The verb forms in these imperative examples are called finite because of their placement; while non-finite main verbs always follow verbal particles and complementives, the unacceptability of the primed examples in (78) shows that the imperative forms under discussion must precede them.

Verbs in finite imperative clauses typically occur in first position, and in this respect they differ markedly from finite verbs in declarative clauses, which are usually preceded by some constituent; cf. the contrast between the two examples in (79).

One could derive the V1-nature of finite imperatives by assuming that the clause-initial position is filled by a phonetically empty second-person subject pronoun. The assumption of such a pronoun would directly account for the fact that second-person reflexive pronouns can occur in imperative constructions; cf. Section 1.4.2, sub IIA, for discussion and further data. The assumption of a phonetically empty second-person subject can also be motivated by the fact that supplementives and floating quantifiers are possible, since these usually have a nominal associate in the clause.

At first glance, the postulation of a phonetically empty second-person pronoun in the sentence-initial position seems to contradict the placement of the subject pronouns in the more special imperative construction in (81), all of which are more or less equivalent to the more regular form Ga eens weg!Get out of the way, please!; cf. Section 1.4.2, sub IIB, for a more detailed discussion of this type. These examples show that subject pronouns must follow the verb in first position.

Barbiers (2013) noticed, however, that the overt expression of subject pronouns requires the presence of a particle like maar or eens: if we replace the particle maar in (81) by the locative proform daarthere, the examples indeed become somewhat marked, while adding the particle eens to the resulting structures makes them perfectly acceptable again; this is shown in (82). Barbiers (2013) concludes from this that the overt subject pronouns are in fact licensed by the particles and not by the verb, and therefore maintains that present-day Dutch imperatives have a phonetically empty subject pronoun in sentence-initial position, licensed by the imperative verb.

The verb-first requirement on finite imperatives does not hold for Middle Dutch, many eastern varieties of Dutch and certain varieties of modern German. Barbiers (2007) suggests that the loss of topicalization in Dutch imperatives is related to the decline of specialized imperative forms in the language (the stem weez-be as in Wees voorzichtig!Be careful! being the only surviving remnant in standard Dutch). In (83), a Middle Dutch example is given from a 14th century manuscript of De reis van Sente Brandane (The Voyage of Saint Brandaen); we refer the reader to Van der Horst (2008:543) for more diachronic data and to Barbiers (2007/2013) for similar data from modern German as well as some Dutch dialects.

Barbiers (2013) claims that the overt number marker -t on the imperative form of the verb renders the covert subject pronoun in the sentence-initial position superfluous, leaving this position available for topicalization; cf. Van Alem (2023) for an account along similar lines, which builds on the observation that all varieties of Dutch that allow topicalization in imperatives also have verbal umlauts (just as German does). For our present goal, it is sufficient to assume that in modern Dutch the sentence-initial position is filled by a phonetically empty element expressing imperative mood and/or second person; for convenience, we will refer to this element as the empty operator (abbreviated as OP in some of the examples below). This empty imperative operator is signaled by the placement of the verb in the sentence-initial position, thus indirectly signaling the imperative mood. That the phonetically empty imperative operator blocks topicalization in examples such as (79b) can be supported by the fact that resumptive pronouns in finite imperative left-dislocation constructions cannot be sentence-initial either, but must occupy a position in the middle field of the clause: the examples in (84) show that while the resumptive pronouns het/dat can be clause-initial or clause-external in declarative clauses, finite imperatives allow them only in clause-internal position. For further discussion, see Barbiers (2007/2013).

The examples in (85) show that finite imperative clauses differ markedly from declarative clauses in that direct objects are often optional and can be placed after verbal particles. The imperative constructions in the primed examples raise a possible problem for the postulation of an empty imperative operator, because some analyses proposed for them are not a priori compatible with it.

Den Dikken (1992) argues that constructions such as (85b') involve a phonetically empty operator (OP) with the function of a direct object. We will tentatively assume that it is base-generated in the canonical position of the direct object and moved into sentence-initial position, which would account for the V1-order; cf. (86a). We will further assume that the operator licenses the apparent object dat boek, which is base-generated not as an argument but as an adjunct to the right of the particle terug. This analysis has the advantage that we can accommodate the fact that the noun phrase dat boek is optional: because the phonetically empty operator OP functions as the direct object of the sentence, the selection restrictions of the verb are also satisfied in (86b) without the noun phrase.

Note that it is not a priori clear that the presence of the empty operator is crucial. Since examples such as (87a) show that topicalized pronouns can also license direct objects in the right periphery of the clause, Koopman (2007) argues that imperative clauses such as (85b') involve an elided object pronoun; we are dealing with a combination of right dislocation and topic drop. That the two operations can indeed be combined is independently supported by the fact that (in the right context) example (87b) is also perfectly acceptable.

