The Syntactic Structure of PPs∗
Petr Biskup
Universität Leipzig
This paper is concerned with the syntactic structure of prepositional phrases in Czech.
Investigating morphological, case and referential properties of different types of PPs, I argue
that PPs have a rich internal structure. I propose that in addition to the standardly assumed
locative and directional projection and in addition to degree adjuncts, PPs contain
T(ense)-head, which bears a valued T-feature and unvalued ϕ-features and which is
responsible for the case assignment. I propose that syntactically, case on the prepositional
complement is a reflection of the Agree operation between T-features and ϕ-features on the
complement and T-head and semantically, it is a reflection of semantic features of the
decomposed preposition. The prepositional complement can be overt as well as covert. In the
case of the covert prepositional complement, case is spelled out on the closest overt element
in PP. I show that the prepositional complement is typically expressed covertly in adverbial
PPs. I also discuss referential properties of the covert noun and argue that they cannot be
determined in the narrow syntax.
1. Data
1.1 Prepositional merger
A closer look at prepositions reveals that they can merge with different categories.
Concretely, they can merge with a preposition, as shown in (1a), with a noun phrase, as in
(1b), with an adverbial whP, as in (1c), (1g), with a verb (1d), an adjective (1e) and with a
clause, as in (1f). Traditionally, lexical items like (1a) or (1c) are called adverbials, items like
(1f) conjunctions and items like (1d) prefixed verbs.
(1) a. do-před-u b. do Prah-y c. do-kdy d. do-jít
to-in.front.of-gen.sg to Prague-gen.sg to-when to-go
‘forward’ ‘to Prague’ ‘till when’ ‘come’
e. do-modr-a f. přes-to-že [TP …] g. z-kam-a
to-blue-gen.sg over-the-that out-where-gen.sg
‘a little blue’ ‘although’ ‘from where’
There are also complex prepositions, as demonstrated by the bimorphemic examples (2a) and
(2b) and the trimorphemic examples (2c) and (2d).
(2) a. z-pod b. ze-za c. z-po-za d. z-po-nad
out-under out-behind out-along-behind out-along-above
‘from under’ ‘from behind’ ‘from behind’ ‘from above’
∗
I would like to thank the audience at the CFG conference for helpful comments and suggestions. Particular
thanks go to the non-anonymous reviewer Jakub Dotlačil.
1
1.2 Prepositions and case
As shown by the following examples, prepositions can assign case to different categories. The
preposition do assigns genitive to the preposition před (3a), ku assigns dative to před (3b) and
na assigns accusative to před (3c). In (3e) do assigns genitive to the adjective modr and in (3f)
na assigns accusative to modr. Do can also assign genitive to the noun in (3d) and the
preposition z assigns genitive to the deictic adverb tam in (3g).
(3) a. do-před-u b. ku-před-u c. na-před
to-in.front.of-gen.sg toward-in.front.of-dat.sg on-in.front.of.acc.sg
‘forward’ ‘forward’ ‘ahead’
d. do Prah-y e. do-modr-a f. na-modr-o g. z-tam-a
to Prague-gen.sg to-blue-gen.sg to-blue-acc.sg out-there-gen.sg
‘to Prague’ ‘a little blue’ ‘become blue’ ‘from there’
It is a well-known fact that certain prepositions assign one case, as shown in (4), and other
prepositions assign more cases, as demonstrated by the accusative/locative prepositions in (5)
and by the accusative/instrumental prepositions in (6).
(4) a. do dom-u b. k dom-u c. při dom-ě
to house-gen.m.sg toward house-dat.m.sg at house-loc.m.sg
d. přes dům e. z dom-u
across house-acc.m.sg out house-gen.m.sg
(5) a. na / o dom-ě b. na / o dům
on about house-loc.m.sg on about house.acc.m.sg
(6) a. mezi / nad / pod / před / za aut-y
between above under in front of behind car-inst.n.pl
b. mezi / nad / pod / před / za aut-a
between above under in front of behind car-acc.n.pl
The cases in examples (5) and (6) express the difference between the stative meaning
(location) and the dynamic meaning (direction or path).1
More concretely, locative and
instrumental express the locative meaning and accusative expresses the directional meaning.
This is evidenced by (in)compatibility of particular Ps with stative verbs. For instance, only
the instrumental Ps in (6a), and not the accusative Ps in (6b), are compatible with the stative
predicate stál, as shown by (7).
1
The preposition o cannot have the locative meaning (only the directional). The locative-assigning o could
probably be described with a more abstract feature like stativity.
2
(7) a. Stál mezi / nad / pod / před / za aut-y
stood between above under in front of behind car-inst.n.pl
‘He stood between/above/under/in front of/behind cars.’
b. * Stál mezi / nad / pod / před / za aut-a
stood between above under in front of behind car-acc.n.pl
2. Analysis
2.1 Decomposition of PPs into DirP and LocP
I begin with the difference between directional and locative prepositions. It has been argued
that directional (dynamic) PPs are more complex than the locative ones (Jackendoff 1983,
Bierwisch 1988, Kracht 2002, van Riemsdijk & Huybregts 2002, Zhang 2002, den Dikken
2006, among others). For instance, according to Jackendoff (1983:164), directionals have
meaning (8c), where the PATH-FUNCTION takes the meaning of the locative preposition
(8a) as its argument. Consequently, the directional PP from on the table is represented as (8d).
(8) a. [Place PLACE-FUNCTION ([THING])]
b. [Place ON ([Thing TABLE])]
c. [Path PATH-FUNCTION ([Place PLACE-FUNCTION ([THING])])]
d. [Path FROM ([Place ON ([Thing TABLE])])]
According to Bierwisch (1988:34), the locative in can be represented as (9a), where the region
of the external argument x is included in the region of the internal argument y. The directional
in is again more complex and contains the path function FIN. Hence, the region of the end of
the path of the external argument x is included in the region of y.
(9) a. Locative in: /in/; [-N, -V, -Dir]; λy λx [ LOC x ⊆ LOC y ]
b. Directional in: /in/; [-N, -V, +Dir]; λy λx [ FIN [ LOC x ] ⊆ LOC y ]
Kracht (2002:159) argues that locative expressions universally consist of two layers. The first
layer relates to the location and the other one to the motion. The first layer is called localiser
(L) and describes the way in which objects are positioned wrt. each other. The other one is
called modaliser (M) and describes the way in which an object moves wrt. the given
configuration; consider (10).
(10) [ M [ L DP ]]
In what follows, I present several empirical arguments for the decomposition of PPs and for
the claim that DirPs are more complex than LocPs. The first argument comes from
3
morphologically complex Ps. As one can observe in (2), there are complex directional Ps
containing a locative P.2
In contrast, there are no complex locative Ps containing a directional
P. Similarly, directional (dynamic) wh-adverbs can be derived from locative (stative)
wh-adverbs, as shown for temporal adverbs in example (11).3
However, there are no examples
of locative wh-adverbs derived from directional wh-adverbs.
