Since the seminal work in Green (1974) and Oehrle (1976), it has usually been assumed that double object constructions and their periphrastic counterparts are semantically similar but not semantically equivalent; cf. also Balk-Smit Duyzentkunst (1968) for a similar claim based on Dutch. Although it has proved no easy task to provide a fully adequate description of the difference in meaning between the two constructions, it is generally described in terms of change of possession and location. The examples in (333) all express that the theme is relocated, but the double object construction additionally expresses that the referent of the indirect object becomes (or is expected/intended to become) the new possessor of the theme. The meanings ascribed to the double object and periphrastic indirect-object constructions can be schematized as in (334), which is a somewhat adapted version of the semantic representations proposed in Gropen et al. (1989:241); we will return in Subsection IV to proposals that provide syntactic structures that can be paired with the semantic representations proposed in (334).

The dual meaning approach has recently been censured in Rappaport Hovav & Levin (2008), which points out that periphrastic indirect objects of certain verbs can in fact also express that the referent of the indirect object becomes the new possessor of the theme. This can be illustrated by the examples in (335): the fact that (335a) is odd and (335b) perfectly acceptable regardless of the form of the indirect object shows that the difference in meaning between the two alternants cannot be expressed by the notion of possession.

Rappaport Hovav & Levin (2008) concludes from this that the choice between the double object and the periphrastic indirect-object construction does not completely determine the interpretation, but that the interpretation is also sensitive to the type of verb that enters the construction, in the way indicated in Table (336).

	dative object	periphrastic object
give-type verb	caused possession	caused possession
throw-type verb	caused motion or caused possession	caused motion or caused possession

Rappaport Hovav & Levin (2008) refers to this proposal as the verb-sensitive approach in order to express that the interpretation of the double object and periphrastic indirect-object construction is largely determined by the verb’s meaning. That the meaning of the verb is relevant is also clear from the fact that only a subset of the verbs exhibiting the dative alternation with aan-PPs inherently express caused possession. Verbs of transfer like gevento give, lenento lend, overhandigento hand, verhurento rent out do have this property, while verbs indicating future possession like belovento promise, nalatento bequeath, aanbiedento offer and toewijzento assign and verbs of communication like vertellento tell, lerento teach and schrijvento write do not; this is illustrated by the contrast between example (335a) and the examples in (337).

Now, let us return to the two verb types in Table (336). From a Dutch point of view, it seems uncontroversial to distinguish between these two types; the examples in (338) show that periphrastic indirect objects take the preposition aanto with give-type verbs, but the preposition naarto with throw-type verbs. This observation supports another claim made in Rappaport Hovav & Levin, namely that constructions with give- and throw-type verbs differ in that the latter, but not the former, involve the notion of path; aan-PPs belong to the set of adpositional phrases that merely indicate a change of location, whereas naar-PPs are always directional; cf. also Schermer-Vermeer (2001:29) and the references cited there. For a more detailed discussion of the distinction between these terms, see Section P32.3.1.1.

Now consider the examples in (339), which are constructions with a spatial complementive. These examples stand in a similar opposition as the examples in (338); the change-of-location construction in (339a) strongly suggests that the referent of the direct object has obtained a new location (i.e. on the wall), whereas this implication clearly does not hold for the directional construction in (339b): Els need not have reached Groningen, as follows from the acceptability of the adverbial phrase in brackets.

This difference between the two examples in (339) is even more conspicuous in their non-causative counterparts in (340): the locational construction implies that the painting was located on the wall during the entire time interval of the event, whereas the directional construction implies that Jan was not in Groningen during that time interval; cf. also P34.1.4, sub II, for relevant discussion.

