Syntax
Case Theory
Modul 04-006-2002
Phonology – Morphology – Syntax
Institut für Linguistik
Universität Leipzig
[Link]/heck
Morphological case
Point of departure:
Many languages involve morphological case marking on noun phrases
(and its dependents, such as determiners and adjectives). How many
different morphological cases there are depends on the language.
German, for instance, distinguishes nominative, genitive, dative, and
accusative (1-a-d).
(1) a. Die blaue Elise schläft.
[Link] [Link] [Link] sleeps
b. Die blaue Elise bedarf der Hilfe (der Ameise).
the blue Elise needs [Link] [Link] [Link] [Link]
c. Die Ameise entkommt der blauen Elise.
the ant escapes [Link] [Link] [Link]
d. Die blaue Elise hat einen Plan.
the blue Elise has [Link] [Link]
Structural and abstract case
The syntactic dependency of case:
(Morphological) case is syntactically dependent. The case on an NP
depends on the abstract syntactic structure the NP shows up in.
In German, for instance, the object of a verb generally bears
accusative, the subject generally bears nominative. Such cases do not
depend on the concrete predicate that is involved (2): structural case.
Even if a case is not always morphologically visible (cf. no
morphological marking on the noun Elise itself in (2-a-c)), it is
assumed to be abstractly present.
(2) a. Die blaue Elise hat einen Plan.
[Link] [Link] [Link] has [Link] [Link]
b. Die blaue Elise hasst die Ameise.
[Link] [Link] [Link] hates [Link] [Link]
c. Die Ameise überlisted die blaue Elise.
[Link] [Link] outwits [Link] [Link] [Link]
Lexical case
Lexical case:
Sometimes NPs do not bear structural case. In German, for instance,
an object may bear a case other than accusative, which depends on
the concrete lexical predicate the NP is an argument of (recall (1-b)
and (1-c)).
In Icelandic, even subjects may bear another case than nominative
(e.g., accusative or dative), which then depends on the concrete lexical
predicate that is involved. (3-a,b) (Sigurðsson 2004).
Since the argument’s case is not determined by the structural
configuration but by a lexical property of the predicate, one speaks of
lexical (or inherent) case.
(3) a. Hana grunar að hann fari.
[Link] suspects that he leaves
‘She suspects that he will leave.’
b. Henni líkuðu hestarnir.
[Link] liked [Link]
‘She liked the horses.’
Structural accusative
Question:
Where does the structural accusative come from?
Two answers (more are possible):
Structural accusative is “assigned” by the lexical verb.
Structural accusative is “assigned” by the v associated with the lexical
verb.
Two arguments for case assignment by v:
As already noted, structural case is not sensitive to the individual
predicate involved. This follows if accusative is not dependent on the
lexical verb but on v.
There is a correlation, called Burzio’s generalization (alternatively: the
Burzio-Perlmutter generalization), that holds between the assignment
of accusative case and the presence of an external argument in Specv.
The Burzio-Perlmutter generalization
Perlmutter (1978), Burzio (1981, 1986):
The object of a predicate P may not receive accusative case unless P
has an external argument.
Unaccusative verbs merge their only argument in object position. Yet,
this argument typically does not receive structural accusative (but,
e.g., nominative). Why? Because there is no external argument.
(4) a. Posy/She/*Her fell.
b. Pottwal/He/*Him collapsed.
Consequence:
Since the external argument is introduced by v, the generalization can be
formulated straightforwardly if v also assigns accusative:
(5) Burzio-Perlmutter generalization:
If v assigns accusative, then v also introduces an external argument.
Case “assignment”
Question:
What exactly does it mean that case is “assigned”?
Idea:
If case assignment is a relation between a functional head and an
argument (e.g., between v and the object), then it seems plausible to
assume that case assignment works like agreement.
Reminder: Agreement
Recall:
Agreement is the result of a probe scanning its c-command domain
for a matching goal. Once the goal is found, agreement applies.
Agreement can be formulated in terms of feature checking (7-a,b) (the
probe already bears a value and gets checked) . . .
(6) Dr. Brumm eat-s the honey.
