The agent argument of the input verb is not realized as a complement of the deverbal noun but is represented by the suffix -er of the noun; the derived noun actually denotes the agent of the input verb. As a result, er-nouns derived from intransitive verbs such as slapento sleep, do not select a PP-complement.

If the er-noun is derived from a transitive verb, the theme argument should be present, either explicitly or implicitly; the theme arguments in the examples in (165), which are realized as van-PPs, can be omitted only if their referents are contextually recoverable.

Theme arguments within the nominal domain typically appear postnominally in the form of a van-PP, but [+human] themes can sometimes also appear prenominally as a possessive pronoun or genitive noun phrase; cf. Section 19.2.2 for a detailed discussion of the restrictions on this option. Note in passing that the prenominal position is not available for inherited PP-complements with the thematic role of theme, as shown in the (b)-examples in (166).

That the prenominal elements are indeed the theme arguments can also be shown by the fact that they cannot co-occur with a van-PP fulfilling the same function: examples such as (167) only allow a possessive interpretation for Jans and mijnmy.

Note that simple person nouns behave exactly like er-nouns derived from a transitive verb, provided that the semantic relation between the head noun and its complement is similar to that between the verb and its complement. Often these nouns are (near-)synonyms of derived deverbal nouns; this is shown for auteurauthor and schrijverwriter in (168a), and architectarchitect and ontwerperdesigner in (168b).

Since the nouns auteur and architect in (168) are not derived, these similarities cannot be accounted for in terms of inheritance: they are simply relational nouns. The inherent relation between the noun and its associated argument thus has its origin in the meaning of the noun itself; cf. Section 16.2.2.

The examples in (169) show that constructions with ditransitive verbs can take two forms: the recipient can appear as a noun phrase preceding the theme, or as an aan-PP generally following the theme.

The theme argument of the corresponding er-noun cannot be expressed prenominally in the form of a prenominal possessive pronoun or genitive noun phrase, but appears as an obligatory postnominal van-PP. This may be related to the fact that the theme of a ditransitive verb is generally inanimate, but even if the theme is [+human], as in (170), prenominal realization of the theme is excluded.

The recipient argument of the er-noun is always realized as an aan-PP, and can often be omitted, like the recipient in the corresponding verbal construction; in fact, constructions with a realized theme feel somewhat heavy, and there is a certain preference not to realize the recipient.

Although recipient arguments are typically [+human], they never appear as prenominal genitive noun phrases/possessive pronouns: the following constructions, headed by er-nouns derived from ditransitive verbs, are unacceptable.

A potential problem with the claim that recipient arguments must be realized as aan-PPs is that the noun donateur seems to occur with a recipient expressed by a van-PP or a possessive pronoun: Jan is donateur van onze voetbalclubJan is a donor of our soccer club; onze donateursour donors. However, this may only appear to be the case because the van-PP functions rather as an adjunct with the role of possessor. In fact, the relation is one of pseudo-possession: although it is possible to say De club heeft donateursThe club has donors, the verb hebbento have cannot be replaced by the lexically more specific verb bezittento own, as is possible in prototypical cases of possession: Jan heeft/bezit een fietsJan has/owns a bike. This may also explain why the construction does not pass the second adjunct/complement test (*De donateur is van de voetbalvereniging), since the occurrence of a van-PP in post-copular predicative position requires a true possession relation.

For the sake of completeness we need to mention that although at first sight the ditransitive verb betalento pay seems to have the same argument structure as other ditransitive verbs, there is a difference with regard to complementation, which also affects the form of the complement of the derived er-noun betaler. First, as with all (di-)transitive verbs, passivization is possible with the theme being assigned nominative case, as in (173b), and the theme can be premodified by the past participle, as in (173c).

However, the examples in (174) show that, in the absence of the theme dit loonthese wages, it is also possible to promote the recipient to subject in the passive construction and to have the recipient argument premodified by the participle. In other words, the recipient in (174) functions as a regular direct object.

The constructions in (174) are perfectly acceptable because the implied theme is fully recoverable: even without further context, the missing theme will be interpreted as the employees’ wages. The most likely analysis of the constructions in (173) and (174) is therefore one that distinguishes between two different forms of the verb betalen. The most common form is the ditransitive verb, with a theme and a recipient complement, and with the general meaning of “to pay”. There is also a less common monotransitive form, with only a theme complement, and with the more specific meaning of “paying wages”; cf. Section V3.2.1.3, sub IIC, for further discussion of this type of verb.