(11) a. kdy b. do-kdy c. od-kdy
when to-when from-when
‘when’ ‘till when’ ‘from when’
As shown in (3a-c) and (12a), the directional adverbial PPs are also derived from locative
PPs. One might object that this argument is problematic because před has both the locative
and the directional meaning. However, the Russian preposition pered has only the locative
meaning (12b) and still can serve as the base of the directional PPs, as shown in (12c,d).
(12) a. v-před b. pered c. v-perёd d. s-pered-i
in-in.front.of-acc.sg in front of in-in.front.of-acc from-in.front.of-gen
‘forward’ ‘in front of’ ‘forward’ ‘from the front’
The next argument is based on Bošković (2004). He discusses floating quantifiers in PPs and
argues that PPs have a layered structure similar to CP. If Bošković is right, then in example
(13), which is a modification of Bošković’s examples, the preposition s is head-moved and
těmi studenty is XP-moved across všemi, which means that Czech PPs can also be
decomposed.4
(13) [S těmi studenty všemi] jsem mluvil o politice.
with the students all am talked about politics.
‘Pavel talked with all the students about politics.’
Having said this, I decompose PPs into the Dir(ectional)P and Loc(ative)P, as shown in (14).
This means that in complex Ps like e.g. (2a) and (2b) the left morpheme spells out the head
Dir and the right one the head Loc. Note that they cannot be reversed because z(e) has only
the directional meaning. This is supported by the contrast between (11) (and other complex
2
Although pod, za, nad are ambiguous between the locative and directional meaning, in the complex directional
Ps in (2), their locative meaning is used. For po, see below.
3
For locative wh-adverbs, compare e.g. the German example (i).
(i) a. wo b.wo-hin c. wo-her
where where-there where-here
‘where’ ‘where’ ‘from where’
4
PPs) and (i) in note 2. In Czech, which is a VO language, Dir precedes Loc and in German,
which is an OV language, Dir follows Loc.
(14) [DirP Dir [LocP Loc [DP N ]]]
It seems that a more fine-grained decomposition is necessary; recall the trimorphemic Ps
zpoza, zponad in example (2). The question arises what the morpheme po expresses. It is
known that in the verbal domain po works as a delimiter; see example (15a) and Filip (2000,
2003), Součková (2004) and Ziková & Dočekal (2007) for discussion of the extensive
measure function of po. Thus, the meaning of zpoza can be paraphrased as ‘from the place
(region) that is located a little behind the Ground argument’. For some speakers, the measure
meaning of po is not evident in examples (2c,d) but there are PPs in which the meaning is
clearly present. Consider e.g. the adverbial PPs in (15b), where po occurring between the two
directionals means short, and the adverbial in (15c), where po can be paraphrased as a
delimiter kousek ‘piece’.
(15) a. po-zpívat si b. z-po-vz-dál-í c. o-po-dál (= o kousek dál)
along-to.sing self.dat out-along-up-distance-gen.sg about-along-distance.acc.sg
‘to sing for a while’ ‘from a short distance’ ‘at a short distance’
The data suggest that there is a degree or quantificational phrase between DirP and LocP. This
seems to be supported by example (16), where the quantificational morphemes ně, ni and k
follow the directional od.5
However, there are also adverbial PPs where po certainly has the
measure meaning and precedes the directional P, as shown by the translations and the
ordering of po wrt. the directional P z in example (17).
(16) a. od-ně-kud b. od-ni-kud c. z-kam-a
from-some-which.way from-no-which.way out-where-gen.sg
‘from some place’ ‘from nowhere’ ‘from where’
(17) a. po-z-voln-a b. po-z-ne-náhl-u
along-out-free-gen.sg along-out-neg-sudden-gen.sg
‘little by little’ ‘little by little’
Thus, there are two basic possibilities. Po is either a degree/quantificational head in the
extended projection of P that can be located above Dir, between Dir and Loc as well as
4
The fact that s těmi studenty všemi precedes the second-position clitic jsem shows that the PP form a
constituent.
5
As we will see in section 2.3.2, k-a-m can also be decomposed. The wh-morpheme k- can be treated as an
existential quantifier.
5
between two Dirs, or it is a degree/quantificational phrase that can be adjoined to Dir as well
as to Loc. I favour the second possibility and analyze po with the measure meaning as
Deg(ree)P that can adjoin to different projections in PP.
The first reason for this is that the delimitative po, in contrast to other Ps, does not
participate in case assigning processes.6
This holds for cases where the measure po is a lower
preposition in PP, as in (2c,d), where zpoza and zponad (in fact, the preposition z) assign
genitive to their complement (note that po assigns locative and accusative) as well as for cases
where po is the highest preposition, as in (17a), where the genitive case is also assigned by z.7
The next reason is that we would have to assume more quantificational heads in the extended
projection of P because there are PPs with several quantificational morphemes, as e.g. (17b),
where the directional z intervenes between the delimitative po and the negative morpheme
ne.8
In addition, po can appear in different positions in the syntactic structure of PP, hence the
adjunct analysis seems to be more appropriate than the head analysis, which would have to
assume different positions for the head po in the extended projection of P.
2.2 The prepositional case and the head T
Prepositions standardly assign case to their complement. A question immediately arises how
it works in decomposed PPs and what the difference between Ps assigning one case and Ps
assigning more cases is. A closer look at data shows that case is determined by the highest P.
For instance, the complex prepositions in example (2) assign genitive because the directional
z(e), which assigns genitive, is higher than (precedes) the other prepositions in the appropriate
PPs. In the case of adverbial PPs like e.g. (3a-c), case is also determined by the higher P, i.e.
by the directional do, ku, na, and not by the lower P před, which assigns the accusative and
instrumental case. As already mentioned above wrt. examples (5) and (6), when a simple P
has the locative meaning, i.e. only the head Loc projects, it assigns the locative case.
However, when the P has the directional meaning, i.e. the head Dir projects as well, than it
assigns the directional case because the head Dir is higher than Loc.
6
For more details, see the next section.
7
The other relevant examples are not conclusive either because of identity of the assigned case (15c) or because
of the case syncretism (15b), (17b). Consider also example (i), with the Russian complex prepositions assigning
the instrumental case, which shows that po does not affect case (Czech counterparts ponad and poza assign
instrumental and accusative).
(i) a. po-nad bereg-om b.po-za les-om
along-above bank-inst.sg along-behind wood-inst.sg
‘above the bank’ ‘behind the wood’
8
Another example is e.g. po-ně-kud ‘in some degree’ with the delimitative po and the existential morpheme ně.
