4.2 The distribution of the P + CP pattern
4.2.4 Temporal adjunct clauses
A third context in which a preposition takes a CP as its complement involves temporal adjunct clauses. The internal syntax of these clauses will be discussed below.
As the examples in (48ab) illustrate, Dutch CPs can also occur as the complement of a temporal or locative preposition:
(48) a. Na [CPdat Jan binnenkwam], ging de telefoon.
after- that John entered rang the phone ‘After John entered, the phone rang.’
b. Om [CPdat Jan binnenkwam], hield iedereen z’n mond.
for- that ‘because’ John entered held everyone his mouth ‘Because John entered, everyone fell quiet.’
In Larson (1990) it is argued that in similar constructions in English, the specifier of the CP contains a temporal operator. Larson’s reasoning is as follows: in sentences of the kind in (49), the temporal preposition before is ambiguous between two readings:
(49) I saw Mary in New York [PP before [CP1 she claimed [CP2 that she
would arrive]]]. Readings
a. I saw Mary in New York before she made a certain claim, namely that she would arrive (some time).
b. I saw Mary in New York prior to some time t that she alleged would be the
time of her arrival. (Larson 1990:170)
In other words, before is construed either with the event denoted by CP1 or with the
event denoted by CP2. The same ambiguity can be found in (50), which contains the
temporal preposition after:
(50) I encountered Alice [PP after [CP1 she swore [CP2 that she had left]]].
Readings
a. I encountered Alice subsequent to her swearing that a certain proposition was true, namely that she had left (sometime).
b. I saw her after some time t that she swore would be the time of her arrival. (Larson 1990:170)
Larson observes that the same sort of ambiguity arises in adverbial clauses involving
when, where when can be construed either with the event in the least embedded
clause, i.e. CP1 in reading (51a), or with the event in the most deeply embedded
clause, i.e. CP2 in reading (51b):
(51) I saw Mary in New York [CP1 when she claimed [CP2 she would
arrive]. Readings
a. I saw her when she uttered the words. b. I saw her at the alleged arrival time.
(Larson 1990:170)
In the case of when-clauses, such ambiguities have standardly been analysed as involving movement (Larson 1990:170-171). In relation to this, Larson proposes
that the ambiguities observed in the context of before and after also involve movement. Thus, Larson assumes the presence of a temporal operator O of the category NP, which is generated in the adjunct position occupied by the bare-NP
when. As a result, the reading in which before is construed with the most deeply
embedded CP (i.e. CP2), for instance, has the structure in (52):
(52) [PP before [CP1 she claimed [CP2 that she would arrive O]]]
In (52) O is base-generated in the most deeply embedded complement of before, and
moves through successive cyclic movement to the specifier of CP1. This results in
the structure in (53) (see also Larson 1990:178):
(53) [PP before [CP1 Oi she claimed [CP2 ti that she would arrive ti ]]]
If the temporal preposition is construed with the event in CP1, then the temporal
operator starts out in the adjunct position in CP1 and subsequently moves to the
specifier of CP1, leaving a trace. The location of the trace(s), either in CP1 or in CP2,
accounts for the ambiguities.
Larson further assumes that the temporal operator in the specifier of CP1 receives
case from before. This case is thus a property of the chain Oi … ti …(ti)… The
empty category must bear case in order to be interpreted as a variable bound by O. Without such a variable, the operator O would bind nothing, and the structure would violate the general ban on vacuous quantification (Larson 1990:177). Before has the property to assign case because it can also take a nominal complement, as in a PP like before the party. Larson’s account is summarised in (54), and exemplified for a reading in which the preposition is construed with the highest CP:
(54) [PP before [CP1Oi … ti …]]
Ô Case Ò
Below, I will follow Larson and assume that a CP that functions as the complement of a temporal preposition contains an operator in its specifier.
As far as Dutch is concerned, it should be noted that Dutch shows the same ambiguities with adverbial clauses introduced by voordat (‘before’). Voordat can be
construed either with the event denoted by CP1 or with the event denoted by CP2.
(55) Ik zag Jan in Den Haag [PP voordat [CP1 ik dacht [CP2 dat
I saw John in The Hague for-that I thought that ik hem zou ontmoeten]]].
I him would meet
Readings
a. I saw John before the moment I thought that that I would meet him. b. I saw John before the moment I would meet him.
Furthermore, Larson observes that the prepositions although, because, unless, in
case and while are not ambiguous between a high and low reading in the way before, after, since and until are. This is not because such ambiguity is conceptually
inaccessible. As (56) shows, the lower reading, while conceptually conceivable, is ungrammatical:
(56) I visited New York [PP because [CP1 Mary dreamed [CP2 that Max was
there]]]
Readings
a. I visited New York because of Mary’s dreaming that Max was in New York.
b. * I visited New York because of what Mary dreamed, namely that Max was in New York.
It is tempting to relate the lack of ambiguity with prepositions such as because to the fact that these prepositions are not temporal, while before and after are. However, as Larson (1990:174) points out, this cannot be correct, given that the temporal preposition while also resists a long-distance reading. Rather, so Larson argues, the explanation for this division lies in the complement-taking properties of the prepositions concerned: while prepositions like before and after can select both CPs
or DPs, prepositions like although and because can select CPs only.16 From this,
Larson concludes that before and after are case-assigning prepositions, whereas
although and because are not.
