It seems that zullenwill differs from epistemic modal verbs like moetenmust and kunnenmay in that it has no intrinsic quantificational force. This will be clear from the examples in (331), in which the quantificational force must be attributed to the modal adverbs: zekercertainly expresses universal quantification over possible worlds, mogelijk/misschienpossibly expresses a low degree of probability, and waarschijnlijkprobably expresses a high degree of probability.

If zullen were inherently quantificational, we would expect the examples in (331) to be degraded or to produce certain special effects (which is indeed the case to varying degrees when we replace zullen with moeten or kunnen). For example, if zullen by itself expressed universal quantification, the modal adverb zeker in (331a) would be tautological and the adverbs mogelijk and waarschijnlijk in (331b&c) would be contradictory; if zullen by itself expressed existential quantification, the reverse would hold. Nevertheless, it should be noted that examples like (326) and (327a), which do not contain any element with quantificational force, are typically used when the speaker has strong reason to believe that eventuality k will occur in all possible worlds; high degree quantification therefore seems to be the default reading of sentences with zullen.

In order to describe the semantic contribution of zullenwill, we need to discuss a semantic aspect of epistemic modality that has only been mentioned in passing. Epistemic modality stands in contrast to what is known as metaphysical modality. The latter is concerned with objective truth and is part of a very long philosophical tradition concerned with the reliability of scientific knowledge. Epistemic modality, on the other hand, is concerned with the degree of certainty that an individual assigns to the truth of a proposition on the basis of his knowledge (cf. Greek epistemeknowledge). Epistemic modal verbs like moetenmust and kunnencan, for example, do not express a degree of probability that is objectively given, but one that results from an individual’s assessment of the situation based on the knowledge available to him. Hence, the difference between the declarative clause without a modal verb in (332a) and the declarative clause with a modal verb in (332b) is that in the former the proposition that Marie is at home is merely asserted “without indicating the reasons for that assertion or the speaker’s commitment to it” (Palmer 2001:64), while in the latter the modal verb indicates “that a judgment has been made or that there is evidence for the proposition” (Palmer 2001:68).

In his Kritik der reinen Vernunft (1781), Immanuel Kant already distinguished three types of epistemic modality, which he called problematic, apodeictic, and assertoric modality. Palmer (2001) makes essentially the same distinctions in Section 2.1; he refers to the three types as speculative, deductive and assumptive modality. Illustrations are given in (333).

By uttering examples such as (333), the speaker provides three different epistemic judgments about (his commitment to the truth of) the proposition that Marie is at home. The use of kunnenmay in (333a) 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 (333b) presents the proposition as the only possible conclusion: on the basis of the information available the speaker concludes that it is true. Finally, the use of zullenwill in (333c) presents the proposition as a reasonable but uncertain conclusion on the basis of the available evidence; cf. also Droste (1958:311) and Janssen (1983/1989). Palmer (2001) further suggests that the evidence may include experience and generally accepted knowledge as in Het is vier uur; Marie kan/moet/zal nu thuis zijnIt is 4.00 p.m.; Marie may/must/will be home now. Contrary to Palmer (2001: §2.1.2), we think that (at least in Dutch) this holds not only for assumptive, but for all kinds of epistemic modality.

The claim that epistemic modality involves some subjective evaluation is fully compatible with our earlier claim that epistemic modality introduces a set of possible worlds. Indeed, the notion of a possible world only makes sense if such a world is accessible in the sense that it can be entered from the world that is taken as the starting point. Thus, the creation of a point of view, however metaphorically expressed, is an essential part of the notion of a possible world; “Suppose now that someone living in w1 is asked whether a certain proposition p is possible (whether p might be true). He will regard this as the question as to whether in some conceivable world (conceivable, that is, from the point of view of his world, w1), p would be true …” (Hughes & Cresswell 1968:77).