6
This means that the case-assigning head should somehow know whether or not the head
Dir projects. It seems that Loc and Dir cannot be the case-assigning heads because in such a
case Dir should assign case exactly when Loc does not assign case and Loc should assign case
when Dir does not project. Given the derivational point of view, the problem with this
dependency is that Loc does not know whether or not Dir will be merged in the derivation,
hence it does not know whether or not it shall assign case. A related problem is that it is not
clear why in certain cases Loc could assign case and in others could not.
To avoid this complication, one could propose that the case-assigning ability of Loc is
connected to the presence of certain features on Loc (e.g. to unvalued uninterpretable ϕfeatures,
as in the case of structural cases) and that the features are optional. However, this
proposal also cannot solve the dependency problem, in this case the dependency between the
presence/absence of ϕ-features on Loc and the presence/absence of Dir (more specifically, the
presence/absence of ϕ-features on Dir, under the assumption that Dir assigns case in the same
way as Loc). Thus, it could e.g. happen that both Loc and Dir bear unvalued uninterpretable
ϕ-features, which would lead to a crashing derivation because ϕ-features on Dir could not be
valued because the case feature on the prepositional complement is already valued by Loc and
the argument is no more active. For all possible scenarios wrt. the presence/absence of Loc,
Dir and their ϕ-features in PP, see the following table.
a. no Dir b. Dir without ϕ-features c. Dir with ϕ-features
1. no Loc not interesting * *
2. Loc without ϕ-features * * OK
3. Loc with ϕ-features OK * *
The two OK cells pose the dependency problem. Cases 1b and 1c are ungrammatical because
the presence of the directional meaning (Dir) presupposes the presence of the locative
meaning (Loc). Cases 2a and 2b violate the Case Filter because the prepositional complement
does not bear a case. Case 3b is bad because the appropriate P would have the directional
meaning but the prepositional complement would bear a locative case. 3c is bad because ϕfeatures
on Dir would be unvalued.
Given the dependency complication, I propose that case is assigned by a higher head,
which can see all the relevant information. More specifically, I follow Biskup (2009), who
proposes that all cases are an unvalued T(ense)-feature on the head D and that prepositions
bear unvalued ϕ-features and a valued T-feature. This is an extension of Pesetsky & Torrego
(2004, 2006), who propose that structural case is an unvalued T-feature on D that is valued by
7
T and T0 and that prepositions bear a valued T-feature. Biskup’s proposal has the advantage
that all cases are treated uniformly as Agree between T-features and ϕ-features of the probe
and goal and that the T-feature on Ps can account for the relation between the prepositional
case, the morphological aspect and the perfective structural accusative. Thus, I propose that
analogously to the verbal domain, there is a head T in PP (called TP) that is placed higher than
Dir and Loc, as demonstrated in (18).
(18) [TpP TP [DirP Dir [LocP Loc [DP N ]]]]
As demonstrated by examples (4)-(6), the directional and locative case is not identical for all
prepositions. The case assigned to the prepositional complement is always determined by the
particular preposition, e.g. the directional do assigns genitive and the directional k assigns
dative. This means that the head TP must somehow get the information which case it shall
assign. This is ensured by incorporation of Loc (and Dir if it projects) into TP; recall that
preposition incorporation is possible in cases like (13).9
There are also cases like e.g. the
prefixed verb in (1d) that show that prepositions can even incorporate into a higher
category.10
As already discussed above, case is a reflection of the operation Agree between
the valued T-feature and unvalued ϕ-features on P, i.e. on TP, and the unvalued T-feature and
valued ϕ-features on the prepositional complement. Then, the whole process in a locative PP
looks like (19).
Agree
(19) [TpP TP-Loc (unvalφ-Fs, valT-F) [LocP Loc [ DP(valφ-Fs, unvalT-F) ]]]
incorporation
That there is an Agree relation between P (i.e. TP) and the prepositional complement is
supported by the fact that there are languages with P agreement, e.g., Hungarian, Irish, Welsh,
Jacaltec, Abaza; see Asbury et al. (2007), Baker (2008), Brennan (2008).
To be more concrete, let us take e.g. the preposition za. Za assigns instrumental as the
locative case and accusative as the goal directional case. The first possibility would be to treat
the relation between a particular P and case as idiosyncratic. Such information is stored in the
9
One might propose that TP can get the information via the selection relation between TP and Loc (Dir).
However, if it is true that selection holds only between sisters, then this proposal might be problematic in the
case of more articulated PPs. Another proposal could be based on Agree between TP and Loc (Dir) but it is not
clear what feature should be involved in this relation and how to solve the dependency between the type of the
feature on TP and the presence/absence of Dir in the derivation.
8
lexicon. Then, the derivation with the locative preposition za would look as follows. Za
merges as Loc with the prepositional complement and LocP merges with the head TP because
there is no Dir in this case. Then, za incorporates into TP. Since there is no Dir incorporated
into TP, the incorporated za would have to switch on the instrumental case on the head TP.
As for the directional za, the derivation would work similarly. What is different is that
Dir projects and za incorporates into Dir before it incorporates into TP. Since Dir incorporates
into TP as well, TP would have to assign the directional case (accusative). However, it is not
clear how it is ensured that TP assigns the directional case when Loc also incorporates into TP.
The same complication arises in the case of the complex P zeza, where the incorporated Dir ze
would have to somehow switch on the directional case (genitive) on TP.
There is another possibility, which does not suffer from the problem above and is
conceptually superior to the first possibility because the relation between Ps and their case(s)
is not idiosyncratic there. This analysis is in the Jakobsonian tradition and treats prepositional
cases as a reflection of semantic features of particular Ps. Concretely, I propose that semantic
features of Loc and Dir incorporated into TP represent the value of the T-feature. I assume that
the T-feature can have more values. The value (or values) then is copied on the prepositional
complement by the Agree operation and is spelled out as a certain case at PF. On might ask
what the T-feature has to do with prepositional semantic features. As we will see below, there
is indeed a relation between the type of the prepositional semantic feature and temporal
properties. Given the fact that all source prepositions assign genitive in Czech, I propose that
if Dir with the source meaning - i.e. with the source-feature - incorporates into TP, then TP
values the unvalued T-feature on the prepositional complement as [source]. Given the
following case rule, the prepositional complement then is spelled out with the genitive case:
[source] → genitive.11
Consequently, we expect an appropriate source temporal interpretation.
Indeed, e.g. in od 1980 ‘since 1980’, there is an opened temporal interval going from 1980
onwards.
