As shown in example (568), the agent argument of a picture noun need not be overtly expressed, even though it will always be semantically implied.

Example (569) shows that theme PPs can sometimes also be omitted, even in cases where the picture noun is derived from a transitive verb that requires a theme complement, such as schilderijpainting or tekeningdrawing.

Omitting the theme arguments does not always yield an acceptable result: with a noun like afbeeldingpicture in (570), omitting the theme often leads to a degraded result.

The difference between (569) and (570) is probably related to the degree of lexicalization: picture nouns like schilderij and tekeningdrawing are highly lexicalized, as a result of which they may have lost their argument structure; the noun afbeeldingpicture, on the other hand, has a lower degree of lexicalization and has retained the argument structure of the verb, even though its denotation has changed; cf. also Section 16.2.5.2 for further discussion on this issue.

In some cases, the complements of story nouns with a verbal counterpart cannot easily be omitted, while in others explicit mention of the complement(s) is not necessary. When the noun has abstract reference, i.e. refers to the content of some object, mention of at least one argument is preferred; example (571a) shows that this argument may be the theme (subject matter) or the agent (creator). When the referent is a concrete object, as in (571b), there is no need for an argument. Finally, example (571c) shows that story nouns can appear without complements when the creator of the story noun is expressed by the agent of the sentence; cf. 16.2.5.2, sub II.

If the story noun has no verbal counterpart, both the agent and the theme arguments can usually be omitted, even if it is the content of the story noun that is relevant. This is illustrated in example (572).

That the identity of the agent or theme is still somehow implied is clear from the fact that sentences such as (573a) typically serve to start a discourse, with the speaker waiting for encouragement from the addressee, and evoke questions such as (573b) concerning the identity of the agent or theme of the story noun. The story noun is usually modified in such cases.

We have seen in Section 16.2.1, sub III, that van-PPs in post-copular predicative constructions are typically interpreted as possessors. So it is not surprising that example (574a) can be interpreted with Jan as the possessor of the painting. What is surprising, however, is that Jan can also be interpreted as the agent of the construction. In fact, the examples in (574b&c) are also ambiguous in terms of the distribution of the roles of agent and possessor: the preferred interpretation indicated by the labels is based entirely on our knowledge of the world.

Example (575a) shows that placing a theme van-PP in post-copular predicative position often yields a degraded result (although some people accept it). However, examples like this improve in contrastive contexts with strong emphasis on the subject noun phrase; (575b) shows that they are especially acceptable when the agent argument is present.

As expected, the possessor in story noun constructions can be used as the predicative part of copular constructions. Similar to picture nouns, however, the van-PP in this position can also be interpreted as the agent argument, as seen in (576). Again, similar to picture nouns, knowledge of the world will affect the preference for one interpretation or the other; cf. example (574).

This test cannot be used to determine whether the theme of story nouns functions as a complement of the noun: it does not have the form of a van-PP. However, example (577) shows that the PP-theme can be used in constructions with the verb gaanto go.

Example (578) again shows that PP-modifiers of picture nouns cannot undergo R-pronominalization: example (578b) is only acceptable if the phrase er ... mee is construed as an independent instrumental adverbial phrase.

The fact, illustrated in (579b), that it is impossible to pronominalize possessive van-PPs thus supports the conclusion from Subsection I that these van-PPs function as modifiers of nouns.

On the other hand, the acceptability of R-pronominalization in the primed examples in (580) with the verb ziento see can be taken as evidence for the claim that van-PPs with the role of agent and theme function as complements of the picture noun. However, the fact that both the unsplit and the split pattern are acceptable suggests that, at least in the latter case, we may be dealing with a restrictive adverbial phrase; cf. Section 16.2.1, sub IV. That the examples may not involve PP-complements of the picture noun is also suggested by the fact that the acceptability of the examples decreases when we use affective verbs like beschadigento damage, which normally do not license the presence of dependent PPs; cf. Section 16.2.5.5. The results of the test are therefore inconclusive.

That PP-adjuncts of story nouns cannot undergo R-pronominalization is illustrated in (581): (581b) is acceptable, but only if the phrase er ... mee is construed as an independent instrumental adverbial phrase.

The finding that it is impossible to pronominalize possessive van-PPs with story nouns, illustrated in (582), supports the conclusion in Subsection I that these van-PPs function as adjuncts.