The discussion of the similarities between the examples in (338) on the one hand and the examples in (339) and (340) on the other hand shows that Table (336) should be adapted as in (341). This strongly suggests that the implications concerning possession are no more than pragmatic inferences based on the aspectual properties of the constructions in question: constructions expressing a change of location imply that the located object occupies the position indicated by the locational PP or the dative noun phrase at the end of the event time interval, whereas directional constructions leave this open.

	dative indirect object	periphrastic object
give-type verb	change of location	change of location
throw-type verb	directional	directional

Rappaport Hovav & Levin’s (2008) critique of the dual meaning approach seems devastating for any attempt to describe the difference between double object and periphrastic indirect-object constructions in terms of possession. However, this does not necessarily mean that the dual meaning approach must be abandoned altogether. Consider again the semantic representations in (334), repeated here as (342).

Leaving aside the precise substantive content of the predicates have and be at, the core intuition is that the double object and periphrastic indirect-object constructions differ with respect to whether the activity performed by the causer affects the referent of the indirect object or the direct object of the construction. The periphrastic construction has an affected direct object in that its referent undergoes a change of location; the double object construction has an affected indirect object in that its referent undergoes a change of state, e.g. by becoming the possessor of the direct object. The remainder of this subsection will show that this intuition may still be close to the mark.

Consider the examples in (343), in which the direct object does not refer to a tangible entity that can be physically located in space. However, since it can still be said that the referents of the indirect objects are affected by the action performed by the subjects of the clauses, we correctly predict that these double object constructions are possible.

The corresponding periphrastic indirect-object constructions in (344), on the other hand, are usually regarded as marked, since they lead to the anomalous interpretation that the referents of the direct objects are located in space (and thus tangible) and actually undergo a change of location by being transferred to the referent of the indirect object; this was the core intuition expressed in Balk-Smit Duyzentkunst (1968), Green (1974), and Oehrle (1976) that underlies the dual meaning approach.

A similar but somewhat sharper contrast is found in the (a)-examples in (345), where it is not the doctor but his treatment that causes Marie to acquire a soft skin; the primed example is unacceptable because the met-PP expressing the cause strongly disfavors the change-of-location reading associated with the periphrastic indirect-object construction. The same holds for the (b)-examples in which the cause is expressed as the subject of the clause.

In the examples in (346), on the other hand, the subject’s action does not directly affect the referent of the indirect object, but the referent of the direct object, by locating the responsibility for the failure of the plan at the referent of the indirect object; cf. Oehrle (1976). The semantic representations proposed in (342) therefore correctly predict that example (346b) is degraded.

The discussion in this subsection suggests that the dual meaning approach was probably wrong by claiming that the difference between the double object construction and the periphrastic indirect-object construction must be expressed by the abstract predicate have (possession). However, this approach may have been correct in assuming that the two constructions differ in terms of which object is (primarily) affected by the subject’s action. This leads to the semantic representations in Table (347), which of course abstract from the semantic contributions of the verbs that enter the constructions. In the periphrastic indirect-object construction it is the direct object that undergoes a change of state: with give-type verbs it is subject to a change of location (here expressed by cause to be at) and with throw-type verbs it traverses a certain path (here expressed by cause to go to). In the double object construction, on the other hand, it is the indirect object that is affected by the action of the subject, which we have indicated by the abstract predicate be affected by, because we have seen that the abstract predicate have may not be the most appropriate choice for expressing the resultant state of the referent of the indirect object (although it may still be one of the options, depending on the semantic contribution of the actual verb used).

	dative indirect object	periphrastic indirect object
give-type verb	[S cause [IO to be affected by DO]]	[S cause [DO to be at IO]]
throw-type verb	[S cause [IO to be affected by DO]]	[S cause [DO to go to IO]]

When we consider the literature on Dutch since Balk-Smit Duyzentkunst (1968), which focuses mainly on give-type verbs, it seems hardly controversial to assume that the dative and periphrastic constructions differ semantically in the way indicated in Table (347). We illustrate this here with a small number of examples, ignoring the details of the individual proposals; cf. Schermer-Vermeer (1991: §9) for a more extensive review. Balk-Smit Duyzentkunst already noted that the periphrastic construction expresses the transfer of the referent of the direct object. Kooij (1975) explains the difference in meaning in terms of “affectedness of IO” and “transfer of DO”. Jansen (1976: §3.3) seems to have a similar contrast in mind: whereas the periphrastic construction is said to focus on the activity/process itself and the periphrastic aan-PP is taken to be a “local destination”, the dative construction expresses that the activity/process has a specific function for the referent of the indirect object. Finally, the abstract predicate be at is clearly related to Schermer-Vermeer’s (1991: §7) term contact, which she rightly claims to be the core meaning of the preposition aan (cf. Section P32.3.1.2.3), and the abstract predicate be affected by seems to come close to what she calls reactiefresponsive.

