This section has discussed the complementation of nouns by noun phrases or PPs. There are only a limited number of relevant noun classes that allow this: (i) relational nouns, which are typically non-derived; (ii) nominalizations derived from verbs or adjectives; and (iii) picture/story nouns, which can be either derived or non-derived. We have applied the four tests described in Section 16.2.1 for distinguishing adjuncts from complements to all these types of nouns. Overall, the application of these tests has proved useful in gaining more insight into the status of the PPs accompanying these nouns, not in the least because in many cases the results are unexpected and the discussion has provided us with interesting new material for further research. At the same time, the fact that it is often not possible to give a clear answer to the question of whether a particular constituent functions as an adjunct or as a complement raises the question of how real this distinction is.

Let us briefly evaluate the results of the four tests. The first test, concerning the obligatoriness of the PP, is inherently problematic, since the context plays a crucial role in deciding whether a particular element can be omitted or not. On the whole, however, intuitions seem to agree on which element should in principle always be present. The second test, according to which complement van-PPs can be used as a predicate in a copular construction, actually works quite well, but only distinguishes possessive van-PPs from other van-PPs. The third test concerning R-pronominalization also works well, but has the disadvantage of not being easily applicable to [+human] constituents. Finally, the results of the fourth test concerning PP-extraction are far from clear: only topicalization seems to provide more or less usable results for determining the status of the PPs in question. Even then, however, no firm conclusion can be drawn, because the supposedly extracted PPs can often, or perhaps even better, be analyzed as independent clausal constituents, especially as restrictive adverbial van/over-PPs.

The systematic application of these tests has revealed a number of other unexpected facts. First, there is a marked difference between verb-inherited arguments that take the form of van-PPs (usually NP-themes in the related verbal constructions) and PPs introduced by other prepositions selected by the input verb. Second, there is massive variation between the different types of deverbal nouns. One would expect the behavior of inherited arguments to be the same for all instances of nominalization, regardless of the type of nominalization and the preposition used. Yet, there is a clear difference between, for example, ing-nominalizations and inf-nominalizations as far as the results of the tests are concerned. This seems to indicate that the degree of verbalness of the constructions (the number of verbal features they have) plays a role in the results of the tests, since inf-nominalizations are by far the most verbal and ing-nominalizations are by far the most nominal of the deverbal nouns. Third, the tests work relatively well for theme arguments, but are not easily applicable to agent or recipient arguments. In particular, although the agent is an obligatory argument in the verbal domain, in the nominal domain agents do not behave as obligatory complements, even though the agent is semantically implied in most cases. Recipients behave more like themes, but their complement status is still less obvious.