Deverbal verbs prefixed with be- come in different kinds. For example, Subsection I has shown that in many cases the accusative object of the derived verb corresponds to the nominal part of a prepositional phrase in constructions with the corresponding simple verb; cf. (471).

The present discussion focuses on the locative alternation in (472), in which the prepositional reference object of (472a) appears as the direct object of the derived verb in (472b). That the noun phrase has the grammatical function of direct object in this example will be clear from its promotion to subject in the corresponding passive construction in (472b'). The accusative located object of (472a) appears as an optional met-PP in the (b)-examples in (472); when it is omitted, the located object is semantically implied in the sense that we can still infer that the reference object is covered with “pastable” objects.

There is a clear difference in meaning between the two examples in (472a&b): while (472a) is compatible with a reading in which the located object covers only part of the reference object, (472b) implies that the reference object is completely (or at least largely) covered by the located object. This comes to the fore by replacing the plural noun phrase de posters in (472) by a singular one; while (473a) is easily possible, example (473b) is only acceptable in the less common case that the poster completely covers the wall.

This contrast suggests that deverbal be-verbs express that their objects are affected as a whole. This could be further supported by the fact that example (472a) also alternates with the construction in (474a), where the notion of total affectedness is expressed by the adjective volfull. The crucial observation is that this adjective is incompatible with deverbal be-verbs; this could be explained by claiming that (474b) is tautological: vol and the prefix be- perform in a sense the same semantic function. We will return to a more formal account of this point in Subsection B.

Note in passing that the notion of total affectedness should not be taken too literally, since the extent to which the reference object is affected can be further specified by attributive modifiers such as heel/halfwhole/half or degree modifiers such as helemaal/gedeeltelijkcompletely/partly; cf. (475). This suggests that the relevant aspect of meaning is simply “affectedness”, with the interpretation of “total affectedness” as a default value that can be overridden by the addition of the modifiers mentioned above.

Table 3 lists a small sample of verbs of the type in (472). Note that not all verbs in this table can be combined with volfull; this is possible with the first five, but not with the last three. This suggests that the prefix be- and the adjective vol are not fully equivalent semantically; cf. Van Hout (1996:48) for a first attempt to describe this difference in meaning.

stem	verb	translation
hangen ‘to hang’	behangen met	to paper with
laden ‘to load’	beladen met	to load with
leggen ‘to put’	beleggen met	to fill (a sandwich) with
plakken ‘to paste’	beplakken met	to paste with
smeren ‘to smear’	besmeren met	to smear with
sproeien ‘to spray’	besproeien met	to spray with
spuiten ‘to spray’	bespuiten met	to spray with
strooien ‘to strew’	bestrooien met	to strew with

The verbs in Table 3 do not form a uniform set, and may exhibit diverging behavior in other respects. For example, while the verb plakken must be prefixed with be- in order for the reference object to appear as an accusative object, this is not the case for the verbs ladento load, (een boterham) smerento butter (a sandwich), (het gazon) sproeiento water (the lawn) and (de auto) spuitento spray (the car); the examples in (476) show for two of these verbs that they alternate not only with the (b) but also with the (c)-examples.

Our judgments in (476) suggest that the met-PP produces a slightly better result in the (b) than in the (c)-examples, but this has not yet been seriously investigated. It is also interesting to note that Dutch deverbal be-verbs differ crucially from their English counterparts in that they always allow the omission of the met-PP. Hoekstra et al. (1987) note that English deverbal be-verbs fall into two subgroups here: verbs corresponding to Dutch verbs that allow the (c)-alternant in (476), like to load and to spray in the (a)-examples in (477), tend to take an optional with-PP; verbs corresponding to Dutch verbs that do not allow this alternant, like to hang and to pack in the (b)-examples, take an obligatory with-PP (cf. Hoekstra & Mulder, 1990:20, for further discussion).

De Haas & Trommelen (1993:68-9) shows that denominal verbs prefixed with be- can be of different types; here we are interested in cases such as (478b). Example (478b) has a similar meaning to (478a), but additionally expresses that the reference object is totally affected; after the completion of the activity, the bread will be completely covered with butter, on one side. Example (478b) also shows that the located object boterbutter has in a sense been incorporated into the verb, i.e. has become an inherent part of the be-verb. The prepositional reference object op het broodon the bread, on the other hand, appears as the accusative object of the denominal verb: it is promoted to subject of the clause in the regular passive construction in (478b').