The interpretations ascribed to the double object and periphrastic indirect-object constructions in Table (347) seem real, but should probably be regarded as tendencies rather than absolute rules, since there are various other factors that may affect the acceptability of the two alternating syntactic structures; cf. Den Hertog (1973a:62) for some early remarks in this regard. That this is the case is immediately clear from the fact, discussed in Section N14.3.1.2, sub IIIC, that det-inf nominalizations of double object constructions in which the theme argument is realized as a postnominal van-PP require the use of the periphrastic indirect object; cf. Van den Toorn (1971b). The examples in (348) show that this is also true for nominalizations of een trap gevento give a kick and een poetsbeurt gevento give a shine in (343) and (344), which normally involve a dative object.

The relative length of the objects can also affect the acceptability of the two constructions. For instance, the degraded primed examples in (344) improve considerably when the nominal part of the aan-PP is a larger noun phrase.

Stowell (1983:333) further pointed out for English that Latinate verbs do not easily enter the double object construction; the contrast between the semantically nearly equivalent (a) and (b)-examples in (350) suggests that the same is true for Dutch.

Other factors that may affect the actual choice between the double object and the periphrastic indirect-object construction are related to the information packaging of the clause; cf. Huddleston & Pullum (2002) for similar claims for English. Although speakers will normally prefer the double object construction in (351a) to the periphrastic indirect-object construction in (351b), the latter is perfectly acceptable if the referent of the direct object is part of the presupposition of the clause, e.g. if (351b) is used as a response to the question Wat deed Jan met het water uit de regenton?What did Jan do with the water from the rain barrel?. Example (351b) is also acceptable if the direct object het water uit de regenton is contrastively focused.

Furthermore, a dative object often seems to be preferred when the direct object is a clause, which is especially common in the case of verbs of communication. A Google search (April 18, 2024) showed that the periphrastic indirect object examples in the primed examples in (352) are much less common than the primeless double object constructions. The numbers in square brackets refer to the number of hits for the search strings [V hem dat ik] and [V aan hem dat ik]; the pronoun ikI was added to keep the count manageable. The hit for [beloofde aan hem dat ik] should be dismissed as irrelevant, as it is from an older source (1682); the same search string without the pronoun, i.e. [V aan hem dat], produced only three contemporary results (mainly from religious contexts).

The contrast between the primeless and primed examples in (352) may again be due to information packaging, since the embedded clause usually contains the relevant new information that is expressed. This seems to be supported by the fact that the use of periphrastic indirect-object constructions is very natural in questions such as (353), in which the indirect object is questioned and thus part of the “new” information of the clause. We have the impression that the periphrastic indirect object is even preferred to the dative one, but unfortunately the results of our Google search on the string [(aan) wie heb je V dat] are insufficient to substantiate this claim here.

Finally, the examples in (354) illustrate the fact that double object constructions with inanimate indirect objects are often less felicitous. This may also depend on the verb: example (354a) improves considerably if we replace geven by the more formal verb schenken ‘to donate’.

Bresnan et al. (2007) investigated the interfering factors in more detail for English and found that inanimate, non-pronominal, indefinite and informational structurally “new” indirect objects are much more likely to appear as PPs than their animate, pronominal, definite or presuppositional counterparts. The nature of the direct object also seems to affect the choice between the two alternants: pronominal, definite and presuppositional direct objects favor the periphrastic indirect-object construction more than their non-pronominal, indefinite or non-presuppositional counterparts. The results are summarized in (355), in which the “>” sign should be interpreted as “is more likely to appear in a double object than in a periphrastic indirect-object construction”.

