1 The dataset

The paradigm in (1) shows that Dutch sometimes exhibits quite a complex system of PP alternations with near-synonymous meanings. The six acceptable forms in this paradigm all translate into English as ‘(up) to the end’. The examples in (1) show an ever increasing wealth of adpositional material, with tot, aan and toe all belonging to the category P; in (1c′) all three adpositions are present and aan even occurs twice. Exploiting the standard use of parentheses, we can collectively refer to all and only the acceptable forms in (1) with the string tot (aan) het einde ((aan) toe).

    1. (1)
    1. a.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. a′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1.  
    1. b.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. b′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1.  
    1. c.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1. c′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1.  
    1. d.
    1. *tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. d′.
    1. *tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on

There may be subtle meaning differences between the acceptable forms in (1), which we will not discuss here but which we take to be related to the fact that the functional make-up of the various PP-forms may be different; cf. Koopman (2010) and den Dikken (2010), as well as some relevant discussion in section 2, below.

The PP-forms in (1) can be used as temporal or as spatial adjuncts: cf. (2a & b). The fact, illustrated in (2c), that the verb lopen cannot take the auxiliary zijn when combining with a PP of the type in (1) shows that the complex PPs in (1) cannot be used as predicates (which are always spatial). We will leave the study of the meaning as well as the external distribution of the PPs in (1) to future research and focus our attention on the structural representations of these formations.

    1. (2)
    1. a.
    1. Jan
    2. Jan
    1. heeft
    2. has
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)
    2. to
    1. geslapen.
    2. slept
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.     [temporal]
    2.  
    1. ‘Jan has slept up to the end (of e.g. the meeting).’
    1.  
    1. b.
    1. Jan
    2. Jan
    1. heeft
    2. has
    1. het
    2. the
    1. gras
    2. grass
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)
    2. to
    1. verwijderd.
    2. removed
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1. [spatial]
    2.  
    1. ‘Jan has removed the grass up to the end (e.g. from the garden path).’
    1.  
    1. c.
    1. Jan
    2. Jan
    1. heeft/*is
    2. has/is
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)
    2. to
    1. gelopen.
    2. walked
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.     [not predicative]
    2.  
    1. ‘Jan has walked up to the end.’

Before we start discussing the internal structure of the acceptable PP-formations in (1), we will first briefly discuss these examples at a more superficial, observational level. The (a)-examples in (1) show that the preposition tot ‘(up) to’ is special in that it can not only take a DP as its complement but is also able to take a PP. The examples in (3) show that this is not possible for other directional prepositions such as naar ‘to’.

    1. (3)
    1. a.
    1. tot/naar
    2. to/to
    1. het
    2. the
    1. einde
    2. end
    1. a′.
    1. tot/*naar
    2. to/to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1.  
    1. b.
    1. tot/naar
    2. to/to
    1. de
    2. the
    1. kerk
    2. church
    1. b′.
    1. tot/*naar
    2. to/to
    1. voor
    2. in.front.of
    1. de
    2. the
    1. kerk
    2. church
    1.  
    1. c.
    1. tot/naar
    2. to/to
    1. de
    2. the
    1. hoek
    2. corner
    1. c′.
    1. tot/*naar
    2. to/to
    1. in
    2. in
    1. de
    2. the
    1. hoek
    2. corner

The (b)-examples in (1) show that the DP and the PP can both be followed by the adpositional element toe. This element is generally taken to be the allomorph of the preposition tot which appears when the preposition is not followed by its complement. This is very clear in cases such as (4b), in which daartoe is the pronominalised counterpart of the PP tot strenge maatregelen in (4a): because the D-word daar, which is a pro-from replacing the DP strenge maatregelen, precedes the preposition, the latter surfaces as toe.

    1. (4)
    1. a.
    1. het
    2. the
    1. schandaal
    2. scandal
    1. dat
    2. that
    1. de
    2. the
    1. president
    2. president
    1. tot
    2. to
    1. strenge
    2. stern
    1. maatregelen
    2. measures
    1. dwong
    2. forced
    1. ‘the scandal that forced the president to take stern measures’
    1.  
    1. b.
    1. het
    2. the
    1. schandaal
    2. scandal
    1. dat
    2. that
    1. de
    2. the
    1. president
    2. president
    1. daar
    2. there
    1. toe
    2. to
    1. dwong
    2. forced
    1. ‘the scandal that forced the president to that’

The adpositional element toe may also be used with other functions, for instance, as a verbal particle in particle-verbs such as toezeggen ‘to promise’ in (5a), or as the second part of a circumposition such as naar … toe ‘to(wards)’ in (5b). This may raise the question as to whether postpositions and particles should be considered different but this is not a topic we will discuss in this paper; see Koopman (2010) for relevant discussion

    1. (5)
    1. a.
    1. De
    2. the
    1. president
    2. president
    1. heeft
    2. has
    1. strenge
    2. stern
    1. maatregelen
    2. measures
    1. toegezegd.
    2. prt-promised
    1. ‘The president has promised stern measures.’
    1.  
    1. b.
    1. Jan
    2. Jan
    1. liep
    2. walked
    1. naar
    2. to
    1. de
    2. the
    1. kerk
    2. church
    1. toe.
    2. to
    1. ‘Jan walked to(wards) the church.’

The (c) and (d)-examples in (1), finally, show that the DP and the PP can also be followed by the adpositional element aan, but only if the adposition toe is also present. The distribution of aan within the complex DP will be a central topic of this paper, and we will present a simple explanation for its doubling in (1c′). This doubling is not mentioned in the Syntax of Dutch (Broekhuis 2013); this reference work does, however, contain discussion of the other patterns, and also discusses the relevant literature on complex adpositional structures, starting with Van Riemsdijk (1978).1

2 The analysis

This section begins the discussion of our analysis medias in res, by laying out what we believe is the right analysis for the (b) and (c)-examples in the paradigm in (1). Sections 3 and 4 will subsequently substantiate the proposed treatments of the individual P-elements of tot, aan and toe. Section 5 concludes, and puts the antonyms of (1) on the agenda and shows that it is not possible to straightforwardly apply the analysis of tot (aan) het einde ((aan) toe) to these cases.

