default case, np-ellipsis, and the theory of control in polish · 2016-08-30 · default case,...
TRANSCRIPT
DOI: 10.1515/qal-2016-0002
ISSN: 2299–8942
QUESTIONS AND ANSWERS IN LINGUISTICS, VOLUME 3, NUMBER 1, 2016, PAGES: 1-26
© 2016 Center for General and Comparative Linguistics, University of Wrocław, Poland
Default case, NP-ellipsis, and the theory of control in Polish*
Patrick Lindert
ABSTRACT The goal of this paper is two-fold. In the first part, I will offer a closer look into the nature of the
instrumental case in Polish. In the literature, the instrumental case has been identified as a lexical, predicational,
and a default case. In this paper, I will review the arguments for these distinctions, and argue that a default usage
of instrumental is empirically not tenable. In the second part, an analysis of obligatory control constructions with
the instrumental and agreeing case on predicates is discussed. It will be proposed that predicates that agree with
their subjects are bare adjectives, whereas instrumental adjectives are situated within a DP with its head noun
being optionally elided. As a last point, I will show how control mechanisms forbid bare adjectives in object
control.
Keywords: case, control, predication, agreement, genitive of negation, nominalization
1 Introduction
The nature of default case is still a matter of debate in the literature. Why languages would
resort to a mechanism that assigns default case is not clear; however, it seems to be agreed on
that such a mechanism exists (Schütze, 2001; McFadden, 2004, 2007; Pesetsky, 2015).
Figuring out what case functions as the default in a certain language has been proven to be a
challenging task. Schütze (2001) proposes a number of syntactic tests that force the default
case to become visible on case bearing elements. His tests have shown that in English the
accusative case is the default.
The goal of this paper is two-fold: First, I want to take a closer look at the distributional
pattern of the instrumental case in Polish. This case has been argued to appear as a lexical
case, a predicative, and, crucially for this paper, a default case. Second, I will apply Schütze’s
default case tests to Polish to independently show that nominative, rather than instrumental, is
the default case in Polish. I will show that the idea of a default instrumental case rests on a
stipulation born theory-internally to explain certain phenomena within Polish control
configurations and I will show that these constructions do not involve a default instrumental
case, but are rather subject to predicative case assignment, thereby unifying two occurrences
of the instrumental case.
* I am grateful to two anonymous reviewers as well as the editors for valuable comments and suggestions. I
would like to thank those in attendance at ConSOLE 24, PLC40, and the conference on ‘Case attraction and
related phenomena’ in Wrocław, 2015, where versions of this paper have been presented. For their insightful
comments I would like to thank Artemis Alexiadou, Elena Anagnostopoulou, Silke Fischer, and Marcel Pitteroff.
All remaining errors are, of course, mine. This work is supported by Deutsche Forschungsgesellschaft (DFG),
grant AL 554/10-1; FI 1959/2-1.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 2
2 The family of case
The rich literature on case has identified (at least) three types of cases, namely structural,
lexical, and inherent (Woolford 1997, 2003a, 2003b, 2006; McGinnis, 1998; Przepiórkowski,
1999; Alexiadou, 2003; Sigurðsson, 2003; Richardson, 2007; Alexiadou, Anagnostopoulou,
and Sevdali, 2014; Baker, 2015; Pesetsky, 2015). Furthermore, the literature identifies
instances of so-called semantic case (Freidin and Babby, 1984; Babby, 1986),1 abstract case
(McFadden, 2004; Legate, 2008)2 and default case (Schütze, 2001), the latter being a major
topic of this paper.
In this paper, I will not distinguish lexical from inherent case and thus they are used
interchangeably.3 The distinction between lexical and structural case is, however, relevant and
will therefore be discussed in more detail. Richardson (2007) states that “the hallmark of
lexical case in the Slavic languages is that it is unpredictable and obligatory” (p. 26). She
notes that nothing about the semantic structure of a verb can give the language learner hints as
to whether it will assign lexical case or not. Apart from unpredictability there is also the
observation that lexical case cannot be overridden by certain syntactic operations, hence it is
obligatory. It seems to be language specific which operations one can apply to tease structural
and lexical cases apart, e.g., passivization seems to be a good test for German; see (1) and (2).
(1) a. Die Maria hilft dem Peter.
the.NOM Mary helps the.DAT Peter
‘Mary helps Peter.’
b. Dem / *Der Peter wird geholfen.
the.DAT / NOM Peter is helped
‘Peter is being helped.’
(2) a. Die Maria küsst den Peter.
the.NOM Mary kisses the.ACC Peter
‘Mary kisses Peter.’
b. *Den / Der Peter wird geküsst.
the.ACC / NOM Peter is kissed
‘Peter is being kissed.’
In (1a) the internal argument of helfen ‘to help’ receives dative. Once the sentence is
passivized, the object’s case is not changed into nominative, but stays dative, as shown in
(1b). A change in structure does not result in a change in case – which is not surprising if we
assume that dative is not assigned structurally, and therefore not sensitive to structural
changes, but already lexically determined by a given lexical element in the clause, here helfen
‘to help’. In (2), the complement of the verb küssen ‘to kiss’ receives accusative; see (2a).
Once the structure is passivized, the case on the object changes to nominative, as illustrated in
1 Semantic case is defined by its contribution to the semantic interpretation of a given clause. Plus, it has been
observed (Babby, 1986) that semantic case can only alternate with structural case, but never with lexical case. I
will not go into the details of semantic case in this paper. 2 There is also the discussion of the nature of morphological case which cuts through the classification of cases.
Morphological case refers to the morphological realization of the above mentioned cases. See McFadden (2004)
for a thorough investigation of the nature of this phenomenon. In this paper, the distinction between
morphological, structural, and abstract case is not important. 3 See Woolford (2006) on how to tease apart lexical from inherent case.
3 PATRICK LINDERT
(2b). This shows that the accusative case is a structural one, as it is sensitive to structural
changes like passivization.
Tests such as passivization are language specific: for example, the passivization test is only of
limited use in Japanese (Woolford, 2006) and Polish (Przepiórkowski, 1999). I will review the
Polish data in light of the structural/lexical case discussion in the following section.
3 Instrumental case in Polish
According to the literature on the case system of Polish, the instrumental case has a unique
position in the grammar. It can be called unique as it can appear as a lexical case (with certain
verbs and prepositions), predicative case (in predication) and default case (in certain control
environments). I will review the data and analyses for these environments in this section and
propose an analysis that can unify the occurrences of predicative and default case, so that the
idea of a stipulative default instrumental must no longer be entertained.
3.1 Instrumental as lexical case
Earlier studies were conducted on the status of the Polish case system. Two of the studies
were done by Willim (1990) and Przepiórkowski (1999) and both arrive at very different
conclusions when it comes to the status of the instrumental case. Consider the data in (3).
(3) a. Jan napisał to ołówk-iem. (Willim, 1990, p. 208)
John.NOM wrote this pencil-INST
‘John wrote this with a pencil.‘
b. Piotr kieruje fabryk-ą.
Peter.NOM leads company-INST
‘Peter runs a company.’
For Willim (1990) the instrumental marking on ołówek ‘pencil’ in (3a) would be lexical,
whereas the instrumental on fabryka ‘company’ in (3b) is structural. The reason for this has to
do with theta-role assignment; she assumes that whenever the instrumental case is associated
with the theta-role source/instrument the occurrence is lexical. In (3a) ołówek ‘pencil’ would
be the bearer of a source (or instrument) theta-role, and therefore this occurrence must be
lexical. All other occurrences with different theta-roles are structural. As the theta-role on
fabryka ‘company’ in (3b) would be theme, and not source, it is structurally assigned. In order
to demonstrate that the distinction between structural and lexical cases boils down to theta-
role assignment, she argues that a sentence like (3b) can actually be passivized and
instrumental marking switches to nominative (see (4)), which would be unexpected if this was
a lexical case.
(4) Fabryka jest kierowana przez Piotr-a.
company.NOM is led by Peter-ACC
‘The company is run by Peter.’
However, Przepiórkowski (1999) already noted that another prediction would also be that an
instrumental marked object should always be passivizable when it is not tied to a source theta-
role. He shows that this prediction is not borne out; cf. (5).
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 4
(5) a. Jan macha chorągiewk-ą. (Przepiórkowski, 1999, p. 102)
Jan.NOM waves banner-INST
‘John is waving a banner.’
b. *Chorągiewka jest machana (przez Jan-a).
banner.NOM is waved (by Jan-ACC)
(Intended meaning: ‘The banner is waved (by John).’4)
The object chorągiewka ‘banner’ is not associated with a source theta-role (but with a theme),
so according to Willim (1990) it must be structural. A structural case object should be easily
passivizable, however for (5a) it is not the case, as illustrated in (5b). The conclusion to be
drawn here is that the status of lexical or structural case is independent of theta-role
assignment.
Przepiórkowski (1999) does not take theta-roles into consideration when discussing the
lexical/structural distinction, but applies the genitive of negation test to tease apart lexical
from structural cases.