The examples in (479a&b) show that there is a striking syntactic difference between the two examples in (478a&b); while the assertion in (478b) can be made more specific by adding a substance-denoting met-PP, the addition of such a PP leads to an incoherent reading of (478a). To express the more specific assertion, we need to replace the direct object boter with margarine, as in (479a'). This shows that the denotation of the nominal part of the be-verb has become less prominent as a result of incorporation.

The examples in (480) further show that the formation of be-verbs is not fully productive; a noun like jam in (480) cannot be used as the stem of a be-verb. This suggests that the attested denominal be-verbs are listed in the lexicon.

A small sample of be-verbs of the type in (478) is given in Table 4. The first column gives the nominal stem of the verb and its English translation, the second column the derived verb, and the third column a translation or paraphrase in English.

stem	verb	translation
bos ‘wood’	bebossen	to afforest
dijk ‘dike’	bedijken	to put dikes around/next to
mest ‘manure’	bemesten	to manure/fertilize
modder ‘mud’	bemodderen	to put mud on
schaduw ‘shadow’	beschaduwen	to shade
vracht ‘load’	bevrachten	to load
water ‘water’	bewateren	to water

Note that it is sometimes difficult to tell whether we are dealing with a denominal or a deverbal be-verb. For instance, the examples in (481) suggest that beplantento plant with can be deverbal or denominal.

The examples discussed in the previous subsections involve a certain change of location; some entity is relocated with respect to some reference object. The examples in (482) are different in that they involve a path: the (a)-examples express that Jan travels a path that has its endpoint inside the hall, and the (b)-examples that Peter travels a path that goes to the top of the mountain.

Levin (1993:43) discusses this alternation as a special case, but it seems that we are basically dealing with the same phenomenon; the verb is prefixed with be-, and the postpositional phrases de zaal binnen and de berg op are replaced by noun phrases that function as direct objects. The fact that the noun phrases in the primeless and primed examples have different syntactic functions is clear from the fact that they behave differently under passivization; the complement of the postpositional phrase in the primeless examples cannot be promoted to subject, whereas the complement of the be-verb in the primed examples can; this is illustrated in (483) for the (b)-examples in (482).

However, there is also an essential difference between the change-of-location cases and the directional cases; the stem of the directional be-verbs typically belongs to the class of unaccusative verbs. The examples in (484) illustrate the inability of verbs of transitive resultative constructions (i.e. constructions in which the complementive is predicated of an accusative noun phrase) to act as the stem of a directional be-verb.

be-verbs denoting a change of location are not restricted in this way, as can be seen from the difference between the (b)-examples in (484) and the examples in (485).

In fact, the stems of deverbal be-verbs denoting a change of location are typically transitive. Unaccusative verbs of change-of-location verbs such as vallento fall cannot be used as input to such be-verbs; the examples in (486b&c) show that the reference object cannot occur as an accusative or nominative noun phrase, and that the located object cannot be realized as a met-PP.

The only possible counterexample we could find is given in (487), but it seems likely that we are dealing here with a directional rather than a change-of-location construction, since (487c) does not necessarily imply that the lion will land on top of the gazelle; the examples in (487a&b) show that this also holds for the directional construction, but not for the change-of-location construction.

Some possible cases of unaccusative verbs that can be used as input for the formation of directional be-verbs denoting a path are given in Table 5; these cases require more in-depth investigation.

stem	verb	translation
naderen ‘approach’	benaderen	to approach (something)
reizen ‘to travel’	bereizen	to travel (across/through)
springen ‘to jump’	bespringen	to pounce
sluipen ‘to steal/prowl’	besluipen	to stalk
stijgen ‘to rise’	bestijgen	to mount/ascend
varen ‘to sail’	bevaren	to sail (the sea)

Denominal ont-verbs like ontharento depilate and ontkurkento uncork in the singly-primed examples in (488) express in a sense the opposite of the denominal be-verbs discussed in Subsection 2; both types denote a change of location, but whereas in the case of the denominal be-verbs the reference object refers to the new position of the moved entity, in the case of the denominal ont-verbs it refers to its original position. The doubly-primed examples show that, as with the be-verbs, the reference object appears as the direct object of the ont-verbs, as can be seen from the fact that it is promoted to the subject in the regular passive.

Table 6 provides some more examples of denominal verbs prefixed with ont-. Sometimes denominal be- and ont-verbs are antonymous, as in bebossen and ontbossen, but in many other cases there is no antonym. This strongly suggests that the formation of be- and ont-verbs is not a productive process, and that the attested cases are listed in the lexicon.

stem	verb	translation
bos ‘forest’	ontbossen	to deforest
grond ‘soil/basis’	ontgronden	to take away the soil/basis
hoofd ‘head’	onthoofden	to decapitate
kalk ‘lime’	ontkalken	to decalcify
volk ‘people’	ontvolken	to depopulate

The examples in (489) denote a metaphorical path from one state of affairs to another. The referent of the noun phrase Krakras (a character from a Dutch series of children’s books) changes from a state in which it has the form of an unpalatable-looking bird to a state in which it looks like a tasty duck that can be used as an ingredient for soup.