2.1 The underlying structure

We propose the underlying representation in (6) for the complex examples in the set, containing all three P-elements: tot, aan and toe. For ease of presentation, the structure in (6) abstracts away from details regarding the functional superstructure of PP2: rather than taking a specific stand (unnecessary here) on the label of its functional extension, we will throughout use the label “xPP2”, standing for “extended projection of PP2”. While the projection of P2 is always extended up to xPP2 (for reasons discussed in section 2.6), that of P3 is either “bare” or extended (whence the parenthesised ‘x’ here). This will play a role in the accounts of (1b′) and (1c) offered in sections 2.2 and 2.3. The functional superstructure of PP1 does not play a role at all, and will therefore be ignored completely in what follows.

    1. (6)

The structure in (6) cannot be pronounced as is because toe is a postposition, hence must receive something to its left, in the specifier position of xPP2. We propose that the requirement that SpecxPP2 be filled can be met in one of two ways in the course of the derivation.

2.2 The derivation of tot aan het einde toe (1bʹ)

The first way of meeting the requirement that SpecxPP2 be filled is based on the version of (6) that features an extended projection of P3, and is sketched out in (7). This derivation involves movement of xPP3, the complement of P2, into SpecxPP2.2 The derivation in (7) delivers tot aan het einde toe in (1b′) as its grammatical output.3

    1. (7)

2.3 The derivation of tot het einde aan toe (1c)

The second way of meeting the requirement that SpecxPP2 be filled starts out from the version of (6) with a “bare” PP3 in the complement of P2. Because P3 now lacks its own extended projection, it cannot be functionally licensed within the confines of its extended projection, and must instead be incorporated into (i.e., left-adjoined to) P2, resulting in the formation of a complex postposition [P3+P2] aan+toe. A corollary of P-incorporation is the raising of the complement DP het einde of P3 into SpecxPP2. Structure (8) illustrates this derivation. It delivers the output tot het einde aan toe in (1c).

    1. (8)

According to the spirit of the Government Transparency Corollary in (9), incorporation of P3aan into P2toe causes DP to come to behave like the complement of P2.4

(9) The Government Transparency Corollary: A lexical category which has an item incorporated into it governs everything which the incorporated item governed in its original structural position (cf. Baker 1988: 64).

Adopting this, we are able to account for the unacceptability of the three examples in (10) in a simple way. First, (10a) is unacceptable because after incorporation of aan into toe, the DP should behave as a complement of toe but fails to occupy the specifier position SpecxPP2 in toe’s extended projection; incorporation of aan thus forces the derivation in (8). Second, the unacceptability of (10b) is due to the fact that aan does not incorporate into toe, so that the DP does not have the licence to extract from the aan-PP, and does not behave as the complement of toe and is thus not a suitable candidate for occupying the specifier position SpecxPP2; in the absence of aan-to-toe movement, this position can only be filled by the full aan-PP, as in the derivation in (7) above. Finally, the otherwise surprising unacceptability of (1d), repeated as (10c), follows from the fact that movement of DP into SpecxPP2 is contingent on incorporation of aan into a postpositional P2, which is not present in this case.

    1. (10)
    1. a.
    1. *tot
    2.   to
    1. aan
    2. on
    1. toe
    2. to
    1. het
    2. the
    1. einde
    2. end
    1.  
    1. b.
    1. *tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. aan
    2. on
    1.  
    1. c.
    1. *tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. aan (=(1d))
    2. on

Note in passing that example (11a) cannot be derived along the lines of (8), with movement of DP to SpecxPP2 and incorporation of aan into toe, because specifiers of postpositions pronominalise as ordinary D-pronouns, not as R-pronouns; cf. de boom die hij is in geklommen (lit: ‘the tree D.PRON he is in climbed’). So (8) delivers (11b), which is somewhat marginal, on a par with ?tot aan dit/dat toe, where the D-pronoun has not moved. To derive (11a), we need to avail ourselves of the derivation in (7), with R-pronominalisation of aan het einde as hier/daar aan ‘on this/that’. This an independently motivated option in Dutch; see also example (23) below.

    1. (11)
    1. a.
    1.   tot
    2.   to
    1. hier/daar
    2. here/there
    1. aan
    2. on
    1. toe
    2. to
    1.  
    1. b.
    1. ?tot
    2.   to
    1. dit/dat
    2. this/that
    1. aan
    2. on
    1. toe
    2. to

2.4 The derivation of tot aan het einde aan toe (1cʹ)

We can derive the pattern in (1c′), tot aan het einde aan toe, by assuming that movement of PP3aan het einde into SpecxPP2 can apply in tandem with incorporation of aan into P2toe, as shown in (12). We propose that this derivation yields a grammatical output with full exponence of both copies of P3 because neither copy c-commands the other — since the two tokens of aan are structurally disconnected neither is required to be deleted; cf. Kayne (1994: 96, (50)).5

    1. (12)

Note that our proposal entails that the PP pattern tot aan het einde aan toe in (1c′) involves syntactic reduplication of aan and not independent selection of two accidentally identical lexical items aan, which is supported by the fact, illustrated in (13), that the P-elements to the right of tot and to the left of toe cannot be chosen independently of one another.

    1. (13)
    1. a.
    1. *tot
    2.   tot
    1. naar
    2. to
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1.  
    1. b.
    1. *tot
    2.   to
    1. na
    2. after
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to

The examples in (13) should be contrasted with the primeless cases in (14). The latter involve accidental identity rather than reduplication, as is supported by the fact that they occur side-by-side to the cases in the primed examples.

    1. (14)
    1. a.
    1. aan
    2. at
    1. de
    2. the
    1. steiger
    2. pier
    1. aanleggen
    2. on.moor
    1. ‘to moor to the pier’
    1.  
    1. a′.
    1. bij
    2. near
    1. de
    2. the
    1. steiger
    2. pier
    1. aanleggen
    2. on.moor
    1. ‘to moor near the pier’
    1.  
    1. b.
    1. iets
    2. something
    1. aan
    2. to
    1. iemand
    2. someone
    1. aangeven
    2. on.give
    1. ‘to give something to someone’
    1.  
    1. b′.
    1. iets
    2. something
    1. aan
    2. to
    1. iemand
    2. someone
    1. doorgeven
    2. on.give
    1. ‘to pass something on to someone’

We thus propose that the reduplication of aan in (1c′) arises thanks to the fact that second occurrence of aan itself binds a copy. This is in agreement with the fact that standard assumptions concerning the successive cyclicity of movement force incorporation of aan into P2toe to precede movement of PP3 into SpecxPP2. Note that the derivation in (12) does not license spell-out of the DP het einde ‘the end’ to the right of the P2-complex aan toe: because the higher copy of PP3 asymmetrically c-commands the lower copy of PP3, the latter must be silenced in its entirety. So we expect that DP can only be spelled out to the left of the complex P2aan toe, and, indeed, *tot (aan (het einde)) aan toe het einde is impossible.