(6) a. Lubię Mari-ę. (Przepiórkowski, 1999, pp. 101-102)
I.like Mary-ACC
‘I like Mary.’
b. Nie lubię Mari-i / *Mari-ę.
NEG I.like Mary-GEN / Mary-ACC
‘I don’t like Mary.’
(7) a. Pomogłem Jan-owi.
I.helped Jan-DAT
‘I helped Jan.’
b Nie pomogłem Jan-owi / *Jan-a.
NEG I.helped Jan-DAT / Jan-GEN
‘I didn’t help Jan.’
(8) a. Kieruję firm-ą.
I.lead company-INST
‘I run a company.’
b. Nie kieruję firm-ą / *firm-y.
NEG I.lead company-INST / company-GEN
‘I don’t run a company.’
In (6) we see the application of the genitive of negation (GoN). When a structural accusative
object occurs below sentential negation, it surfaces with genitive case marking; accusative is
not an option, as shown in (6b). This does not happen when the object bears a lexical case like
dative, as in (7); there, negation does not affect the lexical case marking, the object retains
dative case, as demonstrated in (7b), suggesting that dative is lexical. If we put an
instrumental-case marked object (see (8a)) into that environment, we see that it keeps its case
under negation (cf. (8b)), which strongly suggests that instrumental should be treated as
lexical.5
4 Another possible source for the ungrammaticality of (5b) might not be located in the case marking or the theta-
role of the object, but rather in the impossibility of verb machać ‘wave’ to passivize in the first place. 5 For a more detailed analysis of the GoN, the interested reader is referred to Błaszczak (2001, 2007).
5 PATRICK LINDERT
Another argument was made by Przepiórkowski (1999) building on nominalizations. He notes
that structurally case marked objects change into genitive when the predicate is nominalized;
see (9).
(9) a. Janek lubi Mari-ę. (based on Przepiórkowski, 1999, p. 106)
John.NOM likes Mary-ACC
‘John liked Mary.’
b. lubienie Mari-i / *Mari-ę (przez Jan-ka)
liking Mary-GEN / Mary –ACC (by John-ACC)
‘the/John’s liking of Mary’
In (9a) the object of lubić ‘like’ carries accusative. Once the predicate is nominalized, as in
(9b), the object surfaces with genitive marking, accusative marking is impossible.
Przepiórkowski (1999) observes that this does not happen with objects lexically marked for,
e.g., dative; cf. (10).
(10) a. Janek pomaga Tomk-owi. (based on Przepiórkowski, 1999, p. 108)
John.NOM helps Tom-DAT
‘John is helping Tom.’
b. pomaganie Tomk-owi / *Tomk-a (przez Jank-a)
helping Tom-DAT / Tom -GEN (by John-ACC)
‘the/John’s helping of Tom’
If we now put the instrumental case in the environment of nominalizations, the results are as
follows:
(11) a. Janek kieruje fabryk-ą. (Przepiórkowski, 1999, p. 109)
John.NOM leads company-INST
‘John runs a company.’
b. kierowanie fabryk-ą (przez Jank-a)
leading factory-INST (by John-ACC)
‘the/John’s managing a factory’
(12) a. Maria bawi się lalk-ą.6 (Przepiórkowski, 1999, p. 109)
Mary.NOM plays REFL doll-INST
‘Mary plays with a doll.’
b. bawienie się lalk-ą
playing REFL doll-INST
‘playing with a doll’
In (11a) we can observe that the object of kierować ‘lead’ appears in instrumental. If the
predicate is nominalized, as in (11b), the instrumental marking on the noun fabryka
‘company’ remains. We observe the same with the data in (12). In (12a) the inherently
6 Here one could argue that there is actually a covert preposition z (with) that assigns instrumental. However,
such an analysis would not capture the different semantics. Consider the following example.
(i) Piotr bawi się (z) Mari-ą.
Peter.NOM plays REFL (with) Mary-INST
In the version with the preposition z (with) one gets a reciprocal interpretation, i.e., they are both agents. This
forms a comitative construction. Without the preposition, the dominant reading is that Peter actively plays, while
Mary is a theme that undergoes the playing event (just like the doll in (12)).
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 6
reflexive bawić się ‘to play’ governs the instrumental case. When nominalized, as in (12b), its
object keeps the instrumental case marking, strongly suggesting that instrumental is indeed a
lexical case.
In this subsection I have looked at the arguments for and against the classification of
instrumental as lexical case. We have seen that theta-marking does not seem to play a role in
this process (pace Willim, 1990). Furthermore, Przepiórkowski (1999) has shown that
passivization is not a good test as there are examples of lexical cases that can be passivized in
Polish.7 Applying tests like the genitive of negation and nominalizations left us with the
conclusion that these occurrences of the instrumental case are lexical in Polish.
3.2 Instrumental as predicative case
Apart from the constructions discussed in the previous subsection, there are occurrences of a
predicative instrumental in Polish; see (13).
(13) Piotr jest (dobry-m) pilot-em / aktor-em / lekarz-em.
Peter.NOM is (good-INST) pilot-INST / actor-INST / doctor-INST
‘Peter is a good pilot/actor/doctor.’
In predication, it is obligatory to realize the predicated DP in the instrumental case (Bailyn
and Citko, 1999; Geist and Błaszczak, 2000; Bailyn, 2001).8 In addition, all elements within
the DP (like adjectives) that have case morphology must appear in instrumental, i.e., they
agree in case with the noun they modify.
It would be tempting to conclude that the copula być ‘be’ assigns instrumental, thus making it
a clear instance of lexical case assignment. However, it is nouns that receive instrumental,
adjectival predicates appear in nominative; see (14).
(14) Piotr jest miły / dobry / mądry / pijany.
Peter.NOM is nice.NOM / good.NOM / smart.NOM / drunk.NOM
‘Peter is nice / good / smart / drunk.’
If the copula were an instrumental case assigner, then we would expect instrumental on the
adjectives in (14) as well, contrary to the facts. Instead, the adjectives seem to get case via
agreement with the subject. This goes hand in hand with the well-known observation that
adjectives typically agree (usually with a noun), and do not get case assigned directly.
What has been proposed is that a functional projection is involved that establishes the
predication relation and also assigns a predicative case to its complement. This projection has
been called a PredP (Bowers, 1993; Bailyn, 2001; Bondaruk, 2013; Franks, 2015) or PhiP
(Citko, 2008) in the literature.9 The exact nature of that functional projection is not too
relevant for this discussion; what is, is the general assumption that a functional projection is
7 For more Polish data supporting this idea, the reader is referred to Przepiórkowski (1999, pp. 100-104).
8 See Bondaruk (2014) for a different view. See Bogusławski (2001) who discusses some exceptions to this rule.
Moser (1993) gives a historical overview of the development of Slavic nominative/instrumental variations of
predicative nouns. 9 Bailyn (2001) was concerned with the distribution of the instrumental case in Russian, which is not parallel to
the distribution in Polish. He arrives at the conclusion that instrumental case assignment in Russian must always
be an instance of secondary predication.
7 PATRICK LINDERT
responsible for the assignment of instrumental in the same fashion as T is for nominative, v is
for accusative and Appl is for dative.10
It is still not clear whether the occurrence of the instrumental case as a predicative case
qualifies as structural or lexical. The presented tests have, to my knowledge, not been applied
to predication contexts so far. If we apply the tests discussed in the previous section, we get
the following results.
(15) bycie lekarz-em / *lekarz-a Nominalization
being doctor-INST / doctor -GEN
‘being a doctor’
(16) Piotr nie jest lekarz-em / *lekarz-a. Genitive of Negation
Peter.NOM NEG is doctor-INST / doctor-GEN
‘Peter is not a doctor.’
In (15), the verb być ‘be’ is nominalized but the complement lekarz ‘doctor’ does not change
its case into genitive, i.e., it remains instrumental. In (16), the phrase is negated via nie ‘not’
but the case of lekarz ‘doctor’ remains instrumental and does not change to genitive. The
results of these two tests strongly suggest that this instance of instrumental is also lexical. The
question now is, what lexical element assigns lexical case here. One possibility would be the
copula być ‘be’. However, we have also established that the lexical element być ‘be’ is not
responsible for case marking, as adjectives do not receive instrumental in the same
environment from the copula (see (14)).11
Yet, it is conceivable that the instrumental case is
indeed assigned by a functional projection like a PredP, PhiP, or simply an FP, much in the
same fashion as lexical dative can be assigned by a functional applicative head (McFadden,
2004; Pylkkänen, 2008) without losing its lexical case status.
3.3 Instrumental as a default case
Apart from lexical and predicative occurrences of the instrumental case in Polish, it has also
been argued that it functions as a default case (Przepiórkowski and Rosen, 2005; Witkoś,
2008, 2010; Franks, 2015) or elsewhere case (Bondaruk, 2004) in certain control
constructions.12
Some people disallow nominative in (17), but I am reporting Witkoś’
judgments.
