The vast majority of deverbal er-nouns denote a [+human] object, which acts as the agent in the argument structure of the input verb. A representative set of examples has already been given in Table 15.

Less common are examples like (197), where the deverbal er-noun represents a [-human] agent of the input verb. Nevertheless, these [-human] objects still perform the action denoted by the input verb and are therefore called impersonal agents; cf. Van der Putten (1997) and De Caluwe (1995). As shown in (197a-d), the input verb is usually transitive, although the intransitive input verb zoemento buzz in (197e) is also possible.

Er-nouns of this type do not allow complementation; although the input verbs may obligatorily contain a theme argument, this argument is not inherited by the er-nominalization. As a consequence, deverbal er-nouns denoting impersonal agents cannot be followed by a van-PP denoting the object of the input verb. This means that in the primeless examples of (198) the er-noun gets its prototypical [+human] agentive interpretation; if this is not possible, the result is unacceptable. In (198a), for example, the noun een zender is interpreted as referring to a person who sent the message; forcing a non-agentive interpretation, as in (198a'), leads to unacceptability. Similarly, een wekker in (198b) is interpreted as a wake-up person, i.e. a person who wakes up other people; alternatively, een wekker is interpreted as an alarm clock belonging to lazy people, a reading that is more likely with a specific possessor like Jan in (198b').

If the er-noun cannot receive a [+human] interpretation, then its use with a theme complement is straightforwardly infelicitous. This is illustrated in (199).

Er-nouns denoting impersonal agents are quite common in compounds of the kind illustrated in (200). Given that the first member of the compound functions as an incorporated theme of the input verb (and bearing in mind the discussion of the examples in (198) and (199)), it stands to reason that these nouns do not accept a van-PP expressing a theme. The er-nouns in (200) seem to be fully lexicalized in the sense that the “incorporated” theme is more or less fixed; the compound wasverzachter, for example, does not alternate with something like lakenverzachter, in which the theme (meaning “sheet”) is more specific. Furthermore, the second member often does not occur without the incorporated theme: *verzachter, *stiller, #koker.

There are also cases where it is not the external argument of the input verb that forms the denotation of the derived er-noun. In this case, it is usually the instrument used to perform the state of affairs denoted by the derived noun. Such nouns are especially common as the second member of compounds with an incorporated theme of the input verb as its first member. Thus, the er-nouns in the primeless examples in (201) refer to instruments used to open corked bottles, to tighten screws, to report a fire, and to play a CD. That it is indeed an instrument and not an impersonal agent that is being referred to is clear from the primed examples, which show that these er-nouns cannot themselves perform the action denoted by the input verb.

The unacceptability of the primed examples in (201) probably lies in the fact that the verbs in question normally select a [+human] agent. This selection restriction can be overcome by adding an adverbial phrase such as automatischautomatically, as in (202), which makes these examples acceptable.

The instrumental er-nouns in (201) have more or less the same properties as the er-nouns in (200) denoting [-human] agents: (i) it is impossible to express the theme argument by means of a van-PP; (ii) the first member of the compound is more or less fixed; and (iii) the er-noun often does not occur without the incorporated theme, as shown by the fact that the nouns trekker, draaier, melder, #speler are all agentive in the typical case. The third restriction is not absolute, however, since there are cases of instrumental nouns where incorporation is not necessary; cf. (flessen)openerbottle opener, (was)knijperclothespin, and (vogel)kijkerbinoculars. Note that we are clearly not dealing with impersonal agents in such cases, since the primeless examples in (203) show that the instrument itself cannot normally perform the action denoted by the input verb. The primeless examples show that, as with impersonal agents, explicit mention of the theme as an argument of the derived noun yields an unacceptable result, or forces an often improbable [+human] reading. The examples in (203) that allow a [+human] reading are marked with “#”.

The constructions in (204) seem to be counterexamples to the claim that the instrument itself cannot perform the action denoted by the input verb. However, the instruments do not act as agents in these cases: we are dealing here with so-called adjunct middle constructions, which involve some implicit or generic agent for whom it is easy/pleasant to perform the denoted action with the help of the specified instrument; we refer the reader to Section V3.2.2 for a detailed discussion of these constructions.

Note that it is also possible to have a voor-PP in examples like (203); een opener voor flessenan opener for bottles; knijpers voor wasgoedpegs for laundry; een kijker voor vogels binoculars for watching birds. In such cases the postnominal PP is an adjunct and not the theme of the input verb. This is clear from the fact, illustrated by the examples in (205), that it is possible to place the PP in post-copular position, which is a hallmark of adjunct status; cf. Section 16.2.1, sub III.

Finally, we want to make clear that er-nouns are often ambiguous between an agentive and an instrumental reading, especially when there is no incorporation of the theme. A typical example is kijker in (206), which is used as an instrumental er-noun in (206a), but as an agentive er-noun in (206b).

This subsection concludes with two rare and unproductive types of er-noun. The first type concerns cases in which the -er ending represents the theme argument: martelaar in (207a) denotes the person being tortured, and aanrader in (207b) denotes the thing being recommended. The correct use of the noun gijzelaar in example (207c) is a matter of debate: in its normative reading it is used to refer to the theme of the action of kidnapping (i.e. the hostages), while in colloquial speech it is often used to refer to the agent (i.e. the kidnappers). It is therefore not surprising that gijzelaar is often avoided by using the unambiguous forms gijzelnemer (agent) and gegijzelde (theme) instead.

The primeless constructions in (208) show that it is not possible to express the agent in these constructions with a van or door-PP. Although example (208a') suggests that the agent cannot be expressed by a possessive pronoun/genitive noun phrase either, it is difficult to determine whether Jan functions as the agent or as the possessor of Jans aanrader in (208b'); if the latter, this explains the unexpected acceptability of this construction.