2.5 The derivation of tot het einde toe (1b)

Section 1 already suggested that the (a)-examples in the paradigm in (1) simply involve adpositional structures with, respectively, a DP and a PP complement, as indicated in (15a) and (15b). The structures underlying these cases are thus much reduced compared to the cases discussed above, which were argued to have the underlying structure in (15d). We did not yet discuss the underlying structure of tot het einde toe in (1b).

(15) a. [PP1 P1=tot [DPhet einde]] [example (1a)]
  b. [PP1 P1=tot [xPP3 P3=aan [DPhet einde]]] [example (1a′)]
  c. ?? [example (1b)]
  d. [PP1 P1=tot [xPP2 P2=toe [(x)PP3 P3=aan [DPhet einde]]]] [remaining cases]

For tot het einde toe, the question arises whether its structure is analogous to that of the complex cases, which also contain both tot and toe, or whether it is simpler. The latter would amount to saying that (1b) differs from the other examples with toe in that P3=aan and its projection are absent from the structure, as indicated in (16): P2=toe directly takes the DP het einde as its complement, and, just as in the derivation in (8), forces it to raise into SpecxPP2.

(16) [PP1 P1=tot [xPP2 P2=toe [DPhet einde]]]

A serious worry for the proposal in (16) is that the postposition toe does not otherwise seem to accept DP dependents in the standard language: whenever toe has a DP to its left, it serves as a verbal particle, as was already illustrated by (5a) in section 1. A treatment of toe in (1b) as a particle is impossible, however, because doing so would make the DP het einde case-dependent on P1=tot, while prepositions in Dutch normally do not engage in “exceptional case-marking” (exceptions are met ‘with’ and zonder ‘without’ in absolutive constructions). This wrinkle leads us to regard (16) with serious suspicion. As an alternative outlook on (1b), we suggest (17), which treats (1b) as structurally on a par with the three other examples with toe in (15c) but with a silent P3.

(17) [PP1 P1=tot [xPP2 P2=toe [PP3 P3=Ø [DPhet einde]]]]

The postulation of a silent allomorph of aan is not an innovation conjured up specifically for the purpose of analysing the string in (1b): on the transformational approach to the dative shift alternation pursued in Den Dikken (1995: section 3.9), aan has a silent allomorph in ditransitive constructions, too, which is licensed by incorporation; we return to this issue below example (34) in section 4. If we assume the same for the empty P3 in (17), there are two possible continuations of the derivation: either the DP or the full PP3 can be moved into the specifier of SpecxPP2, along the lines of the representations in (8) and (12), respectively. At this point, we see no clear reason to prefer one of the two analyses, and it may well be the case that both derivations are available. We leave this to future research.

2.6 A note on extended projection and recursion

We have argued that the complement of P2 (toe) in (6) is either a “bare” PP or a functional extension xPP, with the choice between the two options giving rise to different outputs. The complement of P1 (tot), on the other hand, is always an xPP. The bare version of PP2 would not feature the postposition toe but instead its prepositional counterpart tot because the adposition can only be spelled out as toe if something is placed in SpecxPP. This is reflected in the empirical fact, not discussed earlier, that it is clearly impossible to embed immediately below the preposition tot a projection of the same preposition: the string *tot tot (aan) het einde, with two consecutive tokens of tot, is unacceptable. Why should this be?

Directly embedding a bare projection of P2=tot under P1=tot would instantiate a kind of self-embedding recursion which, besides adding a second token of tot and its associated semantics, would not make any contribution to the syntax or semantics of the resulting construct. Syntactically, P2 introduces something as its complement (viz., a projection of aan) that P1 could perfectly well have introduced itself (as was shown in (15b)). Furthermore, the lexical meaning borne by P2 is exactly the same as that of P1, resulting in reduplication. Such reduplication could not contribute anything apart from emphasis – and quite generally, Dutch cannot place emphasis on adpositions by reduplicating them (De kat ligt OP (*op) het bed, niet ONDER (*onder) het bed ‘the cat is lying on the bed, not under the bed’). Directly embedding a bare projection of tot beneath another token of tot thus results in complete redundancy. Embedding an extended projection of postpositional toe below tot does not: it gives rise to a morphosyntactic output that could not have been obtained by forgoing the inclusion of toe and its entourage.6

2.7 Conclusion

The structures in (18) sum up the underlying representations needed for an analysis of the paradigm in (1). We have shown that the structure in (18d) may give rise to three different surface structures with the derivations indicated in (7), (8) and (12). The structure in (18c) was argued to give potentially rise to two different surface structures, with derivations similar to those in (8) and (12), but the difference between their outputs is difficult to demonstrate due to the fact that P3 here is phonetically empty. Observe that a derivation based on (18c) along the lines of (6) is excluded: for licensing purposes, the silent head of the “bare” PP3 must incorporate into P2.

(18) a. [PP1 P1=tot [DPhet einde]] [example (1a)]
  b. [PP1 P1=tot [xPP3 P3=aan [DPhet einde]]] [example (1a′)]
  c. [PP1 P1=tot [xPP2 P2=toe [PP3 P3=Ø [DPhet einde]]]] [example (1b)]
  d. [PP1 P1=tot [xPP2 P2=toe [(x)PP3 P3=aan [DPhet einde]]]] [remaining cases]

3 P1tot: The head of the complex structure

A salient feature of the underlying structure in (6) and the derivations based on it is that the head of the complex structure is P1=tot, NOT P2=toe. This allows us to treat all of the examples in (1) as fundamentally the same at the highest level: all the PPs in (1) are projections of the preposition tot; variation is a function of the internal composition of PP1. This was also the reason why we did not consider in section 2.5 two alternative analyses that easily spring to mind for tot het einde toe in (1b). The first alternative analysis that one might consider takes tot het einde toe to be a circumpositional phrase just like naar het einde toe. The second alternative analysis takes (1b) to be a PP (tot het einde) followed by the verbal particle (toe), which might be feasible for the particle-verb toelaten (tot) ‘to admit (to)’. We will show that these two alternative options can be excluded and we will conclude from this that all PPs in (1) are indeed headed by the preposition tot.