(7) a. TP b. TP
T vP T vP
[u3,usg] [u3,usg]
NP v′ NP v′
[3,sg] [3,sg]
eat+v VP eat+v VP
Reminder: Agreement
(continued from previous page):
. . . or it can be formulated in terms of Agree, which involves
valuation (8-a,b) (the probe is valued by the goal and gets checked).
(8) a. TP
TP b.
" T # vP
" T # vP uπ:3
uπ: u#:sg " NP # v′
u#: " NP # v′ π:3
π:3 eat+v VP
#:sg
eat+v VP
#:sg
Accusative assignment as checking
Case checking:
Little v bears [ucase:acc] (and [uN], see the Burzio-Perlmutter
generalization). It searches its c-command domain for a goal NP that also
bears [ucase:acc]. Once such a goal is found (9-a), case is checked (on both
v and NP), see (9-b).
(9) a. vP b. vP
" v # VP " v # VP
uN uN
ucs:acc V NP ucs:acc V NP
[ucs:acc] [ucs:acc]
Accusative assignment as valuation
Case valuation (Agree):
Little v bears [ucase:acc] (and [uN]). It searches its c-command domain for
a goal NP that bears an unvalued case feature [ucase: ]. Once such a goal
is found (10-a), Agree applies: the goal is case-valued as [ucase:acc] and
case is checked (on both v and NP), see (10-b).
(10) a. vP b. vP
" v # VP " v # VP
uN uN
ucs:acc V NP ucs:acc V NP
[ucs: ] [ucs:acc]
Checking or valuation: a trade-off
Checking:
Checking theory contains a certain redundancy: both elements (probe
and goal) contain the same set of features although, intuitively, the
feature value of one depends on the value of the other.
Moreover, derivations that start with non-matching probe-goal pairs
(e.g., [ucase:acc] on v, [ucase:dat] on NP) are doomed from the
beginning.
Valuation:
The valuation theory accounts for the intuitive asymmetry between
probe and goal (and avoids many crashing derivations).
But it encounters the problem that it apparently involves a violation
of the Inclusiveness Condition: the feature-value of the dependent
element is added in the course of the derivation.
Note:
The same considerations apply to agreement.
Checking or valuation: probe-hood
Note:
A potential issue that is raised by the valuation account of case
assignment is the notion of probe-hood.
If probes are characterized as features that need to be valued, then
the Agree-based case theory implies that Agree can apply “upward”,
in contrast to what been assumed (see Bjorkman and Zeijlstra 2019
for a theory of upward Agree).
This is because the element that is to be case-valued is the downstairs
NP-argument, not the upward functional head.
One can escape this conclusion by assuming that probes are charac-
terized as features that bear a diacritic u-prefix, be they valued or not:
[uF:α]/[uF: ]. (Since case on NP is not interpreted, this means that
the u-prefix does no longer signal uninterpretability in the narrow
sense of the word.)
Transitive little v would then bear [ucase:acc] (a probe), while
NP-arguments would be assumed to be lexically specified as bearing
[case: ] (a non-probe).
Case in ECM-infinitives
Recall:
The ECM-construction involves an overt subject in the embedded
infinitive (11-a). Crucially, the subject of such an infinitive receives
accusative case (only visible with pronouns in English) (11-b).
Analysis: Accusative is checked/valued by the v that is associated
with the embedding predicate (12).
(11) a. Hackenpiep considers [ TP Pottwal to be a thief ].
b. Hackenpiep considers [ TP *he/him to be a thief ].
(12) vP
" v # VP
uN
ucs:acc V TP
Subj T′
[cs:acc]
... ...
Case in for-infinitives
Recall:
For-infinitives also involve an overt subject (13-a). This subject is also
marked as accusative (13-b,c).
Analysis: Accusative is checked/valued by the complementizer for.
(13) a. Pottwal wants [ CP for Dr. Brumm to shave ].
b. Pottwal wants [ CP for *he/him to shave ].
(14) CP
C TP
[ucs:acc]
Subj T′
[cs:acc]
... ...
Structural nominative
Structural nominative:
Structural nominative seems to be connected to finiteness/agreement:
In infinitives, nominative is not checked/valued.
This makes it possible in some cases for accusative to be checked on
the subject of the infinitive by some higher functional head (see ECM
and for-infinitives).
Since finiteness/agreement is associated with T, the hypothesis is that
structural nominative is checked/valued by a finite T-head.