The deverbal nouns in (207) behave similarly to the deverbal nouns ending in -sel, which typically represent the theme argument: verzinselfabrication, bakselbaking, bouwselbuilding/structure; cf. Knopper (1984). For example, the construction with van in (209a) is only acceptable in a possessive reading; this is also clear from the fact, illustrated by (209b), that placement of the van-PP in post-copular position is possible, which is a hallmark of adjunct status; cf. Section 16.2.1, sub III.

The second rare and unproductive type are er-nouns used to refer to the event denoted by the input verb or its result. Two examples are given in (210). The input verbs typically denote actions that the participants in the action cannot control; cf. Van der Putten (1997:147).

The verb uitglijden in (210b) is unaccusative, which is remarkable since Subsection B will show that unaccusative verbs cannot normally be the input for er-nominalization. Other er-nouns derived from unaccusative verbs that may belong to the same unproductive class are meevallerpiece of good luck and tegenvallerdisappointment, which are derived from dyadic unaccusative (nom-dat) verbs: De opkomst viel hem mee/tegenThe turnout was encouraging/disappointing for him. Since Section 15.3.1.1, sub IIID, has shown that nom-dat verbs do not allow any form of nominalization, we can safely assume that the er-nouns of the kind under discussion are fully lexicalized idiomatic expressions.

Er-nouns prototypically denote concrete agentive entities: as a rule, verbs must have an (external) agent argument denoting the person (or thing) performing the action denoted by the verb in order to qualify as input to er-nominalization. After er-nominalization, the agent of the input verb is expressed by the nominalizing affix –er itself, and is therefore not part of the argument structure of the derived noun. This predicts that unaccusative verbs cannot be the input for er-nominalization, since these verbs select an (internal) theme and not an (external) agent argument: they do not denote an action performed by their subject, but a process to which their subject has been submitted (which does not preclude that the argument is actively involved in bringing about the process). This prediction is indeed correct for unaccusative verbs indicating movement or change of state. Some examples are given in (211), all of which seem extremely marked, at least when considered in isolation.

Other unaccusative verbs also lead to questionable results: the er-nominalizations of the unaccusative verbs in (212) are unacceptable (although trouwer can be used with the intended reading in Flemish Dutch, hence the “%” sign).

At the same time, it must be said that even highly marked er-nouns can become acceptable in the right context. For example, the noun sterver in (212c') could conceivably be used in an example such as (213a) to refer to an actor who excels in dying scenes, in which case we are dealing with a repeated and deliberate action. Note that in this usage verbs like stervento die also allow passivization, as shown in (213b). This shows that these unaccusative verbs can behave like intransitive verbs in more than one way (but not all, given that the auxiliary verb cannot be changed to hebbento have).

The nouns in (214) also exhibit unexpected behavior in the sense that compounding can positively affect the acceptability of the derived noun. For example, although the motion verb komento come does not allow the derivation of *komer, the compounds laatkomer and nieuwkomer do exist. Such cases involve a certain degree of lexicalization, as can be seen from the fact that in the case of gaan, the er-noun has the irregular form -ganger.

The derived noun may also exist as a lexicalized form with a specialized meaning. For example, the er-noun beginner denotes an inexperienced person, not just any person starting out. Similarly, the er-nouns stijgerclimber/riser and dalerfaller/descender can be used to denote persons or things that climb or fall in the context of a listing or competition (as in sports, charts or financial indexes), but not just any entity that physically rises or falls. Two more examples are given in (213): the noun blijvertje in (213a) denotes entities that are of a more permanent nature, not just entities that remain in a certain place, and the noun phrase de binnenkomer in (213b) refers to a witty introduction to a topic and cannot be used to refer to an entity that enters a place.

All in all, er-nouns derived from unaccusative verbs are either reinterpreted in the sense of a prototypical er-noun, i.e. as denoting an agent, as in stervera person who repeatedly dies, or they acquire a specialized meaning as in beginnerdebutant/novice. This means that er-nouns derived from at least some unaccusative verbs are acceptable, although in many cases highly marked, provided the context is appropriate. This is confirmed by the fact that many of the nouns in (211) and (212) can be found at least occasionally on the internet.

There is another group of verbs that cannot be the input to the process of er-nominalization, consisting of monadic verbs like those in (216). Although these verbs are generally considered to be intransitive, there are also reasons to consider them to be unaccusative verbs; cf. Section V2.1.2. This view is supported by the fact that these verbs (in their monadic use) cannot easily undergo er-nominalization.

A noun such as brander is possible, but it does not denote burning entities (like a candle), but a [+human] agent (“distiller”) or an instrument with which old paint can be removed; this noun is therefore clearly not derived from the monadic verb we find in De kaars brandtThe candle is burning, but rather from its transitive counterpart. The er-nouns bloederbleeder/hemophiliac and drijverfloat are also used, but again these nouns have specialized meanings and should therefore be considered lexicalized. Another potentially problematic case is the somewhat marginal noun ?stinkerstinker, which can be used to refer to a person who stinks, although there is also a fully acceptable but highly lexicalized version of this noun, stinkerdrascal.

Example (217) shows that inherently reflexive verbs cannot undergo er-nominalization; the simplex reflexive zich cannot occur in either prenominal or postnominal position. We have seen in Section 15.3.1.2, sub IV, that in inf-nominalizations the reflexive must be realized in prenominal position; since er-nominalizations only take postnominal complements, the impossibility of er-nominalization of inherently reflexive verbs is exactly what one would expect. Note that these er-nominalizations are also unacceptable if the PP is omitted.

For completeness’ sake, (218) shows that er-nominalization is possible with optionally reflexive verbs; in these examples zichzelf can be treated as a regular argument, comparable to Bach/het recht op zelfbeschikking.