The acceptability of R-pronominalization in (583a'&b') with the verb lezento read, on the other hand, seems to indicate that PPs with the roles of agent and theme do function as complements of the story noun. However, the fact that both the unsplit and the split pattern are acceptable suggests that, at least in the latter case, we may be dealing with a restrictive adverbial phrase; cf. Section 16.2.1, sub IV. That the examples do not involve PP-complements of the story noun is also suggested by the fact that the acceptability of the examples decreases when we use affective verbs like verscheurento tear up, which do not license the presence of dependent PPs; cf. Section 16.2.5.5. The results of the test are therefore inconclusive.

Topicalizing the possessor van-PP of a picture or story noun always leads to unacceptable or, at best, questionable results. In example (585a), for instance, the preposed van-PP can only be interpreted as the theme or the agent of the picture noun. On the intended reading, the examples in (585b&c) are at best marginally acceptable on a contrastive reading.

In (586) we provide similar examples with the story noun boekbook: in (586a) the preposed van-PP can only be interpreted as the agent (the author), not as the possessor of the book. Examples (586b&c) are again at best marginally acceptable on a contrastive reading.

It has already been shown in Section 16.2.5.5 that it is difficult to exclude with certainty that PP-complements of picture/story nouns can be topicalized when the noun phrase is selected by a non-affective verb such as ziento see. However, several other factors seem to influence the acceptability of extracting postnominal PPs in picture/story noun constructions: we will discuss the role of focus, the choice of determiner, and the presence of numerals/quantifiers or other arguments. Since picture and story nouns exhibit more or less the same behavior, we will discuss them simultaneously.

This subsection shows that relativization and questioning point in the same direction as topicalization. Possessors resist extraction and thus clearly function as adjuncts. Extraction of agent and theme seems possible at first sight but may in fact involve the fronting of an independent restrictive adverbial phrase.

The judgments on relativization and questioning the possessor are perhaps less sharp than those on topicalization, but still the most salient readings of the primed examples in (604) are again those in which the van-PP refers to the agent or the theme. This suggests that possessors cannot be extracted and thus function as adjuncts.

Relativization and questioning the agent and the theme are easily possible when they are the only arguments present. Extraction of the agent does not seem to be sensitive to the presence of the theme argument, as the examples in (605) show.

Relativization of the theme is more restricted in the sense that the results get much worse once the agent is added; examples in which the agent appears in the form of a van-PP may be slightly better than those in which it appears as a prenominal genitive noun phrase. The fact that the differences in judgments for the primeless examples in (605) and the primed examples in (606) are similar to the judgments for the primed examples in (602) and (603) again suggests that we may be dealing with independent restrictive adverbial phrases rather than complements of the noun.

The examples in (607) show that we find similar judgments when the topic is questioned. The differences in the judgments on the primed examples in (605) and the primed examples in (607) are similar to the judgments on the primed examples in (602) and (603). This leads to the by now familiar conclusion that we are dealing with independent restrictive adverbial phrases and not with complements of the noun.

PP-over-V does not seem to be a very good test for establishing the complement or adjunct status of the agent, theme, and possessor; the examples in (608) show that, under certain conditions, adjunct PPs also seem to allow PP-over-V in picture/story noun constructions. For the sake of completeness, we will nevertheless discuss the relevant constructions.

PP-over-V of the possessor argument is possible in picture noun constructions such as (609a), where the agent and the theme are absent. As soon as the theme or the agent is added, the result gets worse, as shown in (609b&c), which is at best marginally possible in a non-appositive reading. Adding the agent argument in the form of a genitive noun phrase, as in (609c'), makes PP-over-V of the possessor completely impossible. In (610) we see the corresponding story noun constructions.

As shown in (611), PP-over-V of the adjunct PPs in (608) is blocked under the same circumstances (unless the adjunct is given an appositive reading). This suggests that we can interpret the degraded status of the (b) and (c)-examples in (609) and (610) as evidence for adjunct status of the possessor.

In (610) we see that PP-over-V of the agent of a picture noun is possible when it is accompanied by the theme, but impossible when the possessor is expressed, regardless of the form and position of the latter.

Example (613a) shows that PP-over-V of the agent of a story noun is possible if it is the only argument expressed; if the theme is also present, the result also seems to be acceptable, although slightly marked. If the possessor is expressed, the result is again highly questionable; the examples in (613b&b') are only possible with an appositive reading of the agent.

PP-over-V of the theme of a picture noun is acceptable when it is the only argument present, as in (614a), and somewhat marked when the agent is expressed in the form of a van-PP, as in (614b). Surprisingly, the result seems acceptable with an agentive or possessive genitive noun phrase, as in (614b').

