Sometimes it's hard to be Coherent
Available from
Elizabeth Coppock's profile on Mendeley.
Page 1
Sometimes it's hard to be Coherent
SOMETIMES IT’S HARD TO BE COHERENT
Elizabeth E. Coppock
Stanford University
Proceedings of the LFG03 Conference
University at Albany, State University of New York
Miriam Butt and Tracy Holloway King (Editors)
2003
CSLI Publications
http://csli-publications.stanford.edu/
126
Elizabeth E. Coppock
Stanford University
Proceedings of the LFG03 Conference
University at Albany, State University of New York
Miriam Butt and Tracy Holloway King (Editors)
2003
CSLI Publications
http://csli-publications.stanford.edu/
126
Page 2
Abstract
The paper provides an analysis of the construction in Hungarian known as focus raising, illustrated below.
Ja´nos-t mond-t-am hogy jo¨n
John-ACC say-PAST-1SG that come.3SG
‘It is John that I said is coming.’
In focus raising, the focus of the matrix clause is identified with a grammatical function in the embed-
ded clause. Focus raising is particularly interesting because of the case marking on focus-raised subjects.
Rather than having nominative case (unmarked), they appear suffixed by the accusative marker, -t, as shown
in the example. This fact is analyzed in the paper as follows.
• Accusative NPs such as Ja´nost in the example function as objects; this is evidenced by verb-object
definiteness agreement, and the possibility of a reflexive pronoun in the same position.
• On the other hand, the object function is argued not to be thematic (i.e., semantically selected by the
matrix predicate), based on its ability to be realized as an expletive pronoun, and the long-distance
nature of focus raising.
A pattern of variation among Hungarian speakers observed by Gervain (2002) shows that this athematic
object function may host either functional or anaphoric binders for the embedded clause subject. Anaphoric
binding from an athematic position leads to a violation of the semantic Coherence condition (Dalrymple
2001, p. 243), but this condition still appears to apply in other cases. Therefore, an Optimality-Theoretic
treatment is proposed, in which the semantic Coherence condition is a violable constraint.
127
The paper provides an analysis of the construction in Hungarian known as focus raising, illustrated below.
Ja´nos-t mond-t-am hogy jo¨n
John-ACC say-PAST-1SG that come.3SG
‘It is John that I said is coming.’
In focus raising, the focus of the matrix clause is identified with a grammatical function in the embed-
ded clause. Focus raising is particularly interesting because of the case marking on focus-raised subjects.
Rather than having nominative case (unmarked), they appear suffixed by the accusative marker, -t, as shown
in the example. This fact is analyzed in the paper as follows.
• Accusative NPs such as Ja´nost in the example function as objects; this is evidenced by verb-object
definiteness agreement, and the possibility of a reflexive pronoun in the same position.
• On the other hand, the object function is argued not to be thematic (i.e., semantically selected by the
matrix predicate), based on its ability to be realized as an expletive pronoun, and the long-distance
nature of focus raising.
A pattern of variation among Hungarian speakers observed by Gervain (2002) shows that this athematic
object function may host either functional or anaphoric binders for the embedded clause subject. Anaphoric
binding from an athematic position leads to a violation of the semantic Coherence condition (Dalrymple
2001, p. 243), but this condition still appears to apply in other cases. Therefore, an Optimality-Theoretic
treatment is proposed, in which the semantic Coherence condition is a violable constraint.
127
Page 3
Focus raising: wh- movment without the wh-
In Hungarian, just as in English, question wh- phrases such as mikor ‘when’ are normally fronted to the left
periphery of the sentence, as in (1). Unlike in English, non-wh- phrases like holnap ‘tomorrow’ may front
to the same position, as shown in (2).1 (The fronted constituent is indicated in bold face.)
(1) Mikor mond-ott, hogy jo¨n?
when say-PAST.3SG that come.3SG
‘When did he say he would come?’
(2) Holnap mond-ott, hogy jo¨n.
tomorrow say-PAST.3SG that come.3SG
‘He said that he would come tomorrow’
(3) *Tomorrow did he say that he would come.
The phenomenon shown in (2) is often called focus raising, and has been analyzed in numerous works,
including Zolnay (1926), Mara´cz (1989), ´E. Kiss (1987, 1990), Kenesei (1994), Horvath (1986, 1995, 1998),
Lipta´k (2001), and perhaps most recently, (Gervain, 2002).
The focus-raising construction shares many properties with the wh- question construction in Hungarian.
In both constructions, the fronted element is expressed in the focus position of the matrix clause, imme-
diately preceding the verb. In addition, the extraction path may cross several finite clause boundaries, as
exemplified in (4), where ket dolgot ‘two things’ fronts across two CPs ( ´E. Kiss, 1987, p. 125).
