An important difference between picture nouns and dyadic ing-nominalizations like vernietigingdestruction is that picture nouns do not denote states of affairs, but concrete entities; even if a noun phrase is headed by a deverbal picture noun like schilderijpainting in (480a), it cannot refer to the action denoted by the verb but only to the result of this action, i.e. the concrete object that has been created. Deverbal picture nouns are lexicalized and are on a par with lexical deverbal nouns like bestratingpavement, verzamelingcollection or uitvindinginvention. This also explains why picture nouns can be used quite felicitously without complements in most cases: since they do not share the denotation of the input verb, they do not inherit the argument structure of the verb. Consequently, a picture noun such as schilderijpainting in (480a) does not require the presence of a complement, even though its verbal counterpart schilderento paint usually does; in this sense, it behaves more like the non-relational noun fietsbicycle in (480a) than like the ing-noun vernietigingdestruction in (480b).

However, example (481) shows that omitting the theme argument does not always lead to a fully acceptable result, suggesting that picture nouns may differ in their degree of lexicalization: since afbeeldingpicture is felicitous only if the theme is expressed, it can be said to have inherited the argument structure of the verb afbeeldento depict; since tekeningdrawing does not require the presence of a theme, it can be said to be lexicalized.

Note, however, that other, more pragmatic factors may come into play. For example, when we are dealing with a work of art, as in (482), the proper use of the noun tekeningdrawing often seems to require the presence of the agent or a theme. This suggests that we are not dealing with inherited arguments, but with contextually evoked adjuncts. This may account for the fact that adding the name of a style period would also make (482) perfectly natural, and that the preferred addition depends on the expertise or interest of the participants in the conversation: the layman will probably want to know what the drawing represents, while someone knowledgeable about art will be interested in the maker or the periodization.

Although the arguments of a picture noun need not be expressed overtly, they can be present implicitly. This is very clear with unexpressed agent arguments, which can influence the form of referentially dependent theme arguments. To see this, first consider the examples in (483). When the theme of a picture noun is bound by (= coreferential with) the agent of the picture noun, as in (483a), it must have the form of a reflexive pronoun like zichzelfhimself, whereas it normally appears as a referential pronoun when it is bound by a constituent outside the noun phrase, as in (483b). Note that coreferentiality is expressed by means of subscripts.

The standard explanation of the examples in (483) is roughly that while anaphors must be bound in the minimal domain containing a potential antecedent, referential pronouns cannot be bound in that domain; cf. Section 23.3 for a more detailed discussion of this version of binding theory. Since the possessive pronoun is a potential antecedent, the relevant binding domain is the noun phrase: the anaphor must and the referential pronoun cannot be bound in this domain. Now consider example (484a), which shows that the binding behavior of the referential pronoun remains unaffected when the agent of the noun phrase is not expressed; it can take the subject of the clause as its antecedent, just as in (483b). If we assume that the structure is as given in (484b), the clause will be the smallest category containing a potential antecedent: it thus functions as the binding domain within which the referential pronoun must be free (= not bound), and we wrongly predict (484a) to be impossible. This has led to the assumption that the noun phrase contains a phonetically empty pronoun PRO, as indicated in (484b'): as a result, the noun phrase functions as the binding domain of the pronoun, and (484a) is correctly predicted to be possible on the intended reading; cf. Chomsky (1986).

Note that (484a) cannot be interpreted with the PRO agent as coreferential with Jan, i.e. with Jan as the maker of the picture, since in this case the referential pronoun would again be incorrectly bound in its binding domain by the agent of the picture noun: Jani zag [NP een PROi foto van ’mi]. This explains why (485a) is unacceptable under the intended idiomatic reading “taking a picture”, because the verb nemen forces a reading in which the implied PRO argument is coreferential with the subject of the clause; the referential pronoun ʼm is incorrectly bound by PRO in representation (485b).

The claim that the agent of the picture noun can be syntactically realized by the phonetically empty pronoun PRO is consistent with the acceptability of example (486a), in which the anaphor zichzelf is coreferential with the subject of the clause: the structure in (486b) shows that the anaphor is bound by the implicit PRO agent, which in turn is bound by the subject of the clause.

