When modal verbs are used to express epistemic judgments, they indicate the likelihood of the actual occurrence of a particular eventuality. Although we will not address this issue here, the notion of actual occurrence should be understood as “actual occurrence within the present/past-tense interval”, with the choice between the present/past-tense interval depending on the tense marking of the clause; cf. Section 1.5.2 for a detailed discussion. This subsection focuses on Palmer’s distinction between three types of epistemic judgments, which he refers to as speculative, deductive and assumptive, and which are prototypically expressed in Dutch by kunnenmay, moetenmust and zullenwill, respectively.

By uttering sentences like (639a-c), the speaker makes three different epistemic judgments about (his commitment to the truth of) the proposition be at home (Marie), as expressed by the verbal projection embedded under the modal verb. The use of kunnenmay in (639a) presents the proposition as a possible conclusion: the speaker is uncertain whether the proposition is true, but on the basis of the information available to him he cannot exclude it. The use of moetenmust in (639b) presents the proposition as the only possible conclusion: on the basis of the information available, the speaker concludes that the proposition is true. The use of zullenwill in (639c) presents the proposition as a reasonable conclusion on the basis of the available evidence. The type of evidence on which the speaker’s evaluation is based is not important; it can consist of any information available to the speaker, including experience and generally accepted knowledge, as in Het is vier uur; Marie kan/moet/zal nu thuis zijnItʼs four oclock; Marie may/must/will be home now’.

It is not immediately clear whether the three subtypes of epistemic modality in (639) are exhaustive. For example, the slightly different constructions with mogen and willen in (640) may be instantiations of epistemic modality, but they also have additional meaning aspects (which can partly be attributed to the particles dan and nog wel eens). The clause with the verb mogen in the first conjunct of (640a) is assumptive in that it indicates that the speaker accepts that the proposition Jan is smart is true, but the central meaning aspect of the sentence as a whole is concessive and somewhat derogatory in nature; the main message is given in the second conjunct, which asserts that Jan is not very clever with his hand; cf. also Haeseryn et al.(1997:1618). Similarly, the construction with the verb willen in (640b) seems speculative in nature, but the main message of the construction as a whole is that the lamp has a tendency to tip over.

Let us return to the judgments concerning the truth of the proposition be at home(Marie). It is clear from the paraphrases of (639) in (641) that the truth values of the embedded propositions are evaluated epistemically: the paraphrases express the epistemic judgment and the proposition in different clauses, with the former expressed in the main clause and the latter expressed in a finite embedded clause. The embedded clause functions as the logical subject of the epistemic predicate in the main clause, which is expressed here by coindexing the anticipatory subject pronoun het and the embedded clause.

That we are dealing with special cases of epistemic modality in (640) is supported by the fact that these examples can be given similar paraphrases as the examples in (639), as shown by the examples in (642).

Note that in (641) and (642) we used the predicate Vmod wel zo zijnV well be the case, but that the modal verb kunnenmay can also function autonomously as an epistemic predicate: cf. Het kan dat Marie nu thuis isIt may be that Marie is at home now. This autonomous use seems less common with moeten, mogen and willen, and practically impossible with zullen.

That epistemic modal verbs are predicated of a propositional complement is also clear from the (b)-examples in (643); the modal verbs are predicated of the demonstrative pronoun datthat, which is interpreted as referring to the proposition expressed by Marie is nu thuisMarie is at home now. However, the use of epistemic mogen and willen is less felicitous in this case: Ja, dat #mag/*wil wel. Note that the number sign indicates that Dat mag (niet) is acceptable with a deontic reading “That is (not) allowed”. For a corpus research on the autonomous intransitive use of modal verbs, see Nuyts (2014) and Nuyts et al. (2023); the latter study considers (643b') as an isolated idiomatic case (with an implicit infinitive); another case might be Het zal weer eens niet! ‘Oh no: not like this again, eh?’. Unfortunately, this corpus study does not systematically take the type of modality into account.

The (b)-examples in (643) shows that epistemic modal verbs are monadic in that they take a single (pronominalized) propositional argument, as argued in Klooster (1986/2001a:259-260). This is also recognized in Nuyts (2014:363-4), but analyzed there as involving an implicit (or elided) infinitive or infinitival complement; cf. also Aelbrecht (2014) for relevant discussion. We will adopt Klooster’s proposal, which implies that the nominative subject Marie in (639) is not selected by the epistemic modal verb, but licensed by the main verb of the bare infinitival clause embedded under the modal verb and subsequently promoted to the subject position of the whole clause by subject raising, as schematically indicated in (644) for the modal verb moeten; Subsection B will show that epistemic modal verbs differ crucially in this respect from modal verbs expressing dynamic and directed (but not non-directed) deontic modality.