3.1 The inadequacy of a circumpositional analysis of tot het einde toe

This section will show that it is not possible to analyse the string tot het einde toe in (1b) as a circumpositional phrase headed by toe comparable to naar het einde toe in (19a): the derivation in (19b) is impossible.

(19) a.   [PP1 toe [PP2 naar het einde]] → [xPP1 [PP2 naar het einde]i [PP1 toe ti]]
  b. *[PP1 toe [PP2 tot het einde]] → [xPP1 [PP2 tot het einde]i [PP1 toe ti]]

We can justify this by pointing to the fact that the syntactic distribution of tot het einde toe fits in perfectly with the other members of the paradigm in (1), which clearly do not allow an analysis as a circumpositional phrase, and does not match that of naar het einde toe. A first indication to this effect is that while the circumpositional phrase naar het einde toe can be used as a postnominal modifier in examples such as (20a), none of the acceptable adpositional tot-phrases in (1) allow this.

    1. (20)
    1. a.
    1. Dit
    2. this
    1. is
    2. is
    1. [de
    2. the
    1. weg
    2. road
    1. [naar
    2. to
    1. het
    2. the
    1. einde
    2. end
    1. toe]].
    2. to
    1. ‘This is the road towards the end.’
    1.  
    1. b.
    1. *Dit
    2.   this
    1. is
    2. is
    1. [de
    2. the
    1. weg
    2. road
    1. [tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)]].
    2. to

Conversely, as illustrated in (21), the tot-phrases in (1) can be used adverbially whereas the circumpositional phrase naar het einde toe cannot easily be used in this function.

    1. (21)
    1. a.
    1. *dat
    2.   that
    1. Jan
    2. Jan
    1. de
    2. the
    1. weg
    2. road
    1. [naar
    2. towards
    1. het
    2. the
    1. einde
    2. end
    1. toe]
    2. to
    1. afliep.
    2. prt-walked
    1.  
    1. b.
    1.   dat
    2.   that
    1. Jan
    2. Jan
    1. de
    2. the
    1. weg
    2. road
    1. [tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)]
    2. to
    1. afliep.
    2. prt-walk
    1. ‘that Jan walked down the road up to the end.’

The proposed distinction between the tot-PPs in (1) and the circumpositional phrase naar het einde toe is also supported by auxiliary selection. Example (22a) shows that when directional PPs combine with a lexically unergative verb such as lopen ‘walk’ that selects hebben ‘have’ as its perfect auxiliary, they normally cause the motion verb to undergo “ergative shift”, resulting in the selection of the auxiliary zijn ‘be’. However, when one of the tot-PPs combines with such a verb, only hebben-selection is acceptable, showing that they cannot serve as complements to lexically unergative motion verbs.

    1. (22)
    1. a.
    1. Jan
    2. Jan
    1. is/?heeft
    2. is/has
    1. [naar
    2. to
    1. Leiden
    2. Leiden
    1. toe]
    2. to
    1. gelopen.
    2. walked
    1. ‘Jan has walked to Leiden walked.’
    1.  
    1. b.
    1. Jan
    2. Jan
    1. heeft/*is
    2. has
    1. [tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)]
    2. to
    1. gelopen.
    2. walked
    1. ‘Jan has walked up to the end.’

Why tot-PPs of the type in (1) can only serve as adjuncts is not a question we will try to answer here. For us what matters is simply the observational fact that all the tot-PPs in (1) behave on a par with respect to this distributional restriction. It is this distributional parallel between the examples in (1) that confirms that they all have tot as their head.

That it is tot and not toe that is the head of the complex PPs in the (b) and (c)-examples in (1) is also indicated by constituency tests of the familiar sort, involving replacement or displacement of a subportion of the complex PPs in question. The examples in (23) show that the PP following tot can be replaced by the locational/temporal proform daar/dan ‘then’; see also Broekhuis (2013). The facts in (23) confirm both the constituency of PP3 and that of xPP2, and they also show that the preposition tot is outside both of these constituents.

    1. (23)
    1. a.
    1. tot
    2. to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. a′.
    1. tot
    2. to
    1. daar/dan
    2. there/then
    1.  
    1. b.
    1. tot
    2. to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. b′.
    1. tot
    2. to
    1. daar/dan
    2. there/then
    1. toe
    2. to
    1.  
    1. c.
    1. tot
    2. to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1. c′.
    1. tot
    2. to
    1. daar/dan
    2. there/then
    1. aan
    2. on
    1. toe
    2. to

3.2 The inadequacy of a verbal-particle analysis of tot het einde toe

The verbal-particle analysis takes the string tot het einde toe in (1b) to be similar to the string in boldface in example in (24a), which like (1b) features both the preposition tot and the element toe. For the example in (24a), it is clear that tot forms a constituent with the following DP to the exclusion of toe (i.e., not [tot [… toe]] but [tot …] toe), which is used as a verbal particle here. A characteristic property of this configuration is that pre- and extraposing the string formed by tot and the DP following it (here, hun therapiegroep) is possible, as shown in (24b & c). The primed examples show that pied piping of the particle is impossible.

    1. (24)
    1. a.
    1.   Ze
    2.   they
    1. willen
    2. want
    1. hem
    2. him
    1. niet
    2. not
    1. [tot
    2. to
    1. hun
    2. their
    1. therapiegroep]
    2. therapy group
    1. toelaten.
    2. prt-let
    1. ‘They don’t want to admit him to their therapy group.’
    1.  
    1. b.
    1.   Tot
    2.   to
    1. hun
    2. their
    1. therapiegroep
    2. therapy group
    1. willen
    2. want
    1. ze
    2. they
    1. hem
    2. him
    1. niet
    2. not
    1. toelaten.
    2. prt-let
    1.  
    2.  
    1. [topicalisation]
    2.  
    1.  
    1. b′.
    1. *Tot
    2.   to
    1. hun
    2. their
    1. therapiegroep
    2. therapy group
    1. toe
    2. prt
    1. willen
    2. want
    1. ze
    2. they
    1. hem
    2. him
    1. niet
    2. not
    1. laten.
    2. let
    1.  
    1. c.
    1.   Ze
    2.   they
    1. willen
    2. want
    1. hem
    2. him
    1. niet
    2. not
    1. toelaten
    2. prt-let
    1. tot
    2. to
    1. hun
    2. their
    1. therapiegroep.
    2. therapy group
    1.  
    2.  
    1.  [extraposition]
    2.  
    1.  
    1. c′.
    1. *Ze
    2.   they
    1. willen
    2. want
    1. hem
    2. him
    1. niet
    2. not
    1. laten
    2. let
    1. tot
    2. to
    1. hun
    2. their
    1. therapiegroep
    2. therapy group
    1. toe.
    2. prt

The string tot het einde toe in (1b) clearly does not involve the verbal particle toe. First, the discussion of the examples in (21) and (22) has already shown that it is like the other strings in (1) in that it cannot be used as a verbal complement — it only serves as an adjunct. Concomitantly, toe is not the adpositional head of (1b), and, as a consequence of this, preposing or extraposing the string tot het einde to the exclusion of toe is impossible, as is shown in (25b & c); the primed examples show that pied piping it is obligatory. Note that we have added aan within brackets in order to show that the same is true for the string tot aan het einde in example (1b′).