There is, however, a complication here concerning the interpretation of example (487a). First, assume that the implicit PRO agent is obligatory: since the anaphor zichzelf must be bound by PRO, and since zichzelf is coreferential with the subject of the clause, it follows that PRO is also bound by the subject of the clause. This gives rise to the structure in (487b), which expresses that Jan saw a picture of his own making depicting himself. Although this is certainly a possible interpretation, the sentence can also be interpreted to mean that the picture was made by someone else. However, we cannot assume that the PRO agent refers to someone else: representation (487b') is ungrammatical because the anaphor is not bound in its binding domain. The intended interpretation can therefore only be accounted for if we assume that the PRO agent is optional: the anaphor in (487b'') has no potential binder within the noun phrase, so that it can take as its antecedent the subject of the clause, which is now the smallest category containing a potential antecedent. The conclusion that there is no PRO agent correctly accounts for the fact that there is no implication concerning the identity of the photographer.

Another prediction of the claim that picture nouns can have a PRO agent is that anaphors can also occur when there is no overtly realized local noun phrase that could act as an antecedent. Consider the examples in (488) where the subject of the matrix clause is too far away from the anaphor to act as a local antecedent.

Example (488a) receives a percentage sign because Vanden Wyngaerd (1994:214ff.) claims it is unacceptable, while others seem to assign it more or less the same status as example (488b), which can be judged as marked but acceptable. The doubtful status of (488a) is further evident from the following facts: (i) Vanden Wyngaerd does not assign it a full asterisk, but “?*”, (ii) Vanden Wyngaerd claims that its status improves if we add a restrictive relative clause as in een foto van zichzelf die hij onlangs had genomena picture of himself that he had recently taken, and (iii) Vanden Wyngaerd goes to some lengths to show that the construction does not have the properties that he takes to be typical of control structures, e.g. that the presumed PRO-subject cannot take a split or a non-c-commanding antecedent.

Vanden Wyngaerd’s rejection of a PRO agent in picture noun constructions is related to the specific control theory he endorses, but it is unclear whether the same conclusion follows from other proposals. However, it seems that the acceptability of (488b) can be used to support the assumption of a PRO agent in picture nouns. Furthermore, it may be that the degraded status of (488a) is due to the fact that this example competes with the fully acceptable structure in (490a) with the referential pronoun henthem: given the fact established earlier that the PRO agent need not be present, this example may be more economical in the sense that co-reference can be established in a more direct way than in (488a). The attractiveness of this is increased by the fact that the referential pronouns can also be used in examples with a split or a non-c-commanding antecedent, as illustrated in (490b&c).

Of course, a similar competition is not found in the case of the reciprocal pronoun elkaareach other in example (488b), because referential pronouns such as henthem cannot express the relation of reciprocity.

It may be interesting to note that the form ’mzelf can also be used in picture noun contexts. The examples in (491) show that this form behaves like the reflexive pronoun zichzelf in that it must have an antecedent, but differs from it in that it cannot be bound by an antecedent within the noun phrase. However, some caution is needed, since the binding behavior of ’mzelf has not been much discussed in the literature; the most detailed discussion is found in Koster (1987:344ff.), and even this discussion is no longer than two pages. Moreover, Koster claims there that ’mzelf can be bound by the agent in ing-nominalizations like Jansi beschrijving van ’mzelfi Janʼs description of himself, so it may be that some people have more liberal judgments concerning (491b). However, since the speakers we consulted agree with the judgments given in (491), we will provisionally adopt them as an idealization of the data, but more research is certainly needed.

The claim that ’mzelf cannot have an antecedent within the noun phrase immediately accounts for the contrast in (492): example (492a) contains an (optional) PRO agent that is disjoint in reference from the subject of the clause and ’mzelf can therefore be correctly bound by an antecedent external to the noun phrase; the noun phrase in the idiomatic example in (492b) contains an (obligatory) PRO agent that is coreferential with the subject of the clause, so that ’mzelf is incorrectly bound in the noun phrase. The use of the number sign indicates that the example is acceptable on a non-idiomatic reading (without a PRO agent).