That the nominative subject of the clause is selected by the embedded main verb is also supported by the fact that it can be part of an idiomatic construction, as in (645a). If the subject were selected by the modal verb, the availability of this idiomatic reading would be quite surprising, since idioms are stored as units in the lexicon.

Evidentiality does not pertain to the truth of the proposition, but to the evidence that supports it. Palmer (2001: §2.2) distinguishes two kinds of evidence. The first type is reported evidence and includes evidence based on second/third-hand accounts, hearsay, etc. Dutch does not seem to have special modal verbs to express this kind of evidential modality, with perhaps one exception: the past-tense form of zullenwill can be used to express that the speaker is not committing himself to the proposition, but is relying on some source of information, which usually remains unidentified, but can in principle be made explicit by an adverbial volgens-PP. This description would also explain the contrast in Hij zou volgens hem/??mij steenrijk zijn in terms of internal (in)consistency regarding the speaker’s commitment to the truth of the proposition.

The possibilities in Dutch are more limited than in German, which can use both the present and past tense of the verb sollen as well as the verb wollen to express this kind of evidential modality; cf. Palmer (2001: §2.2.2) and Erb (2001:82) for discussion and examples. The fact that Dutch has no specialized modal verbs to express this kind of evidentiality does not mean that it lacks the means to express it: communication verbs such as zeggento say are of course tailor-made for this function; cf. Peter zegt dat hij steenrijk isPeter says that he is immensely rich.

Palmer calls the second type of evidential modality sensory, and this refers to evidence obtained by means of the senses. It can be argued that this type of modality is expressed in Dutch by the perception verbs when they take a bare infinitival clause. For instance, example (647a) expresses that the speaker has direct, auditory evidence that the proposition Jan vertrokJan has left is true. In this respect, (647a) differs crucially from (647b), which states that the speaker has no direct evidence for the truth of the proposition Jan vertrokJan has left; he may have heard something from which he concludes that the proposition is true, but he may also have been told so by another person.

Similar examples are given in (648) with the perception verb zien ‘to see’. Example (648a) indicates that the speaker is observing Jan and evaluating his actions as part of a leaving event, such as packing his suitcase or getting into his car. The same may be true for (648b), but such examples are also possible if the speaker sees something that leads him to infer that Jan is leaving, such as his suitcase in the hallway or Marie looking sad.

There are several facts that support the idea that perception verbs can function as markers of evidential/sensory modality. First, perception verbs are like the undisputed modal verbs moeten, kunnen and zullen in that they take bare infinitivals as complements, albeit that these infinitival complements can contain an (optional) overt subject. Second, it seems that the verbs ziento see and horento hear are the ones that most often occur with a bare infinitival, which is consistent with the fact that, cross-linguistically, sensory evidential modality is also most often expressed by markers pertaining to visual and auditory stimuli. Third, it may explain the acceptability of examples like (649b&c) with the verb vindento consider: like constructions with perception verbs, the vinden-construction has a bare infinitival complement that typically refers to an eventuality that can be perceived by the senses, while further expressing that the truth assignment to the proposition denoted by the bare infinitival clause is based on the (subjective) sensory perception of the subject of the clause.

By assuming that Dutch has a set of modal verbs expressing sensory evidentiality, we avoid the need to assume a separate class of verbs consisting only of the verb vindento consider, which has properties virtually identical to those of the class of perception verbs (although one notable difference is that the subject of the bare infinitival complement of vinden cannot be omitted). However, since perception verbs are usually not treated as a subtype of modal verbs, we will not pursue this option here, but will discuss them in their own right in Section 5.2.3.3.

Finally, it should be noted that Dutch verbs like blijkento turn out, lijkento appear, and schijnento seem in (650) are evidential in the sense that they can be used to indicate whether there is direct evidence for the truth of the proposition, whether there are identifiable persons who can be held responsible for the truth of the proposition, or whether we are dealing with hearsay/rumors; cf. Vliegen (2011). Since blijken, lijken, and schijnen do not select bare infinitival complements, they are not discussed here, but in Section 5.2.2.2.