There are also directional Ps of the goal type. They mostly assign accusative. Hence, if
Dir with the goal-feature incorporates into TP, then TP values the unvalued T-feature on the
prepositional complement as [goal], which - given the following case rule [goal] →
accusative - results in the accusative marker on the argument. However, there are also goal Ps
like k and do, which assign dative and genitive, respectively. Do differs from other goal Ps in
10
The complex P-head incorporates into V and v and then interacts with the T(ense)-feature on the aspectual
head. This brings about perfectivity (definiteness of the reference time) and the definiteness effect with the
structural accusative (see Biskup (to appear)).
9
the fact that the Figure argument ends in the Ground argument. Therefore, I assume that do
also bears a contain-feature. Consequently, when Loc and Dir with the contain-feature and the
goal-feature incorporate into TP, TP values the unvalued T-feature on the prepositional
complement as [goal, contain], which given the following rule results in the genitive marker
on the complement: [goal, contain] → genitive. I assume that application of case rules is
determined by the Subset Principle (similarly as insertion of vocabulary items in Distributed
Morphology). This means that the case rule [goal, contain] → genitive is more specific (there
are two values) than the case rule [goal] → accusative (there is only one value). Therefore the
prepositional complement bears genitive, and not accusative. Values [goal, contain] of the Tfeature
should again bring about an appropriate temporal interpretation. As expected, e.g. in
do 1980 ‘till 1980’, there is a temporal interval going to 1980 and included in it. It is obvious
now why this analysis does not suffer from the problem of the first possibility. Since the
prepositional case is a spellout of several different features, it does not play a role how many
heads incorporate into TP.
As for k, I assume that Dir bears an oriented-feature because according to Mluvnice
češtiny (1986:200), in the case of k, the Figure is oriented wrt. the non-oriented Ground. Thus,
TP values the unvalued T-feature on the prepositional complement as [goal, oriented], which
given the following rule results in the dative marker on the complement: [goal, oriented] →
dative.
Since all Ps, more specifically, their meaning, can be decomposed into certain features,
such an analysis can be proposed for other types of Ps as well. In this approach, case is not
determined by particular Ps but rather particular submeanings of Ps, i.e. by heads with
appropriate features incorporated into TP. Such a generalized way is certainly more attractive
than association of Ps with particular cases in the lexicon. Therefore e.g. in the case of the
locative za, the analysis should work in a parallel fashion. Since za assigns instrumental as the
locative case, Loc should bear a feature that specifies the meaning of instrumental locative
prepositions. This can be the projective-feature; see Zwarts & Winter (2000), Gehrke (2008),
Caha (2009).12
Thus, TP values the unvalued T-feature on the prepositional complement as
[projective] and the complement is spelled out with the instrumental marker because of the
following rule: [projective] → instrumental.
11
This rule says that if DP bears the source value, it must receive genitive.
12
Alternatively, as proposed by Emonds (2007), instrumental could be the default case and the other locative
cases (more precisely, Loc of the appropriate Ps) would need a special feature.
10
The directional za is more complex; Dir with the goal-feature projects, hence TP values
the T-feature on the prepositional argument as [goal, projective]. However, there is a conflict
between the case rule [projective] → instrumental and the case rule [goal] → accusative
because both rules (their values) are a subset of the featural set on the prepositional
complement and neither of them is more specific. Therefore I modify the case rule [goal] →
accusative as follows: [goal, x] → accusative. X is a variable over values and this ensures that
the case rule [goal, x] → accusative (with two values) is more specific than [projective] →
instrumental (with one value). Hence the complement receives accusative. Since x is only a
variable, the case rule [goal, x] → accusative is still less specific than e.g. the case rule [goal,
contain] → genitive, hence it cannot block genitive with the goal P do.
In the case of the complex P zeza, Dir (ze) projects and bears the source-feature; hence
TP values the T-feature on the prepositional complement as [source, projective], which gives a
conflict between the case rules [projective] → instrumental and [source] → genitive.
Therefore I modify the case rule [source] → genitive as follows: [source, x] → genitive. This
rule is then more specific than [projective] → instrumental; hence the complement DP
receives genitive, and not instrumental.
Having the incorporation analysis in place, we can come back to the delimitative po. In
the preceding section, we saw that the delimitative po does not affect case and I proposed that
po is DegP adjoined to LocP or DirP. It is a well-known fact that adjuncts have peculiar
properties (e.g. they are not visible for syntactic processes like Binding Condition C);
therefore it has been proposed that they are merged acyclically (Lebeaux 1988), that they are
merged in a third dimension (Åfarli 1997) or that they are merged by the pair-merge
operation, in contrast to other elements (Chomsky 2004). I take a similar point of view here
and assume that features of the delimitative adjunct po, in contrast to features of Loc and Dir,
are not visible for the head TP. This is the reason why the delimitative po cannot affect case
assignment properties of TP.
This analysis predicts that po with a non-adjunct status can affect case. This prediction
is right, as shown by example (20a), where po brings about the manner reading and assigns
locative (of the paradigm hrad ‘castle’). The case is indeed assigned by po, and not by před,
because před assigns either instrumental (20b) or accusative (20c).13
That -u in (20a) is really
13
In (20b) I mean the temporal předem ‘in advance’. Since this adverbial PP is locative (stative) – as shown by
the fact that it can answer the stative question kdy ‘when’, in contrast to the dynamic odkdy ‘from when’ or
dokdy ‘till when’ – Dir is not projected. And since the locative P před assigns instrumental, the instrumental
marker -em is determined by Loc před. This contrasts with the locative path adverbial předem ‘through the
front’, which is directional (dynamic), as shown by the fact that it can answer the dynamic question kudy ‘which
11
the locative case ending is evidenced by the locative PP ve-před-u ‘in the front’, in which v(e)
assigns the locative case.
(20) a. běhání po-před-u b. před-em c. před dům
running along-in.front.of-loc.sg in.front.of-inst.sg in.front.of house.acc
‘forward running’ ‘in advance’ ‘out in front of the house’
2.3 Prepositional complements
2.3.1 Overt complements
Now, I turn to the bottom part of decomposed PPs, concretely, to overt prepositional
complements. We already saw that the complement of Loc can be a bare DP, as in (1b),
repeated here as (21a). The complement can also be expressed by a modified DP, as in (21b).
(21) a. do Prah-y b. do (toho) (velkého) měst-a
to Prague-gen.sg to the big city-gen.sg
‘to Prague’ ‘to the big city’
The prepositional complement can also be expressed by a sentential DP, as demonstrated in
(22). The question arises why the demonstrative pronoun to is necessary in sentences like
(22a); consider also conjunctions in (22b-d).14
The reason probably is not interpretational
because in examples (23a,b) to is optional and its presence does not change the meaning of
the sentence. Rather, to must be present in (22) because of case. The contrast between the
accusative and nominative clause in (23a,b) and the instrumental clause in (23c) suggests that
oblique cases, in contrast to structural ones, must be spelled out overtly in Czech. Since Ps in
(22) assign oblique cases and since clauses do not have a special case marker, their case is
marked by the case ending on the demonstrative to located between P and the
complementizer.