A similar distinction can be seen in the derived noun betalerpayer in (175). Example (175a) corresponds in meaning to (173), where the verb is used ditransitively, and the theme and the recipient argument are expressed by a van and an aan-PP, respectively. Example (175b), on the other hand, corresponds in meaning to (174), where the (apparent) recipient acts as the direct object of the verb, which implies that the recipient argument must appear in the form of a van-PP; the aan-PP is not acceptable in this example.

The verb voerento feed seems to behave similarly to betalento pay: both the theme and the recipient of the verb can appear as a van-PP in the corresponding er-nominalization, as shown in the primed examples in (176). However, the primed examples will be less acceptable to most speakers, because the noun voerder is not often used with the meaning “feeder”, due to competition from the more common (but also obsolete) form voeder with the same meaning.

In (177) we see some examples of er-nouns derived from verbs that select a PP-theme. Whether or not the presence of the PP is required seems to be largely determined by the behavior of the input verb in this respect: as shown in the primed examples, the verb lijdento suffer seems to prefer the presence of a complement, whereas klagento complain can be used comfortably without one.

The assumption that the presence of the PP-complement is the result of inheritance and as such part of the argument structure of the derived noun is supported by the examples in (178) and (179), where the PPs are adjuncts and not complements of the verb. Since the verb schilderen in (178a) is not subcategorized for an instrument-PP, this PP cannot be inherited; example (178b) can therefore only be interpreted as “a painter who has no brushes” (in which case the PP is a modifier of the noun), not as “a person who paints without brushes” (in which case the PP would be inherited).

Similarly, the PP met de treinby train in (179a) is an adjunct and not a PP-complement of the verb reizento travel; consequently, it cannot appear as the complement of the derived noun reizigertraveler in (179b) either.

Er-nominalization is not possible in constructions with a complementive. This is illustrated by the transitive resultative constructions in example (180). That it is indeed the presence of the predicative adjective that causes the unacceptability of the er-noun is clear from the fact that in (180b) the verb in question can be the input for er-nominalization when the complementive is not present.

Example (181) shows that the restriction also applies to intransitive verbs: example (180b) is only acceptable if the predicate is not expressed; the fact that the noun phrase zijn schoenen cannot be expressed in (180b) either is due to the fact that it is not an argument of the verb in (180a), but only semantically licensed as the logical subject of the complementive; cf. Section V2.2 for a more detailed discussion of verbs taking a complementive.

The previous subsections dealt with the inheritance by agentive er-nouns of the argument structure of their input verb. In general, the internal arguments of the input verb turn out to be complements of the derived noun; the external (agent) argument is not inherited, but is denoted by the er-noun itself. This means that in the case of a transitive input verb, er-nouns have an argument structure with a slot for a theme argument, which is typically realized as a van-PP, or alternatively as a possessive pronoun or a genitive noun phrase (with the suffix -s) in prenominal position. If the input verb is ditransitive, the recipient argument is added (usually optionally) as a postnominal aan-PP. PP-themes can also be inherited, in which case the preposition selected by the input verb is also used in the er-nominalization. These findings are summarized in Table 7.

Er-nouns derived from transitive verbs are usually not interpretable without the addition of the inherited theme argument: the inf-nouns makermaker and bedenkerdesigner in (182) usually require the overt realization of a theme complement. The number sign in the primed examples indicates that the theme arguments can be omitted in certain contexts, to which we return below.

Er-nouns derived from ditransitive verbs inherit both complements of the verb: the noun schenkerdonor in (183) must be related to the theme and the recipient argument in order to be interpretable. However, example (183a) shows that, as in the corresponding verbal construction, the recipient can often be left implicit. The number sign in (183b) again indicates that the theme argument can be omitted in certain contexts. Finally, example (183c) shows that the theme argument cannot be left unexpressed when the recipient is overtly expressed.

If the er-noun is derived from a verb that selects a PP-object, the inherited PP is usually obligatory as well, as shown in example (184a&b). Exceptions are fully lexicalized er-nouns such as jagerhunter in example (184c).

Although the inherited argument must normally be expressed overtly by a PP, there are a number of contexts in which the PP-complement can (or even must) be omitted. Each of these situations will be briefly discussed below.

According to the second test, only adjunct van-PPs can occur in post-copular predicative position; complement PPs in this position lead to unacceptable results. The unacceptability of the transitive examples in (208) thus suggests that the postnominal van-PPs of agentive er-nouns are indeed arguments.

Applying this test to er-nominalizations derived from ditransitive verbs yields similar results. Again, the placement of the van-PP in post-copular position is excluded, as exemplified in (209).