Dynamic/dispositional modality describes some moving force internal to the nominative subject of the construction as a whole that favors the realization of the potential event denoted by the main verb of the verbal projection embedded under the modal verb. Two verbs prototypically used in this modal function are kunnento be able and willento want, expressing ability and volition respectively.

That the modal verbs in (652) function as main verbs is clear, as we saw earlier, from the fact illustrated in (653) that the bare infinitival clauses can be pronominalized. The fact that the nominative subjects in these examples are realized shows that they are external arguments of the modal verbs and not (introduced as) part of the bare infinitival clauses; this shows that dynamic/dispositional verbs differ from epistemic modal verbs in that they are not monadic but dyadic predicates.

To account for the fact that the nominative subject of the construction as a whole is also construed as the subject of the infinitival clause, Klooster (1986) proposes a control analysis for constructions of this type: the external argument of the modal verb functions as the controller of the implied subject PRO of the embedded bare infinitival clause. This is shown schematically in (654), where coindexing indicates coreference.

The modal verbs moetenmust and zullenwill can also express a dynamic/dispositional modality, in which case they express a strong desire and a great determination, respectively. The primeless examples in (655) show that this use of moeten and zullen is special in that it requires an emphatic accent on the modal.

The examples in (656) show that this use of moeten and zullen is also special in that pronominalization of the bare infinitival clause leads to a less felicitous result. The degraded status of (656a) in the intended reading may be due to the fact that the directed deontic (obligation) reading of this example is simply the more salient one, but a similar explanation is not available for the degraded status of (656b).

The modal verb zullen is often used in coordinate structures with the other dynamic/dispositional modal verbs to express determination in addition to ability, volition, desire, etc. In particular, the combination moet en zal is very common and has the fixed meaning “nothing will stop me from ...”. All examples in (657) are taken from the internet and require emphatic accents on the modal verbs. Pronominalization of the bare infinitival clause is not illustrated here, but again gives degraded results in these cases.

The modal verb mogen, which is normally used as a deontic verb, is occasionally found with a dynamic/dispositional function when it is accompanied by the adverbial phrase graaggladly; this is shown in (658), which is obsolete for some speakers.

Directed deontic modality differs from dynamic/dispositional modality in that the moving force is external to the subject of the sentence rather than internal to it. Two verbs prototypically used with this modal function are moetento be obliged and mogento be allowed, expressing obligation and permission respectively. The examples in (659) show that when the external force is exerted by a person in authority, this can be made explicit by an adverbial van-PP; cf. Barbiers (1995a) and Aelbrecht (2014).

The external force may also be impersonal (e.g. laws and other regulations), in which case the force can often be made explicit by a volgens-P. See Nuyts et al. (2007) for a discussion of the role of certain standards., i.e. in terms of admissibility, desirability or acceptability.

That the modal verbs in (659) function as main verbs is clear from the fact that the bare infinitival clause lends itself quite easily to pronominalization. The examples in (661) show that the subject of the sentence is not part of the infinitival clause, from which we can conclude that modal verbs expressing directed deontic modality are similar to modal verbs expressing dynamic/dispositional modality in that they are not monadic but dyadic predicates.

Pronominalization may be a bit marked in the case of examples such as (660); it seems preferable to replace the infinitival clause by the verb phrase doen + datto do that. Note, however, that negative clauses with the deontic modal mogento be allowed are perfectly normal without the verb doen: Volgens de regels mag hij dat niet (doen)According to the rules, he is not allowed to do that.

The fact that the nominative subject of the construction as a whole is not affected by pronominalization shows that directed deontic constructions are like dynamic/dispositional modal constructions in that they are amenable to a control analysis. This is shown in (663).

The verb zullen can also be used to express directed deontic modality when the speaker wants to express that he is committed to the actualization of the proposition denoted by the main verb of the bare infinitival clause; with an example such as (664a) the speaker indicates that he has the authority to instruct the technical department and can therefore promise that everything will be fixed. For the same reason, examples such as (664b) are perceived as rude (or even threatening), since the speaker indicates that he has the authority to order the addressee about (and to take measures if he does not obey).