(22) a. [PP Přes[DP -to[CP -že Pavel chodí pozdě]]], Jana ho miluje.
over the that Pavel goes late Jana him loves.
‘Despite the fact that Pavel is always late, Jana loves him.’
b. po-*(tom) [C’ co [TP …]] c. mezi-*(tím) [C’ co [TP …]] d. pro-*(to)-[C’že [TP …]]
after-the what between-the what for-the-that
‘after’ ‘while’ ‘because’
way’ and not the stative question kde ‘where’. Hence the marker -em must be determined by a covert Dir
because the directional před assigns accusative.
14
The complementizer co, originally a declinable wh-pronoun, does not decline.
12
(23) a. Řekl (to), že Pavel odešel.
said the that Pavel left
‘He said that Pavel had left.’
b. (To) Že odešel, naštvalo Jirku.
the that left made angry Jirka
‘That he had left made Jirka angry.’
c. Potrestal Jirku *(tím), že ukradl knihu z jeho obchodu.
punished Jirka the.inst that stole book from his shop
‘He punished Jirka by stealing a book from his shop.’
Another argument for the case analysis comes from co-relatives. The following example
shows that in relative clauses with a predicate assigning structural cases (24a,b), the
resumptive pronoun, which is case marked, is optional with the accusative argument. In
contrast, in relatives with a predicate assigning an oblique case (24c,d), the resumptive is
obligatory when it refers to the oblique argument. Thus, since co does not decline and oblique
cases must be expressed overtly, the resumptive bearing the appropriate oblique case must
appear in the relative clause. The control example (24e) shows that when the declinable whP
které is used instead of co, the resumptive does not have to (in fact, must not) be present.
(24) a. To je ta žena, co miluje Pavl-a.
it is the woman what loves Pavel-acc
‘This is the woman who loves Pavel.’
b. To je ta žena, co (ji) Pavel miluje.
it is the woman what she.acc.f.sg Pavel.nom loves
‘This is the woman who Pavel loves.’
c. To je ta žena, co Pavl-ovi pomohla.
it is the woman what Pavel-dat helped
‘This is the woman who helped Pavel.’
d. To je ta žena, co *(jí) Pavel pomohl.
it is the woman what she.dat.f.sg Pavel helped
‘This is the woman who Pavel helped.’
e. To je ta žena, kter-é Pavel pomohl.
it is the woman who-dat.f.sg Pavel helped
‘This is the woman who Pavel helped.’
13
Thus, we arrive at the syntactic structure of PPs as shown in (25).15
Since covert prepositional
complements are more interesting than the overt ones, I leave the overt complements and in
the next section I turn to the covert complements.
(25) [TpP TP [DirP (DegP) Dir [LocP (DegP) Loc [DP (Modifier) N/CP]]]]
2.3.2 Covert complements
In section 1.2, we saw that prepositions assign case to different overt categories. In this
section, I argue that there is a covert noun complement in PPs like dopředu ‘forward’, napřed
‘ahead’, ztama ‘from there’ and zřídka ‘seldom’. I begin with the type dopředu.
Dopředu type
The first argument for the presence of a covert noun comes from the consistent case behavior
of dopředu PPs. A closer look at (26a-c) (repeated from (3)) and (26d) reveals that PPs of the
dopředu type bear nominal cases of the masculine paradigm hrad ‘castle’. Thus, there must be
some information present in the PPs that ensures that the genitive, dative and accusative case
is always spelled out as a case ending of the paradigm hrad. Note also that if it is correct that
case is a reflection of the Agree operation between ϕ-features (and T-features) of the probe
and goal, there must be an element in the syntactic structure of PP bearing valued
interpretable ϕ-features (of the paradigm hrad).
(26) a. do-před-u
to-in.front.of-gen.sg
‘forward’ d.
nom hrad
gen hrad-u/a
dat hrad-u
acc hrad
loc hrad-u/e
inst hrad-em
b. ku-před-u
toward-in.front.of-dat.sg
‘forward’
c. na-před
on-in.front.of-acc.sg
‘ahead’
There are also prepositional combinations where the noun is visible, as demonstrated by the
following example, where (e)k spells out N. PPs like do-před-u, do-zad-u etc. refer to a
certain place, hence there should be a referential element (e.g. something like Kayne’s (2004)
15
Quantificational morphemes like e.g. the existential ně are not considered here.
14
covert noun PLACE) in such PPs. For instance, předek means ‘the place in the front’ and
dopředu means ‘to the place in the front’.
(27) a. před-ek b. do před-k-u c. na před-ek
in.front.of-N.nom.sg to in.front.of-N-gen.sg on in.front.of-N.acc.sg
‘the front’ ‘to the front’ ‘on the front’
We know that in dopředu PPs, given Agree between TP and the prepositional complement, the
covert N must bear valued interpretable ϕ-features and that it is lexicosemantically deficient
(covert). How is then reference obtained in cases like dopředu?
The verbal head T is related to subject reference – it agrees with the subject – and to the
speech time (Demirdache & Uribe-Etxebarria 1997), i.e. to the speaker. As shown by the
following examples, do in (28) and (29) brings about speaker/subject anchoring. Concretely,
there is no speaker anchoring wrt. the Ground argument with the unprefixed verb in (28a).
However, the prefix do brings about speaker anchoring if no Ground is present (28b).16
The
prefix brings about subject anchoring when the Ground is an anaphor, as shown in (28c). And
if there is an overt Ground PP, the Ground reference emerges, as shown in (28d).
(28) a. Pavel nesl knihy.
Pavel carried books
‘Pavel was carrying books.’
b. Pavel do-nesl knihy.
Pavel to-carried books
‘Pavel brought books (to the speaker).’
c. Pavel si do-nesl knihy.
Pavel self.dat to-carried books
‘Pavel brought books to his place.’
d. Pavel do-nesl knihy do knihovny.
Pavel to-carried books to library
‘Pavel brought books to the library.’
The contrast in (29) shows that do in dopředu, in contrast to ze in zepředu, can bring about
both subject anchoring and speaker anchoring. While in (29a) both readings are possible, in
16
Consider question (i). This question can be answered by (28b). This shows that if a certain place appears in the
context, as tam in (i), that is related to the event of the appropriate sentence ((28b) in our case), then the Ground
is determined by the context place. This place then does not have to coincide with speaker’s place, as in our
example.
(i) Co tam dělal?
what there did
‘What did he (Pavel) do there?’