(17) Piotr marzy, [żeby być bogaty / bogaty-m]. (Witkoś, 2008, p. 265)
Peter dreams [so-that be.INF rich.NOM / rich-INST]
‘Peter dreams to be rich.’
10
For predicative adjectives, it is assumed that the functional projection that establishes predication is defective
and cannot assign instrumental case (see Bondaruk, 2013). 11
Geist (1999) has argued for Russian that there are two copulae byt ‘be’; one that assigns nominative, and one
that assigns instrumental. She shows that these two elements behave differently w.r.t different phenomena like
scope and focus. 12
There is one more environment where it has been claimed that a default instrumental appears and that is in
non-obligatory control contexts as in (i):
(i) Być miły-m / *miły to być głupi-m / *głupi. (Przepiórkowski, 2004a, p. 106)
be.INF nice-INST / nice.NOM that be.INF stupid-INST / stupid.NOM
‘To be nice is to be stupid.’
In this paper, I am not going to discuss cases of non-obligatory control.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 8
(18) Kazałem mu być trzeźwy-m / *trzeźw-emu. (Witkoś, 2008, p. 257)
I.told him.DAT be.INF sober-INST / sober-DAT
‘I told him to be sober.’
The data in (17) shows obligatory subject control constructions where the predicate bogaty
‘rich’ can either appear in nominative, thus matching the case of the controller, or in
instrumental. In an object control construction in (18) only instrumental is licit on the
adjective, agreement with the dative controller is impossible. Witkoś (2008, 2010) analyses
the cases where instrumental can appear as instances of a default mechanism. He relates that
to the notion of phases, stating that whenever a (strong) phase is completed, no case valuation
by finite T (subject control) or v (object control) is possible in the non-finite clause and thus
default instrumental appears as last resort. In (18) the CP constitutes a strong phase, and
therefore valuation of dative case on the predicate is impossible. As a result, the default case
is assigned. The interesting case is (17), where we get optionality between nominative and
instrumental. Witkoś states that this has to do with the individual speakers’ grammar – if for
them the CP is a strong phase, instrumental becomes licit, for those speakers who allow
agreeing nominative, the CP is a weak phase and finite T can penetrate the phase, and licence
nominative on the predicate. This proposal becomes problematic, however, when a speaker
accepts both forms in (17). My informants and I accept both versions. If both versions are
fine, the proposed analysis of strong and weak phases becomes unsatisfactory. In addition, we
can get optionality in subject control even when there is no overt complementizer; see (19).
These data are analysed by Witkoś as having no CP, and therefore no phase boundary making
it possible for finite T to assign nominative to the embedded predicate in any case.13
This
analysis predicts that instrumental should never be possible there, contrary to fact.
(19) Piotr chce być pijany / pijany-m.
Peter.NOM wants be.INF drunk.NOM / drunk-INST
‘Peter wants to be drunk.’
The data in (19) is taken from Witkoś (2010, p. 193). He states that in an informal study
people accepted both versions of the predicate. In a formal study instrumental was rated lower
in acceptability than nominative. I conducted a small pilot study with ten Polish native
speakers, and all informants reported that the instrumental case marking in examples like (19)
was good. I conclude that both versions are part of the grammar and an account should be
able to explain the optionality.
Based on all the data presented in this section, Witkoś (2010) concludes that we do find three
default cases in Polish; “nominative for nominal projections, dative for semi-predicates and
instrumental for adjectival predicates” (pp. 209-210).14
Nevertheless, the question on how to test or falsify such a generalization remains. Schütze
(2001) offers independent tests for default case environments for nominal expressions to
13
See Błaszczak (2007) who mentions that certain scope interactions are unexpected if there is no CP. 14
I do not discuss semi-predicates in this paper. These predicates can appear in the dative case, where other
predicates appear in instrumental. Sam ‘alone’ constitutes, to my knowledge, the only semi-predicate in Polish.
(i) Piotr zrobił to wczoraj sam / sam-emu. (Witkoś, 2010, p. 210)
Peter.NOM did it yesterday alone.NOM / alone-DAT
‘Peter did it yesterday himself.’
In (i) we see that there is also optionality on the predicate, this time however between agreeing nominative and
dative. The observations that semi-predicates behave differently when it comes to case marking has also been
observed for Russian (Landau, 2008).
9 PATRICK LINDERT
which I will turn to in the next section. Witkoś (2010) uses certain data as a test for default
case for which he also wants to propose an analysis. According to Witkoś, the data in (17)-
(19) show that instrumental is a default case for adjectives, and at the same time these
adjectives appear in instrumental there because it is the default case. This approach seems
circular and descriptive, and it is not clear whether it can be falsified.
In this section I have reviewed the arguments for the different usages of the instrumental case
in Polish. I have argued that the default case usage is hard to falsify and stipulative. In the
next section, I will apply default case tests to Polish to show that nominative, and not
instrumental, behaves like a default case. I will then continue with an analysis for the
obligatory control cases where it will be argued that instrumental marks DPs in predication
and, by extension, that adjectival predicates in instrumental are actually situated in a DP with
their head noun being optionally elided in a process of NP-ellipsis. That means that
instrumental adjectives are never bare adjectives to begin with, but are rather projected as
attributive adjectives of DPs.
4 Default case in Polish
In this section, I will explore the notion of default case in Polish. I am especially interested in
finding and applying independent tests to find out which case acts as the default case in
Polish. There is a rich discussion in the literature exploring the exact nature of default case
(McCloskey, 1985; Chung and McCloskey, 1987; Zwart, 1988; Duffield, 1989; McFadden,
2007) and the question of why exactly natural languages would employ such a mechanism;
the discussion on the latter aspect is still ongoing and I have nothing to add there. First, let me
clarify what a default case is supposed to be. I follow Schütze (2001), who offers the
following definition, which in turn is inspired by Marantz (2000).
(20) DEFAULT CASE
The default case forms of a language are those that are used to spell out nominal
expressions (e.g. DPs) that are not associated with any case feature assigned or
otherwise determined by syntactic mechanisms. (Schütze, 2001, p. 206)
Schütze makes it clear that he assumes default case not to be part of the syntactic component,
i.e., it is basically invisible in the syntax. Therefore, the presence of a default case form on a
nominal element can never save a derivation – no matter whether one assumes that case has
that kind of power in syntax to begin with – as the ingredients for default case are simply not
present in the syntactic structure. This leads to the conclusion that case and licensing must be
dissociated from one another (in the spirit of Marantz, 2000).
Many researchers have been investigating the default case of a certain language; some of the
results are nominative for German (McFadden, 2007), Icelandic (Boeckx and Hornstein,
2006), Dutch (Havranová, 2014), and Russian (Matushansky, 2010),15
accusative for Irish
(Bondaruk, 2004), and dative for Hungarian (Szécsényi, 2002). Often Schütze’s tests were
used in their respective research. For Polish, no thorough investigation has been conducted so
far.
15
Pesetsky (2015) argues that genitive is a default for Russian by stating that nouns seem to be ‘born genitive.’
He bases this on data involving numerals.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 10
In the remaining part of this section, I will show how Schütze came to the conclusion that
accusative is actually the default form in English. I will then apply the tests to Polish to show
that it is nominative that functions as the default, and never the instrumental case in Polish.
Schütze (2001) proposes a number of tests to disclose the default case. I will quickly
demonstrate these for English, before turning to the Polish data in the next section; all the
English data are taken from Schütze (2001). Due to space reasons, I will not go into the
details of how they work. The interested reader is referred to Schütze (2001).
(21) a. Me / *I, I like beans. Left Dislocation
b. The best athlete, her / *she, should win.
(22) a. We can’t eat caviar and him / *he (eat) beans. Gapping
b. For Mary to be the winner and us /?? we the losers is unfair.
(23) a. Us and them / *We and they are gonna rumble tonight. Coordination
b. Her and us / *She and we have been friends for ages.
(24) a. Lucky me / *I gets to clean the toilets. Modified Pronouns
b. The real me / *I is finally emerging.
c. Dear me / *I!
(25) Q: Who wants to try this game? Ellipsis
A: a. Me / *I.
b. Not us / *we.
Schütze (2001) offered a number of syntactic environments where DPs universally surface
with case morphology that does not seem to be assigned by a functional projection during the
syntactic derivation. These environments include left dislocation, gapping, coordination,
modified pronouns, and ellipsis. The conclusion is that the observed case must come from
somewhere else, namely from a last resort mechanism assigning default case. The data in
(21)-(25) suggests that the default case in English is accusative.
Turning to Polish, I will show that in all of the proposed environments only nominative is licit
in Polish, instrumental is out in all of them. I will start with the left dislocation test the
corresponding data is presented in (26).16
16
An anonymous reviewer is concerned whether left dislocation actually exists in Polish at all, citing the
following example as being more natural:
(i) Jeśli chodzi o mnie, to ja lubię fasol-ę.
when goes about me.ACC TO I.NOM like beans-ACC
‘When it comes to me, I like beans.’