(15) He/*him eats the honey.
(16) TP
T vP
[ucs:nom]
NP v′
[cs:nom]
eat+v VP
... ...
Case in raising infinitives
Recall:
In raising infinitives, the subject of the raising verb is merged in the
embedded infinitive. From there, it undergoes movement (via the
embedded SpecT) to the matrix SpecT-position.
Since only the matrix T-head is finite/agreeing, nominative case is
assigned to the subject by the matrix T-head (before raising) (17).
(17) Pottwal appears [ TP to be sick ].
(18) TP
T vP
[ucs:nom]
v VP
V TP
Subj T′
[cs:nom]
... ...
Lexical case
Lexical case:
Some verbs determine a particular case on their object as a lexical
property. In such a case (e.g., in German), the object is not marked
with accusative but with some other case, such as dative (19-a) or
genitive (19-b).
Other languages allow for the assignment of further lexical cases
(Fanselow and Felix 1987): Latin (20-a) (ablative) or Russian (20-b)
(instrumental).
(19) a. Die Ameise entkommt der blauen Elise.
the ant escapes [Link] [Link] [Link]
b. Dr. Brumm bedient sich einer List.
Dr. Brumm uses self [Link] [Link]
(20) a. variis instrumentis ad lanificia untuntur
[Link] [Link] for weaving [Link]
b. Ivan rukovodit otdelom.
Ivan leads [Link]
Lexical case
Analysis of lexical case:
Lexical case is either checked/valued by the lexical verb itself (21-a)
...
. . . or by a functional head that is associated with this particular type
of lexical verb (for instance a special variant of v, say vdat ), see (21-b).
(21) a. VP b. vP
NP V VP " vdat #
[cs:dat] [ucs:dat] uN
NP V ucs:dat
[cs:dat]
Structural vs. lexical case: absorption
Absorption of structural accusative:
In many languages (but not all) structural accusative is “absorbed” in
the passive construction.
In a passive, there is no external argument (except in form of an
optional element such as a by-phrase in English) (22-a). According to
the Burzio-Perlmutter generalization, v then does not assign
accusative. Consequently, the internal argument can receive
nominative case by T (22-b).
In languages that require SpecT to be filled (such as English), the
internal argument also moves to become the subject of the clause
(22-c).
(22) a. Dr. Brumm was bitten (by a carnivorous plant).
b. He/*Him was bitten.
c. *Was bitten Dr. Brumm.
Structural vs. lexical case: absorption
No absorption of lexical dative:
If the internal argument of a passivization is marked by a lexical case,
no absorption takes place.
In German, for instance, the dative that is assigned to the object by
the verb helfen ‘help’ (23-a) is preserved under passivization (23-b,c).
(23) a. Die Ameise entkommt der blauen Elise.
the ant escapes [Link] [Link] [Link]
b. *Die blaue Elise wird geholfen.
[Link] [Link] [Link] becomes helped
c. Der blauen Elise wird geholfen.
[Link] [Link] [Link] becomes helped
Structural case and double object verbs
Double object verbs:
In some languages, both cases associated with double object verbs are
structural. In some, the dative is lexical while the accusative is
structural.
German exhibits both cases (shown by absorption possibilities),
depending on whether passive is formed with the auxiliary werden
‘become’ (25) or bekommen ‘get’ (26).
(24) Die blaue Elise stellt der Ameise eine Falle.
the blue Elise sets [Link] [Link] [Link] [Link]
(25) a. Der Ameise wird eine Falle gestellt.
[Link] [Link] becomes [Link] [Link] set
b. *Die Ameise wird eine Falle gestellt.
[Link] [Link] becomes [Link] [Link] set
(26) a. *Der Ameise bekommt eine Falle gestellt.
[Link] [Link] gets [Link] [Link] set
b. Die Ameise bekommt eine Falle gestellt.
[Link] [Link] gets [Link] [Link] set
Structural case and double object verbs
Question:
Which head assigns dative and which assigns accusative in a double object
construction? Assuming that the indirect object in double object
constructions is introduced by the functional head Appl (“applicative”),
which takes VP as its complement, the following analytical options arise.
Different analyses:
Dative is assigned by Appl, accusative is assigned by v (as in
transitives), (27-a) (see Georgala and Whitman 2009, Georgala 2011
for different lexical cases assigned by Appl).