15
(29b) only speaker anchoring wrt. to the Ground is possible. Since reference of zepředu and
the moving event are related to the same time, then in our world, where people do not split,
Pavel cannot be the moving object (the Figure argument) and at the same time determine
reference of the Ground argument (of the reference object).
(29) a. Pavel šel do-před-u.
Pavel went to-in.front.of-gen.sg
‘Pavel was going forward.’
b. Pavel šel ze-před-u.
Pavel went out-in.front.of-gen.sg
‘Pavel was going from the front.’
Given these facts and Agree between TP and the prepositional complement in PPs, one can
ask whether referential properties of the covert noun in PPs are determined by a connection
between TP and the verbal head T (according to Biskup (2009, to appear), verbal prefixes are
incorporated prepositions, hence there is a covert doPP in (28b)). Such an analysis is not on
the right track because there are cases like (30), with no speaker or subject anchoring.
Sentence (30) can refer to a situation where the speaker stood behind Pavel and Pavel moved
the piece behind his back toward the speaker. This means that in this situation the reference of
dopředu cannot relate to Pavel or to the speaker and the only anchoring we get is the
chessboard anchoring.
(30) Pavel táhl za svými zády figurkou do-před-u.
Pavel moved behind self back piece to-in.front.of-gen.sg
‘Pavel moved the piece forward behind his back.’
Another possibility would be to determine referential properties of the covert noun by
locality, by the closest potential coreferent. However, this proposal must also be refuted
because we find examples where referential properties of the covert noun certainly are not
determined by the closest element. Consider example (31), which can refer to a situation
where Pavel moves having a pawn in his hand and there is no chessboard at hand. In this
situation only subject or speaker anchoring is possible despite of the fact that figurku is the
closest potential coreferent of dopředu. Since there is no chessboard and pawns do not have a
front or back, reference of the covert noun in dopředu cannot be determined by figurku.17
17
In a situation where Pavel is playing chess with big pieces, moving on a huge chessboard, reference of
dopředu can be determined by the pawn. Note that ambiguity of dopředu also poses a problem for the locality
analysis.
16
(31) Pavel nesl figurku do-před-u.
Pavel brought piece to-in.front.of-gen.sg
‘Pavel was carrying the piece forward.’
From this, I conclude that reference of the covert noun in PPs cannot be determined by the
narrow syntax and as is obvious from the examples above, pragmatics must be involved.
Now, let us go back to arguments supporting the presence of the noun in adverbial PPs.
The next argument comes from namísto. As shown in (32), in this preposition, the noun
PLACE is expressed overtly.
(32) a. na-míst-o b. na-míst-o vět-y
on-place-acc.sg on-place-acc.sg clause-gen.sg
‘in place of/instead of’ ‘instead of the clause’
Another argument for the presence of the covert noun is based on case marking. Adverbial
PPs mostly assign genitive (if they can assign case), as demonstrated in (33), compare also
(32b). And it is a well-known fact that genitive is the case of nominal complements.18
(33) ze-zad-u místnost-i
out-behind-gen.sg room-gen.sg
‘from the back of the room’
Some data wrt. decomposed PPs of the dopředu type are summarized in (34). (34), which
linearizes the syntactic structure of appropriate PPs, shows that there is a dependency between
the presence of the full lexical noun (N) and the presence of the complement of N (the only
exception is zezadu). As already mentioned above, although the covert N bears valued
interpretable ϕ-features, it is lexicosemantically deficient. Thus, the ability of adverbial PPs to
select a complement depends on the non-deficient lexicosemantic status of N.
(34)
Dir Loc Loc N case complement of N
do před u
na před
ve před u
před ek místnosti
v zad u
v zad
do zad u
ze zad u místnosti
zad ek místnosti
18
A nice contrast can be found in Russian, see (12b-d). Pered ‘in front of’ assigns instrumental and complex
prepositions v-perёd ‘forward’ and s-pered-i ‘from the front’ assign genitive.
17
Zřídka type
This type of PPs contains a covert noun of the neuter paradigm město ‘city’ with appropriate
ϕ-features. This is obvious from the fact that this type of PPs, similarly as the preceding type,
behaves consistently wrt. case, as shown by paradigm (35e) and examples (35a-d).
(35) a. z-řídk-a e.
out-rare-gen.sg nom měst-o
gen měst-a
dat měst-u
acc měst-o
loc měst-ě/u
inst měst-em
‘seldom’
b. na-řídk-o
on-rare-acc.sg
‘to thinness’
c. řídc-e
rare-loc.sg
‘rarely’
d. řídk-o
rare-nom/acc.sg
‘thinness’
Supporting arguments come from Caha & Medová (2009), who argue that manner adverbs
derived from adjectives like e.g. rychl-e ‘fast’ bear the locative case of the paradigm město
‘city’ and that adverbs like smutn-o ‘sadness’ bear nominative or accusative (they correspond
to (35c) and (35d), respectively). Their arguments are based on the parallel behavior of the
adverbs and resultatives in the active and passive, on the allomorphy of the locative ending ě
and on crosslinguistic data.
The following examples containing ráno ‘morning/in the morning’ demonstrate that
there is indeed a relation between nouns of the paradigm město, as shown in (36a), and
adverbials of the paradigm město (36b).
(36) a. Ráno miluju. b. Pavel přišel ráno.
morning like Pavel.nom came morning
‘I like mornings.’ ‘Pavel came in the morning.’
In adverbial PPs like zřídka, the noun is not visible but there are adverbial PPs in which the
noun of the paradigm město is expressed overtly; consider kol in okolo (37) and kolem (38).
The P o in okolo is directional because it assigns accusative and this is a dynamic case.
Hence, (37a) with the dynamic predicate is grammatical, in contrast to (37b) with the stative
predicate.
18
(37) a. Jan šel o-kol-o aut-a.
Jan went about-circle-acc.sg car-gen.sg
‘Jan was going around the car.’
b. * Jan stojí o-kol-o aut-a.
Jan stands about-circle-acc.sg car-gen.sg
In the PP kol-em, in addition to the noun kol, there is a covert Dir (and Loc) because
prepositionless instrumental expresses the path meaning, as shown in (38c). For this reason,
(38a) with the dynamic šel is grammatical and (38b) with the stative stojí ungrammatical.
(38) a. Jan šel kol-em aut-a. b. * Jan stojí kol-em aut-a.
Jan went circle-inst.sg car-gen.sg Jan stands circle-inst.sg car-gen.sg
‘Jan was going around the car.’
c. Jan šel les-em
Jan went wood-inst
‘Jan was going through the wood.’
It seems that given its path lexicosemantic properties, kolo ‘circle’ can be combined only with
directional Ps. Although o also assigns locative, there is no P o-kol-e in Czech, with the
locative e because locative is a stative case.
