The introduction to this section above has shown that while VP adverbials modify the predicative part of the clause, clause adverbials minimally modify the propositional part of the clause. Moreover, the labels VP adverbial and clause adverbial correctly suggest that the two types of adverbials apply to different syntactic domains, which we will assume to correspond to the so-called lexical and functional domain of the clause. We will briefly introduce these notions in this subsection, and refer the reader to Chapter 9 for a more detailed discussion.
      The lexical domain of the clause consists of the main verb and its arguments and (optional) VP modifiers, which together form a proposition. In (2a), for example, the verb kopen'to buy' takes a direct object as its internal argument and is subsequently modified by the manner adverb snel'quickly', while the resulting complex predicate is finally predicated of the verb’s external argument Jan. The complex phrase thus formed expresses the proposition represented by the logical formula in (2b).

As it is not likely that the linking of semantic and syntactic structure varies arbitrarily across languages, it is often assumed that the hierarchical structure of the lexical domain is more of less invariant across languages, and that the surface differences in word order between languages are superficial phenomena due to, e.g., differences in linearization or movement. Adopting a movement approach, we may assume that the lexical domain is hierarchically structured as in (3), where NP and Clause stand for the internal theme argument of the verb: we can then account for the word order difference between VO-languages such as English and OV-languages such as Dutch by assuming that the former but not the latter has obligatory V-to-v movement; see Section 9.4, sub IC, for more detailed discussion.

The structure in (2a) can now be made more explicit as in (4): internal arguments such as the theme het boek'the book' are generated within VP, VP adverbials such as the manner adverb snel'quickly' are adjoined to VP, and external arguments such as the agent Jan are generated as the specifier of the “light” verb v. For concreteness’ sake, we have assumed that the manner adverb is adjoined to the maximal projection VP within the lexical domain; we will return to this assumption shortly.

      Clause adverbials are generated external to the lexical domain, that is, within the functional domain which contains various functional heads that add information to the proposition expressed by the lexical domain (vP). For instance, the functional head T in (5) adds the tense feature ±past and the functional head C indicates illocutionary force (declarative, interrogative, etc.), as is clear from the fact that the complementizers dat'that' and of'if/whether' introduce embedded declarative and interrogative clauses, respectively. In addition to these functional heads there may be other functional heads, indicated by X in (5), which introduce other features.

Modal adverbs and negation seem to be located at the boundary between the functional and the lexical domain. On the assumption that adverbial phrases are introduced into the structure by adjunction to the various maximal projections found in representation (5), we should conclude that they are adjoined to vP (or, alternatively, some low functional projection XP). This is illustrated in (6b), where we have assumed that the subject is moved from its vP-internal position into the regular subject position, the specifier of TP. It should be noted, however, that the adjunction analysis is not uncontroversial; Cinque (1999), for example, made a very strong case for assuming that the various subtypes of clause adverbials are generated as specifiers of designated functional heads. If we accept such an approach, the adverb waarschijnlijk would be located in the specifier position of a functional head EM expressing epistemic modality, as indicated in (6b').

Because the choice between the two analyses will not be crucial for the discussion of the Dutch data in this chapter, we refer the reader to Cinque (1999/2003), Ernst (2002), and the references cited there for extensive discussion of the pros and cons of the two approaches. We also refer the reader to Section 13.3.1 on Neg-movement, where we will show that there are strong empirical reasons for adopting Cinque’s analysis for the negative adverb niet'not' at least.

The hypothesis that clause adverbials are external while VP adverbials are internal to the lexical domain of the clause correctly predicts that the former precede the latter in the middle field of the clause; cf. Cinque (1999) and Zwart (2011: section 4.3.2). This generalization is illustrated by the two (b)-examples in (7) for the modal adverb waarschijnlijk'probably' and the manner adverb hard'loudly'.

However, the assumptions so far wrongly predict that VP adverbials precede the internal arguments of the verb. Example (8a) shows that it is possible for the direct object de handleiding'the manual' to follow the manner adverbial zorgvuldig'meticulously', but example (8b) shows that the object may also precede the adverb. In fact, example (8c) shows that the object may even precede clause adverbials such as waarschijnlijk'probably'. The examples in (8) thus show that there is no strict order between the adverbials and the arguments of the verb in Dutch, a phenomenon that has become known as scrambling. This word order variation is discussed extensively in Section 13.2, where we will argue that it results from optional leftward movement of the nominal arguments of the verb across the adverbials.

Note in passing that there are reasons for assuming that the movement which derives example (8b) is (virtually) obligatory in English, since it accounts for the fact that objects normally precede the manner adverbials in English; see Broekhuis (2008:ch.2) for detailed discussion. An alternative approach to this problem can be found in Ernst (2002:ch.4).

