The distribution of the standard Dutch possessive dative construction is fairly restricted and usually requires the possessum to be embedded in a complementive locational PP, as in (379); if the locational PP has an adverbial function, as in (382), the possessive dative is excluded. This does not hold for the corresponding possessive bij-PP or the NP-internal possessor, which are perfectly acceptable in such cases.

Double object constructions such as (383a), in which the indirect object functions as the possessor of a direct object, are also normally excluded in standard Dutch; since the same holds for the possessive bij-PP in (383b), the usual way of expressing the intended meaning is by using an NP-internal possessor, as in (383c). The percentage sign in (383a) is used to indicate that such possessive dative constructions are common in various southern and eastern dialects of Dutch; see Van Bree (1981), Cornips (1994a), and Scholten (2018) for a description of the dialect data, and also Barbiers et al. (2005:78), which describes the distribution of this possessive construction with a reflexive indirect object. The number sign in example (383b) indicates that it is marginally acceptable when the bij-PP functions as an adverbial locational phrase (in which case the example becomes perfectly acceptable when the direct object has an internal possessor such as zijn handenhis hands); cf. Subsection II for this adverbial use of the bij-PP.

There are exceptions to the general rule that an indirect object cannot act as an inalienable possessor of a direct object; for instance, the examples in (384) show that possessive constructions of the type in (383a) are possible in certain idiomatic expressions. Possessive datives in examples of this type do not usually alternate with possessive bij-PPs.

In other cases, the possessive relation between the indirect and the direct object may be triggered by our knowledge of the world. In (385a), the dative phrase functions as the syntactically encoded possessor of the nominal part of the predicative locational PP op de rug, but the fact that the dative phrase is also construed as the possessor of the direct object de handen is related to our knowledge of the world; cf. Schermer-Vermeer (1991:205ff) for a more general discussion. Knowledge of the world may also be relevant for (385b) with an optional adverbial PP; this example is given as a case of (inalienable) possession in Janssen (1976:43), but we think that the hotel context evoked by the noun piccolobellhop simply favors the interpretation that the room in question is the room rented by Marie.

Standard Dutch possessive datives are associated with entities that are inalienably possessed, like body parts or certain items of clothing (provided that they are actually worn at the time of the event); the primeless examples in (386) illustrate that the use of possessive datives leads to degraded sentences when the possessum is not inalienably possessed. The singly-primed and doubly-primed examples show that possessive datives differ crucially in this respect from periphrastic bij-PPs and NP-internal possessors. The percentage signs in the primeless examples again indicate that these examples are perfectly acceptable in some southern and eastern varieties of Dutch; cf. Cornips (1994a:153).

Some standard Dutch examples that may be on the borderline between alienable and inalienable possession are given in (387), where the possessed entity is a location that is in some sense inherently associated with the possessor.

Note in passing that it has been claimed that dative objects cannot be interpreted as inalienable possessors if the possessed noun phrase is modified by a non-restrictive modifier; cf. Vergnaud & Zubizarreta (1992:603) and the references cited there{Vergnaud, 1992 #1036|{Vergnaud, 1992 #1129|{Vergnaud, 1992 #1129|{Vergnaud, 1992 #1129|. However, the examples in (388) show that this is not true for Dutch: the dative phrase can be interpreted as the possessor, regardless of whether the modifier of the possessum is restrictive or non-restrictive.

The examples in (389) show that dative possessors differ from their corresponding possessive bij-PPs and NP-internal possessors in that they must be animate.

The set of examples in (390) simply illustrates the same point.

Corver (1990/1992) has argued on the basis of a large number of tests that the possessive bij-PP and the locational PP containing the possessum form a constituent. The first argument is based on the standard constituency test, according to which the sentence-initial position preceding the finite verb in main clauses can be occupied by at most a single constituent. Consider the examples in (400), in which the possessive bij-PPs are construed with the nominal parts of the adverbial phrases in de tuin and op de schouder. The fact illustrated in the singly-primed examples that these bij-PPs can be pied-piped by topicalization of the locational PPs immediately establishes that the bij-PPs can be part of the adverbial phrases. The fact illustrated in the doubly-primed examples that the pied piping of the bij-PPs is obligatory shows that we can even conclude that the possessive bij-PPs must be part of the adverbial phrases; note that some speakers report that they marginally accept the doubly-primed examples with a contrastive (adverbial) reading of the bij-PPs.

