Home » Articles » NG54. Attaching preposition phrases

NG54. Attaching preposition phrases

 Recent Network Grammar pieces showed how attachment of simple adverbs can be explained by NG.

A crucial point was that the C (category) concept for the parent of an adjunct can change as incremental processing of the sentence proceeds.  This means that the adjunct may be grammatical in some positions but not in others.

An account of how PPs attach was promised in NG30 but what follows is no more than a start on a very big topic.


NG51 to 53 showed that, while some sentences with simple adverbs are clearly grammatical and others clearly ungrammatical, for a significant number grammaticality is unclear.  The suggestion was that the debatable sentences may have the dependent adverb in a position where the word that ought to be its parent does not have a C (category) for which a C / R / C rule-of-combination exists.  The meaning of the putative junction is clear and so cognition, downstream from automatic language processing, can deal with the situation.

NG allows the relevant rule to be idiosyncratic.  Thus two native speakers may disagree about whether the sentence feels awkward or not, while agreeing that the meaning is clear.

My guess is that I am relatively tolerant, finding more of the sentences unexceptionable than many other English speakers will.

Attachment of PPs is a bit different.  The relevant junction involves the preposition, not its complement.  Prepositions form a word-class much smaller than adverbs but they allow a larger range of possibilities.  PPs may act as valents or as adjuncts.  Furthermore some of the words may act as particles or as prepositions.  Therefore language processing may need to keep open the possibility of (parent word)__(particle) even if (parent word)__(preposition) is impossible.

In view of these potential complications, my suggestion is that the reader revisit NG51 to 53 to try out variants of the sentences there using PPs in place of the adverbs.  The adverbs discussed in NG51 to 53 can be replaced by roughly equivalent PPs, all with in.  Gladly becomes in joy; sadly becomes in sorrow; probably becomes in my opinion; yesterday becomes in September; and today becomes in October.  This should make a good test because in can also act as a particle.

Presented here are my thoughts about differences in grammaticality and the reasons for those differences.  This avoids a detailed discussion of sentences here and may avoid alienating the reader if my grammaticality judgements are controversial.

If you’ve followed my suggestion …

Now read on

My conclusion is that broadly PPs are allowed or disallowed in the same way as simple adverbs.  The exception to this is where the adverbial precedes any argument of the verb.  In such cases a PP seems to me less objectionable than a simple adverb – for example, in variants of (5d) and (7d).

(244) Nero is giving in joy Olivia …

How can we explain this?  Unlike giving__gladly, giving__in can’t change giving1 to giving2.  This is because in may be a particle.  If that is the case, the following NP – joy in (244) – provides THEME of GIVE IN rather than the complement of the preposition.  (The meaning is weird but not impossible: giving__in can be used for returning something, a library book for example; so if Joy were a colleague of Olivia, Nero could be returning her to Rent-a-Slave.)

If this analysis is correct, how are the options – particle or preposition – kept open?  The reader should now revisit LS30 and 31 on disambiguating particle and preposition forms.  The parallel with sentence (104) should be obvious.

(104) Nero is giving up Vesuvius

This looks ambiguous even at sentence-end.  But LS31 claimed that there is a prosodic distinction between up as particle and as preposition, the latter being marked with a preceding pause and/or some emphasis.

Sentence (244) however continues with Olivia.  Only the preposition reading is possible.  Pause/emphasis is unnecessary and actually seems a bit stilted.  The tabular analysis starts in a similar way to that for (104) which has distinctive UP rather than up.

(244) Nero is giving IN joy Olivia

The analysis for (104 ) showed UP as REL in proposition 2.  Here REL is generalised to ADVERB because IN can be used for LOCATIVE as well as MANNER adverbials.  Perhaps the generalisation is better, with SIC carrying all the semantic load.

LS51’s discussion of ungrammatical sentence (20d) explained why the double-object form doesn’t work with an adjunct immediately after the verb: there are two changes of C and no giving3__(noun) junction is possible.  Therefore for giving2__Olivia in (244), there is no reason not to split OLIVIA initially across THEME and GOAL propositions as has been seen in analyses from LS13 onwards.

We’re getting there

Unlike the previous three Network Grammar pieces, this one is mercifully short.  However you have been asked to look at a lot of earlier material. Hard work for you but encouraging to me.  Increasingly, new sentences are being explained using existing analyses – which is what ought to happen as we approach the target of 57 pieces.

Before then we should take a look at production because up to now everything has been on comprehension.  Next time.