The analysis of the preposition mè in terms of an iT feature would gain further support if there is cross-linguistic evidence for prepositional elements with temporal, aspectual or modal properties. In this section I will show that such evidence can indeed be found.
The first point to note is that tense is not a unique property of verbs. In Irish, for instance, tense is also a property of complementisers. (80a) contains the uninflected complementiser go; in (80b) the complementiser gur is inflected for past tense: (80) a. Deir sé go dtógfaidh sé an peann.
say-PRES he that take-FUT he the pen
‘He says that he will take the pen.’
b. Deir sé gur thóg sé an peann.
say-PRES he that-PAST take-FUT he the pen
‘He says that he took the pen.’ (Cottell 1995:108)
In Malagasy, an Austronesian language of Madagascar, prepositions are inflected for past and non-past. Past is marked with t-; non-past is unmarked. The prefix t- attaches to a closed class of elements, which Pearson (forthc.) refers to as “oblique phrases”. These elements include locative, instrumental and manner PPs, spatial deictics (i.e. words equivalent to ‘here’ and ‘there’), and the operator ‘where’. In (81) the phrase ao anatin’ny ala functions as a matrix predicate. In (81a), the deictic element ao is unmarked, and the sentence receives a present tense interpretation; in
(81b) the deictic element contains the t-prefix, and the sentence receives a past tense interpretation:
(81) a. Ao anatin’ny ala ny gidro.
there inside of-DET woods DET lemur
‘The lemur is in the woods.’
b. Tao anatin’ny ala ny gidro.
PAST-there inside of-DET woods DET lemur
‘The lemur was in the woods.’ (Pearson forthc.:2)
Consider next (82), where t- has been combined with an argument PP:
(82) Napetrako tamin’ny latabatra sy tamin’ny seza ny
boky.
PAST-put1SG PAST-on-DET table and PAST-on-DET chair DET
books
‘I put the books on the table and on the chair.’ (Pearson forthc.:7)
This last example therefore demonstrates that prepositions are one category that can be inflected for tense.
The cross-linguistic data discussed above show that the dialect of Wambeek is not unique in having a preposition with a tense feature. Needless to say, there are many differences between the languages that I have discussed, but the general picture is nevertheless clear: prepositions can carry tense features.
There is also evidence for the relation between prepositions and tense in Dutch. Barbiers (2002) argues that, depending on the context, the preposition van (‘of’) can have an iT feature. To appreciate this point, consider first of all the verbal structure in (83), taken from Barbiers (2002:7), where the DP argument and the CP argument occupy different structural positions:
(83) [VP DP [V’ [V thinks [CP ]]]]
This difference is based on the observation that a DP and a CP complement of a verb like think are in complementary distribution (see also §4.3, where I will discuss the approach of Barbiers 2000). If the complement position (i.e. CP) is filled, then the specifier position (i.e. DP) can be filled only if there is an expletive (84a) or the negation niet, (84b):
(84) a. Ik betreur het dat Jan gaat verhuizen.
I regret it that John goes move-house ‘I regret that John is moving house.’
b. Ik denk niet dat Jan komt. I think not that John comes ‘I don’t think that John will come.’
The complementary distribution of DP and CP complements follows from theta theory. The verb think assigns only one internal theta-role. If this role is assigned to a DP argument, then another CP argument would remain theta-less and thus violate the projection principle (or vice versa). The expletive and negative XP in (84a,b) do not require a theta-role, and can therefore be combined with a CP complement, which in turn receives the internal theta-role.
As regards case, Barbiers follows Pesetsky & Torrego (2001), who argue that case is an uninterpretable tense feature uT. Barbiers further assumes that argument status implies presence of an (un)interpretable T feature, a theta-role, or both. This predicts four types of arguments, which Barbiers (2002:8) interprets as follows:
(85) a. Arguments with an uT-feature and a theta-role : DP arguments b. Arguments with an uT-feature and no theta-role : expletives/negation c. Arguments with an iT-feature and a theta-role : CP arguments d. Arguments with an iT-feature and no theta-role : root CPs
With this in mind, consider next (86), where the PP [van niet] occupies the internal argument position of the verb denk:
(86) Ik denk van niet (*dat Jan komt).
I think of not that John comes ‘I don’t think so.’