The theme of a story noun behaves somewhat differently. As with the picture nouns, PP-over-V of the theme is possible if the theme is the only argument, and may be slightly marked if the agent is expressed as a postnominal van-PP. However, if an agentive or possessive genitive noun phrase is present, the result is highly marked. This can be seen in (615b').

Finally, extraposing both the theme and the agent/possessor of a picture noun is possible, although the acceptability depends on the order of the elements in extraposed position: (616a) seems acceptable on a non-appositive reading, but an appositive reading of the theme is perhaps more preferred (hence the question mark); example (616b), on the other hand, is unacceptable on a non-appositive reading — the only available reading is that with Rembrandt as the possessor of the Westertoren.

Extraposing both the theme and the agent/possessor of a story noun is also possible, again depending on the order of the elements in extraposed position: example (617a) is acceptable, although an appositive reading of the theme is more likely (hence the question mark); example (617b), with the theme preceding the agent, on the other hand, is unacceptable on a non-appositive reading. Note that (617b) is possible on the non-intended reading, with Huizinga as the possessor in a complex noun phrase de middeleeuwen van Huizinga, referring to, say, the medieval period as described by Huizinga.

For completeness’ sake, note that (617b) seems possible on the non-intended reading with Huizinga as the possessor in a complex noun phrase de middeleeuwen van Huizinga referring to the medieval period as described by Huizinga.

Judgments on scrambling of the possessor are again less sharp than those on topicalization. A possessive reading of the primeless examples in (618) is not easily obtained; the most salient reading of these examples is the one in which the van-PP refers to the agent. This is consistent with the fact that these examples become completely unacceptable when we add the agent in the form of a postnominal van-PP or prenominal genitive noun phrase, as in the primed examples. From this, we conclude that scrambling of possessors is impossible.

Scrambling of the agent leads to acceptable results when it is the only argument expressed or accompanied by the theme, as in the primeless examples of (619). However, as shown in the primed and doubly-primed examples, scrambling of the agent is impossible when the possessor is expressed, regardless of the form and position of the possessor.

Examples (620a) and (621a) show that scrambling of the theme leads to an acceptable (though slightly marked) result when the theme is the only argument expressed. When the agent or possessor is also present, scrambling of the theme leads to questionable or unacceptable results, depending on the presence and form of the agent/possessor. This is shown in the (b)-examples.

Unlike with PP-over-V, scrambling of more than one argument is impossible. This is illustrated in (622). Note that the primed examples are acceptable on the irrelevant reading in which the second van-PP functions as a modifier of the noun phrase embedded in the first van-PP. In (622a') this leads to the unlikely interpretation of Rembrandt as the possessor of the Amstel, and in (622b') to the more readily available reading “the Middle Ages as described by Huizinga”.

The results of the four tests for distinguishing between adjuncts and complements of the noun are summarized in Table 16. They clearly show that possessors behave like adjuncts. The results for the agent van-PP point in the same direction: the first two tests provide clear evidence against assuming the status of complement; the results of the R-pronominalization test are less clear, as the relevant cases can perhaps be reanalyzed as involving a restrictive adverbial phrase.

The results for the theme argument are far from clear. The first test points in the direction of complement status in some but not all cases: while a picture noun such as afbeeldingpicture must have a complement, other picture nouns such as tekeningpicture can easily be used without one. In the case of story nouns, only nouns referring to abstract content obligatorily take a complement; nouns referring to the physical object do not. The second test is only relevant for picture nouns, since story nouns do not take a van-PP but an over-PP. Even then, the results for picture nouns are far from conclusive: although in neutral contexts the use of a theme PP in post-copular position is marked, the judgments are definitely not such that they would constitute a firm basis for assuming complement status. R-pronominalization seems possible, but the availability of the split pattern may indicate that we are actually dealing with an independent restrictive adverbial phrase.

	possessor		agent		theme
Test 1: PP obligatory	—	negative	—	negative	+/—	?
Test 2: Post-copular position	+	negative	+	negative	?; n/a	?; n/a
Test 3: R-pronominalization	—	negative	?	?	?	?
Test 4A: Topicalization	—	negative	+	positive	+	positive
Test 4B: Relativization/Questioning	—		+		+
Test 4C: PP-over-V	+		+		+
Test 4D: Scrambling	—		?		?

The results of the extraction tests again contradict the results of the other tests in the case of agents but it seems that we can ignore this because there are several reasons to assume that the apparent agent PPs in the test examples are actually independent restrictive adverbial PPs. If we assume the same for the theme PPs, we must conclude that there is no conclusive evidence for the claim that picture/story nouns select complements.