The distinction between dependent and independent van-PPs relates to a difference in interpretation, which can be made more conspicuous by adding the ordinal numeral eerstefirst; cf. also (475) and (478) in Section 16.2.5.1. The examples in (530a&b) turn out to be ambiguous: if the van-PP depends on the picture noun, as in the singly-primed examples, we are not dealing with the first drawings the speaker ever made or bought, but merely with the first drawing of/by Rembrandt he made/bought; if the PP is independent of the noun, as in the doubly-primed examples, we are dealing with the first drawings the speaker made or bought. The second reading is not available for (530c), which can only mean that it is the first picture made by or depicting Rembrandt that is damaged. From this we can conclude that the van-PP cannot be used independently with verbs like beschadigen.

The insertion of the negator nietnot can also be used to disambiguate these sentences, provided that we assume that it can only be placed between the noun and an independent van-PP, i.e. that it cannot occur internal to the noun phrase. This correctly predicts that (531a&b) is acceptable only on the reading that the picture in question is the first one the speaker has ever made/bought, whereas the construction in (531c) is unacceptable, since neither of the two possible structures is available.

Topicalization of the noun phrase can also be used to distinguish the two readings. Of course, since van-PPs can be dependent on the noun with all verbs, it stands to reason that it is always possible to topicalize the van-PP together with the noun phrase. Since in Dutch only a single constituent can be placed in sentence-initial position (the constituency test), the prediction is that only the structures in the singly-primed examples are acceptable. This prediction is again borne out: the primeless examples can only mean that the speaker made/bought/damaged a drawing of/by Rembrandt for the first time in 1982; the reading that should correspond to the doubly-primed examples, i.e. that the drawing in question was the first the speaker ever made/bought/damaged, is not available.

If we consider the possibilities of topicalizing only the picture noun, i.e. of stranding the PP in its original position, we would expect only the doubly-primed structures to yield acceptable results. The examples in (533) show that this is also confirmed: the examples in (533a&b) can only express that the drawing in question is the very first one the speaker made or bought, which means that these sentences must be assigned the structures in the doubly primed examples. The fact that (533c) does not allow this kind of topicalization again shows that the PP cannot be used independently with the verb beschadigento damage.

The examples in (534) involve topicalization of the van-PP, which, not surprisingly, seems to be possible only with independent van-PPs. This is clear from the interpretation of (534a&b): example (534a) expresses that the first drawing the speaker ever made was one of Rembrandt, and example (534b) expresses that the first drawing the speaker ever bought was bought from Rembrandt (source). As expected, example (534c) is degraded: the reading associated with (534c') is at best marginally acceptable in the (unlikely) context that the speaker is planning to damage several of his drawings of/by Rembrandt.

Note, however, that it is possible to force the reading associated with (534a'), according to which the drawing in question is not the very first one the speaker made, but his first one of Rembrandt, by adding an adverb like nunow or inmiddels ookby now also. Obviously, this cannot be immediately accounted for by assuming that only PPs functioning as independent constituents can be topicalized.

The fact that (535) can express the reading associated with (534a') can thus be taken as an argument for the claim that theme arguments of picture nouns can be extracted from the noun phrase by topicalization; however, this would come at the cost of being unable to explain the robust syntactic pattern established in (532) to (534). We might therefore consider a non-syntactic account, related to the observation that the use of adverbial phrases such as nunow and inmiddels ookby now also entails a context in which the speaker has planned to produce a series of drawings of famous painters. The function of the restrictive van-PP seems to be to make a selection from a presupposed set of painters: all this in turn would give the participants in the discourse sufficient reason to assume a phonetically empty but syntactically expressed theme argument of the picture noun.

The distinction between dependent and independent over-PPs relates to a difference in interpretation, which can be made more prominent by adding the ordinal numeral eerstefirst. Consider the ambiguous example in (545a). If the over-PP is dependent on the story noun, as in the primed example, we are not dealing with the first book Jan has ever written; it is merely the first book about World War II that he has produced. If the over-PP is independent of the noun, as in the doubly-primed example, the book in question is the very first one Jan wrote. Sentence (545b) does not allow this kind of ambiguity: the noun phrase must refer to the first book on World War II that Jan wrote or acquired (depending on the interpretation of the possessive pronoun zijn as agent or possessor).

Insertion of the negator nietnot can also be used to disambiguate these sentences, if we assume that negation cannot occur within the noun phrase, i.e. that it can only occur between the noun and an independent over-PP. This correctly predicts that (546a) is acceptable on the reading that the first book Jan ever wrote was not about World War II, whereas the construction in (546b) is unacceptable, since neither of the two possible structures is available.

Topicalization can also distinguish between the two readings. Since over-PPs can be dependent on the story noun with both verb types, it should not surprise us that it is always possible to topicalize the noun phrase and the PP as a whole. Since in Dutch only a single constituent can be placed before the verb in sentence-initial position (the constituency test), we predict that only the constructions in the singly-primed examples are acceptable. This prediction is correct: (547a) can only mean that Jan wrote his first book about the World War II in 1982, and not that this book was the first one he ever wrote.

