Even without inflection, English has some flexibility in the order of predicate and arguments. Recent Network Grammar pieces have shown that there is less than total flexibility in the placement of particles and adverbs. This piece reviews the situation in relation to NG, in particular the idea of alternative C (category) concepts that was developed in NG26.
One interesting conjecture about the attachment of adverbs deserves further investigation. If sustainable, it will provide compelling support for NG.
Position or property?
Earlier we saw how bare nouns can appear in various sequences in a sentence. Grammatical sequences are those where the thematic roles of the nouns can be determined correctly.
We’ve also seen how a particle or adverb has a role which is a property of the word itself, not a consequence of its position within the sentence. The same is essentially true of a preposition, although the phrase it heads may be GOAL, AGENT or adverbial depending on the main verb.
Variants with a purpose
(5) Nero is giving Olivia to Poppaea
(8) Nero is giving to Poppaea Olivia
(9) To Poppaea Nero is giving Olivia
These are all easy to analyse with NG. Sentence (5) seems to be the unmarked form; the sequence as in (8) would allow a heavy noun phrase as THEME; and (9) topicalises GOAL. Why is there not a fourth possibility?
(101) * Nero to Poppaea is giving Olivia
The curious thing is that NG could analyse (101) perfectly well. Language acquisition has somehow left us with Nero__is disallowing to__is but not vice versa as in (9).
In this case, what-allows-what is quite subtle. Remember that in (9) Nero_is is processed before to__is; in (101), to__is is processed first but disallowed. In (9) Nero__is must activate a C that allows the otherwise disallowed to__is. This is an example of broadening the range of junctions that readers of LS26 were asked to think about.
It’s understandable that there are different word orders for different purposes and that other orders may be redundant. Redundant ones are perceived as ungrammatical but are quite comprehensible.
Less obvious cases
It may be difficult to see ‘different purposes’ in (5) and (20).
(5) Nero is giving Olivia to Poppaea
(20) Nero is giving Poppaea Olivia
Perhaps the two forms emerged originally for focus (although it’s not obvious why two forms should not be available for all ditransitives). One might use (5) if the previous sentence mentioned Olivia but not Poppaea, and (20) if the previous mentioned Poppaea but not Olivia.
As discussed in LS24, there could be a good reason for the existence of both (88) and (90).
(88) Nero gave fiddling up
(90) Nero gave up fiddling
Adverbs can attach more freely but not quite everywhere.
(42a) Sadly Nero gave Olivia to Poppaea
(42b) Nero sadly gave Olivia to Poppaea
(42c) * Nero gave sadly Olivia to Poppaea
(42d) Nero gave Olivia sadly to Poppaea
(42e) Nero gave Olivia to Poppaea sadly
(44a) Sadly Nero gave Poppaea Olivia
(44b) Nero sadly gave Poppaea Olivia
(44c) * Nero gave sadly Poppaea Olivia
(44d) * Nero gave Poppaea sadly Olivia
(44e) Nero gave Poppaea Olivia sadly
Adverb sadly immediately after the verb is bad in both to-dative and double-object sentences. The junction gave__sadly must be available because Nero gave sadly is an OK sentence. But it must activate a C that disallows gave__(noun). This also makes (44d) sound awkward.
But there are still several other possibilities for placing sadly. My hunch is that adverbs qualify what is nearest. Often that’s the verb but it can be a noun phrase. Here’s some evidence.
(43) Olivia Nero gave to Poppaea
(43a) Olivia Nero gladly gave to Poppaea
(43b) * Olivia gladly Nero gave to Poppaea
(43c) * Gladly Olivia Nero gave to Poppaea
In (43b) and (43c), I sense ungrammaticality before gave is reached and I assume native speakers would mostly agree. No junction with gave can yet have occurred and gladly must be in a junction with Olivia. I suspect this causes OLIVIA to be treated as AGENT, making it impossible then to deal with Nero.
You might argue that, while the sequence Olivia gladly Nero in (43b) can’t be rescued, Gladly Olivia Nero in (43c) can be.
(102) Gladly Olivia Nero and Poppaea gave alms to the poor
True, but OLIVIA is AGENT in (102) also and my hunch is not invalidated.
Although gladly is a manner adverb and strongly AGENT-related, my comments on the (43) sentences would still apply if yesterday were used instead.
Generally in the sentences we’ve looked at, the position of an adverb doesn’t affect semantics much. Even if the freedom is because attachment is to nouns, sadly/gladly seem to apply to Nero and his act of giving, not to Olivia or Poppaea; while yesterday applies purely to the event.
The only exception is sentence-initial Sadly, which can be used instead of The speaker is sad that… Gladly doesn’t work that way. Interestingly.
That the semantics of such an adverb doesn’t vary with its position in the sentence supports the idea that it qualifies a noun. But what proposition is created for, say, Poppaea__sadly in (42e)? It shouldn’t be POPPAEA / HASP / SAD because that would come from sad__Poppaea. What is needed is a QUO concept that covers the clause generally.
If the adverb qualified the main verb, activating alternative Cs to limit possible word-orders would be straightforward. With three nouns in play, all offering (noun)__(adverb) or (adverb)(noun) junctions, the ‘alternative C’ mechanism could not work. Changing the C for an earlier noun could not affect a later noun because there are no (noun)(noun) junctions here.
Down to Earth
Attachment of adverbs is a big, complicated topic so it’s too early to get excited about the effect of alternative Cs. Besides, you may detect a whiff of circularity in my argument. Please don’t hesitate to comment.
Whatever, I’ll be back on give…up next week – and hoping to finish it.