Constructions such as (489) often alternate with constructions using denominal ver-verbs; an example is given in the (a)-examples of (490). The (b)-examples show that a ver-verb corresponding to the causative counterparts of veranderen sounds a bit awkward.

Similar cases are listed in Table 7. Sometimes the meaning of the denominal ver-verb has narrowed down to the paraphrase given after the arrow “⇒”.

	verb	translation
film ‘movie’	verfilmen	change into a movie ⇒ adapt (a story) for the screen
gas ‘gas’	vergassen	change into gas
gras ‘grass’	vergrassen	change into grassland
kool ‘coal’	verkolen	carbonize
snoep ‘sweets’	versnoepen	change into sweets ⇒ spend money on sweets
water ‘water’	verwateren	change into water ⇒ dilute

Note in passing that the deadjectival verbs prefixed by ver- in the primed examples in (491) express a semantic aspect similar to those in Table 7, but are related to the inchoative copular or resultative constructions in the primeless examples.

The examples in (492a&b) show again that be-verbs can sometimes be paraphrased by a resultative construction with the adjectival complementive vol; cf. Subsection A1. Example (492c) further shows that be- and vol are in complementary distribution.

Following an earlier proposal in Dik (1980:36), Hoekstra et al. (1987) have argued that the pattern in (492) shows that be- functions syntactically as a complementive comparable to vol. However, it has the special property that it incorporates into the verb; if we assume that the complementive and the noun phrase of which it is predicated form a small clause, the analysis of the examples in (492a&b) looks as shown in (493); while vol remains inside the small clause, be- is moved out to incorporate into the verb plant.

Example (494) shows that this analysis can be applied more generally. The fact that the simplex reflexive zich can be used with the complex verb bedrinkenget drunk in (494a) can be seen as an empirical argument for the claim that the element be- functions as a complementive; example (494b) shows that, unlike internal theme arguments of verbs, logical subjects of complementives can normally be realized by such a reflexive (cf. Sections 2.5.2, sub I, and N22.4 for discussion).

The following subsections will show that a considerable subset of complex verbs prefixed by be-, ver-, and ont- can be derived in a similar way, i.e. by incorporation into the verb. The extent to which this type of analysis can be applied to the class as a whole is not a priori clear. For instance, the semantic correspondence between the examples in (495a&b) might lead to the idea that they have a similar underlying structure, in which the adjective vuildirty acts as a complementive, and that be- is thus not the head of a small cause, but a causative element that attracts the predicative head of the small clause, as in (495c).

However, this alternative analysis may be less attractive, since it reintroduces the problem that the prefix be- is exceptional in that it determines the category of the complex form in violation of the righthand head rule. It is therefore not surprising that it has been suggested that (495c) is in fact not the correct analysis. Hoekstra (2004:365ff) argues that the derivation of (495b) proceeds in essentially the same way as in (493), with the difference that the verb into which be- is incorporated is an abstract (phonetically empty) causative verb: the adjective must also be incorporated in order to satisfy the requirement that the prefix be- be morphologically supported. See Mulder (1992: §9) for an alternative proposal.

Note, however, that this analysis implies that be- is polysemous: in examples like (492c) and (494b) it is a monadic predicate expressing some notion of total affectedness, whereas in (496) it functions as a dyadic predicate with a meaning comparable to the copular verb zijnto be. In fact, Hoekstra suggests that this does not exhaust the possibilities, and proposes a derivation for “ornative” be- verbs such as bewapenento arm along the lines of (497b), in which be- is again a dyadic predicate, but now with a meaning comparable to the verb hebbento have.

Of course, derivations similar to those in (496) and (497) can also be used to derive denominal and deadjectival ver- and ont-verbs. Another case discussed in Hoekstra (2004) is the construction in (498). He claims that this is actually an applicative (= preposition incorporation) construction of the type described in Baker (1988) for languages like Chichewa; cf. also Voskuyl (1996). Hoekstra’s analysis is given in (498b).

As the discussion above shows, it seems possible to explain a large variety of be-, ver-, and ont-verbs by syntactic incorporation. This proposal is motivated not only by the fact that it can account for the exceptional behavior of these prefixes with respect to the righthand head rule, but also by a larger set of empirical data that will be discussed in the following subsections.