I do agree that the data in (i) is more natural than the data in (26). However, the data in (i) cannot be used as a
default case environment as the two pronouns are assigned case syntactically. The accusative case on ja ‘I’,
realized as mnie, is assigned by the preposition o ‘about’. Nominative on ja ‘I’ is presumably assigned by finite
T. The data in (26) is probably absent in formal writing, but examples have been reported (Citko, 2008) and my
informants report that they do not have problems with the data, albeit mention that it is not very formal.
11 PATRICK LINDERT
(26) a. Ja / *mną, ja lubię fasol-ę.
I.NOM / I.INST I.NOM like beans-ACC
‘Me, I like beans.’
b. My / *nami, my idziemy do dom-u.
we.NOM / INST we.NOM go to home-GEN
‘Us, we go home.’
c. Najlepszy sportowiec / *Najlepszy-m sportowc-em,
best.NOM sportsman.NOM / best-INST sportsman-INST
on powinien wygrać.
he.NOM should win.INF
‘The best sportsman, him, should win.’
The dislocated element – Ja ‘I’ in (26a), My ‘we’ in (26b) and najlepszy sportowiec ‘the best
sportsman’ in (26c) – can only surface with nominative case morphology, instrumental is
infelicitous. This is a strong indication for nominative in a default case environment.
However, one could argue that the left dislocated element in Polish must bear the case of its
associate as in the examples in (26) the case on the left-dislocated element mirrors the case of
the subject pronoun. However, we see that this is only superficially so if we consider the data
below.
(27) Jan, nikt go nie lubi. (Citko, 2008, p. 284)
Jan.NOM nobody him.GEN not likes
‘Jan, nobody likes him.’
(28) Gosia, ja ją lubię.
Gosia.NOM I.NOM her.ACC like
‘Gosia, I like her.’
In (27), the dislocated element Jan carries nominative. The corresponding pronoun carries
genitive case morphology.17
We can observe the same effect in (28). This time the
corresponding pronoun ją ‘her’ is clearly accusative, however, the dislocated element carries
nominative (Gosia) and not accusative (Gosię).
Let us now turn to the next test, namely gapping.
(29) a. Nie możemy jeść kawior-u a on / *nim (jeść) fasol-i.
NEG we.can eat.INF caviar-GEN and he.NOM / he.INST (eat.INF ) beans-GEN
‘We cannot eat caviar and him beans.’
b. Maria jest zwycięzcą a on / *nim (jest) przegrany-m.
Mary.NOM is winner-INST and he.NOM / he.INST (is) loser-INST
‘Mary is the winner and he is the loser.’
In gapping contexts, like the one in (29), we can also clearly see nominative markings on the
pronouns. Instrumental is not possible in these environments. The next environment used by
Schütze was coordination. The application of this test to Polish is displayed in (30).
17
Based on morphology, it could also be accusative as in this instance (for male, animate DPs) accusative and
genitive case marking is syncretic. However, negation triggers genitive case on structurally case-marked objects.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 12
(30) a. *Nimi i nami idziemy do dom-u.
they.INST and we.INST go to home-GEN
b. Oni i my idziemy do dom-u.
they.NOM and we.NOM go to home-GEN
‘Them and us go home.’
In (30) two pronouns have been coordinated. In (30a) the pronouns are in the instrumental
case, and the sentence is ungrammatical. In (30b) the pronouns are in nominative and the
result is a licit utterance in Polish. We see again, that nominative behaves like the default case
in Polish. One more environment proposed by Schütze was pronoun modification. Applied to
Polish, it gives the following results.
(31) a. My / *Nami trzy musimy teraz iść.
we.NOM / we.INST three have now go.INF
‘Us three have to go now.’18
b. ?Prawdziwa ja / *prawdziwy-ą mną pojawiła się.
real.NOM I.NOM / real-INST I.INST surfaced REFL
‘The real me surfaced.’
In (31a) a modified pronoun (my trzy ‘us three’) has been inserted in the derivation. As can be
seen, that pronoun must surface with nominative, instrumental marking results in
ungrammaticality. We can also see that in (31b) the pronoun can never appear in instrumental
case, the only possible way to express this sentence in Polish is to have the pronoun – and the
corresponding adjective – in nominative. It should be noted that constructions like (31b) are
not perfect in Polish to begin with, even when they are marked with nominative. However,
there is a clear contrast between the nominative and the instrumental variant, with the latter
being definitely rated worse. Also consider the examples in (32).
(32) a. Prawdziwa ja zaczęła się teraz.
real.NOM I.NOM began REFL now
‘The real me starts now.’
18
An anonymous reviewer points out that numerals in Slavic usually complicate matters, citing the following
example.
(i) Nas / *My pięc-iu musi teraz iść.
we.GEN / we.NOM five-GEN have now go
‘Us five have to go now.’
As is well-known, numerals higher than five assign genitive to their complements, here to my ‘we’ surfacing as
nas. What could be said is that in these cases the numeral assigns case, whereas numerals lower than five do not
assign any kind of case; therefore a default case emerges on the modified pronoun. However, the matter of
Polish numerals is indeed complex as (ii) further shows.
(ii) Nas trz-ech idzie teraz do dom-u.
we.GEN three-GEN go now to home-GEN
‘Us three go home now.’
In (ii), we have a numeral lower than five, thus assumed not to assign any case at all, yet we have genitive case
marking on the pronoun, as well as on the numeral itself. There are three competing hypotheses as to how to
analyze the numeral – the nominative hypothesis, the nominative-genitive hypothesis, and the accusative
hypothesis – see Przepiórkowski (2004b) for an overview and arguments for the numeral actually being
inherently accusative) It is evident that the topic of Polish numerals deserves a much closer look, and due to
space limitations I cannot do it justice here. The interested reader is referred to the rich literature on Slavic
numerals (see, among others, Franks, 1994, 1995; Przepiórkowski, 2000, 2004b; Rutkowski, 2002, 2007;
Miechowicz-Matthiasen, 2012).
13 PATRICK LINDERT
b. Ta prawdziwa ja jest gdzieś we mnie
that.NOM real.NOM I.NOM is somewhere in me.LOC
‘The real me is somewhere inside of me.’
(Renata Górska, Przyciąganie niebieskie, p. 89)
(32a) is an example taken from an online search, and (32b) is a quote from a Polish novel.
While naturally occurring examples of these modifying pronouns are rare in Polish, when
they appear, it is always with nominative case marking, and never with the instrumental case.
The last environment to test default case is ellipsis. However, this test is not applicable in
Polish due to case markings on wh-questions.
(33) Kto chce jeść fasolę? Ja! / *Mną! / On! / *Nim!
who.NOM wants eat.inf beans? I.NOM / I.INST / he.NOM / he.INST
‘Who wants to eat beans? Me! Him!’
In (33) we can see that the elliptical answer to the question Who wants to eat beans? must be
in nominative (Ja/On ‘I/He’) and never in instrumental (Mną/Nim ‘I/He’). However, this
might be connected to the interrogative pronoun kto, which already implies a nominative
answer. Unlike in English, where wh-elements do not show case markings,19
Polish has a full
fletched paradigm for wh-words.
Table 1: Wh-elements
Case Wh-element
who / what
Nominative kto / co
Genitive kogo / czego
Dative komu / czemu
Accusative kogo / co
Instrumental kim / czym
Locative o kim / o czym (o = about)
Vocative -----
Forming questions with these different question words yields an elliptical answer in the
corresponding case; see (34).
(34) a. Kim on jest? *Kucharz. / Kucharz-em.
who.INST he is? cook.NOM / cook-INST
‘What is his profession? Cook.’
b. Kogo lubisz? *Ona. / Ją.
who.ACC like.you? she.NOM / she.ACC
‘Who do you like? Her.’
c. Komu pomagasz? *On. / Jemu.
who.DAT help.you? he.NOM / he.DAT
‘Who(m) are you helping? Him.’
19
An exception would be whom that still maintains its objective marking from older stages of the language.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 14
d. Kogo to jest dom? *Ojciec. / Ojca.
who.GEN this is house? father.NOM / father.GEN
‘Whose house is this? Father’s.’
In (34a) only instrumental is licit, as nominal predicates always appear in instrumental in
Polish. In (34b) the accusative marking presumably comes from the elided verb, dative in
(34c) probably has the same source, namely the verb. In (34d) with a genitive pronoun, it is
likely allocated within the NP of the non-elliptical structure. In sum, ellipsis in Polish cannot
be used as a test for default case.
In this section, I have shown that the default case of Polish can only be nominative. I applied
Schütze’s (2001) default case tests and the conclusion was unequivocal. The results are
summarized in Table 2.
Table 2: Results of default case tests in Polish
Default Case test Nominative Instrumental
Left Dislocation *
Gapping *
Coordination *
Modified Pronouns *
Ellipsis not applicable not applicable
5 An alternative to default instrumental
5.1 The proposal in a nutshell
In the previous sections I have reviewed the usages of the instrumental case in Polish, which
were of a lexical, predicative, and default nature. I have furthermore shown that instrumental
does not behave like a default case in Polish at all, and therefore the occurrences which have
been attributed to a default or elsewhere mechanism must be revisited. These environments
involve control constructions. In this section, I will outline a possible unification of these
occurrences and predication structures showing that the case properties in Polish control
follow the case distribution of predication.