For the sake of presentation, we will tentatively adopt the empty operator movement analysis in what follows, without the intention of immediately dismissing the topic-drop analysis. Den Dikken motivates the involvement of leftward movement in imperative examples such as (85b') by pointing out that parasitic gaps are licensed in such constructions. Consider first the declarative sentences in (88). Example (88a) shows that the adjunct clause introduced by zonderwithout must have an overtly realized object pronoun; the indices indicate that this pronoun can be interpreted as coreferential with the direct object of the main clause. The examples in (88b&c) show that the leftward movement of the object of the main clause allows the object pronoun to remain unexpressed; the interpretive gap in the adjunct clause must then be interpreted as coreferential with the object of the main clause. Since the occurrence of the interpretive gap is “parasitic” on the leftward movement of the object, such gaps are known as parasitic gaps: cf. Section 11.3.8 for a detailed discussion. Note that acceptability judgments on parasitic gaps vary from speaker to speaker: while some consider the use of a parasitic gap (pg) in (88b&c) perfectly acceptable, others prefer the use of an overt pronoun.

The crucial observation is that parasitic gaps also occur in imperative constructions such as (85b'); Den Dikken takes this as evidence for leftward operator movement: on our assumption that the empty operator is moved into the clause-initial position, example (89a) has the formal representation in (89b). Although Den Dikken’s judgment has been accepted in the literature, example (89a) may be somewhat more marked than the (b)-examples in (88). We will return to this at the end of this section.

The analysis in (89b) seems to be incompatible with Barbiers’ (2007) hypothesis if we maintain that finite verbs in main clauses can be preceded by at most one constituent, since Barbiers’ imperative operator would then block leftward movement of Den Dikken’s empty operator in examples such as Geef terug, dat boek!Give back, that book!. Accepting both proposals is only viable if we are willing to stipulate that Den Dikken’s empty operator targets a position in the middle field of the clause, which is indeed what is proposed in both Den Dikken (1992) and Barbiers (2007). The topic-drop analysis proposed in Koopman (2007), on the other hand, cannot be made compatible with Barbiers’ hypothesis, because Section 11.2.2 has shown that topic drop is only possible when the pronoun is in sentence-initial position; cf. also Van Kampen (2020a: §6). Den Dikken (1992) has already pointed out that the empty operator movement analysis of examples such as (85b') encounters a number of possible problems. First, examples such as (90) show that while it is possible to place an indirect object in the right-peripheral position of declarative clauses regardless of whether the pronoun diethat is overtly expressed, this is only possible in the imperative construction when a resumptive pronoun is present.

A second, related problem is that the indirect object is preferably omitted when the direct object is placed in the right periphery of the imperative clauses, while realization of the indirect object is perfectly acceptable in the case of right dislocation in declarative clauses. This is illustrated by the acceptability contrast between the two primed examples in (91). We used the diacritic % in example (91b') because Den Dikken claims that such examples are marked but grammatical, while we have found that at least some speakers find the realization of the indirect object quite awkward.

Den Dikken (1992) claims that the contrasts found in (90) and (91) are to be expected, by showing that similar restrictions as in finite imperative constructions are found in other constructions involving empty operator movement. Again, this line of reasoning would be impossible if we accept Koopman’s (2007) proposal that the relevant declarative and imperative constructions both involve a combination of right dislocation and topic drop.

A third problem that needs to be mentioned is that the proposed empty operator movement cannot strand prepositions: while the declarative example in (92a) is acceptable both with and without the R-pronoun daar, the R-pronoun must be overtly realized in the corresponding finite imperative construction, as is clear from the contrast between the two (b)-examples in (92).

Den Dikken provides a special account of the unacceptability of (92b') because preposition stranding is possible in the infinitival imperative in (93), for which he also proposes an empty operator movement analysis; he therefore cannot appeal to special properties of operator movement in this case.

Various solutions for the acceptability contrast between (92b') and (93) have been proposed, all of them of a highly theory-internal nature: for completeness, we will briefly summarize them. Den Dikken proposes that R-extraction must be licensed by the main verb, and that finite imperatives are special in that they cannot do this. This somewhat ad hoc proposal is not needed according to Barbiers (2007), as he claims that empty operator movement in imperatives is parasitic on case assignment: since the complement of the preposition is not assigned case by the verb, it follows that it cannot undergo empty operator movement either. Of course, this still does not explain why the infinitival imperative in (93) is acceptable. Visser (1996) suggests that this contrast between finite and infinitival imperatives follows from the fact that only the latter allow empty operator movement into the sentence-initial position: one way of formally expressing this (not proposed by Visser) is by saying that finite imperatives do not have an empty imperative operator. Note in passing, that (92b') is acceptable without the R-pronoun daar if the particle maar is added: Denk maar opnieuw over na, (over) dat probleem. Barbiers (2007/2013) suggests that this may be due to the fact that such particles create an additional position accessible for an empty operator that functions as the complement of a preposition; cf. Visser (1996) for a proposal that is similar in spirit.

A completely different approach to imperatives of the type Leg terug, dat boek!Put back, that book! in (85b') can be found in Postma & Rooryck (2007) and especially in Postma (2013). It is claimed that such imperatives are forum imperatives, which are formally characterized by the fact that they are always accompanied by a particle. The examples in (94) show that such particles are not ordinary verbal particles: while the verb gevento give cannot normally be combined with the particle op, the particle is possible in the imperative construction in (94b). Postma (2013) calls such particles forum particles.