That we are dealing with subjective assessments is clear: examples such as (334a) are definitely strange. The modals moeten and kunnen express that the suggested probability of the sun rising is only the result of an assessment by the speaker, who thereby suggests that the alternative view that the sun will not rise tomorrow could in principle also be viable. Example (334b) shows that the modal zullen also leads to an odd result; such examples are only possible if stating the obvious has some rhetorical function, as in Maak je niet druk, de zon zal morgen heus wel opkomenDo not get upset, the sun will still rise tomorrow. Janssen (1983) suggests that the markedness of the examples in (334) follows from Grice’s maxim of quantity: the expression of doubt makes the utterances more informative than necessary.

That epistemic modals entail an assessment by an individual may also receive support from the fact that examples like (335a&b) are perfectly acceptable when uttered by an amateur astronomer who has calculated the time at which the sun will rise on a given day for the first time in his life; in these cases the possibility that the sun will rise at a different time than stated is viable indeed, since the speaker may have made some miscalculation. The expression of doubt in these examples is thus consistent with the maxim of quantity.

Subjective assessment constitutes an essential part of the meaning of epistemic modal verbs. This is perhaps clearer in English than in Dutch, since epistemic clauses require the use of a modal verb in the English simple present, but not in Dutch. This difference can be formulated as in (336): English obeys the material implication in (336a), from which we can deduce (336a') by modus tollens (i.e. the valid inference: if P → Q ∧ ¬Q, than ¬P); Dutch, on the other hand, has the material implication in (336b), from which we cannot deduce (336b'), since that would be a formal fallacy.

It follows from this difference that the Dutch simple present can be used in a wider range of “future” constructions than the English simple present. Comrie (1985:118) has claimed that, under normal interpretations, the English simple present construction can only be used to refer to future states of affairs if we are dealing with what he calls scheduled events (such as the rising of the sun, the departure of a train, etc.); the number sign in (337a) is used to indicate that this example may be more acceptable if it is part of a planned series of events, as in Jan leaves tomorrow, Betty the day after and the rest of the family over the weekend. Under the reasonable assumption that scheduled events do not involve subjective assessment, this is correctly predicted by the valid inference in (336a'). However, see Quirk et al. (1985/1991: §4.9) for some syntactic contexts that also allow the future use of the simple present in English.

The invalidity of the inference in (336b'), on the other hand, expresses that Dutch is not restricted in the same way as English, but can freely use clauses in the simple present to refer to any future event; cf. Section 1.5.4 for further discussion.

Although the presence of an epistemic modal is not mandatory in contexts of subjective assessment in Dutch, the discussion above has shown that subjective assessment is an inherent part of the meaning of epistemic modals. Note that the person whose assessment is given can be made explicit by an adverbial PP. In accordance with the generalizations in (336) such PPs normally require an epistemic modal verb to be present in English present-tense constructions (Carole Boster, p.c.), while in Dutch they can be used both with and without such a modal.

The previous subsection has shown that epistemic modals are used to provide a subjective assessment of the degree of probability that the proposition expressed by the lexical projection of the embedded verb is true. The person making this assessment will henceforth be referred to as the source. Since the source does not need to be syntactically expressed by an adverbial volgens-PP, and does not even need to be identified by the context, it looks as if language users assign certain default values to the source. When uttered “out of the blue”, the assessment expressed by epistemic modals in present tense sentences such as (340a) is attributed to the speaker (who may of course rely either on his own judgment or on some other source). A default interpretation can only be overridden by explicitly assigning a value to the source by adding a volgens-PP, as in (340b). Note that it is also possible for speakers to explicitly present themselves as the source.

In past-tense constructions with the universal modal verb moetenmust, the default interpretation of the source again seems to be the speaker. As in the present tense, the default interpretation can be overridden or made explicit by adding a volgens-PP.

