Verbs that allow the dative alternation with aan-PPs differ semantically from verbs that allow the dative alternation with naar-PPs: the former denote an actual, intended, or future change of location of the referent of the theme argument, whereas the latter are merely directional. The difference can be made explicit by considering the implication relations. The change-of-location construction in the first conjunct of (363a) refers to the act of the actual transfer of the referent of the direct object to the referent of the indirect object, and thus contradicts the second conjunct which expresses that the transfer has not taken place. The directional construction in the first conjunct of (363b), on the other hand, expresses that the referent of the direct object traverses a certain path, but does not imply that it actually reaches the intended goal, as is clear from the fact that (363b) is perfectly coherent; see also Schermer-Vermeer (2001:29), where it is claimed that the notion of contact, which constitutes the core meaning of the preposition aan, is missing in naar. In the following, we will use the term recipient to refer to the indirect object in the change-of-location construction and the term goal to refer to the indirect object in the directional construction.

The double object and the periphrastic indirect-object constructions in (362) seem to differ in a way similar to those discussed in Section 3.3.1.1: whereas the periphrastic construction in (362b) seems to be particularly concerned with the way the action of the subject affects the referent of the direct object, the double object construction in (362a) seems to be more concerned with the way it affects the referent of the indirect object. However, Section 3.3.1.1 has already shown that this difference cannot be adequately expressed in terms of possession: neither the periphrastic nor the double object construction in (362) necessarily implies that Jan will come into possession of the ball. Nevertheless, it seems plausible that some notion of affectedness is relevant, as implied by the semantic interpretations proposed in Table (347) for throw-type verbs, repeated here as (364).

The semantic representation in (364a) expresses that the referent of the indirect object in the double object construction is somehow (potentially) affected by the action of the subject. Since this may be true for the referent of the animate indirect object Jan, but clearly not for the inanimate indirect objects Amsterdam/de korf in the primeless examples in (365), the contrasts indicated there provide additional support for the semantic representations in (364).

Directional PPs are invariably used as complementives; cf. Section P32.1.2.2, sub III. This means that examples such as (362b) provide strong evidence for the hypothesis discussed in Section 3.3.1.1, sub IV, that periphrastic indirect objects function syntactically as complementives. This hypothesis is also supported by the examples in (366), which show that the naar-PP in (362b) behaves like other PP-complementives in that it can only be in extraposed position when a verbal particle such as over is present.

The examples in (367) further show that the use of particles such as over blocks the dative alternation; we will return to this, but first we need to discuss the function of the element toe.

This subsection discusses the element toe found in the examples in (362), repeated here as (368). The starting point of our discussion will be the observation that this element is optional in the periphrastic indirect-object construction, but obligatory in the double object construction.

There are seeming counterexamples to the claim that the element toe must be realized in the double object construction with a goal argument, but it seems that these can usually be traced back to the fact that the indirect objects of the constructions in question can alternate with either an aan- or a naar-PP, i.e. are ambiguous between a recipient or a goal. A clear example is the verb sturento send in (369); example (369a) without toe corresponds to (369b), while example (369a) with toe corresponds to (369b').

The contrast in (368) regarding the distribution of toe is surprising and therefore in need of an explanation. The proposal below supports the transformational approach to the dative/PP alternation by implying that the double object construction is derived from a structure more or less identical to that assigned to the periphrastic indirect-object construction; the reader is referred to Janssen (1976:12) for an early proposal of this kind, to Den Dikken (1995a) for a detailed analysis fully compatible with our findings here, and to Schermer-Vermeer (2001) for an alternative lexico-grammatical approach. The first step in our argument is to establish that the element toe is not always optional in the periphrastic construction, as is borne out by the examples in (370); they show that the element toe must be realized when the nominal complement of the naar-PP is moved into the clause-initial position; cf. Section P36.2 for a more detailed discussion.

Now suppose that the double object construction is derived from a structure similar to that of the periphrastic indirect-object construction by eliminating the preposition naar: [PP naar Jan (toe)]. Den Dikken (1995a) claims that this is the result of so-called incorporation of the preposition into the verb, but the exact technical means are not important here; all that matters is that as a result of the elimination of the preposition, the noun phrase Jan can no longer be assigned case within the PP and must therefore be promoted to indirect object (in the same way that the direct object of a verb must be promoted to subject in the passive construction; cf. Section 3.2.1). To make this possible, the noun phrase must be moved out of the PP and into the canonical position of the indirect object preceding the direct object: IOi DO [PP naar ti (toe)]. If so, we can account for the obligatory presence of toe in the double object construction by appealing to the fact that the extraction of the proform waar in (370b) also triggers the obligatory presence of toe.

This hypothesis might also be interesting in the light of the problem noted in Subsection II, i.e. that the double object construction is excluded when the verb is preceded by a verbal particle. The relevant example is repeated as (371a). If the hypothesis proposed here is on the right track, we expect that periphrastic indirect objects of particle verbs are also impossible when toe is present; example (371b) shows that such cases are indeed degraded.

The unacceptability of the verbal particle over can be explained by assuming that toe also functions as a verbal particle; verbs never combine with two particles at the same time, which in turn follows from the more general restriction that clauses can contain at most one complementive; cf. Section 2.2.1, sub IV, for discussion.

Since we have seen that the periphrastic naar-PP syntactically functions as a directional complementive, it is not surprising that the set of double object verbs in which the indirect object functions as a goal is a subset of the verbs that can take a directional PP:

There are also a number of verbs that allow the double object but not the periphrastic indirect-object construction. As with verbs that take a recipient, this is especially true for verbs that express the transfer of propositional content like toebijten/toeblaffento snarl at, toefluisterento whisper to, toejuichento cheer at. When the particle toe is not present, these verbs do sometimes allow a PP-complement headed by naar, as shown by the following counterpart to example (373b); Ze fluisterde naar hem [dat ze geslaagd was] (373b). An example involving the transfer of a physical entity is toestoppento slip in (373c).

For completeness’ sake, note that there are also double object constructions with toe that do not allow the periphrastic indirect object with naar, but instead take periphrastic indirect objects with aan. This simply shows that a large number of (non-directional) particle verbs with the verbal particle toe take a recipient. Some examples are: iemand iets toestaanto grant someone something, iemand iets toevertrouwento entrust/confide something to someone, iets toewijzen aan iemandto assign something to someone, iemand iets toezeggento promise something to someone, etc. As expected, such double object constructions do alternate with periphrastic indirect-object constructions with aan.

Since the dative alternation with naar-PPs has hardly been studied so far, future research will have to clarify which double object constructions with toe do or do not belong to the class of constructions discussed in this section.

The previous subsections have discussed a second type of dative/PP alternation, in which the periphrastic indirect object appears as a naar-PP, and which seems to have gone largely unnoticed in the otherwise vast literature on dative shift. We have seen that this alternation can shed new light on the analysis of the dative/PP alternation in the sense that it supports the following two hypotheses put forward in Den Dikken (1995a): (i) the periphrastic indirect object functions syntactically as a complementive, and (ii) the double object and the periphrastic indirect-object construction are derived from similar underlying structures. It therefore seems worth studying this alternation more closely in the future.