Forum imperatives also have specific semantic properties. The participants involved in the eventuality expressed by the verb are always discourse-linked (i.e. part of the forum): (i) the implied subject refers to the addressee (as in all imperatives), (ii) an implied indirect object typically refers to the speaker, and (iii) the implied direct object typically refers to some entity in the vicinity of the addressee. Property (i) is clear from the fact that the pronoun in (95a) must be second person; any other pronoun would lead to an unacceptable result. For completeness’ sake, note that the pronoun cannot occur in the middle field of the clause, which suggests that we are not dealing with a subject pronoun but with a pronoun functioning as a vocative, which must also be nominative in Dutch.

Property (ii) is illustrated by the (a)-examples in (96): while the regular finite imperative allows a goal to be expressed overtly, this is excluded in the case of the forum imperative. Property (iii) is illustrated by the (b)-examples: the noun phrase dit boekthat book, referring to a book in the vicinity of the speaker but not of the addressee, can be used in the regular imperative, but not in the forum imperative.

Furthermore, the desired eventuality should take place in the speaker’s here-and-now, as illustrated by the fact that the forum particle cannot be used in examples such as (97b), where morgentomorrow locates the intended eventuality after speech time.

Crucial for our discussion here is that the implied direct object cannot be overtly expressed in the middle field of the clause, but must appear after the forum particle.

A quick look at the examples in (94) to (98) shows that the middle field of forum imperatives is often empty: the arguments of the verb are typically left implicit, and spatiotemporal adverbial phrases are impossible due to the fact that the desired eventuality should be located in the here-and-now of the speaker. However, the examples in (99) show that it is not the case that the middle field must be radically empty: certain adverbs and particles are possible.

Now consider the possibility that particle verbs such as teruggevento give back can also be used in forum imperatives, but that the verbal particle prevents the realization of an additional forum particle such as op in Geef op!Give it to me!. This would imply that particle verbs can be used in both regular and forum imperative constructions: in the former case the nominal objects must be expressed overtly in the middle field of the clause, whereas in the latter case the nominal objects are omitted or expressed in the right periphery of the clause, after the particle.

This would also explain why there is no problem in expressing the indirect object overtly in examples such as (100a), but that this leads to an infelicitous result in the (b)-examples. This has already been illustrated by the (b)-examples in (91), which are repeated in a slightly different form as (101); example (101b) is marked with a number sign to indicate that it would be perfectly acceptable as a regular imperative if the noun phrase were interpreted as a theme (direct object); the percentage sign in example (101b') is used for the same reason as in (91b').

The fact, illustrated earlier in (90), that the indirect object can only be placed after the particle if a resumptive pronoun is present, also follows from this proposal: example (102a) can be analyzed as a regular imperative with right dislocation of the indirect object; example (102b) should be analyzed as a forum imperative, but is unacceptable because the middle field of the construction is not empty and the indirect object refers to someone other than the speaker.

The contrast with respect to preposition stranding in (103) again follows from the fact that arguments of the verb cannot be expressed in the middle field of a forum imperative. Note in passing that on the assumption that (103b) is a forum imperative, there is reason to assume that the pronominal part of the preposition over is missing, which would of course be another source of ungrammaticality.

This section discussed Barbiers’ (2007) hypothesis that the V1-order of Dutch imperatives is due to the fact that the sentence-initial position is occupied by an empty imperative operator; this operator blocks the topicalization of other constituents. At first glance, his hypothesis seems incompatible with the hypothesis that examples such as Geef terug, dat boek!Give back, that book! involve empty operator movement into the sentence-initial position, because this position cannot accommodate two empty operators at the same time. Barbiers solved this problem by assuming (in accordance with Den Dikken’s, 1992, original proposal) that the empty operator targets a position other than the specifier of CP. Whether taking this step is really necessary is not entirely clear, given that Geef terug, dat boek! may be an instantiation of the so-called forum imperative, which is characterized by the fact that the participants in the eventuality expressed by the verb are discourse-linked (and typically left unexpressed). However, this would leave us with the problem that parasitic gaps are claimed to be possible in examples such as (89): Berg [zonder pgi te lezen] op, dat boeki! Put that book away without reading it. In our discussion of this, we have already indicated that we are not entirely sure whether parasitic gaps in such examples are as acceptable as parasitic gaps licensed by an overtly realized noun phrase, although they are certainly better than expected if we analyze such cases as forum imperatives. Since we have nothing illuminating to say about the precise status of such examples at the moment, we will not digress on this issue here; future research will have to show which of the two competing analyses of Geef terug, dat boek!’ is more promising. Whatever the outcome of this research, we can conclude for the moment that the V1-nature of finite imperative constructions can be explained by adopting Barbiers’ (2007) proposal that the clause-initial position of such constructions is filled by an empty imperative operator. This is consistent with the hypothesis that V1-order is usually a superficial phonetic phenomenon.