    1. (25)
    1. a.
    1.   Ze
    2.   they
    1. hebben
    2. have
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. gerend.
    2. run
    1.  
    1. b.
    1. *Tot
    2.   to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. hebben
    2. have
    1. ze
    2. they
    1. toe
    2. to
    1. gerend.
    2. run
    1.  
    2.  
    1. [topicalisation]
    2.  
    1.  
    1. b′.
    1.   Tot
    2.   to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. hebben
    2. have
    1. ze
    2. they
    1. gerend.
    2. run
    1.  
    1. c.
    1. *Ze
    2.   they
    1. hebben
    2. have
    1. toe
    2. to
    1. gerend
    2. run
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde.
    2. end
    1.  
    2.  
    1. [extraposition]
    2.  
    1.  
    1. c′.
    1.   Ze
    2.   they
    1. hebben
    2. have
    1. gerend
    2. run
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe.
    2. to

It should further be noted that all versions of tot (aan) het einde ((aan) toe) including (1b) can easily be combined with a particle-verb such as tegenwerken ‘to thwart’ or toestaan ‘allowed’. Given that verbs cannot combine with more than one particle (let alone two identical ones), the examples in (26) show that the verbal-particle analysis of tot het einde toe is not viable.

    1. (26)
    1. a.
    1. Ze
    2. they
    1. hebben
    2. have
    1. Peter
    2. Peter
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)
    2. to
    1. tegengewerkt.
    2. prt-thwarted
    1. ‘The have thwarted Peter (up) to the end.’
    1.  
    1. b.
    1. Mobieltjes
    2. cell.phones
    1. werden
    2. were
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)
    2. to
    1. toegestaan.
    2. prt-allowed
    1. ‘Cell phones were allowed (up) to the end.’

This section has shown that there can be no doubt that the constituent structures of (24a) and (25a) are very different: while in (24a) the verbal particle toe heads the complex structure and tot+DP is a constituent, in (25a) it is tot that heads the structure and the entire string tot (aan) het einde toe is one structural unit.

3.3 All phrases in (1) are headed by tot

To close this discussion of the headedness of the PPs in (1), let us return to the examples in (22), which are repeated here as (27) in a slightly adapted version for convenience. Though we did not make a point of this up until now, the reader will have noted that the element toe can legitimately occur in both these sentences, which gives them a piece of morphological matter in common. We have further noted that there is a difference between the two examples in (27) with respect to auxiliary selection.

    1. (27)
    1. a.
    1. Jan
    2. Jan
    1. is
    2. is
    1. [naar
    2. to
    1. Leiden
    2. Leiden
    1. toe]
    2. to
    1. gelopen.
    2. walked
    1. ‘Jan has walked to Leiden walked.’
    1.  
    1. b.
    1. Jan
    2. Jan
    1. heeft
    2. has
    1. [tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. ((aan)
    2. on
    1. toe)]
    2. to
    1. gelopen.
    2. walked
    1. ‘He has walked up to the end.’

Here we add that (27a) and (27b) also differ with respect to constituency. Example (28a) first shows that the string naar Leiden toe can easily be split by topicalisation, while the (b)-examples show that this split is not possible for tot (aan) het einde ((aan) toe): topicalisation cannot strand toe. These examples thus show very clearly that the strings naar x toe and tot (aan) x (aan) toe behave radically differently with respect to the possibility of fronting the substring following the initial P-element as a unit.7

    1. (28)
    1. a.
    1.   Naar
    2.   to
    1. Leiden
    2. Leiden
    1. is
    2. is
    1. Jan
    2. Jan
    1. toe
    2. prt
    1. gelopen.
    2. walked
    1.  
    1. b.
    1. *Tot
    2.   to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. (aan)
    2. on
    1. heeft
    2. has
    1. Jan
    2. Jan
    1. toe
    2. to
    1. gelopen.
    2. walked
    1.  
    1. b′.
    1.   Tot
    2.   to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. (aan)
    2. on
    1. toe
    2. to
    1. heeft
    2. has
    1. Jan
    2. Jan
    1. gelopen.
    2. walked

The ungrammaticality of (28b) versus the grammaticality of (28a) could be taken to indicate that naar LeidenIS, while tot (aan) einde (aan) is NOT, a constituent to the exclusion of toe. Interpreted this way, the facts in (28) are certainly compatible with the underlying representation in (6), which denies the string tot (aan) x constituent status to the exclusion of toe. But unfortunately, we cannot chalk these data up as evidence for (6) because, as it turns out, the ungrammaticality of (28b) could also be derived in another way. A logically plausible alternative explanation would capitalise on our earlier observation that complex PPs of the type represented by (6) only distribute as adjuncts when combined with a lexically unergative motion verb: even if tot (aan) x were a constituent, one would expect it to be prevented from movement stranding toe by the Huang’s (1982) Adjunct Condition. Indeed, (28a) only allows fronting of naar Leiden when naar Leiden toe is construed as the directional complement of lopen, causing “ergative shift” and concomitant selection of the auxiliary zijn ‘be’ in the perfect: this is shown by the fact that the pattern in (29) matches that of the (b)-examples in (28).8

    1. (29)
    1. a.
    1. *Naar
    2.   to
    1. Leiden
    2. Leiden
    1. heeft
    2. has
    1. hij
    2. he
    1. toe
    2. to
    1. gelopen.
    2. walked
    1.  
    1. b.
    1.   Naar
    2.   to
    1. Leiden
    2. Leiden
    1. toe
    2. to
    1. heeft
    2. has
    1. hij
    2. he
    1. gelopen.
    2. walked

So to some extent, comparing (28a) and (28b & b′) is a case of comparing apples and oranges. This severely diminishes the strength of the constituency test applied in (28). But all the facts remain perfectly compatible with the claim that all PPs in (1) are headed by the preposition tot. And since we had already discovered some unequivocal support for the underlying structure in (6), we can safely maintain that it is also correct for the string tot het einde toe in (1b), with the additional assumption that P3 is phonetically empty.

