Example (383a) shows that the agent argument of an intransitive ge-nominalization appears postnominally as a van-PP; the use of an agentive door-phrase is questionable. The agent can also appear prenominally in the form of a possessive pronoun or genitive noun phrase, as in (383b). That the postnominal van-PP and the prenominal genitive form both express the agent argument is shown by the fact, illustrated in (383c), that they cannot co-occur. The agent is normally obligatory: only in a generic sentence like (383d) can it be left unexpressed; cf. Section 16.2.1, sub IIA, for implicit arguments.

In some cases the agent can be expressed by an attributively used relational adjective of the geographical type, like AmerikaansAmerican and RussischRussian in (384); cf. Section A24.3.3. This does not mean, however, that this adjective is to be interpreted as the inherited agent argument of the verbs huichelento feign and blunderento blunder; it can simply have the non-agentive interpretation, as in de Amerikaanse dollarthe American dollar, and allow the agent argument to remain unexpressed by making it contextually recoverable.

Unaccusative verbs cannot be used as input for ge-nominalization; this is discussed in Section 15.3.1.4, sub IIID.

When the ge-nominalization is based on a transitive verb, three scenarios can be distinguished: one in which only the theme argument is expressed, one in which the two arguments are expressed, and one in which only the agent is expressed.

The number of triadic ge-nouns is fairly restricted because many ditransitive verbs are preceded by a particle or a prefix, and as such are excluded from ge-nominalization. Examples are given in (398).

However, ge-nouns can be derived from ditransitive verbs like gevento give, donerento donate etc. It is possible for such ge-nominalizations to occur with all three arguments expressed, although such occurrences are very rare in practice. More often, one (typically the agent) or two (agent and recipient) of the arguments are left unexpressed; in generic contexts, none of the arguments need be expressed, as e.g. in example (399). In the following subsections, we will consider those cases in which at least one argument appears.

ge-nominalizations can also inherit PP-themes from verbs like jagen opto hunt for and zoeken naarto search for. This is shown for the ge-noun gejaag in (405a), which inherits the preposition selected by the input verb jagen. The agent can be realized postnominally as either a door-PP or a van-PP, and prenominally as a genitive noun phrase/possessive pronoun. The agent can also be expressed by a relational adjective such as NoorsNorwegian.

Like ing-nominalizations, ge-nominalizations do not occur with complementive adjectives. This is illustrated by the examples in (406), which show that these constructions are unacceptable regardless of the position (post or prenominal) of the predicate.

When the complementive is introduced by a preposition like tot or als, the ge-nominalization is marked but still more or less acceptable when the complementive occurs postnominally. This is illustrated in examples (407a&b).

The previous subsections discussed the form and distribution of the various arguments of ge-nominalizations. As with inf and ing-nominalizations, the theme argument is usually obligatory; it must appear as a postnominal van-PP, preferably in the position immediately adjacent to the head. Recipients may (but need not) be expressed as a postnominal aan-PP, which typically follows the theme. Agents may also be expressed by a postnominal PP, which typically follows the theme and the recipient, if present. The form of the agentive PP depends on the type of the input verb: if the input verb is intransitive, the agent is obligatorily realized as a van-PP; if the input verb is (di-)transitive, it is realized as a door-PP; if the input verb takes a PP-complement, the agent can be expressed by either a van- or a door-PP. The agent can also appear as a genitive noun phrase or possessive pronoun, provided it has a [+human] referent. Table 13 summarizes the discussion of ge-nominalizations derived from intransitive, transitive, and ditransitive verbs.

A comparison of Table 13 with Table 11, which lists the basic patterns of ing-nominalizations, reveals two important differences. First, the monadic verbs are unaccusative in ing-nominalizations but intransitive in ge-nominalizations. Second, prenominal realization of the theme as a possessive pronoun or genitive noun phrase is possible in ing- but not in ge-nominalizations.

Ge-nominalizations inherit the argument structure of the input verb, with the nominal construction resembling the verbal construction in terms of the number of arguments and their thematic functions. However, while the arguments of verbs must be explicitly expressed, this is not true to the same degree for the inherited arguments of the corresponding ge-nouns. If the input verb is intransitive, the agent is usually expressed, but it can still be left implicit if it is somehow implied, as in the primed examples of (408); in (408b'), for example, it is clear from the context that the giggling was done by students attending the class.