In Polish subject control with predicative adjectives, two possible case markings are attested
for the predicate.
(35) a. Piotr próbuje być miły.
Peter.NOM tries be.INF nice.NOM
‘Peter tries to be nice.’
b. Piotr próbuje być miły-m.
Peter.NOM tries be.INF nice-INST
‘Peter tries to be a nice one.’
In (35a) the predicative adjective appears in nominative, just as the subject Piotr ‘Peter’. This
is an instance of case concord. In (35b), the adjective appears in instrumental, which seems to
display some kind of case mismatch. The possibility of this case reflecting some default case
is hard to entertain here, as the adjective seems to be projected in a position where it can get
15 PATRICK LINDERT
case via concord. Besides, instrumental does not act as a default case, as I have argued in the
previous section. If we take a closer look at Polish in general and look for a similar
distribution of nominative/instrumental, we find predicational structures as in (36).
(36) a. Piotr jest miły. AP Predication
Peter.NOM is nice.NOM
‘Peter is nice.’
b. Piotr jest (miły-m) lekarz-em. DP Predication
Peter.NOM is (nice-INST) doctor-INST
‘Peter is a (nice) doctor.’
As is well-known, predicative adjectives, as in (36a), agree in case with their subject, here
nominative. Predicative nouns, on the other hand, must appear in instrumental; see (36b). In
addition, all elements that appear inside that DP, like adjectives, must agree with their head
noun (here: lekarz ‘doctor’) in case.20
I propose that the data in (36) actually mirrors the data
in (35) one to one. It is only that in predication (36) the distribution of cases is less opaque –
adjectives agree, nouns get instrumental. In the control constructions, the distribution of case
is not so clear on the surface as we seem to only have bare adjectives in predicative position
in (35). According to the rules of predication, these adjectives should agree with their
corresponding subject. I submit that only data like (35a) involve predication over an adjective
(AP predication), whereas data like (35b) do not predicate over adjectives, but other full-
fledged DPs (DP predication). The head noun of the NP has been elided in (35b) in a process
of NP-ellipsis and therefore, on the surface, looks like a bare adjective, when it is in fact a DP.
This would explain the case optionality, and would build a bridge to the already existing rules
of predication without stipulating an exception.
There are some reasons to believe that the underlying structure of (35b) involves a more
elaborate construction, which I will briefly discuss here. One indicator for having two
syntactic structures is the difference in interpretation that the data in (35) receive. In (35a) the
interpretation is Peter wants to be nice, whereas in (35b) it is more like Peter wants to be a
nice ONE. The latter interpretation is consistent with a DP structure, whereas the former gets
an interpretation not involving any kind of nominal structure or ellipsis.
Another indicator that NP-ellipsis has taken place in (35b) is that the head noun may be
overtly realized in the structure. If it is realized, we get the following distribution.
(37) a. Piotr próbuje być miły-m człowiek-iem / *miły człowiek.
Peter.NOM tries be.INF nice-INST person-INST / nice.NOM person.NOM
‘Peter tries to be a nice person.’
b. Piotr próbuje być nim / *on.
Peter.NOM tries be.INF he.INST / he.NOM
‘Peter tries to be him.’
In (37a), the overt realization of the head noun człowiek ‘person’ must be in instrumental, a
case marking which is predicted if these structures follow simple predication forms as in (36).
In addition, we see that nominative on the noun is not even an option in this structure, even
though nominative is the default case in Polish, as shown in the previous section. This already
shows that this position in (37a) is not a position for default case as otherwise one should at
least have nominative as a possibility. Furthermore, (37b) shows that if the DP is realized as a
20
Adjectives also agree in number and gender, but these features are not the focus of the current paper.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 16
pronoun, this element needs to surface in instrumental marking, too, never in nominative. As
is well-known, pronouns constitute unambiguously DPs, hence the instrumental marking is
expected if one plugs in simple DPs predication there.
With this approach, a default instrumental (Witkoś, 2008, 2010), an elsewhere case
(Bondaruk, 2004) or the stipulation of simply stating that instrumental is available
(Przepiórkowski, 2004a) is not needed. In my proposal, there is nothing special or
extraordinary connected to the instrumental case. Rather, its appearance is predicted and
expected.
The proposal of a predication relation in control is visualized in Figure 1 below.
[finite clause [CP [AP/DP Predication]]]
AP Predication DP Predication
[NPNOM [be [APNOM]]] [NPNOM [be [PredP/PhiP/FP [DPINST]]]]
Figure 1: Polish control with embedded predication
In the next section, I will outline how NP-ellipsis is licenced in these constructions, how the
two structures of subject control are derived and why object control does not allow agreeing
adjectives, i.e., AP predication.
5.2 The process of NP-ellipsis
According to the present proposal, predicative instrumental adjectives are analysed and
projected as attributive adjectives of optionally elided nouns. A prediction this analysis makes
is that adjectives that can only appear in an attributive position like rzekomy ‘alleged’ or były
‘former’ should be possible with instrumental marking but not agreeing with the subject as in
the latter case they are actual predicative adjectives, while in the former case they are
attributive adjectives in disguise. The prediction is only partially borne out.
(38) a. Piotr nie chce być rzekomy-m *(morderc-ą).
Peter.NOM NEG wants be.INF alleged-INST murderer-INST
‘Peter does not want to be an alleged murderer.’
b. *Piotr nie chce być rzekomy.
Peter.NOM NEG wants be.INF alleged.NOM
Intended meaning: ‘Peter does not want to be alleged.’
In (38b) the predicative adjective rzekomy ‘alleged’ is not licit. This is predicted, as predicates
of this type are not allowed to appear as predicative adjectives (Alexiadou and Wilder, 1998).
In (38a), the same predicate appears now as an attributive adjective, however, NP-ellipsis is
not licenced.
In order to explain the distribution, I will follow Alexiadou and Gengel’s (2012) (henceforth
A&G) approach of NP-ellipsis, which I will outline next.
After looking at a number of Germanic (Dutch, German, English) and Romance (Italian,
Spanish) languages, A&G (2012) conclude that classifiers seem to be a licencing factor when
it comes to NP-ellipsis in general. I will illustrate the idea for English only, for the other
languages the interested reader is referred to their paper.
17 PATRICK LINDERT
(39) Talking about cars, I prefer a red *(one).
In English, one licensed NP-ellipsis, i.e., the head noun car in (39) may be elided, but only if
one in inserted. This is probably due to the fact that English has lost adjectival inflection and
therefore needs an element that can recover material from the elided structure. Following
Borer (2005), who projects one in a ClassP, A&G assume that ClassP is then the licencing
factor for NP-ellipsis. See (40) for an illustration.
(40)
A&G (2012) further follow Sleeman (1996) regarding the idea that there seems to be an
additional interplay when it comes to NP-ellipsis. They propose that the concept of partitivity
is also an important licensing factor. Partitivity is defined as follows:
(41) Partitivity
(i) Partitivity means properly or improperly included within a set.21
(ii) Partitive also means potentially but not necessarily specific.
(Sleeman, 1996, p. 34)
The definition of partitivity in (41) goes back to Sleeman (1996), who bases her
generalizations on the structure of French adjectival modification. Furthermore, she suggests
that partitivity is a requirement for NP-ellipsis to take place. A&G (2008) in addition show
that classifiers licence partitive constructions.
(42) a. Un grand vassoio è già in tavola. (Italian)
a big tray is already in table
‘A big tray is already on the table.’
b. Uno grande _ è già in tavola.
‘A big one is already on the table.’
For the data in (42), A&G argue that the theme vowel o in (42b) licences ellipsis, much like
English one, meaning that it is projected as the head of ClassP. In addition, the construction
21
The exact difference between proper and improper inclusion is not of great importance here. In her
dissertation, Sleeman (1996) also discusses specificity (see (ii)). This aspect is not important for the current
discussion either.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 18
shows a partitive construction where ellipsis is only possible if the theme vowel o is attached
to the determiner. Eliding the noun in (42a) without a theme vowel is impossible.
The interplay between partitivity and NP-ellipsis is now the following: Classifiers are
assumed to be the licencing factor for NP-ellipsis. Furthermore, A&G have shown that
classifiers are projected in partitive constructions, therefore partitive constructions should be
able to licence NP-ellipsis. The prediction is borne out. Consider the data in (38a), repeated as
(43), and (44).
(43) Piotr nie chce być rzekomy-m *(morderc-ą).
Peter.NOM NEG wants be.INF alleged-INST murderer-INST
‘Peter does not want to be an alleged murderer.’
(44) Piotr nie chce być okropny-m człowiek-iem.
Peter.NOM NEG wants be.INF awful-INST person-INST
‘Peter does not want to be an awful one.’