The R-pronominalization test suggests that the van-PPs following er-nouns derived from transitive verbs behave like complements: the examples in (210) show that they allow R-pronominalization.

Section 16.2.1, sub IV, has shown that using the split version of the pronominal PP usually leads to a marked result. However, the examples in (211) show that with er-nouns, the use of the split version is often perfectly acceptable. This may support the suggestion in Subsection D that many of the apparent cases of extraction of van-PP are actually cases with an independent restrictive adverbial phrase.

Applying R-pronominalization to constructions with a ditransitive input verb yields similar results: example (212a) is acceptable with both the split and unsplit patterns, provided that the recipient is not expressed. Example (212b) shows that R-pronominalization is only marginally possible with recipient arguments, and requires that the pronominal PP be unsplit.

Applying the R-pronominalization test to inherited PPs with prepositions other than van yields somewhat ambiguous results. The examples in (213) show that R-pronominalization can lead to marked constructions, although it is certainly not impossible given the right context (i.e. one in which the pronominalized part is the discourse topic), especially when the head noun is given contrastive accent.

Note that only the constructions with the unsplit form are acceptable; this supports the suggestion above that the split cases in (211) may involve independent restrictive adverbial phrases.

The cases in (214) and (215) suggest that PP-extraction of a theme argument is possible with er-nouns, although the results seem to be slightly marked (and seem to differ from speaker to speaker). The less acceptable status resulting from extraction is surprising because the first three tests give a positive result as far as the complement status of the van-PP is concerned; we would therefore expect the extraction of this PP to be fully acceptable in all cases.

PP-over-V is perfectly acceptable if the PP is preceded by an intonation break. However, this results in a reading in which the PP is presented as an afterthought. If such an intonation break is missing, as is usually the case in PP-over-V constructions, the result is marked. The same applies to the scrambling examples in (217).

Now consider argument extraction from a noun phrase headed by an er-noun such as schenker in (218), derived from the ditransitive verb schenkento donate.

The examples in (219)-(222) suggest that extraction of the inherited theme argument is possible, provided that the recipient is left unexpressed, while extraction of its recipient argument always leads to a severely degraded result.

However, the fact that the presence of a recipient makes the (a)-examples unacceptable can also be taken to show that we are not actually dealing here with the extraction of PP-themes from the noun phrase, but with independently generated restrictive adverbial phrases: the unacceptability of the (a)-examples with the recipient present would then follow from the fact that the noun phrase does not contain a theme argument; cf. the discussion of (183d) in Subsection A.

Just as in the case of the recipient aan-PPs, the PP-extraction tests do not yield the expected, felicitous results for cases in which the theme complement is headed by a preposition other than the functional preposition van. Examples (223) show that topicalization leads to an unacceptable result. The examples in (224) and (225) show that the same is true for the other forms of extraction.

Given these facts, one possible conclusion would be that theme arguments headed by prepositions other than van are not complements of the noun, but adjuncts. However, it is clear that the PPs under discussion behave differently from undisputed adjuncts and more like the PP-complements of the input verb: they are obligatory, headed by the same preposition as the PP selected by the input verb, and their semantic relation to the er-noun is similar to that between the input verb and its PP-complement. This may lead to the conclusion that the PP-extraction test is in fact not a good test for determining complement status of the PP, and that complement PPs are like adjunct PPs in that they cannot be extracted from noun phrases: the “displaced” van-PPs are not arguments of the noun but independent restrictive adverbial phrases; cf. the discussion in Section 16.2.1, sub VC.

Table 9 summarizes the results from the previous subsections of the four tests for inherited theme arguments of agentive er-nouns. The third and fifth columns indicate whether the results provide evidence for or against the assumption that we are dealing with complements.

The results show that it is justified to consider inherited theme arguments appearing as van-PPs as complements of the derived er-noun. However, recipients and theme PPs with prepositions other than functional van are more problematic. Only test 1 on the obligatoriness of the PP provides clear evidence for the complement status of these PPs. Test 2 on the post-copular placement of van-PPs does not apply to these cases. The results of test 3 on R-pronominalization seem to point in the direction of complement status, but the results are still not entirely convincing. The results of test 4 are decidedly negative.

Although the results of the test do not clearly show that PPs introduced by a preposition other than van are complements, we will consider them as such. A first reason for this is that we have seen that test 4 may not be a good test for distinguishing between adjuncts and complements: putative cases of extraction may actually involve independent restrictive adverbial phrases. A second reason is that at least the theme PPs clearly function as complements for the other forms of deverbal nominalization discussed in Sections 16.2.3.2-16.2.3.4.