For completeness’ sake, note that examples (487a) and (492a) allow for an alternative reading in which the reflexive form refers not to the theme but to the agent of the picture noun. This reading is excluded with referential pronouns, which follows directly from standard binding theory, since the noun phrase does not contain a PRO agent (this function is performed by the pronoun itself); the clause therefore constitutes the binding domain within which the referential pronoun must be free. In (493) coreference is indicated by italics.

The binding facts discussed in this subsection suggest that in many cases where the agent of the picture noun is not visible, it is nevertheless syntactically present as a phonetically empty PRO argument. However, the PRO agent is usually not obligatory with the exception of the idiomatic construction een foto nemento take (i.e. make) a picture.

The question whether complements of deverbal story nouns can be felicitously left out seems to depend on the interpretation of the noun. Example (494) shows that explicit mention of at least one of the complements is preferred when the noun phrase has an abstract reference, i.e. refers to the content of an object.

When the referent is a concrete object, on the other hand, there is no need to express either argument. This suggests that the concrete interpretation of the story nouns in (495) involves a higher level of lexicalization: the examples in (494) behave like nominalizations in that they inherit the argument structure of the verbal stem, while those in (495) are more like lexicalized forms. Recall that the discussion of the examples in (471) already led to a similar conclusion.

The implied agent in the nominalizations in the primed examples of (494) is necessarily disjoint in reference with the subject: the speaker/writer cannot be Jan. These examples differ in this respect from those in (496), which also have abstract reference, but in which the agent is necessarily coreferential with the subject.

The fact that the theme can easily be omitted in (496) can be accounted for by claiming that the agent is syntactically expressed by a phonetically empty pronoun PRO: since the examples in (494) have already shown that we do not need to express both arguments, it is sufficient to have a PRO agent. However, assuming a PRO agent for the nominalizations in (496) raises the question of why it is not easily possible to leave the theme argument unexpressed in the primed examples of (494); it suggests that PRO need not be present in this case, which will be independently supported by our discussion of the binding data in (499) and (500) below.

If the story noun is not derived from a verb, the complements need not be expressed, i.e. both the agent and the theme can easily be left out, even if it is the content of the story noun that is relevant; cf. the examples in (497). This shows that story nouns are similar to picture nouns in that they behave more like non-relational nouns such as fietsbicycle than like ing-nominalization such as vernietigingdestruction.

However, unexpressed agent arguments can be implicit: as in the case of picture nouns, they can influence the form of a referentially dependent theme argument. To see this, first consider the examples in (498), which again show that anaphors must be bound in the minimal domain containing a potential antecedent, whereas referential pronouns must be free in that domain.

The fact that the referential pronoun in (499a) can also be used to refer to the subject of the clause gives rise to the idea that the agent of the story noun is realized as a phonetically empty PRO argument. The fact that this PRO argument cannot be interpreted as coreferential with the subject of the clause without invoking a violation of the binding condition on the referential pronoun correctly predicts that the story was written by someone else. It also follows that examples such as (499b), in which the verb forces a reading according to which PRO is interpreted as coreferential with the subject of the clause, are unacceptable. This account of the examples in (499) is thus completely parallel to that of (484) and (485).

The anaphor can also be construed as coreferential with the subject of the clause, as in (500). These examples can be explained along the lines of those in (486) and (487): the fact that (500a) is acceptable is consistent with the idea that the noun phrase contains a PRO agent that is coreferential with the subject of the clause; the fact that (500b) is acceptable despite the fact that the agent of the story noun is not coreferential with the subject of the clause shows that the PRO agent of the story noun need not be syntactically present.

The use of the reflexive form ’mzelf is unproblematic in constructions with the verb lezenread in (501a): this is as predicted, since ’mzelf can be bound by the subject of the clause while remaining free in its noun phrase. Our discussion so far predicts that (501b) should be excluded, since’mzelf is bound by the PRO agent within its noun phrase; the fact that this example seems marginally possible may therefore be unexpected. However, more research on the grammaticality status of examples like this is needed before we can draw any serious conclusions.

As in the case of picture nouns, the most important conclusion to be drawn from the binding facts discussed in this subsection is that in many cases where the agent of the picture noun is not visible, it can still be syntactically present as an (optional) PRO argument.