Okolo and kolem can function not only as an adverbial but also as a preposition. As
already discussed above, prepositional complements with intrinsic lexical content can take a
complement. In this respect, PPs okolo and kolem behave as expected. Since they contain a
full lexical noun kolo, they can take a genitive complement.
The next argument for the presence of a covert noun is based on Doetjes (1997), who
argues that quantificational adverbs contain nominal material which forms the restrictor of the
tripartite quantificational structure: Q [restrictor noun][nucleus VP]. According to Doetjes, this is
the reason why Q-adverbs cannot combine with nouns. Example (39) shows that this holds for
zřídka, which is a Q-adverb. Zřídka cannot combine with the noun in (39b), only with vP
(39a). However, if zřídka is replaced by the adjective řídká, the phrase becomes grammatical.
(39) a. Zřídka zpívá nahatý. b. * zřídka píseň c. (v rádiu) řídká píseň
seldom sings naked seldom song (in radio) rare song
‘He seldom sings naked.’
The same also holds for the Q-adverb čast-o, which probably contains a covert noun of the
paradigm město as well; consider the ending o. In the Czech často, the noun is not visible but
19
it is expressed overtly e.g. in Russian, as shown by the masculine raz in (41a), and in German,
as shown by the noun mal with the genitive -s in (41b).19
(40) a. Často zpívá doma. b. * často píseň c. častá píseň
often sings at home often song frequent song
‘He often sings at home.’
(41) a. mnogo raz b. oft-mals
many time often-times
‘often’ ‘often’
Table (42) summarizes PPs of the type zřídka and linearizes their syntactic structure. P marks
the covert preposition. The table again shows that the ability of N to select a complement
depends on the presence of intrinsic lexicosemantic features on N.
(42)
Dir Loc A N case complement of N
na řídk o
P řídc e
z řídk a
P kol em auta
o kol o auta
na míst o auta
P větš in ou aut
Ztama type
In this type of PPs, the covert noun of the neuter paradigm město is combined with adverbs,
concretely, with the wh-adverb kam ‘where’ in z-kam-a ‘from where’, and with the deictic
adverb tam ‘there’ in z-tam-a ‘from there’. These PPs bear the genitive ending a of the
paradigm město, which is assigned by z.
Some evidence that there is a noun of the paradigm město comes from Old Czech data.
In Old Czech, there were adverbs sem-o, tam-o (see Rusínová 1984), bearing the ending o,
which is identical with the nominative and accusative case of the paradigm město.
An argument for the presence of a covert noun in wh-adverbs can be found in
crosslinguistic data. As noted by Caha & Medová (2009), Vangsnes (2008) shows that
Scandinavian wh-phrases ‘how’ contain an abstract nominal morpheme, dubbed ‘WAY’.
19
In Czech, there are also instrumental Q-adverbs, as (ia), derived from feminine nouns (ib).
(i) a. většin-ou b.většin-a
most-inst.f.sg most-nom.f.sg
‘mostly’ ‘most (of)’
20
According to Kayne (2004), here and there are demonstratives that modify the empty
noun PLACE which has a null determiner, as shown in (43a). PLACE can be overt in some
dialects of English. The same also holds for Czech, as demonstrated in example (43b). Thus,
the PP ztama may have structure (43c).
(43) a. [DP THAT [NP there PLACE]]
b. tam-to místo
there-that place
c. [TpP TP-Dir-z [DirP Dir-z [LocP z [DP [tam] N-a ]]]]
Alternatively, one can decompose tam into t-a-m and analyze t as a deictic morpheme and m
as the noun N (the place). In a similar fashion, one can decompose e.g. the wh-adverb k-a-m
‘where’, where k is a question morpheme, s-e-m ‘here’, where s is a deictic morpheme
expressing proximity, or on-a-m ‘over there’, where the pronominal on expresses distality.
There are two arguments supporting the view that m spells out the noun in PPs. First, in Old
Czech, m formed m-participles from verbs (Rusínová 1984) and it is well known that
participles exhibit nominal properties. Second, according to Rejzek (2001), the noun pásmo
‘zone/tape’ containing m is possibly derived from the Indo-European verbal morpheme pes-,
pēs- ‘fly/flap’.
How is case spelled out in the case of the covert prepositional complement? We know
that the prepositional complement agrees with TP and that if it is overt, it bears the case
ending. This is illustrated for the proper name Praha in PP do Prahy in example (44a) and for
the nominal morpheme (e)k in PP do předku in (44b). However, in the case of the covert
noun, case cannot be spelled out in this way. I propose that in such a case, case is spelled out
in accordance with the linearized syntactic structure, i.e. on the closest overt element.
Concretely, if there is no modifier in DP, case is spelled out on the preposition because it is
the closest overt element, as shown for PP dopředu by structure (44c). If a modifier is present
in DP, e.g. an adjective, case is spelled out on the modifier, as shown for PP zřídka in (44d).
Nothing changes on it, if the closest overt element is an adverbial modifier, as demonstrated
for ztama in (44e). If tam is decomposed, as suggested above, the spellout will be identical to
the spellout in (44e) because m represents the noun and this is the closest overt element for
the case ending; consider (44f).
(44) a. [TpP TP-Dir2-do1 [DirP t2-t1 [LocP t1 [DP Prah-y ]]]] =(1b)
b. [TpP TP-do2-před1 [DirP t2-t1 [LocP t1 [DP k-u ]]]] =(27b)
21
c. [TpP TP-do2-před1 [DirP t2-t1 [LocP t1 [DP -u ]]]] =(1a)
d. [TpP TP-Dir2-z1 [DirP t2-t1 [LocP t1 [DP [řídk] -a ]]]] =(35a)
e. [TpP TP-Dir2-z1 [DirP t2-t1 [LocP t1 [DP [tam] -a ]]]] =(3g)
f. [TpP TP-Dir2-z1 [DirP t2-t1 [LocP t1 [DP t-a-m-a ]]]] =(3g)
3. Conclusion
In order to accommodate complexity of Czech PPs, I have argued that PPs have a layered
internal structure. Concretely, the structure looks as follows:
[TpP TP (unvalφ-Fs, valT-F) [DirP (DegP) Dir [LocP (DegP) Loc [DP (valφ-Fs, unvalT-F) (Modifier) N/CP]]]]
On the top of the extended projection of P, the prepositional tense head TP is placed. This
head bears a valued T-feature and unvalued ϕ-features. Since the prepositional complement
bears an unvalued T-feature and valued ϕ-features, the prepositional case is a reflection of the
operation Agree between T-features and ϕ-features of the complement and the head TP. The
resulting prepositional case is determined by semantic features of heads incorporated into TP.
I have shown that the prepositional complement can be overt as well as covert and provided a
few arguments for the covert noun in Czech PPs.