A useful test for recognizing VP adverbials is the paraphrase with a conjoined pronoun doet dat + adverb pronoun does that + adverb’ clause; cf. Van den Hoek (1972). This test is schematized in (9a), where the arrow should be read as “can be paraphrased as”: the first conjunct consists of the clause without the VP adverbial, which is used in the second conjunct as a modifier of the phrase doet dat, which replaces the verbal projection VP in the first conjunct. The test is applied in (9b) to example (1a).

The (a)-examples in (10) show that the test does not only work for (in)transitive, but also for unaccusative constructions. The result is sometimes less felicitous in the latter case, but in such cases it is often possible to use an en dat gebeurde + adverb paraphrase instead. This is illustrated in the (b)-examples for the time adverbial plotseling'suddenly: 'the paraphrase in (10b') contrasts sharply with the paraphrase ??De theepot is gebroken, en hij deed dat plotseling.

Unfortunately, the test cannot be applied to all clauses with a VP adverbial, often for reasons not well understood, but it generally gives reliable results for clauses with an agentive subject and a non-stative/dynamic predicate.
      Another test is based on the fact that VP adverbials restrict the denotation of the verbal predicate. As a result of this, the modified predicate will entail the bare predicate, but not vice versa. This is illustrated in (11) for the intransitive verb lachen'to laugh' and the unaccusative verb vertrekken'to leave'. For convenience, we will use the arrow → in the remainder of this chapter to indicate that the entailment is unidirectional.

That clause adverbials like modal adverbs such as waarschijnlijk or the negative adverb niet do not restrict the denotation of the verbal predicate but perform some other function is clear from the fact that they cannot be paraphrased by means of a conjoined pronoun doet dat clause, as shown in (12) for the examples in (1b); the arrow with a slash should be read here as “cannot be paraphrased as”.

The examples in (13) show furthermore that the clause with the clause adverbial does not entail the clause without it, nor vice versa.

Clause adverbials may have several functions: waarschijnlijk and niet, for instance, can be equated with the logical operators ◊ and ¬, which scope over the entire proposition, as in the predicate calculus equivalents of (1b). This is illustrated in (14), where the arrow indicates that the sentence and the logical formula express the same core meaning.

That clause adverbials are external to the lexical domain of the clause is also made clear by the clause-adverbial test in (15), which shows that clause adverbials can even be external to the entire clause.

For the cases in which the VP-adverbial and clause-adverbial tests do not provide satisfactory results, we can appeal to the generalization (7a) from Subsection II that clause adverbials precede VP adverbials in the middle field of the clause: if an adverbial precedes an independently established clause adverbial, it cannot be a VP adverbial; if an adverbial follows a VP adverbial, it cannot be a clause adverbial. For example, all adverbials that precede the modal adverb waarschijnlijk can be considered clause adverbials.
      The tests discussed above should be approached with caution, due to the fact that specific clause adverbials may sometimes be used with a more restricted scope. A well-known example is the negative adverb niet 'not,' which can be used to express sentence negation, that is, with scope over the complete proposition expressed by the lexical domain of the clause, or as constituent negation, that is, with scope over a smaller constituent within the clause; cf. Section 13.3.2, sub IC. The (a)-examples in (16) show that in the latter case, negation may occur in a conjoined pronoun doet dat -clause as a modifier of the negated constituent. Whether or not Jans advent is indeed entailed by a sentence such as Jan komt niet volgende week may be a matter of debate, but it is clear that there is a strong tendency to accept it. The main point is, however, that negation does not function as a VP adverbial in (16a) but as a modifier of the time adverbial; the paraphrase shows that the full constituent niet volgende week functions as a VP adverbial. The (b)-examples show that more or less the same observations can be made for modal adverbs such as waarschijnlijk 'probably' ; the paraphrase shows that waarschijnlijk morgen can function as a complex VP adverbial if morgen is assigned contrastive accent.

Some adverbials can be used either as a clause adverbial or as a VP adverbial, depending on their position in the middle field of the clause. We illustrate this here by means of temporal adverbials. Consider the punctual adverbial om drie uur'at 3 oʼclock' in (17a); the fact that the pronoun doet dat + adverb paraphrase in (17b) is possible and the entailment in (17c) is valid shows that we are dealing with a VP adverbial.

That we are dealing with a VP adverbial in (17a) is also consistent with the fact that it follows the modal adverb waarschijnlijk'probably'. Example (18a) shows, however, that it is not always the case that temporal adverbs must follow the clause adverb. According to the generalization in (7a) that VP adverbials cannot precede clause adverbials, the adverb morgen'tomorrow' must be a clause adverbial, which is confirmed by the fact that the scope paraphrase in (18b) is acceptable.

The hypothesis that the temporal adverbials in (17a) and (18a) perform different syntactic/semantic functions is supported by the fact illustrated in (19a) that they can co-occur in a single clause. Example (19b) shows that we find similar facts for spatial adverbials.

The discussion above shows that we should be aware that adverbials may in principle perform multiple syntactic/semantic functions in a clause, and that we should not jump to conclusions on the basis of the application of a single test.