Another constituency test showing that the bij-PPs can be part of the adverbial phrases is pronominalization: example (392) in Subsection II has already shown that, while adverbial phrases and complementive bij-PPs can be pronominalized by an adverbial proform, possessive bij-PPs cannot. However, it is possible to pronominalize the string consisting of both the locational PP and the possessive bij-PP. We illustrate this with the question-answer pairs in (401); the complex phrase [bij-PP loc-PP] is given as an answer, and therefore clearly has the same syntactic function as the interrogative pronoun waarwhere. Other tests that lead to a similar result concern clefting and pseudo-clefting, but are not illustrated here; cf. Corver (1990/1992) for examples.

The examples in (402) also support the claim that the string [bij-PP loc-PP] is a constituent; conjuncts of coordinate structures are always phrases.

The examples in (403) provide two more cases in which the string [bij-PP loc-PP] is found in a position where we usually find a single constituent. In (403a) the string functions as a postnominal modifier and in (403b) as a complement to the preposition totuntil.

The final and perhaps most conclusive evidence for the claim that the string [bij-PP loc-PP] forms a constituent is that the bij-PP can intervene between the locational PP and its modifiers. This is illustrated in the examples in (404), in which the modifiers of the locational PPs are italicized; cf. Chapter P34 for a detailed discussion of this kind of modification.

Since the previous subsection has established that the string [bij-PP loc-PP] is a constituent, we have to consider the question as to what constitutes the internal structure of this constituent. In principle, we can assume at least the four structures presented in (405), in which the prepositional head of the construction is indicated by italics and the functions of the substrings are indicated by subscripts in small caps; cf. Corver (1990/1992) and the references cited there. Note that there is yet another, somewhat more complex analysis, the discussion of which will be postponed to Subsection C.

The first three structures are all characterized by the fact that the preposition bij is the head of the complete string, We have already seen in Subsection II that such structures are less plausible, since there are cases in which the verb selects the preposition of the locational PP; this is clear from the fact that, while the possessive bij-PP is optional in examples such as (406a), the locational PP cannot be omitted.

The structure in (405a) can also be rejected on semantic grounds; since the locational PP modifies the noun meisje, we wrongly expect the interpretation “near the girl who is in the garden” instead of “in the girl’s garden”.

Structures such as (405b) are typically found in absolute met-constructions such as (407). Again, such an analysis leads to the wrong interpretation. Since the locational PP is predicated of the noun phrase, the absolute met-construction in (407) expresses that the referent of the noun phrase Peter is in a specific place (i.e. in the goal). This interpretation is not found in the possessive construction, which is especially clear in examples such as (406), in which the interpretation that Marie is around the neck would of course be incoherent.

The structure in (405c) leads to a kind of appositional interpretation, in which the locational PP further specifies the bij-PP; this again conflicts with the fact that in examples such as (406) the presumed appositive phrase, the locational PP, cannot be omitted.

This leaves us with the option (405d), in which the bij-PP functions as a modifier of the locational PP; evidence in favor of this analysis is that the possessive bij-PP can easily be omitted (with the concomitant loss of the possessive reading). Another virtue of analyzing the bij-PP as a modifier of the locational PP is that it accounts for the extraction facts in (408), which show that adjectival measure phrases such as diepdeep and possessive bij-PPs are similar in that they can both be extracted from the locational PP by wh-movement. This similarity in behavior follows immediately when both are analyzed as modifiers of the locational PP.

Consider again the analysis in (405d), taken from Corver (1990/1992), according to which the bij-PP functions as a modifier of the locational PP: [PP [PP bij DP]MOD [P DP]]. This structure makes a number of predictions about R-extraction. Consider the examples in (409), which show that modifiers such as vlakjust and directdirectly do not prevent R-extraction from the locational phrase.

If possessive bij-PPs are also modifiers of the locational phrase, we would expect to see the same in examples such as (410). However, the status of (410b) is somewhat unclear: such examples are given as grammatical in Corver (1990/1992), but rejected in Broekhuis & Cornips (1997). Note that it is crucial that the bij-PP follows the modifier diep; if it precedes it, the result is perfectly acceptable, but then the bij-PP probably functions as an adverbial phrase modifying the whole clause.