4 P2toe and P3aan, and their interrelationship

The adpositional complex tot het einde aan toe in (1b′) has the substring aan DP toe in common with the primeless examples in (30), but this section will show that the parallel is merely superficial. More specifically, there is no structural connection between the two complexes: while the substring in (1b′) is part of an adverbial modifier headed by tot, the substrings in (30) occupy the complement position of the verbs zijn ‘be’ and komen ‘come’, and serve as predicates of the subject of the clause, ik ‘I’.

    1. (30)
    1. a.
    1. Ik
    2. I
    1. ben
    2. am
    1. aan
    2. on
    1. vakantie
    2. vacation
    1. toe.
    2. to
    1. ‘I need a vacation.’
    1.  
    1. b.
    1. Ik
    2. I
    1. ben
    2. am
    1. niet
    2. not
    1. aan
    2. to
    1. die
    2. that
    1. review
    2. review
    1. toe
    2. to
    1. gekomen.
    2. come
    1. ‘I haven’t been able to do that review.’

Confirmation for the claim that aan DP toe in (30) is a predicative complement as well as illustration of the fact that toe here is a verbal particle comes from the grammaticality of (31a), the non-root counterpart to (30b): the adpositional element toe is freely included in the verbal cluster, which would have been impossible if aan DP toe had been an adverbial adjunct. Indeed, with respect to particle incorporation, (31a) differs starkly from (31b); not surprisingly in light of the adverbial use of tot (aan) DP toe in (31b) as well as the fact that not toe but tot is the head of the complex PP; cf. (6).

    1. (31)
    1. a.
    1. dat
    2. that
    1. ik
    2. I
    1. niet
    2. not
    1. aan
    2. on
    1. die
    2. that
    1. review
    2. review
    1. <toe>
    2.   to
    1. ben
    2. am
    1. <toe>
    2.   to
    1. gekomen.
    2. come
    1. ‘that I haven’t been able to do that review.’
    1.  
    1. b.
    1. dat
    2. that
    1. ik
    2. I
    1. tot
    2. to
    1. (aan)
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. <toe>
    2.   to
    1. heb
    2. have
    1. <*toe>
    2.     to
    1. gerend.
    2. run

That the examples in (30) have a structure in which the aan-PP serves as the complement of the particle toe and thus forms a constituent to the exclusion of toe is also clear from (32), illustrating pre- and extraposing of the substring aan+DP: we refer the reader to (25) for examples showing that the substring aan+DP in (1b′) cannot be topicalised or extraposed while stranding toe.

    1. (32)
    1. a.
    1. Aan
    2. on
    1. vakantie
    2. vacation
    1. <??toe>
    2.       to
    1. ben
    2. am
    1. ik
    2. I
    1. nu
    2. now
    1. wel
    2. AFF
    1. <toe>.
    2.   to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1. [topicalisation]
    2.  
    1. ‘I could do with a holiday.’
    1.  
    1. a′.
    1. Ik
    2. I
    1. geloof
    2. believe
    1. dat
    2. that
    1. ik
    2. I
    1. nu
    2. now
    1. wel
    2. AFF
    1. toe
    2. to
    1. ben
    2. am
    1. aan
    2. on
    1. vakantie.
    2. vacation
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  [extraposition]
    2.  
    1.  
    1. b.
    1. Aan
    2. on
    1. die
    2. that
    1. review
    2. review
    1. <*toe>
    2.     to
    1. kom
    2. come
    1. ik
    2. I
    1. helaas
    2. unfortunately
    1. niet
    2. not
    1. <toe>.
    2.   to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  [topicalisation]
    2.  
    1. ‘I won’t make it to work on that review.’
    1.  
    1. b′.
    1. Ik
    2. I
    1. geloof
    2. believe
    1. dat
    2. that
    1. ik
    2. I
    1. helaas
    2. unfortunately
    1. niet
    2. not
    1. toe
    2. to
    1. kom
    2. come
    1. aan
    2. on
    1. die
    2. that
    1. review.
    2. review
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1. [extraposition]
    2.  

The examples in (33) and (34) are similar to those in (32) in that aan+DP acts as a constituent to the exclusion of toe, with the latter serving as a verbal particle, as is illustrated by the fact, illustrated in the (b) and (c)-examples, that the string aan+DP can extrapose or topicalise as a unit. The (d)-examples add to this that aan can be omitted from the (a)-examples.

    1. (33)
    1. a.
    1. Alle
    2. all
    1. lof
    2. praise
    1. komt
    2. comes
    1. aan
    2. on
    1. Allah
    2. Allah
    1. toe.
    2. to
    1.  
    1. b.
    1. Aan
    2. to
    1. Allah
    2. Allah
    1. komt
    2. comes
    1. alle
    2. all
    1. lof
    2. praise
    1. toe.
    2. to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  [topicalisation]
    2.  
    1.  
    1. c.
    1. Alle
    2. all
    1. lof
    2. praise
    1. komt
    2. comes
    1. toe
    2. to
    1. aan
    2. on
    1. Allah.
    2. Allah
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  [extraposition]
    2.  
    1.  
    1. d.
    1. Alle
    2. all
    1. lof
    2. praise
    1. komt
    2. comes
    1. Allah
    2. Allah
    1. toe.
    2. to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.    [dative shift]
    2.  
    1. (34)
    1. a.
    1. Dit
    2. this
    1. behoort
    2. belongs
    1. aan
    2. on
    1. hem
    2. him
    1. toe.
    2. to
    1.  
    1. b.
    1. Aan
    2. to
    1. hem
    2. him
    1. behoort
    2. belongs
    1. dit
    2. this
    1. toe.
    2. to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.   [topicalisation]
    2.  
    1.  
    1. c.
    1. Dit
    2. this
    1. behoort
    2. belongs
    1. toe
    2. to
    1. aan
    2. on
    1. hem.
    2. him
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.     [extraposition]
    2.  
    1.  
    1. d.
    1. Dit
    2. this
    1. behoort
    2. belongs
    1. hem
    2. him
    1. toe.
    2. to
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.  
    2.  
    1.       [dative shift]
    2.  