Recall from Subsection I that examples such as (341a) are typically used to indicate that a certain eventuality that occurred before speech time n was inevitable. Furthermore, example (342) shows that it is impossible to cancel the universal quantification expressed by the modal. The reason is that the first and the second conjunct in (342) have the same source, the speaker, so that the usual interpretation that the past-tense interval precedes speech time n leads to a contradiction: according to the first conjunct the eventuality occurs in all possible worlds in the past-tense interval, but according to the second conjunct the eventuality did not occur in the actualized part of the present-tense interval.

A possible problem for this account is that the past-tense interval can in principle include speech time n; cf. Section 1.5.1, sub I. Consequently, the first conjunct of (342) should be true if the collapse of the house takes place after speech time n. However, this reading of (342) is blocked by Grice’s maxim of quantity, since the speaker can express this situation more accurately by using the present-tense counterpart of (341a): Dit huis moet (wel) instortenThis house needs to collapse.

Cases like in (343), which explicitly mention the source by using a volgens-PP, are different in that they do not imply that the eventuality denoted by the lexical projection of the embedded main verb occurred before speech time n; this is clear from the fact that examples such as (343a) do not lead to a contradiction, but are perfectly acceptable. This is because the sources of the first and the second conjunct have different values: the former has Els as source and the latter has the speaker as source. This leads to the coherent interpretation that Els’ previous assessment has been proven wrong. In fact, example (343b) can receive a similar interpretation, provided that we interpret the pronoun mij as referring to the speaker-in-the-past; in (343b) the speaker claims that his earlier assessment was wrong. If we interpret the pronoun as referring to the speaker-in-the-present, the example becomes incoherent again.

In the past-tense example with the existential modal verb kunnen in (344a), the default interpretation of the source is again the speaker; as usual, this default interpretation can be overridden or made explicit by adding a volgens-PP.

Subsection I has shown that examples such as (344a) are typically used when the event denoted by the lexical projection of the embedded main verb has not yet taken place in the actual world; (344a) suggests that certain measures have prevented the eventuality from happening, that we have had a lucky escape, etc. The source of this example is the speaker, as is clear from the fact that the addition of ... maar dat was onzin, as in (345a), leads to an incoherent result. The reason for this is that the first conjunct asserts the speaker’s currently held belief that there are possible worlds, accessible from a point in time in the present-tense interval, in which the house would have collapsed (e.g. if the measures that prevented the eventuality from occurring in the speaker’s actual world had not been taken), and in the second conjunct the speaker characterizes this belief as nonsense. Of course, example (345b) does not suffer from this defect, because it is perfectly coherent to characterize as nonsense a belief held by someone else or by the speaker-in-the-past.

The discussion above might lead to the expectation that in past-tense examples with zullen, the default interpretation of the source is again the speaker, but this is not confirmed: such examples typically involve some other source, as will be clear from the fact that the examples in (346) are both fully coherent: (346a) expresses that the prediction from some source did not come true, and (346b) expresses that someone’s belief was badly motivated.

That past-tense examples with zullen have a default interpretation in which the source is not the speaker may explain why constructions with zullen are frequently found in counterfactuals such as (346a) and conditionals such as (347). We will return to such constructions in Section 1.5.4.2.

The verb zullen differs from moeten and kunnen in that the speaker is the default value of the source in the present tense but not in the past tense. This contrast in interpretation can also be brought to the fore by the contrast between (348) and (349). The fact that the speaker is the default value of the source in present-tense examples with zullen accounts for the fact that the examples (348) are usually interpreted as promises made by the speaker, who can be held responsible for the truth of the assertions.

The fact that the default value of the source in past-tense examples with zullen is someone other than the speaker explains why examples such as (349) are interpreted as promises made by the (explicit or implicit) agent of the clause, who can of course also be the speaker-in-the-past. Examples like (349) often have a counterfactual interpretation: they suggest that, to the knowledge of the speaker-in-the-present, the promise has not been fulfilled, which is also clear from the fact that they are typically followed by a conjunct connected with the adversative coordinator maarbut.