Recall at this point that Larson assumes that the preposition before in (54) assigns
case to the operator in [spec, CP1] to avoid vacuous quantification. The trace in the
lowest CP is in adjunct position, and so does not receive case from any element
inside CP2; hence, it must receive case by forming a chain with the operator. In other
words, the principle of vacuous quantification is violated if a preposition is unable to assign case to the operator. This leads Larson to conclude that prepositions like
although and because do not allow a long-distance reading, since if they did, they
would violate vacuous quantification.
There is a general discussion going on about the relation between case assignment and argument status that must be mentioned at this point. Larson assumes that case must be assigned to the temporal operator (or, more precisely, to the trace that is part of the chain headed by the temporal operator) in order to avoid vacuous quantification. This begs the question of why an adjunct should receive case. Larson assumes that the trace occupies an adjunct position of the bare NP-adverb when. I surmise that Larson’s assumes that all nouns should receive case, and therefore also
when. Consider in this respect sentences like the one in (57):
16Because and in case can select a DP complement if of-insertion takes place, as in because of the
(57) Jan heeft [DP de hele winter] hout gehakt.
John has the whole winter wood chopped
‘John has been chopping wood for the whole winter.’
Given the case filter, the adverbial DP [DP de hele winter] must have received
case, but there is no thematic relation between DP and the predicate hakken (‘chop’). Note that there is no case-assigning preposition involved either. This kind of case assignment to an adverbial DP is problematic with respect to the traditional T-model that is used in the Government & Binding framework (see Chomsky 1981). In this T-model, there is a high amount of overlap between case theory and theta-theory. Given that an argument must have both case and a thematic role, the class of argument positions that is defined by case theory is largely identical to the one that is defined by theta-theory. In order to reduce this unnecessary machinery, Neeleman & Weerman (2001) propose a theory of case that is not intertwined with theta- theory. As a consequence, the licensing of case in their framework is no longer restricted to D-structure (inherent case) or S-structure (structural case), but it is more flexible instead. They propose a theory of case in which the PF and LF interface play a role. At PF, so-called “unspecified case” is licensed, i.e. case on constituents that cannot be associated with semantic functions. At the semantic interface (LF), on the other hand, case is interpreted by theta-theory. This means that at this level, there is a set of LF principles by which constituents (syntactic arguments) can be associated with semantic functions (semantic arguments) (see Neeleman & Weerman 1999:3-4). Within this kind of approach, case assignment to the adverbial DP [DP de hele winter] probably takes place at PF.
Despite these theoretical complications, I will follow Larson’s claim that the specifier of a CP that is the complement of a preposition that introduces a temporal adverbial clause contains a temporal operator:
(58) PP P CP voor Spec. C’ Oi C …ti... dat
The presence of this operator in spec. CP enables the predicate, in this example the preposition voor (‘for’), to assign case to the complement clause. This is necessary given the principle of predicate saturation, which requires the syntactic saturation (or discharge) of obligatory functional features, such as case (see Radford 1990:236). Consider for instance the assignment of nominative case. Case features that are obligatorily assigned by a case assigner (let us assume it is the functional head I for nominative case, but see also chapter 3, §3.4 above), must be syntactically saturated.
This means that they must be projected onto an appropriate constituent that is projected in the syntactic structure of a sentence. If there is no argument available to receive the discharged features, a dummy element is needed. This can be demonstrated on the basis of the easy to please construction:
(59) a. This book is easy to read. b.* Is easy to read this book. c. It is easy to read this book.
In (59a), the nominative case features can be discharged onto the DP this book. In (59b), however, there is no appropriate constituent available to receive the nominative case features, and the sentence is ungrammatical. Finally, in (59c), the sentence is fine because the presence of a dummy element, in this case the expletive
it, enables the predicate to discharge its features.
Regular CPs headed by the complementiser dat lack this kind of temporal operator in their specifier, and I would like to propose that this is why they are not able to receive case. On the assumption that P is a case assigning category, it follows that regular CPs cannot appear in the complement position of a P.
This analysis raises the question of why regular CPs seem to be able to occur in the complement position of a VP, since V is also a case assigning category and syntactic saturation must take place. I will come back to this issue in §4.3 where I will discuss a proposal by Barbiers (2000). Barbiers argues that DPs and CPs occupy different base positions in the syntactic structure because they bear a different
semantic relation to the verb. This difference in semantic status leads to the claim
that DPs are arguments whereas CPs are (mostly) predicates. In this thesis, I will remain agnostic about whether case is always assigned to adjuncts, and the level of interaction between case theory and theta theory.
So far, I have argued that two kinds of CPs in the complement position of P can be distinguished: (1) CPs that are in fact complex DPs. This holds for FRs and hoe- clauses, (2) CPs that have a temporal operator in their specifier, which can absorb the case feature that is assigned by the P. In the remainder of this chapter, I will focus on P+CP constructions that are part of a verbal collocation.