(Barbiers 2002:9)
Note that the PP is in complementary distribution with the CP complement [dat Jan
komt]. The question, then, is why the negation element niet can be in an argument
position when van is present. As (85b) shows, Barbiers analyzes negation as having an uT feature but lacking a theta-role. This means that the uT feature on niet must be deleted and, furthermore, that niet should not receive a theta-role. According to Barbiers, this entails (1) that van must have an iT feature, and (2) that van does not
assign a theta-role. Rather, Barbiers argues, it is the entire constituent [PP van niet]
which receives a theta-role from the verb. This leads Barbiers (2002:10) to propose the following lexical specification for van:
(87) van Feature=iT
Argument structure: assigns no internal theta-role
Note that the lexical specification of van is very similar to the properties of mè in the
with-infinitive. The observation that the subject in the with-infinitive has nominative
the infinitive, and by doing so assigns nominative Case to it (see §3.4.1). Note, too, that mè does not assign a theta-role to the subject of the with-infinitive. Rather, the subject receives this theta-role from the infinitival verb.
Barbiers provides three arguments in support of the specification of van. First, van is incompatible with embedded clauses. The reason is that such clauses require a theta-role, which van is unable to assign. Declarative root clauses, on the other hand, can function as the complement of van, since such clauses do not require a theta- role. Compare (88a) with (88b):
(88) a. * Ik denk [van [ dat je morgen moet stoppen.]]
I think of that you tomorrow must stop ‘I think that you should stop tomorrow.’
b. Dan denk ik [van [ ik stop vandaag]] then think I of I stop today ‘I think I should call it a day.’ (Barbiers: 2002:10)
Second, when van introduces a temporal adjunct it does so obligatorily, and it gives this adjunct a specific temporal reference. This is shown in (89a,b) for the temporal adjunct vanavond (‘of-evening’), which contains van. When this adjunct is used, it must refer to the evening of the day of the utterance. This means that vanavond is incompatible with an adverb which presupposes that there are more evenings, e.g.
altijd (‘always’). The examples in (89c,d) illustrate that van contrasts in this respect
with a preposition such as in (‘in’). The latter lacks an iT feature, which we can infer from the fact that in is compatible with altijd:
(89) a. We gaan *(van)avond.
we go of evening ‘We’ll go this evening.’
b. * We gaan altijd vanavond.
we go always of evening
‘We’ll always go this evening.’
c. We gaan in de avond.
we go in the evening ‘We’ll go in the evening.’
d. We gaan altijd in de avond.
we go always in the evening
‘We always go in the evening.’ (Barbiers 2002:11)
Third, van introduces a temporally opaque domain, which strongly suggests that van has an iT feature (see also Barbiers 1995). The adverb gisteren (‘yesterday’), when
part of a DP, forces past tense on the finite verb. However, when van is present, the verb can also appear with present tense (see also §2.7.2):
(90) a. Die jongen gisteren vertelde/*vertelt een goed verhaal.
that boy yesterday told tells a good story
‘That boy yesterday told a good story.’
b. Die jongen van gisteren vertelde/vertelt een goed verhaal.
that boy of yesterday told tells a good story
(Barbiers 2002:12)
Barbiers’ claim that van has an iT feature provides indirect support for my analysis of the with-infinitive, which rests on the assumption that the preposition mè has an iT feature in its lexical specification.
There is no obvious reason why the presence of an iT feature should be limited to just the two prepositions van and mè. Rather, we expect other prepositions to allow this option as well. Consider in this light omdat (‘because’). Omdat consists of the prepositional element om (‘for’), and the complementiser dat (‘that’). Traditionally,
omdat is analysed as a (complex) complementiser. However, there is some evidence
that omdat is more appropriately viewed as a (complex) tensed preposition. Note first of all that omdat, though morphologically complex, functions as a syntactic atom, since it cannot be coordinated with the -dat part only:
(91) Omdat Jan geen zin heeft en *(om)dat hij geen tijd heeft.
because John no sense- has and because he no time has
‘Because John doesn’t feel like it and because he doesn’t have time.’ Note, too, that dat introduces finite complements only:
(92) a. Ik weet dat Jan geen boeken leest.
I know that John no books read-3-SG-PRES
‘I know that John doesn’t read any books’
b. * Ik weet dat Jan geen boeken lezen.
I know that John no books read-INF
These observations could be taken to suggest that omdat is another preposition that has an iT feature. The difference with van and met would then be that the tense specification of omdat is “spelled-out” by the complementiser part dat, whereas the iT feature of van and met does not have a separate morphological spell-out. I leave this issue for further research.