The unacceptability of example (411b), on the other hand, is crystal clear; the possessive bij-PP blocks R-pronominalization of the locational PP. Of course, it is crucial to note that (411b) would be perfectly acceptable if the bij-PP were omitted.

If we let the clear case in (411b) decide, we can conclude that the possessive bij-PP blocks R-pronominalization and thus R-extraction from the locational phrase. This poses a problem for the hypothesis that the bij-PP acts as a modifier of the locational PP. Another problem is that R-extraction is easily possible from the bij-PP, as shown in (412), which would be unexpected if the bij-PP were an adverbial modifier of the locational PP. Corver answers this objection by pointing out that R-extraction is possible from various adverbial phrases, but such phrases are always modifiers of the verbal projection, and it remains to be seen whether modifiers of other phrases also allow R-extraction.

Broekhuis & Cornips (1997) tried to explain why possessive bij-PPs block R-extraction from locational PPs by assuming that the former are not really base-generated as modifiers of the latter; possessive bij-PPs are supposed to originate within the locational PPs. Following up on a proposal in Categories and Arguments, posthumously published in Hoekstra’s (2004), they assume that the possessive meaning is syntactically encoded by placing the possessor and possessum in a local structural relation; more specifically, they propose that the preposition bij is a two-place predicate expressing possession, which forms the head of a PredP, as shown in (413a). The structure proposed by Corver is then derived by extracting the predicative bij-PP to a position within some higher functional position associated with the PP, which we have called FP for convenience, as in (413b).

This derivation makes it possible to explain why R-extraction of the possessum, as in the (b)-examples in (410) and (411), is excluded by appealing to the more general fact, i.e. that it is normally not possible to extract more than one constituent from a single phrase (here: the PP headed by in); cf. Broekhuis & Cornips (1997) for further details.

A first group of verbs that take a locational PP-complementive are transitive verbs denoting a change of location. The primeless and singly-primed examples in (414) show for the verbs zettento put and trekkento pull that such verbs do indeed allow the possessive dative/bij-PP alternation. The doubly-primed examples are added to show that the possessive dative/bij-PP is optional and can be replaced by an NP-internal possessor.

Although verbs like zetten and trekken are normally used as monotransitive verbs, as in the doubly-primed examples in (414), the primeless examples behave in all respects like ditransitive verbs. For instance, the (a)-examples in (415) show that the direct object is promoted to subject in the regular passive, while the dative possessor is promoted to subject in the krijgen-passive. The (b)-examples are less suited to illustrate this, since the dative possessor also functions as a source, and Section 3.2.1.4 has shown that this blocks krijgen-passivization of ditransitive constructions.

Note that the possessive alternation does not occur in examples such as (416), where a verbal particle such as neerdown is present: in such constructions possessive datives are excluded, while possessive bij-PPs and NP-internal possessors are possible. This suggests that the particle neer functions as a complementive in this construction, not the PP op de knie; cf. Section 2.2.1, sub IV, for relevant discussion.

Note that the fact that possessive datives can be promoted to subject under krijgen-passivization shows that nominal possessors can be assigned nominative case. We therefore expect that nominative possessors can occur with undative verbs; Subsection VI will show that this expectation is indeed borne out.

Locational PP-complementives also occur with causative (transitive) motion verbs such as rijdento drive. It is not easy to construct semantically plausible examples, but the examples in (417), which are all somewhat strange because of the implied purposefulness, show that we can find possessive datives/bij-PPs with such verbs.

The (a)-examples in (418) confirm this by showing that the unaccusative counterparts of the causative motion verbs in (417) readily allow the possessive dative/bij-PP alternation. The (b)-examples show the same for change-of-location verbs such as springento jump.

In some cases, verbs of sound transmission can also be used as unaccusative motion verbs with a locational complementive. This is illustrated for fluiten in example (419); we think that there is a preference for the double object construction with such verbs, but the other two constructions are easy to find on the internet.