This is the kind of alternation familiar from give-type constructions, usually referred to as the dative (shift) alternation. As already mentioned in section 2.5, Den Dikken (1995) argues at length that this alternation involves a silent allomorph PØ of the dative preposition (aan in Dutch), whose projection must move into a position structurally adjacent to the verb. From this position incorporation of PØ into the verb (necessarily for licensing PØ) becomes possible. Den Dikken (1995) also shows in detail, based on data taken mostly from English (but carrying over to Dutch), that in dative constructions with particle-verbs, the particle is structurally higher than the dative PP. When dative shift happens, the particle must reanalyse with the verb to facilitate the movement of the silent-headed PP that makes incorporation of PØ possible. Against this background, the (d)-examples in (33) and (34), which illustrate dative shift, confirm that the aan-PP in the (a)-examples feature a hierarchical structure in which toe is selected by the verb, and in turn takes the aan-PP as its complement. This, as we have seen, is a structure that is very different from the one we need for examples of the type in (1b′). The superficial similarity between (1b′), on the one hand, and the examples in (30), (33a) and (34a), on the other, is merely accidental.

In line with what was argued in the previous paragraph, Broekhuis (2013: 56) treats toe in strings of the type in (30a) as a verbal particle, with the particle-verb in turn selecting the aan-PP. In support of this, he mentions the ungrammaticality of (35a), where the string aan DP toe combines with a noun, making it impossible for toe to serve as a verbal particle. Example (35a) can be compared with the grammatical case in (35b), where the strings from (1) including toe are being used adnominally. Again we see clearly that despite the surface similarity, (1b′) cannot be treated on a par with the examples in (30), where toe is a verbal particle, determining the external distribution of the phrase; by contrast, the element toe in (1) is not a verbal particle, and it does not head the structure.

    1. (35)
    1. a.
    1. de
    2. the
    1. behoefte
    2. need
    1. aan
    2. on
    1. vakantie
    2. vacation
    1. (*toe)
    2.     to
    1.  
    1. b.
    1. de
    2. the
    1. reis
    2. journey
    1. tot
    2. to
    1. (aan)
    2. on
    1. Leiden
    2. Leiden
    1. (aan)
    2. on
    1. toe
    2. to

In the derivations in (8) and (12), the preposition aan incorporates into toe and forms a complex unit with it. Unfortunately, because of the adjunct status of the complex PPs in (28b & b′), we cannot bring this unit to light by trying to include aan+toe into a verbal cluster. But it is still significant that in (36), where we are not dealing with an adjunct, such incorporation fails completely, even though aan and toe are in fact linearly adjacent (thanks to the fact that aan’s complement is an R-word, obligatorily shifted to the left of aan). Clearly, in (36a) the string aan toe is not a head-level complex: if it were, inclusion of aan+toe in the verbal cluster should at least have been marginally well-formed.9

    1. (36)
    1. a.
    1.   dat
    2.   that
    1. ik
    2. I
    1. daar
    2. there
    1. niet
    2. not
    1. aan
    2. on
    1. toe
    2. to
    1. ben
    2. am
    1. gekomen.
    2. come
    1. ‘that I haven’t been able to get to that.’
    1.  
    1. b.
    1.   dat
    2.   that
    1. ik
    2. I
    1. daar
    2. there
    1. niet
    2. not
    1. aan
    2. on
    1. ben
    2. am
    1. toe
    2. to
    1. gekomen.
    2. come
    1.  
    1. c.
    1. *dat
    2.   that
    1. ik
    2. I
    1. daar
    2. there
    1. niet
    2. not
    1. ben
    2. am
    1. aan
    2. on
    1. toe
    2. to
    1. gekomen.
    2. come

If indeed aan and toe can form a complex P together, this may also give us a handle on (37). Expressions of this type (where a wide range of swear terms can be substituted for verdomme) are the Dutch equivalent of German Verdammt noch mal! ‘damned once more’, with Dutch nog (eens) being transparently the counterpart to German noch mal — but what follows nog (eens) in (37) finds no match in the German expression. It is not entirely clear what the function of aan toe is in (37); but it seems intuitively plausible that a culmination marker of the sort found in the (c)-examples in (1) (‘all the way to the end’) would be a natural ingredient for the kind of expression that (37) represents: ‘Dammit, I’m done with it/I’ve had enough of it’.

    1. (37)
    1. Verdomme
    2. damned
    1. nog
    2. yet
    1. (eens)
    2. once
    1. aan
    2. on
    1. toe!
    2. to

This suggestion is not intended as an analysis of (37) (this is plainly the topic for a different paper), nor does it incontrovertibly confirm as such the hypothesis that (1c) and (1c′) feature a complex P-head aan+toe. But if the details of (37) turn out to call for such a complex head, it can readily be thought of as a grammaticalisation of the morphosyntactically and semantically transparent aan+toe found in the structures in (8) and (12).

With reference to the complex strings in (1), Broekhuis (2013: 178) confesses that “the function of the elements aan and toe is … not clear to us”. We have not cleared the mystery up completely here, but we have found places in the tree for them.

5 The antonym of tot (aan) het einde ((aan) toe)

The topic of this paper has been Dutch expressions corresponding to English (right) up to the end, repeated here as (38). We have proposed a syntax for these adpositional phrases built on (6), in which tot is the head of the structure, with toe (when present) projecting an extended PP (xPP) in tot’s complement, and aan (when present) being the complement of toe. This captures all the facts canvassed in this paper.

    1. (38)
    1. a.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. a′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1.  
    1. b.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1. b′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. toe
    2. to
    1.  
    1. c.
    1.   tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1. c′.
    1.   tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. toe
    2. to
    1.  
    1. d.
    1. *tot
    2.   to
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on
    1. d′.
    1. *tot
    2.   to
    1. aan
    2. on
    1. het
    2. the
    1. einde
    2. end
    1. aan
    2. on

A natural follow-up to this piece would be a study of the antonyms of the examples in (1), that is, of adpositional constructions expressing the same thing as English (right) from the beginning (on). In Dutch, these, too, show a remarkable surface variability that is mostly unexplored in the literature, and, to our knowledge, never fully laid out; see Broekhuis (2013: 153–4; 175–6) for some relevant discussion.

