SACRIFICE and from presence/absence of THEME and GOAL. Therefore THEME and GOAL must each be able to act as a switch determining the direction of subsequent progression depending on whether or not that concept is already activated.
The diagram shows one particular case. Grey circles are concepts not activated and grey arrows are paths consequently not followed. The sentence has THEME but not GOAL and therefore QUIT is invoked.
None of this is covered by the mechanisms described in LS21. But if the additional mechanisms can be explained, then we will have a way to account for the semantic subtlety of give up. It’s speculative but this topic deserves to be picked up again later in LanguidSlog.
Context is important
Anyway the table above is not totally precise. For example:
(98) Nero gave up his throne
(99) Nero gave up his seat
Sentence (98) has the sense of QUIT if throne is a metonym for the job of emperor. But (99) would be more like GRUDGE if the emperor’s purpose was to allow someone else to sit down.
Only some of the semantics can be worked out in sentence processing. The propositions delivered to cognition must invoke meaning from associations, already in the mind, between the concepts in those propositions. This conclusion can be supported by several observations.
First, give without up also varies in meaning depending on the presence or absence of propositions. Specifically, GIVE with –THEME and no particle usually indicates benevolent donation.
Second, other particles that work with give have idiosyncratic behaviours. For example, give away with AGENT and THEME has distinctly different meanings where THEME is not human (= GIVE UNCONDITIONALLY SOMETHING OF VALUE) and human (= BETRAY), and the human case has a totally different meaning if a bride.
Third, other verbs in the give class behave idiosyncratically with particles. With up, sell has a particular meaning, often achieved with AGENT alone but unaffected by the presence of THEME and/or GOAL. And feed doesn’t work at all (in Yvette fed the goose up with maize, feed is from a different verb class where it’s the preposition phrase rather than the direct object that specifies the foodstuff).
Despite LS7’s extravagant claims, P / C / M words and C / R / C rules can’t be stretched to cover all the subtleties of English phrasal verbs. I didn’t actually deny that there is a ‘syntax/semantics interface’ – but you can expect NG to bring it into sharper focus than other theories manage.
NG analysis of give…up sentences will start next time.