Location verbs like zittento sit, staanto stand, liggento lie, and hangento hang are also unaccusative and the examples in (420) show that the possessive dative/bij-PP alternation is possible with these verbs.

However, there are restrictions that are not well understood. For instance, while all examples in (420) are acceptable, the structurally parallel (a)-examples in (421) do not allow the possessive dative. The idiomatic (b)-examples, on the other hand, clearly prefer the possessive dative.

A similar contrast is found in (422); while the literal construction in the (a)-examples at least marginally allows all alternants, the metaphorical (b)-examples seem to require a possessive dative.

Possessive indirect objects occur not only with verbs that normally select a PP-complementive, but also with verbs that optionally take such a PP; this is illustrated for the ditransitive verb gevento give in (423). These examples also show that the dative noun phrase, being a recipient, normally alternates with an aan-PP, but that this alternation is blocked when the locational PP-complementive in de armeninto the arms is present; the indirect object must then be realized as a dative noun phrase, which now also acts as an inalienable possessor, or as a possessive bij-PP.

Note that the unacceptability of the aan-PP in (423b) follows directly from the claim in Section 3.3.1.1, sub IV, that recipients PPs are in fact complementives. Since clauses can contain at most one complementive and since the locational PP in de armen already has this function, the complementive aan-PP is blocked; cf. Section 2.2.1, sub IV, for discussion. A similar problem does not arise with the bij-PP as it originates as part of the locational PP; cf. the discussion in Subsection IV.

Another case is presented in (424) with the unaccusative verb vallen. Example (424a) shows again that the locational PP is optional, and (424b) shows that the alternation is at least marginally possible with a locational PP. The percentage sign is used to indicate that our informants give different judgments about the acceptability of the bij-PP.

There are numerous more or less idiomatic examples of inalienable possession with unaccusative verbs of this type. These constructions often do not easily allow alternants with a possessive bij-PP or an NP-internal possessor. The judgments on the primed examples again vary from case to case and probably from speaker to speaker.

The acceptability of the krijgen-passive in (426c) leads to the expectation that subjects of undative verbs can also function as inalienable possessors of the nominal part of a predicative locational PP. The acceptability of the examples in (428) shows that this expectation is indeed confirmed.

In fact, Section 2.1.4 used the acceptability of the inalienable possession reading of examples such as (428) to argue for the existence of undative verbs: the subject is not an external but an internal argument of the verb and thus capable of acting as an inalienable possessor. Indeed, on the basis of the fact that the examples in (429) also have an inalienable possession reading, we concluded that verbs of cognition such as kennen/wetento know also belong to the class of undative verbs.

The possessive nominatives in examples like (428) and (429) never alternate with a possessive bij-PP, which is of course due to the fact that PPs are not normally used as subjects of clauses. This can be seen by comparing the examples in (429) with *Bij Jan kent het gedicht uit zijn hoofd and *Bij Jan weet het uit zijn hoofd.

Although nominal possessors are usually assigned dative case, there are a number of verbs that seem to take a direct/accusative object that can act as an inalienable possessor. These verbs seem to be characterized by the fact that they involve some form of bodily contact. A small sample of these verbs is given in (430); note that most of these verbs can also be used as regular transitive verbs.

Two examples of inalienable possession constructions with these verbs are given in (431). That the inalienable possessors are direct/accusative objects is clear from the primed examples, which show that they can be promoted to subject under regular passivization; krijgen-passivization, on the other hand, yields a marked result. That the inalienable possessors of the verbs in (430) are direct objects is also shown by the fact, illustrated in the doubly-primed examples, that the possessor can be attributively modified by the past participle forms of the verbs in the corresponding active clauses; attributive modification requires that the modified noun be the internal theme argument of the input verb of the participle; cf. Section A31.2.

The fact that the locational PP is optional might lead us to think that it is simply an adjunct and therefore different from the predicative PPs in the possessive dative constructions discussed earlier. However, there are reasons for thinking that this is not the case and that we are actually dealing with constructions that are very similar to these possessive dative constructions. A first reason to reject the idea that the locational PPs in (431) are adjuncts is that they do not pass the adverbial test: the examples in (432) show that the PPs cannot be analyzed as VP adverbials, since the paraphrases with en hij doet dat ... clauses lead to semantically incoherent results (although (432b) is somewhat better because of the contrast between *Jan tikte Peter and Jan kust Marie).