Example (409a) shows that in ge-nominalizations derived from a transitive verb the theme argument must normally be present, while the agent can quite felicitously be omitted. However, if the theme is somehow implied, it need not be expressed: in (409b), for example, it is clear from the context that at least one student in the class is being bullied by some other person(s) in the class, and this makes it possible to leave the theme implicit.

If the input verb is ditransitive, as in (410), the recipient can normally also be left unexpressed, as in the corresponding verbal construction. Note that if they are omitted, the presence of the agent and recipient arguments is still implied and must be recoverable or inferable from the context.

Ge-nominalizations derived from verbs selecting a PP-complement pattern with those derived from transitive verbs; the PP-theme can only be omitted if it is recoverable or inferable from the context. This is illustrated in (411).

In short, it seems that the arguments of the ge-nouns can only be omitted if they are recoverable or inferable from the context. If this is not possible, omitting these arguments is likely to lead to marked results, unless the construction in question is generic; cf. Section 16.2.1, sub II, for these and other exceptions.

The examples in (412) show that the van-PPs found in ge-nominalizations cannot occur in post-copular position. This is hardly surprising, because van-PPs in post-copular position are interpreted as possessive elements, and states of affairs, the denotation of ge-nominalizations, cannot be possessed. This is also true for inherited PP-arguments, as shown in example (412f).

Note that constructions such as (412b), in which the ge-noun is derived from a verb of sound emission, may be acceptable under a slightly different interpretation, namely one in which the post-copular van-PP provides the source of the sound in question. In such cases, we are no longer dealing with theme arguments of the ge-nominalization, but with modifiers: the examples seem to involve an abridgment of is afkomstig vanemanates from.

Example (414a) shows that R-pronominalization of a van-PP expressing the theme of a ge-noun derived from a transitive verb leads to a perfectly acceptable result. R-pronominalization of the theme argument of a ge-noun derived from a ditransitive verb seems somewhat marked, but is acceptable, and the same holds for R-pronominalization of theme arguments headed by prepositions other than van.

Example (415a) shows that R-pronominalization of van-PPs expressing an agent leads to a marked result; this may be due to the fact that agents are typically [+animate], since R-pronominalization of PPs with an animate noun phrase is often disfavored. Examples (415b&c) show that R-pronominalization of agentive door-PPs or aan-PPs expressing a recipient leads to completely unacceptable results.

R-pronominalization is possible with inherited subjects of sound-emission verbs when the ge-nouns are preceded by the article hetthe (but not with the expressive demonstrative dat). However, it is not clear what we can conclude from this, as our discussion of the examples in (418) and (422) below suggests that ge-nouns derived from these verbs have a special status.

The PP-extraction tests again yield results that are far from clear. The acceptability of extraction often depends on the ease with which a contrastive interpretation can be construed and on the type of input verb.

Table 14 summarizes the results of the four tests for inherited theme arguments of ge-nouns. The third and fifth columns indicate whether the results provide evidence for or against our assumption that we are dealing with complements. The first three tests provide unequivocal evidence for complement status of both van-PPs and PP-themes headed by other prepositions. There is a marked difference in the behavior of van-PPs and PPs headed by other prepositions with respect to the possibility of extraction: the conclusion that inherited theme PPs function as complements receives at best weak support from the extraction facts. However, since we have seen that the PP-extraction tests are problematic in various ways and may not be suitable for determining the complement status of PPs, it seems that we can still safely conclude that both types of PP-theme function as arguments of the derived noun.

For recipient aan-PPs and agentive door-PPs it is more difficult to establish whether they are arguments of the noun. Only the first test is relevant for them, and it seems that this test provides only weak evidence for assuming argument status: recipients and agents are semantically implied, but need not be syntactically expressed. However, since recipients and agentive door-phrases are often optional even in verbal constructions, this is not conclusive: we will assume that they have a status similar to that of the theme, which clearly behaves as an argument.