Accusative is assigned by Appl and dative is assigned by v, (27-b)
(Ura 1996, 2000, McGinnis 1998, Doggett 2004).
Structural case and double object verbs
(27) a. vP b. vP
v ApplP v ApplP
[ucs:acc] [ucs:dat]
NPIO Appl′ NPIO Appl′
[cs:dat] [cs:dat]
Appl VP Appl VP
[ucs:dat] [ucs:acc]
V NPDO V NPDO
[cs:acc] [cs:acc]
Structural case and double object verbs
Ad (27-a):
The analysis allows a uniform treatment of the case capacities of v in
both transitives and ditransitives: in both cases v checks/values
accusative.
However, it involves a complication with respect to locality theory
(“minimality”): accusative checking/valuation by v must cross the
indirect object, which is closer to v than the direct object. We come
back to minimality later.
Note that case can be checked/valued under c-command if the probe
on Appl projects onto Appl′ . However, it needs to be said why Appl
cannot assign its case downward (e.g., before the indirect object is
merged).
Ad (27-b):
Here, no minimality issue arises because each head checks/values
case on the closest NP in its c-command domain.
However, v in ditransitives must now be different from v in transitives
with respect to the exact case feature it checks/values.
Accusative vs. ergative case alignment
Two types of languages (others exist):
Many languages (English, Japanese, Latvian, Hungarian, . . . ) assign
the same case (called nominative) to the external argument of
transitives and the sole argument to intransitives (be they unergative
or unaccusative), while the internal argument of a transitive gets a
special case: accusative.
Fewer languages (Hunzib, Tukang-Bezi, Chukchi, Yup’ik, . . . ) assign
the same case (called absolutive) to the internal argument of
transitives and the sole argument to intransitives (be they unergative
or unaccusative), while the external argument of a transitive gets a
special case: ergative.
Accusative vs. ergative case alignment
(28) Latvian (Mathiassen 1997):
a. Putn-s lidoja.
bird-nom [Link].3
‘The bird was flying.’
b. Bērn-s zı̄mē sun-i.
child-nom [Link].3 dog-acc
‘The child is drawing a dog.’
(29) Hunzib (van den Berg 1995):
a. kid-Ø y-ut’-ur
girl-abs cl2-sleep-pst
‘The girl slept.’
b. oždi-l kid-Ø hehe-r
boy-erg girl-abs hit-pst
‘The boy hit the girl.’
Accusative vs. ergative case alignment
Question:
Which head assigns ergative and which assigns absolutive in the transitive
clause of an ergative-aligning language? Interestingly, similar analytical
options arise as with double objects.
Different analyses:
Ergative is checked/valued by v, absolutive by T, (30-a)
(Levin and Massam 1985, Bobaljik 1993, Chomsky 1993).
Ergative is checked/valued by T, and absolutive by v, (30-b)
(Murasugi 1992, Jelinek 1993, Ura 2000).
Ergative-accusative parameter in first analysis:
In ergative languages, v checks/values case on the external argument, in
accusative languages it checks/values case on the internal argument. (T
then just values that remaining argument in the transitive case.)
Ergative-accusative parameter in second analysis:
In ergative languages, the head that checks/values case in intransitive
scenarios (the “active” head) is v; in accusative languages, it is T.
Accusative vs. ergative case alignment
(30) a. TP b. TP
T vP T vP
[ucs:abs] [ucs:erg]
NPext v′ NPext v′
[cs:erg] [cs:erg]
v VP v VP
[ucs:erg] [ucs:abs]
V NPint V NPint
[cs:abs] [cs:abs]
Accusative vs. ergative case alignment
Intransitives:
For intransitives, it must be ensured that the functional head that
assigns absolutive remains active (while the other one is deactivated).
In the first theory, v is inactive (as in accusative languages). T checks/
values absolutive on the only argument (NPext or NPint ) (31-a). In the
second theory, T is inactive, and v checks absolutive case (31-b).
(31) a. TP b. TP
T vP T vP
[ucs:abs]
NPext v′ NPext v′
[cs:abs] [cs:abs]
v VP v VP
[ucs:abs]
V NPint V NPint
[cs:abs] [cs:abs]
Accusative vs. ergative case alignment
Ad (30-a)/(31-a):
The questions that arise are: a) Why doesn’t v check/value ergative
case on the internal argument (e.g., before the external argument is
merged)? b) How can v check/value absolutive across the external
argument (“minimality”) and across v′ (“phase theory”; we return to
phases later).