Second, the PPs behave like locational complementives in the sense that they seem to resist extraposition (cf. Section 2.2.1, sub III): it is strongly preferred that they precede the clause-final verb.

Third, the examples in (434) show that the accusative noun phrases can at least marginally be replaced by possessive bij-PPs; such examples are normally used in contrastive contexts. The primed examples show that this alternation is completely excluded when the locational PP is not present.

We have established in (432) and (433) that the locational phrases in (431) behave not as adjuncts but as complementives; this is quite remarkable, since these locational PPs do not seem to have an argument of which they can be predicated. This problem can be solved by accepting the assumption in Broekhuis et al. (1996) that, despite appearances, there is in fact such an argument, realized not as a noun phrase but as part of the verb; cf. also Bos (1972). The line of reasoning is that verbs of bodily contact are derived from so-called light verbs, i.e. phonetically empty verbs meaning “to give”, which have morphologically merged with their direct object; cf. the examples in (435).

Note that for some verbs from this semantic field it is not possible to give a synchronic paraphrase: for example, the presumed input noun kietel for the verb kietelento tickle is given in the Woordenboek der Nederlandse Taal with the meaning dartele zinneprikkelfrolicsome stimulation of the senses, but will probably not be recognized as part of the current Dutch vocabulary by many speakers.

The merging hypothesis means that an example such as (431a) has an underlying structure that is quite close to the corresponding possessive double object construction with the lexical verb gevento give in (436a). The main difference is that (431a) involves the syntactic incorporation of the direct object into a phonetically empty light verb v with the meaning “to give”; e.g. tikkento hit is a syntactically created morphological complex verb [v N-v], in which N stands for the incorporated noun and v stands for the postulated light verb. A comparison of the (a)-examples in (432) to (434) with those in (436b-d) shows that this incorporation hypothesis accounts for most of the core data.

What does not yet seem to follow from the incorporation approach are the passivization and attributive modification facts in the primed examples in (431). However, if Baker’s (1988: §3.4.1) claim is correct that incorporation of the direct object makes it unnecessary for the direct object to be assigned accusative case, then these facts also fall into a more general pattern; cf. Schermer-Vermeer (1996:276) for essentially the same proposal phrased in slightly different terms. Consider first the examples in (437) with the ditransitive verb voerento feed, in which the noun phrase brood functions as direct object (theme) and the phrase (aan) de eendjes as indirect object (recipient).

Example (438a) shows that the verb voeren is like the transitive verb etento eat in that it takes a cognate direct object which can be left implicit. The acceptability of regular passivization in (438b) shows that this makes it possible for the verb to assign accusative case to the recipient; cf. Section 3.2.1.3, sub IIC, for a more detailed discussion. The acceptability of (438b) thus strongly suggests that the fact that the inalienable possessors in the primeless examples in (431) are assigned accusative case simply follows from Baker’s claim; because the incorporated direct objects need not be assigned case, accusative case becomes available for the recipients. Example (438c) further shows that leaving the cognate object implicit also allows the past participle gevoerd to be used as an attributive modifier of a noun corresponding to its recipient; cf. Section A31.2.1, sub I-I.

The doubly-primed examples in (431) will now follow, assuming that incorporation has an effect similar to the suppression of a cognate object.

The examples in (439) show once again that subjects of active constructions do not normally function as inalienable possessors in standard Dutch: while the indirect object in (439a) can easily function as the inalienable possessor of the nominal part of the locational phrase, this is not possible for the subject in example (439b), which is acceptable, but only if the beard in question is not Jan’s beard.

However, there are several ways of syntactically expressing that a subject is to be understood as an inalienable possessor. The first way, illustrated in (440a), involves the addition of a reflexive dative object; the reflexive then functions as the actual possessor, but since it is bound by the subject of the clause, the referent of the latter is construed as the possessor by transitivity. In (440b) we find essentially the same thing, due to the fact that the bij-PP contains a reflexive bound by the subject of the clause. The use of a possessive pronoun in (440c) in principle leaves open whether the referent of the subject is the possessor, but this reading can be enforced by adding the modifier eigenown.