It is also possible to topicalize only the story noun, i.e. to strand the PP in its original position. Here we would expect only the doubly-primed examples to yield acceptable results. The examples in (548) show that this prediction is borne out: example (548a) can only express that the book in question is the very first one that Jan wrote, and example (548b) does not allow this kind of topicalization at all, since the only structure that would allow this kind of topicalization (the doubly-primed one) is not available.

The examples in (549) involve topicalization of the PP, which, not surprisingly, seems to be possible only with independent over-PPs. This is clear from the fact that (549a) seems to have the interpretation associated with the independent use of the PP: it expresses that the very first book Jan wrote was about the World War II. As expected, example (549b) is degraded because the over-PP can only be interpreted as a dependent of the story noun.

However, it should be noted that it is possible to force the reading associated with (549a'), according to which the book in question is not the first Jan wrote, but the first he wrote about World War II, by adding an adverb like pas onlangsonly recently.

A similar effect can be achieved in constructions with R-extraction. Again, the assumption would be that extraction is possible from a PP-complement of the verb, but not from a PP-complement of the noun. The preferred interpretation of (551a) is in accordance with this assumption: the question concerns the first book that Jan ever wrote. The alternative reading becomes available if we add the adverb onlangs: now the question refers to the new subject about which Jan recently wrote a book.

The diverging interpretations of (550a) and (551b) can be taken as an argument for the claim that extraction from noun phrases is possible, but only at the cost of no longer being able to maintain the robust syntactic pattern established in (547) to (549). As an alternative, therefore, we can consider a non-syntactic account along the lines proposed for example (535).

The distinction between dependent and independent van-PPs relates to a difference in interpretation, which can be made more conspicuous by adding the ordinal numeral eerstefirst. Consider the ambiguous example in (553a). If the van-PP is dependent on the story noun, as in (553b), we are not dealing with the first book Jan has ever read; the sentence states that it is the first time Jan has read a book by Huizinga. If the van-PP is independent of the story noun, as in (553b'), the resulting sentence is marked, but it is still clear that the first book Jan read was written by Huizinga.

Sentence (554a) does not allow this kind of ambiguity: it allows only an interpretation associated with (554b) according to which Jan burned a book written by Huizinga for the first time (and more are likely to follow); the reading associated with (554b'), according to which the very first book that Jan ever burned was a book by Huizinga, is not available.

The insertion of the negator nietnot can also be used to disambiguate these sentences, on the assumption that it can only be placed between the noun and an independent van-PP. We therefore expect niet in this position in (555a) to be acceptable in the reading associated with the structures in (555b'). This expectation seems to be borne out: (555a) can only be interpreted in such a way that the book in question is the very first book about the Middle Ages that the speaker has ever read, i.e. with the van-PP as an independent constituent. Note that the use of such sentences is somewhat restricted: they are only acceptable in contrastive contexts, and omitting the theme argument will degrade the result.

The assumption that the negator nietnot can only be placed between the noun and an independent van-PP also correctly predicts that (556a) is unacceptable, since only the structure with a dependent PP is available.

Topicalization can also be used to distinguish the two readings. Since van-PPs can be dependent on the story noun with both verb types, it stands to reason that it is always possible to topicalize the noun phrase and the PP as a whole. Since in Dutch only a single constituent can be placed before the verb in sentence-initial position (the constituency test), we predict that only the constructions in the singly-primed examples are acceptable. This prediction is borne out: sentence (557a) means that the speaker read his first book by Huizinga, not that this was the first book he read.

It is also possible to topicalize only the story noun, stranding the van-PP in the original position. Here we would expect only the doubly-primed examples to yield acceptable results. The instances in (558) show that this prediction comes true: (558a) can only express that the book in question is the very first one the speaker has ever read; (558b) does not allow this kind of topicalization, since the only structure that would allow this kind of topicalization (the doubly-primed one) is not available.

The examples in (559) involve topicalization of the van-PP, which, not surprisingly, seems to be possible only with independent van-PPs. This is clear from the fact that (559a) seems to have the interpretation associated with the independent use of the PP: it expresses that the very first book the speaker read was about the Middle Ages. As expected, example (559b) is degraded because the van-PP can only be interpreted as a dependent of the story noun.

Note, however, that it is possible to force the reading associated with the structure in (559a'), according to which the book in question is not the very first one the speaker read, but the first one he read written by Huizinga, by adding an adverb like onlangsrecently.

The interpretational ambiguity of (560a) can be construed as an argument for the claim that extraction from noun phrases is possible, but only at the cost of no longer being able to maintain the robust syntactic pattern established in (557) to (559). We may need to consider a non-syntactic account along the lines proposed for example (535).