In (43), we see that NP-ellipsis in not licenced. The head noun morderca ‘murderer’ must be
overtly realized. Even though we do have an attributive adjective here, the core for the
unacceptability for NP-ellipsis lies in the fact that there is no partitive structure in (43). The
non-intersective adjective rzekomy ‘alleged’ does not allow partitive readings.22
The idea is
that one cannot take all the things that are alleged and the things that are murderers and
intersect them as this would lead to a paradox. If a murderer is an alleged murderer, there is a
possibility that the said murderer is not a murderer in the first place, and therefore a proper set
containing alleged and murderer is not possible.
This problem does not arise in example (44). There NP-ellipsis is possible due to the
underlying partitive construction. If we apply the same logic as in (43), we will see that we
will not encounter the same problem. If we take all the things that are awful, and all the things
that are persons, we can intersect between the two sets. There will be a set where both entities
are included, namely awful and persons. Both can be included within one set, and given the
partitivity rule in (41), we should get ellipsis. This prediction is borne out.
There is one issue with this account which I want to briefly address at this point. The proposal
so far predicts that NP-ellipsis should also be possible in simple predication like in (45).
(45) Piotr jest okropny-m ??(człowiek-iem).
Peter.NOM is awful-INST (person-INST)
‘Peter is an awful one.’23
The option of NP-ellipsis in (45) is restricted in Polish, however, not entirely ruled out by the
grammar. There are some attested examples in the National Corpus of the Polish language,
though they are rare. What (45) shows is that NP-ellipsis seems to be conditioned by other
factors as well, not just purely syntactic ones. It is well-attested that some speakers of
languages that have NP-ellipsis still do not accept structures with elided elements as freely as
others. There might be pragmatic factors. In addition, I propose that locality also plays a role
22
The difference between intersective and non-intersective adjectives is thoroughly investigated in Larson
(1998). In addition, there is a finer-grained classification in the literature when it comes to non-intersective
adjectives. Some of them are called privative adjectives (Partee, 2010) like fałszywy ‘fake’, modal adjectives
(Landman, 2001) like rzekomy ‘alleged’ and aspectual adjectives (Szabó, 2015) like były ‘former’. 23
I have two question marks here and not a star, as usually more people accept NP-ellipsis if more context is
added to the structure, i.e., the elided noun becomes more easily recoverable. Uttered out of the blue, the
structure might be rated worse.
19 PATRICK LINDERT
in this. The relationship in (45) between the subject and the predicate is a very local one, thus
deeming the structure with the elided noun uneconomic or unnecessary. However, in control
structures, people seem to be more permissive with NP-ellipsis. If we now create more
distance between the subject and the predicate, we see that instrumental seems to even be
preferred over the agreeing option.
(46) Jan bał się nawet chcieć spróbować wydawać się
Jan.NOM feared REFL even want.INF try.INF seem.INF REFL
??szczęśliwy / szczęśliwy-m (człowiek-iem).
happy.NOM / happy-INST (person-INST)
‘John was afraid to even want to try to seem happy.’
The data in (46) is taken from Przepiórkowski (2004a, p. 107); however, I added the part in
brackets. The data show that agreeing becomes worse in this structure. Presumably, this is
because it is more economical to project the structure with DP predication than establishing
long dependency chains with AP predication across multiple phrases.
In this section, I have shown how the mechanism with NP-ellipsis works on a formal level.
The possibility of NP-ellipsis is encoded in the DP-domain as a classifier phrase. This
classifier phrase is additionally available in partitive constructions, so that only those
constructions should licence NP-ellipsis.
5.3 The challenge of object control and the details of subject control
While the proposal outlined in this paper straightforwardly accounts for a variation in case
marking of adjectives in subject control, it runs into problems when dealing with object
control constructions as in (47).
(47) a. Janek kazał Tom-kowi być miły-m / *mił-emu.
Jan.NOM ordered Tom-DAT be.INF nice-INST / nice -DAT
‘Jan ordered Tom to be nice.’
b. Tom uczył Piotr-a być pijany-m / *pijan-ego.
Tom.NOM taught Peter-ACC be.INF drunk-INST / drunk -ACC
‘Tom taught Peter how to be drunk.’
In (47a), we see that the adjective miły ‘nice’ must appear in instrumental. It is not possible
for the adjective to agree with the controller Tom, which is marked dative. One could argue
that the impossibility of the agreeing case might be connected to the lexical case status of the
dative case. It has been reported that languages that allow case transmission in object control,
do not allow this mechanism when lexical cases are involved. Some of the reported languages
are Russian (Landau, 2008) and Icelandic (Sigurðsson, 2008).24
However, Polish seems to exhibit a stricter rule as structural cases are also exempt from case
transmission, as can be seen in (47b). There, the adjective cannot agree with its controller
24
The case of Icelandic is a bit more complex. Transmission of lexical cases does not seem to be possible when
adjectives are involved. However, floating quantifiers may inherit a lexical case in the same structure, see
Sigurðsson (2008) for discussion and data.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 20
Piotr ‘Peter’ even though the controller is marked with accusative, a structural case in
Polish.25
Thus, the proposal outlined so far overgenerates and predicts that both routes of predication,
DP and AP predication, should work here. To put it differently, there is nothing in the
predicational structure that would prevent the non-nominative case from being transmitted to
the adjective, yet it is impossible. Descriptively, Polish simply does not allow case
transmission in obligatory object control. How can we account for this? The answer might lie
in the underlying syntactic structure of object control.
(48) *[ApplP Tom Appl [CP [FinP PRO [TP PRO T [CopP PRO be [AP PRO niceDAT]]]]]]
The derivation in (48) corresponds to the illicit structure in (47a) with the agreeing adjective.
For convenience, English words are inserted into the derivation. I will go through the
derivation step by step to see at which point we encounter a potential problem that would lead
to a crash. The derivation starts out with PRO and nice sharing their phi-features (Frampton
and Gutmann, 2000). PRO moves up to CopP where it establishes the predication relation
(den Dikken, 2006). PRO is attracted by T, checks the EPP and enters an Agree relation with
T (Landau, 2000). PRO moves further up to FinP (Landau, 2015). In the finite clause, ApplP
introduces the dative marked argument (Pylkkänen, 2008). Tom enters an Agree relation with
the Appl head. PRO finds its controller Tom and they establish an Agree relation. However,
when the eventual structure is shipped to PF, PRO’s case status is not clear. PRO has moved
to Spec,TP, so it might be associated with nominative as it moved into a subject position. In
addition, due to its Agree relation with Tom, case transmission would lead to dative on PRO.
This is problematic as the case feature of the adjective must also be valued, and it is not clear
whether non-nominative cases may be transmitted in this way (Bondaruk, 2004). In addition,
one could say that from a PF view, it is not clear what kind of case should be transmitted.
PRO moves via Spec,TP making it a subject. However, PRO would also receive dative from
its controller. In Polish, there are no quirky dative subjects (Bondaruk and Szymanek, 2007);
as a result, at PF one encounters a contradiction. In order to avoid this, case would not be
transmitted, leaving the case feature on the adjective unvalued – the derivation crashes.
The proposed mechanism (for Polish) is summarized in (49):
(49) PRO and case
At PF, when PRO’s case needs to be morphologically realized, this must not signal
conflicts with independent requirements of the language (such as the ban against
quirky subjects).
In order to see that (49) does not get in the way of deriving a good obligatory object control
sentence, let us consider the data with instrumental marking in more detail.
(50) a. Janek kazał Tom-kowi być miły-m.
Jan.NOM ordered Tom-DAT be.INF nice-INST
‘Jan ordered Tom to be nice.’
b. [ApplP Tom Appl [CP [FinP PRO [TP PRO T [CopP PRO be [FP F
[DP niceINST personINST]]]]]]]26
25
Przepiórkowski (1999, p. 104) notes that there might be some cases of lexical accusative in Polish. However,
no matter what kind of verb one takes here, the pattern in object control is always the same.
26 I simplify the internal structure of the DP here. As this constitutes a partitive construction, a ClassP would be
projected inside the DP to enable NP-ellipsis.
21 PATRICK LINDERT
For (50a), the proposed analysis involves DP predication. Therefore, DP predication structure
(DP, FP, CopP) is projected. Within the DP, nice and person share phi-features. As this is a
predicative structure involving DPs, a predicational functional projection is projected atop of
it. I am staying neutral as to the exact nature of this projection, therefore I label it FP.27
F
assigns case to its complement; both elements appear in instrumental. PRO is projected in
CopP establishing the DP predication. It further moves up to TP to check the EPP and further
to FinP. As in (48), PRO picks Tom as its controller. Thereby, PRO carries dative even though
it moved through the subject position Spec,TP. However, at PF, this is not a problem, as the
case properties of PRO are never morphologically reflected, and the case feature of the
adjective is already independently valued. The derivation of (50) thus does not violate the
restriction in (49) and the sentence is correctly predicted to be acceptable.