The incorporation analysis can immediately account for the complementarity in the distribution of be- and the adjectival complementive volfull in example (492c), by appealing to the more general restriction that a clause can contain at most one complementive. Further examples showing that verbs prefixed by be-, ver-, and ont- cannot be combined with a complementive are given in (499); cf. Section 2.2.1, sub IV, for discussion.

At first glance, the examples in (500) seem to provide counterexamples to the claim that complex verbs prefixed by be- cannot take a complementive; the tot/als-phrases seem to be predicated of the accusative noun phrases and thus function as complementives. However, there are at least two reasons to reject this conclusion: (i) the tot/als-phrases are optional; (i) they can occur in postverbal position. These facts would follow immediately if it is the prefix that functions as a complementive: the tot-phrases would then have a different function, and therefore would not be expected to exhibit the behavior of standard complementives.

This proposal is very close to the one proposed for the constructions with and without the verbal particle neerdown in (501); Section 2.2.1, sub IV, has shown that the contrast between the (a) and the (b)-examples is due to the fact that the particle neerdown must be analyzed as a complementive; this means that the PP op de tafel cannot be analyzed as a complementive in the (b)-examples; it is therefore to be expected that the (b)-examples differ from the (a)-examples in that they allow the PP to be omitted or to occur in postverbal position.

If the prefixes be-, ver-, and ont- do indeed originate as predicative heads of small clauses, we would expect them to have an effect on argument structure similar to that of complementives and verbal particles; cf. Section 2.2. The examples in (502a&b) show that the use of an adjectival complementive can add an argument to the otherwise impersonal verb vriezento freeze, and (502c) shows that prefixation with be- can have a similar effect. The fact that (502b&c) both take the perfect auxiliary zijnto be shows that we are dealing with unaccusative structures, and this is of course to be expected, since the additional argument is introduced as the subject of the small-clause heads dood and be-.

Prefixation with ver- and ont- can also add an argument to the otherwise impersonal verbs waaiento blow and dooiento thaw, as shown in (503)

The primeless examples in (504) show that the use of the adjectival complementive platflat adds an argument to the otherwise intransitive verb lopento walk and that prefixation with be- again has a similar effect. The primed examples show the same for the prefix ver-.

Nice minimal pairs are less easy to find for verbs prefixed with ont-, but possible cases are Jan loopt zijn straf misJan misses his punishment and Jan ontloopt zijn strafJan escapes his punishment can be used to illustrate the main point.

The (a)-examples in (505) show that adding a locational complementive to an intransitive verb can also produce an unaccusative verb; whereas (505a) takes the auxiliary hebbento have, (505b) with the complementive wegaway takes the auxiliary zijnto be, which is sufficient to assume unaccusative status. The (b)-examples show that prefixing with ver- can have a similar effect; other unaccusative verbs prefixed with ver- are vertrekkento leave and vertoevento stay but these do not have an intransitive counterpart in the modern language . We have not found any cases with the prefixes be- and ont- that have this effect.

Example (506a) further shows that the addition of a complementive to an unaccusative verb normally has no effect on the number of arguments. However, the nominative argument is now no longer licensed by the verb but by the complementive, as can be seen from the fact that the complementive cannot be omitted. Example (506b) shows that the nominative argument can also be licensed by the prefix ver-, which supports the underlying small-clause analysis. The primed examples show that the same holds for transitive verbs; the number of arguments is not affected, but the direct object is semantically licensed not by the verb but by the adjectival complementive or verbal prefix.

As with verbal particles, prefixation can affect the aspectual properties of the construction; cf. Van Hout (1996:176ff). We show this here with the unaccusative verb brandento burn; whereas the construction in (507a) is atelic, the constructions in (507b&c) with a particle verb and a verb prefixed by ver- are telic. This aspectual difference is clear from the fact that the former takes the perfect auxiliary hebbento have and the latter the perfect auxiliary zijnto be; cf. Section 2.1.2, sub III, for further discussion of the relation between auxiliary selection and telicity.

The subsections above discussed the hypothesis proposed in Hoekstra et al. (1987) that the prefixes be-, ver-, and ont- function syntactically as complementives, and provided empirical evidence in support of this claim. We should be cautious, however, because the derivation of deverbal verbs prefixed by these prefixes is not a fully productive process, raising complex questions about the relationship between syntax and morphology. Furthermore, many of the putative input verbs are obsolete or no longer used with the intended meaning, and the output forms often exhibit idiosyncratic behavior. Given the complexity of the issue, this hypothesis needs to be investigated more thoroughly.