The antonym of tot het einde aan toe in (38c) is van het begin af aan ‘right from the beginning on’. The preposition van corresponds to tot and functions as the head of the full proposition phrase; the adpositional element af corresponds to the adposition aan but differs from it in that it is postpositional, not prepositional; the closing adpositional element aan corresponds to the postposition toe in that is always the final element in the full adpositional phrase. In light of this, we might expect that the underlying structure of the adpositional phrases with all three adpositional elements present will look as in (39), with the DP het begin ‘the beginning’ undergoing obligatory movement into a specifier position in the extended projection of P3 (af), a postposition. The representation in (39) is still not a possible output because aan is a postposition and must therefore receive something to its left, in the specifier position of xPP2, in order to deliver a well-formed phrase.

    1. (39)

Apart from the fact that PP3 is a postpositional phrase, there are more differences between the adpositional phrases in (38) and their antonyms. This becomes immediately clear when we consider the expected counterparts to the (a)-examples in (38), given in (40). We have given these in a full clause: the reason for this is that, although the PPs van het begin and van het begin af are both impeccable as such, these examples show that only the latter can be used as an adverbial. The use of the percentage sign in (40b) indicates that in an informal questionnaire, all our informants (both linguists and non-linguists) accept this example, but that some of them consider it marked compared to the cases to be discussed below.10

    1. (40)
    1. a.
    1. *Van
    2.   from
    1. het
    2. the
    1. begin
    2. beginning
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1.  
    1. b.
    1. %Van
    2.   from
    1. het
    2. the
    1. begin
    2. beginning
    1. af
    2. off
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1. ‘Right from the start he was nervous.’

On the assumption that representation (39) underlies all other cases, we correctly predict example (41a) to be acceptable. This example can in fact be derived in two possible ways: either postpositional PP3 is moved into SpecxPP2 directly or P3 is incorporated into P2, after which the DP is moved into SpecxPP2 (we leave aside the question as to whether DP is moved from its complement position within PP3 directly or whether it is moved via SpecxPP3). If the incorporation option is indeed available, we correctly predict that af can be reduplicated, leading to the adpositional phrase in (41b).

    1. (41)
    1. a.
    1. Van
    2. from
    1. het
    2. the
    1. begin
    2. beginning
    1. af
    2. off
    1. aan
    2. on
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1.  
    2.  
    1. [structurally ambiguous]
    2.  
    1.  
    1. b.
    1. Van
    2. from
    1. af
    2. off
    1. het
    2. the
    1. begin
    2. beginning
    1. af
    2. off
    1. aan
    2. on
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1. ‘Right from the start he was nervous.’

If the suggested analysis is correct, the examples in (41) are the structural antonyms of, respectively, (38b′)/(38c) and (38c′). Example (38b), however, does not seem to have a structural antonym: the relevant string would be as given in (42), but this string is judged unacceptable by us and all our informants (although we have found various cases on the internet including two cases from the 1977 bible translation produced by Het Nederlands Bijbelgenootschap, which, however, did not return in the new 2004 translation).

    1. (42)
    1. *Van
    2.   for
    1. het
    2. the
    1. begin
    2. beginning
    1. aan
    2. on
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous

The reason for the contrast between (41a) and (42) might be that while aan can easily be used in circumpositional phrases in examples such as De kinderen liepen achter de optocht aan ‘The children followed the parade’, it is not possible to find postpositional phrases with aan; see the relevant lists in Broekhuis (2013: 33/50). If the same restriction holds for aan in (39), the contrast between (41a) and (42) is as expected. This leaves us with the antonyms of the (d)-examples in (38) in (43), which are predicted to be unacceptable. The fact that (43a) is marginally acceptable is not a problem given that it is homonymous to example (40b) with the structure [van [het begin af]], but the relative acceptability of (43b) is problematic for the analysis given above; we return to this case shortly.

    1. (43)
    1. a.
    1.  ?Van
    2.   from
    1. het
    2. the
    1. begin
    2. beginning
    1. af
    2. off
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1.    ‘Right from the start he was nervous.’
    1.  
    1. b.
    1. ??Van
    2.   from
    1. af
    2. off
    1. het
    2. the
    1. begin
    2. beginning
    1. af
    2. off
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous
    1. ‘Right from the start he was nervous.’

The fact that not all acceptable forms in (38) have a structural antonym could be accounted for by assuming that these adverbial tot-PPs differ from adverbial van-PPs with the meaning ‘(right) from the beginning (on)’ in that (i) tot but not van can select a DP complement and (ii) the postposition toe but not the postposition aan can select a DP complement. There are, however, reasons not to accept these conclusions. The first reason is that the examples in (41b) would then be special in that the first occurrence of af is followed by its complement, which is arguably impossible, af being a postposition. A bigger problem is that there is one perfectly acceptable form, given below as (44), which cannot be syntactically derived given the assumptions so far: the only feasible option would be saying that van takes a prepositional phrase af het begin as its complement but this would be expected to lead to a severely degraded result because af is a postposition; example (44) shows that this expectation is not borne out.

    1. (44)
    1. Van
    2. From
    1. af
    2. off
    1. het
    2. the
    1. begin
    2. beginning
    1. was
    2. was
    1. hij
    2. he
    1. nerveus.
    2. nervous

Examples of this type seem to be a relatively recent innovation in the Dutch-speaking world (the first attestations date back to the early 19th century) and have given rise to a lot of opposition from prescriptive grammarians; cf. Woordenboek der Nederlandsche Taal (lemma VANAF) and Van der Sijs (2005: 113). What we might suggest is that vanaf is a compound which differs from van in (40) in that is able to select a DP-complement. In fact, vanaf may also be able to select the postpositional phrase het begin af, which would then account for the problematic example in (43b) as well.

The discussion above has shown that it is not possible to mechanically transpose the analysis developed for tot (aan) het einde (aan (toe)) to their antonyms meaning ‘(right) from the beginning (on)’, due primarily to the fact that the preposition van heading the adpositional phrase differs from tot in that it imposes additional restrictions on its complement. Obtaining a clear view on the internal structure of these adpositional van-phrases is further hampered by the fact that they appear to have a competing form headed by the compound vanaf. The syntax of (40) to (44) thus remains on the agenda. But what we hope to have shown is that for (38) a comprehensive structural perspective can be upheld based on the underlying representation in (6).