The case that is cross-linguistically often morphologically unmarked
(nominative/absolutive) and the case that is often morphologically
marked (accusative/ergative) are checked/valued by the same head
(namely T and v, respectively) in accusative and in ergative languages.
Ad (30-b)/(31-b):
No locality issues arise (both minimality and phase theory are
respected).
The morphologically unmarked case is checked/valued by the head
that remains active in the intransitive scenario in accusative and
ergative languages alike (while the morphologically marked case is
checked/valued by the head that becomes inactive in intransitives).
Bibliography I
Berg, H. v. d. (1995). A Grammar of Hunzib. Lincom Studies in Caucasian
Linguistics. Lincom Europa, München.
Bjorkman, B. M. and Zeijlstra, H. (2019). Checking up on (φ-)agree.
Linguistic Inquiry, 50:527–569.
Bobaljik, J. (1993). Ergativity and ergative unergatives. In Phillips, C.,
editor, Papers on case and agreement II, volume 19 of MITWPL, pages
45–88. MIT Press, Cambridge, Mass.
Burzio, L. (1981). Intransitive Verbs and Italian Auxiliaries. PhD thesis, MIT.
Burzio, L. (1986). Italian Syntax: A Government-Binding Approach. Reidel,
Dordrecht.
Chomsky, N. (1993). A minimalist program for linguistic theory. In Hale, K.
and Keyser, S. J., editors, The View from Building 20, pages 1–52. MIT
Press, Cambridge, Massachusetts.
Doggett, T. B. (2004). All Things Being Unequal: Locality in Movement. PhD
thesis, MIT.
Fanselow, G. and Felix, S. (1987). Sprachtheorie. Band 1: Grundlagen und
Zielsetzungen. UTB Francke, Tübingen.
Bibliography II
Georgala, E. (2011). Why German is not an exception to the universal base
order of double object constructions. In Washburn, M. B.,
McKinney-Bock, K., Varis, E., Sawyer, A., and Tomaszewicz, B., editors,
Proceedings of the 28th West Coast Conference on Formal Linguistics,
pages 96–105, Somerville, MA. Cascadilla Proceedings Project.
Georgala, E. and Whitman, J. (2009). Ditransitives and applicative
structure in Greek. In Adams, N., Cooper, A., Parrill, F., and Wier, T.,
editors, Proceedings from the Annual Meeting of the Chicago Linguistic
Society 43, pages 77–91, Chicago. Chicago Linguistic Society.
Jelinek, E. (1993). Ergative ‘splits’ and argument type. In Bobaljik, J. and
Phillips, C., editors, Papers on Case and Agreement I, Vol. 18 of MIT
Working Papers in Linguistics, pages 15–42. MIT Press, Cambridge,
Massachusetts.
Levin, J. and Massam, D. (1985). Surface ergativity: Case/theta relations
reexamined. In Berman, S., editor, Proceedings of NELS 15, pages
286–301, Amherst, Mass. BLS.
Mathiassen, T. (1997). A Short Grammar of Latvian. Slavica Publishers,
Columbus, Ohio.
Bibliography III
McGinnis, M. (1998). Locality in A-Movement. PhD thesis, MIT, Cambridge,
Massachusetts.
Murasugi, K. (1992). Crossing and Nested Paths: NP Movement in Accusative
and Ergative Languages. PhD thesis, MIT, Cambridge, Massachusetts.
Perlmutter, D. (1978). Impersonal passives and the unaccusative
hypothesis. In Forth Annual Meeting of the Berkeley Linguistics Society,
pages 157–190, Berkeley CA. BLS.
Sigurðsson, H. A. (2004). Icelandic non-nominative subjects, pages 137–159.
John Benjamins, Amsterdam.
Ura, H. (1996). Multiple Feature-Checking: A Theory of Grammatical
Function Splitting. PhD thesis, MIT, Cambridge, Massachusetts.
Ura, H. (2000). Checking Theory and Grammatical Functions in Universal
Grammar. Oxford University Press, Oxford.