We have the impression that the same tendencies apply to Dutch, but to our knowledge this has not yet been tested and we therefore leave it to future research.

The discussion in the previous subsections has shown that it is difficult to provide lists of verbs that allow or disallow the dative noun phrase to alternate with a periphrastic aan-PP. Nevertheless, example (356) shows this for a small sample of double object verbs: in order to eliminate as much as possible the influence of the information structure of the clause, we have checked all cases with interrogative clauses of the type in (353), in which the indirect object is questioned and thus part of the “new” information of the clause: Aan wie/Wie heeft hij dat Vparticiple? To whom/Whom has he V that?. The number sign # indicates that the periphrastic form is not possible in such questions, and a percentage mark indicates that we are not sure of our judgments and that speaker variation is to be expected.

Since this will become relevant in Subsection IV, note that most of the verbs in (356) are particle verbs or verbs prefixed by be- or ver-, although in the prefixation cases the verb is often the result of a historical process, as can be seen by the fact that the original input verb is often no longer used: bevelento order - *velen; vertellento tell - #tellen. Many of the remaining simple verbs in (356) also occur as double object verbs with a verbal particle. The examples in (357) show that the use of the particle terugback is particularly productive in this respect; although not all the cases in (357) are listed in Dutch dictionaries, they are all abundantly used on the internet. The percentage marks in (357) again indicate that we are not sure of our judgments and that we expect speakers to vary on the question as to whether or not they allow periphrastic forms.

This subsection discusses the syntactic status of the aan-PP in the periphrastic indirect-object construction. Early generative grammar followed traditional grammar in assuming that this PP is an alternative realization of the dative object and thus a PP-complement of the verb. However, Den Dikken (1995a) has shown that there are reasons to assume that this is not correct and that the aan-PP behaves more like a complementive, i.e. a predicative locational PP. An important argument for this claim is based on the interpretation of the periphrastic indirect-object construction; the semantic representations given in Table (347) of give-type verbs, repeated here as (358), suggest that the aan-PP has a function similar to that of a locational PP in a copular construction. Since Section 2.2.1, sub IV, has shown that locational PPs function as complementives in constructions like Jan is op schoolJan is at school, it seems natural to assume the same for aan-PPs in periphrastic indirect-object constructions.

The hypothesis that periphrastic aan-PPs function as complementives is perhaps not the most obvious one to formulate, since it predicts that aan-PPs exhibit syntactic behavior similar to that of complementives. This is clearly wrong: unlike the prepositional complementives in (359a&b), the aan-PP in (359c) can easily be in extraposed position.

However, the problem seems to be less serious than it appears at first glance. First, consider the examples in (360), which show that the ban on extraposition of prepositional complementives is lifted if the clause is headed by a particle verb or a verb prefixed with be- or ver-.

Since Subsection III has shown that many, if not most, periphrastic indirect-object constructions are headed by particle verbs or verbs prefixed with be- or ont-, the fact that these constructions allow extraposition of the aan-PP simply fits into a more general pattern; cf. Section 2.2.1, sub IV. This means that we are left with a small subset of verbs without a particle or prefix, and in this context the observation in (357) that most of these verbs can also be used with the particle terugback may become relevant. We can explain problematic examples such as (359c) by assuming that such examples contain a phonetically empty verbal particle, which perhaps functions as a counterpart of the particle terugback; cf. Den Dikken (1995a: §3) for a detailed motivation of this assumption. If so, the examples in (361) are structurally parallel to those in (360), and we thus correctly predict that extraposition is possible in both cases.

The discussion above has made it clear that the hypothesis that periphrastic indirect objects function syntactically as complementives cannot be rejected on a priori grounds, but it has not shown that this hypothesis is in fact the correct one. It is not an easy task to provide theory-independent evidence for this hypothesis on the basis of the alternation between recipient objects and aan-PPs, but Section 3.3.1.2 on the shift between indirect object goals and naar-PPs will show that there is much to recommend this hypothesis.