Kim and Sandy met in Baltimore in the Hyatt hotel in
the lobby.
[P&S 1987, (257a)]
Complements, on the other hand, cannot be iterated. Thus in d12-d14 the prepositional phrases seem to be complements rather than adjuncts.
*Adam gave a book to Debbie to Frank. *Adam told a story to the kids to the adults. *Adam sent a letter to Mary to Diane.
By the same logic, however, it would appear that the prepositional phrases in d15-d17 are complements as well, in contrast to the results of the ``do so'' test reported above.
*Sam kicked a ball to Bill to Frank.
*Sam sent a letter to Bill to Frank.
*Adam baked a cake for Debbie for Susan.
The problem here is that the iterability criterion presupposes a particular semantic relation between the iterated constituents. Adjuncts are not in general iterable if the semantic contribution of each adjunct contradicts the semantic contribution of a previous adjunct. Consider d51a in contrast to d51b: the first sentence is ungrammatical while the second is not.
*Sam kicked a ball at 10 o'clock at 8 o'clock. Sam kicked a ball in the morning at 10 o'clock.
The difference here has nothing to do with a difference in status between the PPs, but rather with the fact that certain semantic roles can be multiply specified via a particular semantic relation. In this example, we see that temporal adjuncts can only be iterated if the information conveyed by a given adjunct is contained in the information conveyed by previous adjuncts. One point in time does not contain another and so the adjuncts in d51a are contradictory, but a point of time is contained in a span of time and so the adjuncts in d51b are considered together to make precise the time at which the event occurred.
So adjuncts can really only be iterated if the semantic (meaning)
contribution each makes is in a relation of containment to the
previous adjuncts. Thus in d18 above, the adjuncts can be
iterated because each one can be interpreted as being contained within
the location specified by the previous adjunct, making more precise
the locative information, rather than providing an overriding semantic
contribution. In d12-d17, the prepositional phrase specifies
the (intended) recipient of some object. The containment relation does
not apply to distinct recipients and therefore these adjuncts are
incompatible with iteration. This analysis is confirmed by the data
in d77 and d78, which contrast with d18 and d16
respectively. Sentence d77 is ungrammatical because Chicago
cannot be contained within Baltimore, while d78 is grammatical
because the head office of the Times is contained within New
York.
*Kim and Sandy met in Baltimore in Chicago. Adam sent a letter to New York to the head office of the Times.
Furthermore, some of the examples Pollard and Sag provide of adjunct iteration rely on pragmatic factors and do not seem to be wholly grammatical. For example, sentence d19, introduced as grammatical and felicitous by Pollard and Sag, in my opinion is not entirely felicitous, and certainly can only be interpreted with the two prepositional phrases as adjuncts if the comma indicates a conjunction such as ``and''.
Heather opened the rusty lock with a key, with a pair of pliers. [P&S 1987, (257e)]
Applying this conjunctive interpretation requirement to the ungrammatical sentences above improves their acceptability, as shown in d20-d23. These sentences seem to display ellipsis, rather than providing a sense of the underlying argument structure. The analysis of such sentences must be discourse-based, rather than purely based on the syntax/semantics interface, in order to appropriately identify the ellided meaning. In these examples the ellision is intra-sentential, but it is still governed by discourse factors controlling reference resolution.
Adam gave a book to Debbie and to Frank. Sam kicked a ball to Bill and to Frank. Sam sent a letter to Bill and to Frank. Adam baked a cake for Debbie and for Susan.
We must conclude on the basis of the examples above that the possibility of iteration of PPs is not a reliable indicator of argument structure, and in fact cannot be viewed as purely syntactic since the phenomenon of iterability seems to interact with semantic and discourse-level factors. The evidence it provides for treating both types of dative prepositional phrases as complements rather than adjuncts will not be taken as definitive.