Although the verb willenwant cannot be used to express directed deontic modality, Barbiers (1995a) suggests that the verb kunnencan in examples such as (665a) can, and indeed such examples can be construed as speech acts of granting permission. However, it is not so clear whether we are really dealing with directed deontic modality, since Palmer (2001:77) notes that examples such as (665b) may simply express that there is nothing to prevent Jan from leaving, and suggests that we are dealing with dynamic/dispositional (ability) modality here. A similar interpretation might be possible for (665a), if we assume that the speaker indicates with this sentence that in his opinion all conditions for Jan’s leaving are fulfilled.

The discussion of the examples in (665) shows that it is not always easy to determine the type of modality we are dealing with, but we will see in the next subsection that there may be reason to assume that kunnen can indeed be used as a directed deontic modal.

This subsection discusses a type of modal construction first discussed in Klooster (1986) and called non-directed modality in Barbiers (1995a). To introduce this type of event modality, we will begin with a brief digression on the passivization of clauses expressing event modality. First, consider example (666a), which expresses dynamic/dispositional modality: the agent het meisjethe girl has the wish to pet the cat. The passive counterpart of this example in (666b) also expresses dynamic/dispositional modality, although now it is the patient de katthe cat that wants to be petted. In both cases, however, the moving force is internal to the nominative subject of the construction as a whole.

That we are dealing with regular dynamic/dispositional modality in the passive construction (666b) is supported by the fact, illustrated in (667), that the modal verb remains dyadic under passivization; pronominalization of the infinitival clauses does not affect the nominative subject, regardless of the voice of the embedded clause. This shows that the nominative subject cannot originate in the embedded infinitival clause, but must be selected by the modal verb itself; we are dealing with control structures in both the active and the passive case.

That the nominative subject of the construction as a whole is selected by willen is also supported by the examples in (668), which show that the subject of the passive construction must have volition. If not, the construction is semantically incoherent.

Things are quite different in the case of directed deontic constructions. This is immediately clear from the fact that examples such as (669a) can easily be passivized, with the result that the nominative subject of the passive construction is an inanimate entity without control over the proposition expressed by the infinitival clause; the fact that (669b) is nevertheless semantically coherent shows that the obligation expressed by the modal verb moeten cannot be directed towards the subject of the clause, but must be directed towards the implicit agent of the infinitival verb vervento paint.

Barbiers (1995a) refers to the examples in (669a) and (669b) as directed and non-directed deontic modality, respectively, where (non-)directed should be interpreted as “(not) directed towards the subject of the sentence”. The examples in (670) show that the active and passive constructions in (669) differ not only with respect to the directional force of the modal, but also with respect to the pronominalization of the embedded infinitival clause; while the nominative subject of the active construction is not affected by pronominalization, the nominative subject of the passive construction is.

As already indicated by the comments in square brackets in (667) and (670), dynamic/dispositional and deontic constructions differ with respect to the origin of the nominative subject of the construction as a whole. In dynamic/dispositional constructions, the subject originates as an argument of the modal verb, regardless of whether the embedded infinitival clause is in the active or passive voice; the schematic representations in (671a&a') show that in both cases we are dealing with control structures. In deontic constructions, on the other hand, the origin of the nominative subject depends on the voice of the embedded infinitival clause: if the infinitival clause is active, the subject is an argument of the modal verb, but if it is passive, the subject originates as an internal argument of the infinitive; the schematic representations in (671b&b') show that in the former case we are dealing with a control structure and in the latter with a subject-raising structure. In short, non-directed deontic modality is special in that it patterns with epistemic modality in requiring a subject-raising analysis.

The discussion above provides us with a test to answer the question raised at the end of the previous subsection, whether the verb kunnen can be used to express directed deontic modality. If so, we expect that example (672a) can be passivized, and that the resulting construction need not involve subject-oriented modality. This seems to be confirmed, since example (672b) must be interpreted as attributing the ability to the implicit agent of the infinitival verb, and not to the inanimate subject dat boekthat book. That (672a) and (672b) are directed and non-directed deontic, respectively, is also supported by the fact that pronominalization of the infinitival verbs affects the nominative subject of the construction as a whole only in the latter case; this shows that kunnen is dyadic in (672a) but monadic in (672b).

The fact that the moving force in non-directed deontic constructions is directed at an entity other than the nominative subject also means that this type of modality differs from the other types of event modality in that the nominative subject need not be able to control the eventuality expressed by the infinitival clause. The examples in (673) show that, as a consequence, the infinitival clause can be a copular construction or headed by an unaccusative/undative verb; all examples are taken from the internet.