Bondaruk (2004) proposed a similar mechanism to (49). However, it differs in one crucial
respect. She assumes as well that PRO does get accusative/dative (‘objective case’ for her) via
the controller. The difference is that in her system, the adjective cannot receive a non-
nominative case, and therefore the case feature of the adjective is satisfied with the elsewhere
case, namely instrumental. The assignment of instrumental is thus a rescue mechanism; a
default option. In my proposal, instrumental arises as the second route of predication, namely
DP predication where a whole DP is projected and case is assigned by the rules of predication
and not as a default option due to failure of Agree.
Having presented how object control is derived within the proposed analysis, let us now turn
to the subject control data and see that both versions – agreeing and instrumental – are
derived. Consider the data in (51).
(51) a. Janek próbuje być miły.
John.NOM tries be.INF nice.NOM
‘John tries to be nice.’
b. [TP John T [VP John tries [CP [FinP PRO [TP PRO T [… [AP PRO niceNOM]]]]]]
The derivation in (51b) is the same as in (48) up to the projection of CP.28
The finite clause
then introduces the verb and the external argument. John enters an Agree relation with T,
moves up to TP, and checks the EPP. Once John receives nominative, the case is transmitted
to all copies and elements that John agreed with. Thus, PRO receives nominative and can
transmit it further to the adjective with which it shared its features. There are no conflicting
case properties of PRO – PRO is in a subject position and only receives nominative. The
derivation is successful.
To complete the picture, the respective data for subject control with instrumental adjectives is
repeated in (52a) and the analysis is sketched in (52b).
(52) a. Piotr próbuje być miły-m człowiek-iem.
Peter.NOM tries be.INF nice-INST person-INST
‘Peter tries to be a nice person.’
27
For more arguments in favour of a functional projection FP in DP predication, but not in AP predication, the
reader is referred to Lindert (to appear).
28 I remain agnostic w.r.t. the phasal status of CP. My approach is compatible with CP being a weak phase
(Landau, 2000, 2008), as well as movement of PRO to the phase edge (Landau, 2015). The potential problem
that nominative is assigned by the matrix T, which is separated from PRO by the vP phase, can be overcome
under the assumption that nominative is determined vP-internally (see Sigurðsson, 2000, 2003; Schäfer, 2008;
see also Błaszczak, 2007, 2008, for a proposal along similar lines).
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 22
b. [TP John T [VP John tries [CP [FinP PRO [TP PRO T [CopP PRO be [FP F
[DP niceINST personINST]]]]]]]]
In (52), we see the structure with an instrumental adjective. I have argued that these are not
predicative adjectives, but rather attributive ones situated in a DP. The derivation would thus
start with the adjective and its head noun, here a generic noun like człowiek ‘person’, in the
same phrase. FP assigns instrumental to its complement. The noun thus receives instrumental
and the adjective agrees with it. CopP is projected, establishes the predication relation and
PRO is inserted. PRO agrees with non-finite T and moves into the specifier of TP to check the
EPP. Next PRO establishes a control relation with John. Eventually, John’s nominative case
may be transmitted to PRO. Even though this would not pose a problem for the rule in (49),
the case is not morphologically realized anyway, as PRO has never established an Agree
relation with the adjective. The latter receives its case from the DP predication structure and
thus its case feature has already been satisfied. The derivation may continue, all features are
checked and deleted.
According to this analysis, one does not need to stipulate a default case mechanism, but can
rather assume that the rules that govern predication are also to be followed in control. This
simplifies the grammar and it also builds on the intuition that similar structures should be
unified. As both structures involve the copula być ‘be’ and the same case markings, it is only
natural to assume that they share the same underlying syntactic structure.
In this section, I have outlined a possible explanation for the apparent puzzling behaviour of
adjectives in obligatory control constructions. I have argued that while agreeing adjectives
follow the rules of AP predication in this language, instrumental adjectives are actually
modifiers of DPs therefore following the rules of DP predication. Thus, I have argued for
analysing these adjectives as DPs with optional NP-ellipsis. I have also offered an account of
how NP-ellipsis is licenced in this construction, and why it is restricted in other types of
predication. As a last point, the striking behaviour of adjectives in obligatory object control
was used to show that not only predicational rules must be taken into account to explain the
distribution, but that the control module also adds further restrictions to it.
6 Conclusion
In this paper, I have taken a closer look at the distribution of the instrumental case in Polish. I
have rejected the idea of classifying the instrumental case as a default or elsewhere case. I
have applied Schütze’s (2001) default case tests to show that Polish indeed takes nominative
as default. In addition, I have proposed an analysis for some of the environments that have
been claimed to have a default or elsewhere instrumental, namely obligatory subject and
object control. I have submitted that instrumental arises as a form of DP predication where the
adjective is actually a DP in disguise. For this, I have shown how NP-ellipsis works in these
constructions, and also why DP predication is the only way to go for object control.
References
Alexiadou, A. (2003). On nominative case features and split agreement. In E. Brandner and
H. Zinsmeister (Eds.), New perspectives on case theory (pp. 23-52). Stanford, CA: CSLI.
23 PATRICK LINDERT
Alexiadou, A., Anagnostopoulou, E., and Sevdali, C. (2014). Opaque and transparent datives,
and how they behave in passives. The Journal of Comparative Germanic Linguistics,
17(1), 1-34.
Alexiadou, A. and Gengel, K. (2012). NP ellipsis without focus movement/projections: The
role of classifiers. In I. Kučerova and A. Neeleman (Eds.), Contrasts and positions
in information structure (pp. 177-205). Cambridge: CUP.
Alexiadou, A. and Wilder, C. (1998). Adjectival modification and multiple determiners. In A.
Alexiadou and C. Wilder (Eds.), Possessors, predicates and movement in the DP (pp.
303-332). Amsterdam: Benjamins.
Babby, L. (1986). The locus of case assignment and the direction of percolation: Case theory
and Russian. In R. Brecht and J. Levine (Eds.), Case in Slavic (pp. 170-219). Columbus,
OH: Slavica Publishers.
Bailyn, J. (2001). The syntax of Slavic predicative case. ZAS Papers in Linguistics, 22, 1-23.
Bailyn, J. and Citko, B. (1999). Case and agreement in Slavic predicates. In K. Dziwirek, H.
Coats, and C. Vakareliyska (Eds.), Annual Workshop on Formal Approaches to Slavic
Linguistics: The Seattle Meeting, 1998 (pp. 17-37). Ann Arbor, MI: Michigan Slavic
Publications.
Baker, M. (2015). Case: Its principles and its parameters. Cambridge: CUP.
Błaszczak, J. (2001). Covert movement and the genitive of negation in Polish. (Linguistics in
Potsdam, 15). Potsdam: Potsdam University Press.
Błaszczak, J. (2007). Phase syntax: The Polish genitive of negation. (Habilitation thesis).
University of Potsdam.
Błaszczak, J. (2008). Case assignment and ‘long-distance’ agreement in a phase model of
syntax revisited. In Y.-S. Kim (Ed.), Minimalist explorations of the syntax-lexicon
interface: Proceedings of the 10th Seoul International Conference on Generative
Grammar, July 17-20, 2008, Dongguk University, Seoul (pp. 101-120). Seoul: The
Korean Generative Grammar Circle.
Boeckx, C. and Hornstein, N. (2006). Control in Icelandic and theories of control. Linguistic
Inquiry, 37(4), 591-606.
Bogusławski, A. (2001). Über Nominative-Instrumental-Variationen im Polnischen. In W.
Boeder and G. Hentschel (Eds.), Variierende Markierung von Nominalgruppen in
Sprachen unterschiedlichen Typs (pp. 101-134). Oldenburg: BIS.
Bondaruk, A. (2004). PRO and control in English, Irish and Polish: A minimalist analysis.
Lublin: Wydawnictwo KUL.
Bondaruk, A. (2013). Copula clauses in English and Polish: Structure, derivation and
interpretation. Lublin: Wydawnictwo KUL.
Bondaruk, A. (2014). Characterizing and defining predicational clauses in Polish. In L.
Veselovská and M. Janebová (Eds.), Complex visibles out there. Proceedings of the
Olomouc Linguistics Colloquium 2014: Language Use and Linguistic Structure (pp. 333-
348). Olomouc: Palacký University.
Bondaruk, A. and Szymanek, B. (2007). Polish nominativeless constructions with dative
experiencers: Form, meaning and structure. Studies in Polish Linguistics, 4, 61-97.
Borer, H. (2005). In name only. Oxford: OUP.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 24
Bowers, J. (1993). The syntax of predication. Linguistic Inquiry, 24(4), 591-656.
Chung, S. and McCloskey, J. (1987). Government, barriers, and small clauses in Modern
Irish. Linguistic Inquiry, 18(2), 173-237.
Citko, B. (2008). Small clauses reconsidered: Not so small and not all alike. Lingua, 118(3),
261-295.
Den Dikken, M. (2006). Relators and linkers. Cambridge, MA: MIT Press.
Duffield, N. (1989). A case for default values? In F. H. Brengelman, V. Samiian, and W.
Wilkins (Eds.), Proceedings of the Western Conference on Linguistics 2 (pp. 82-93).
Fresno: CSU.