References
Åfarli, Tor A. 1997. Dimensions of phrase structure: the representation of sentence
adverbials. Motskrift. Institutt for nordistikk og litteraturvitenskap, NTNU. 91–113.
Asbury, Anna, Berit Gehrke & Veronika Hegedűs. 2007. One Size Fits All: Prefixes,
Particles, Adpositions and Cases as Members of the Category P. In C. Keskin (ed.) UiL
OTS Yearbook 2006. 1–17.
Baker, Mark C. 2008. The Syntax of Agreement and Concord. Cambridge: Cambridge
University Press.
Bierwisch, Manfred. 1988. On the Grammar of Local Prepositions. In M. Bierwisch, W.
Motsch & I. Zimmermann (eds.) Studia grammatica 29. Berlin: Akademie-Verlag. 1–
65.
Biskup, Petr. to appear. P(refixe)s and P(reposition)s. In B. Dvořák (ed.) Proceedings of the
2nd Congress of Slavic Linguistic Society.
Biskup, Petr. 2009. Prefixes as Prepositions and Multiple Cases. In G. Zybatow, U.
Junghanns, D. Lenertová & P. Biskup (eds.) Studies in Formal Slavic Phonology,
Morphology, Syntax, Semantics and Information Structure. Proceedings of FDSL 7,
Leipzig 2007. Frankfurt am Main: Peter Lang. 3-17.
Bošković, Željko. 2004. Object Shift and the Clause/PP Parallelism Hypothesis. In B.
Schmeiser, V. Chand & A. Kelleher (eds.) Proceedings of the 23th West Coast
Conference on Formal Linguistics. Somerville, MA: Cascadilla Press. 101–104.
22
Brennan, Jonathan (2008) Irish Prepositions: Agreement and Impoverishment. In Ch. B.
Chang & H. J. Haynie (eds.) Proceedings of the 26th West Coast Conference on Formal
Linguistics. Somerville, MA: Cascadilla Proceedings Project. 105–113.
Caha, Pavel. 2009. The jungle of the Czech local cases. Ms., handout from Workshop on
Spatial Expressions, University of Tromsø.
Caha, Pavel & Lucie Medová. 2009. Czech Adverbs as Case-marked Adjectives. In G.
Zybatow, U. Junghanns, D. Lenertová & P. Biskup (eds.) Studies in Formal Slavic
Phonology, Morphology, Syntax, Semantics and Information Structure. Proceedings of
FDSL 7, Leipzig 2007. Frankfurt am Main: Peter Lang. 31-42.
Chomsky, Noam. 2004. Beyond Explanatory Adequacy. In A. Belletti (ed.) Structures and
beyond. Oxford, New York: Oxford University Press. 104–131.
Demirdache, Hamida & Miriam Uribe-Etxebarria. 1997. The syntax of temporal relations. A
uniform approach to tense and aspect. In E. Curtis, J. Lyle & G. Webster (eds.)
Proceedings of the 16th West Coast Conference on Formal Linguistics. Stanford: CSLI
Publications. 145–159.
den Dikken, Marcel. 2006. On the functional structure of locative and directional PPs. Ms.,
CUNY.
Doetjes, Jenny. 1997. Quantifiers and Selection. On the distribution of quantifying
expressions in French, Dutch and English. The Hague: Holland Academic Graphics.
Emonds, Joseph E. 2007. Czech Cases and the Syntacticon: Poznámky k, o, okolo, nad něčím
a pro někoho. In M. Dočekal, P. Karlík, & J. Zmrzlíková (eds.) Czech in Generative
Grammar. München: Lincom. 81–103.
Filip, Hana. 2000. The Quantization Puzzle. In C. Tenny & J. Pustejovsky (eds.) Events As
Grammatical Objects. Stanford: CSLI. 39–96.
Filip, Hana. 2003. Prefixes and the Delimitation of Events. Journal of Slavic Linguistics 11:
55–101.
Gehrke, Berit. 2008. Ps in motion. Doctoral dissertation, University of Utrecht.
Jackendoff, Ray. 1983. Semantics and Cognition. Cambridge, Mass.: The MIT Press.
Kayne, Richard S. 2004. Here and there. In C. Leclère, E. Laporte, M. Piot, & M. Silberztein
(eds.) Lexique, Syntaxe et Lexique-Grammaire/Syntax, Lexis & Lexicon-Grammar:
Papers in Honor of Maurice Gross. Amsterdam: John Benjamins. 253–275.
Kracht, Marcus. 2002. On the semantics of locatives. Linguistics and Philosophy 25: 157–
232.
Lebeaux, David. 1988. Language Acquisition and the Form of Grammar. Doctoral
dissertation, University of Massachusetts at Amherst.
Mluvnice češtiny 2 (1986). Praha: Academia.
Pesetsky, David & Ester Torrego. 2004. Tense, Case, and the Nature of Syntactic Categories.
In J. Guéron & J. Lecarme (eds.) The syntax of Time. Cambridge, Mass.: The MIT
Press. 495–539.
Pesetsky, David & Ester Torrego. 2006. Probes, Goals and Syntactic Categories. In Y. Otsu
(ed.) Proceedings of the Seventh Tokyo Conference on Psycholinguistics. Tokyo: Hituzi
Syobo Publishing Company. 25–61.
Rejzek, Jiří. 2001. Český etymologický slovník. Voznice: Leda.
Rusínová, Zdena. 1984. Tvoření staročeských adverbií. Brno: UJEP.
Součková, Kateřina. 2004. Measure prefixes in Czech: Cumulative na- and delimitative po-.
Master thesis, University of Tromsø.
Vangsnes, Øystein Alexander. 2008. Decomposing manner how in colloquial Scandinavian.
Studia Linguistica 62: 119–141.
van Riemsdijk, Henk & Riny Huybregts. 2002. Location and locality. In M. van Oostendorp
& E. Anagnostopoulou (eds.) Progress in Grammar: Articles at the 20th Anniversary of
23
the Comparison of Grammatical Models Group in Tilburg. Amsterdam: Meertens
Instituut. 1–23.
Zhang, Niina Ning. 2002. Movement within a spatial phrase. In H. Cuyckens & G. Radden
(eds.) Perspectives on Prepositions. Tübingen: Max Niemeyer. 47–63.
Ziková, Markéta & Mojmír Dočekal. 2007. Prostorové prefixy v češtině a omezení na
dvojitou prefixaci. Paper presented at the conference Grammar & corpora 2, Liblice u
Prahy, September 25–27 2007.
Zwarts, Joost & Yoad Winter. 2000. Vector Space Semantics: A Model-Theoretic Analysis of
Locative Prepositions. Journal of Logic, Language, and Information 9: 169–211.
24