Frampton, J. and Gutmann, S. (2000). Agreement is feature sharing. (Manuscript).
Northeastern University.
Franks, S. (1994). Parametric properties of numeral phrases in Slavic. Natural Language and
Linguistic Theory, 12, 597-674.
Franks, S. (1995). Parameters of Slavic morphosyntax. New York and Oxford: OUP.
Franks, S. (2015). The overgeneration problem and the case of semipredicates in Russian. In
A. Bondaruk, G. Dalmi, and A. Grosu (Eds.), Advances in the syntax of DPs (pp. 13-60).
Amsterdam: Benjamins.
Freidin, R. and Babby, L. (1984). On the interaction of lexical and syntactic properties: Case
structure in Russian. In W. Harbert (Ed.), Cornell Working Papers in Linguistics VI (pp.
71-103). Ithaca: Cornell University.
Geist, L. (1999). Russisch byt’ (‘sein’) als funktionale und / oder lexikalische Kategorie. In E.
Lang and L. Geist (Eds.), Kopula-Prädikativ-Konstruktionen als Syntax/Semantik-
Schnittstelle (pp. 1-39). (ZAS Papers in Linguistics, 14). Berlin: Zentrum für Allgemeine
Sprachwissenschaft.
Geist, L. and Błaszczak, J. (2000). Kopulasätze mit dem pronominalen Elementen to/èto im
Polnischen und Russischen. ZAS Papers in Linguistics, 16, 115-139.
Havranová, K. (2014). Default case in Dutch: A comparative study of Dutch and English case
systems. In L. Veselovská and M. Janebová (Eds.), Complex visibles out there.
Proceedings of the Olomouc Linguistics Colloquium 2014: Language Use and Linguistic
Structure (pp. 163-174) Olomouc: Palacky University.
Landau, I. (2000). Elements of control. Dordrecht: Kluwer.
Landau, I. (2008). Two routes of control: Evidence from case transmission in Russian.
Natural Language and Linguistic Theory, 26(4), 877-924.
Landau, I. (2015). A two-tiered theory of control. Cambridge, MA: MIT Press.
Landman, M. (2001). Adjectival modification restricted. (Manuscript). University of
Massachusetts, Amherst.
Larson, R. K. (1998). Events and modification in nominals. Semantics and Linguistic Theory,
8, 145-168.
Legate, J. A. (2008). Morphological and abstract case. Linguistic Inquiry, 39(1), 55-101.
Lindert, P. (to appear). Predication under control: The case of Polish. Proceedings of
ConSOLE XXIV.
25 PATRICK LINDERT
Marantz, A. (2000). Case and licensing. In E. Reuland (Ed.), Arguments and case: Explaining
Burzio’s Generalization (pp. 11-30). Amsterdam: Benjamins.
Matushansky, O. (2010). Russian predicate case, encore. In G. Zybatow, P. Dudchuk, S.
Minor, and E. Pshehotskaya (Eds.), Formal Studies in Slavic Linguistics, Proceedings of
FDSL 7.5 (pp. 117-135). Frankfurt: Peter Lang.
McCloskey, J. (1985). Case, movement and raising in modern Irish. In J. Goldberg, S.
MacKaye, and M. Wescoat (Eds.), Proceedings of the Fourth West Coast Conference on
Formal Linguistics (pp. 190-205). Stanford, CA: CSLI.
McFadden, T. (2004). The position of morphological case in the derivation: A study on the
syntax-morphology interface. (Doctoral dissertation). University of Pennsylvania.
McFadden, T. (2007). Default case and the status of compound categories in Distributed
Morphology. University of Pennsylvania Working Papers in Linguistics, 13(1), 18-32.
McGinnis, M. (1998). Case and locality in L-syntax: Evidence from Georgian. In H. Harley
(Ed.), Papers from the UPenn/MIT Roundtable on Argument Structure and Aspect, MIT
Working Papers in Linguistics 32 (pp. 139-158). Cambridge, MA: MIT.
Miechowicz-Mathiasen, K. (2012). Licensing Polish higher numerals: An account of the
Accusative Hypothesis. In J. Błaszczak, B. Rozwadowska, and W. Witkowski (Eds.),
Current Issues in Generative Linguistics: Syntax, Semantics, and Phonology (pp. 58-75).
(Generative Linguistics in Wrocław, vol. 2). Published online 2 August 2012.
Moser, M. (1993). Der prädikative Instrumental. Frankfurt: Peter Lang.
Partee, B. (2010). Privative adjectives: Subsective plus coercion. In R. Bäuerle, U. Reyle, and
T. E. Zimmermann (Eds.), Presuppositions and discourse: Essays offered to Hans Kamp
(pp. 273-285). Bingley: Emerald.
Pesetsky, D. (2015). Russian case morphology and the syntactic categories. Cambridge, MA:
MIT Press.
Przepiórkowski, A. (1999). Case assignment and the complement/adjunct dichotomy.
(Doctoral dissertation). University of Tübingen.
Przepiórkowski, A. (2000). Predicative case agreement with Quantifier Phrases in Polish. In
A. Okrent and J. Boyle (Eds.), The Proceedings from the Main Session of the Chicago
Linguistic Society's Thirty-sixth Meeting, volume 36-1 (pp. 343–354). Chicago, IL:
Chicago Linguistic Society.
Przepiórkowski, A. (2004a). On case transmission in Polish control and raising constructions.
Poznań Studies in Contemporary Linguistics, 39, 103-123.
Przepiórkowski, A. (2004b). O wartości przypadka podmiotów liczebnikowych. Bulletin de la
Société Polonaise de Linguistique, LX, 133-143.
Przepiórkowski, A. and Rosen, A. (2005). Czech and Polish raising/control with or without
structure sharing. Research in Language, 3, 33-66.
Pylkkänen, L. (2008). Introducing arguments. Cambridge, MA: MIT Press.
Richardson, K. (2007). Case and aspect in Slavic. Oxford: OUP.
Rutkowski, P. (2002). Numerals as grammaticalised nouns: A generative approach.
Interlingüística, 13(3), 317-328.
DEFAULT CASE, NP-ELLIPSIS, AND THE THEORY OF CONTROL IN POLISH 26
Rutkowski, P. (2007). Hipoteza frazy przedimkowej jako narzędzie opisu składniowego
polskich grup imiennych. (Doctoral dissertation). Warsaw University.
Schäfer, F. (2008). The syntax of (anti-)causatives: External arguments in change-of-state
contexts. Amsterdam: Benjamins.
Schütze, C. (2001). On the nature of default case. Syntax, 4(3), 205-238.
Sigurðsson, H. A. (2000). The locus of case and agreement. Working Papers in Scandinavian
Syntax, 65, 65-108.
Sigurðsson, H. A. (2003). Case: Abstract vs. morphological. In E. Brandner and H.
Zinsmeister (Eds.), New perspectives on case theory (pp. 223-268). Stanford, CA: CSLI.
Sigurðsson, H. Á. (2008). The case of PRO. Natural Language & Linguistic Theory, 26(2),
403-450.
Sleeman, P. (1996). Licensing empty nouns in French. The Hague: Holland Academic
Graphics.
Szabó, Z. G. (2015). Major parts of speech. Erkenntis, 80(1), 3-29.
Szécsénsy, K. (2002). Default case in Hungarian. Rivsta di grammatical generative, 27, 139-
149.
Willim, E. (1990). On case-marking in Polish. Papers and Studies in Contrastive Linguistics,
XXV, 203-220.
Witkoś, J. (2008). Control and predicative adjectives in Polish. In J. Witkoś and G. Fanselow
(Eds.), Elements of Slavic and Germanic grammars: A Comparative view (pp. 255-277).
Frankfurt: Peter Lang.
Witkoś, J. (2010). Movement theory of control and CP-infinitives in Polish. In N. Hornstein
and M. Polinsky (Eds.), Movement theory of control (pp. 45-66). Amsterdam:
Benjamins.
Woolford, E. (1997). Four-way case systems: Ergative, nominative, objective and accusative.
Natural Language & Linguistic Theory, 15(1), 181-227.
Woolford, E. (2003a). Burzio’s Generalization, markedness, and constraints on nominative
objects. In E. Brandner and H. Zinsmeister (Eds.), New perspectives on case theory (pp.
301-329). Stanford, CA: CSLI.
Woolford, E. (2003b). Nominative objects and case locality. In W. Browne, J.-Y. Kim, B. H.
Partee, and R. A. Rothstein (Eds.), Formal Approaches to Slavic Linguistics: The
Amherst Meeting 2002 (pp. 539-568). Ann Arbor, MI: Michigan Slavic Publications.
Woolford, E. (2006). Lexical case, inherent case, and argument structure. Linguistic Inquiry,
37(1), 111-130.
Zwart, C. J.-W. (1988). The first case: The nominative as a default case and the consequences
for control theory. (Master’s thesis). University of Groningen.
Patrick Lindert
University of Stuttgart
Institute of English Linguistics
Azenbergstr. 12
70174 Stuttgart, Germany
e-mail: [email protected]