(4) Ke´t dolgot hallottam, [CP hogy Ja´nos megı´ge´rt Ma´ria´nak, [CP hogy megtesz]]
two thing.ACC hear.PAST.1.SG. that John promised Mary.DAT that does
‘It was two things that I heard that John promised Mary that he would do.’
Moreover, when the matrix predicate in a focus-raising or question construction is a verb, it must be a bridge
verb, such as the translational equivalents of want and say; as shown in (6), extraction may not cross the
complement to a non-bridge verb such as figyel ‘warn’ ( ´E. Kiss, 2002).2
(6) *A kolle´ga´im egy dia´k figyelmeztettek hogy keres-ett
the my:colleagues a student.NOM warned:me.3PL that seek.3SG-PAST
‘My colleagues warned me that a student was looking for me.’
Finally, focus raising, like question formation, is sensitive to island constraints, such as the Complex Noun
Phrase Constraint ( ´E. Kiss, 1987). For example, (7b) is ungrammatical because extraction may not cross the
complex noun phrase azt, hogy e´rkezik ‘it, that he would arrive’.
(7) a. Ja´nos mondta [ azt, [ hogy holnap e´rkezik ]]
John said it-ACC that tomorrow arrives
‘It John who said it, that he would arrive tomorrow.’
1David Beaver (p.c.) points out that Tomorrow, he said that he would come is, on the other hand, grammatical, so English does
have a type of long-distance focus movement. My point is not to compare Hungarian and English in this paper, but the fact remains
that Hungarian and English differ in that long-distance focus-movement and question formation have identical syntax in Hungarian,
but not in English.
2Focus-raising may also cross adjectival predicates with clausal complements, however (Kenesei, 1994, p. 317):
(5) Emma´-t fontos hogy megl a´togas-s-ad
Emma-ACC important that visit-SBJ-2SGDEF
‘It is Emma whom it is important that you visit.’
128
In Hungarian, just as in English, question wh- phrases such as mikor ‘when’ are normally fronted to the left
periphery of the sentence, as in (1). Unlike in English, non-wh- phrases like holnap ‘tomorrow’ may front
to the same position, as shown in (2).1 (The fronted constituent is indicated in bold face.)
(1) Mikor mond-ott, hogy jo¨n?
when say-PAST.3SG that come.3SG
‘When did he say he would come?’
(2) Holnap mond-ott, hogy jo¨n.
tomorrow say-PAST.3SG that come.3SG
‘He said that he would come tomorrow’
(3) *Tomorrow did he say that he would come.
The phenomenon shown in (2) is often called focus raising, and has been analyzed in numerous works,
including Zolnay (1926), Mara´cz (1989), ´E. Kiss (1987, 1990), Kenesei (1994), Horvath (1986, 1995, 1998),
Lipta´k (2001), and perhaps most recently, (Gervain, 2002).
The focus-raising construction shares many properties with the wh- question construction in Hungarian.
In both constructions, the fronted element is expressed in the focus position of the matrix clause, imme-
diately preceding the verb. In addition, the extraction path may cross several finite clause boundaries, as
exemplified in (4), where ket dolgot ‘two things’ fronts across two CPs ( ´E. Kiss, 1987, p. 125).
(4) Ke´t dolgot hallottam, [CP hogy Ja´nos megı´ge´rt Ma´ria´nak, [CP hogy megtesz]]
two thing.ACC hear.PAST.1.SG. that John promised Mary.DAT that does
‘It was two things that I heard that John promised Mary that he would do.’
Moreover, when the matrix predicate in a focus-raising or question construction is a verb, it must be a bridge
verb, such as the translational equivalents of want and say; as shown in (6), extraction may not cross the
complement to a non-bridge verb such as figyel ‘warn’ ( ´E. Kiss, 2002).2
(6) *A kolle´ga´im egy dia´k figyelmeztettek hogy keres-ett
the my:colleagues a student.NOM warned:me.3PL that seek.3SG-PAST
‘My colleagues warned me that a student was looking for me.’
Finally, focus raising, like question formation, is sensitive to island constraints, such as the Complex Noun
Phrase Constraint ( ´E. Kiss, 1987). For example, (7b) is ungrammatical because extraction may not cross the
complex noun phrase azt, hogy e´rkezik ‘it, that he would arrive’.
(7) a. Ja´nos mondta [ azt, [ hogy holnap e´rkezik ]]
John said it-ACC that tomorrow arrives
‘It John who said it, that he would arrive tomorrow.’
1David Beaver (p.c.) points out that Tomorrow, he said that he would come is, on the other hand, grammatical, so English does
have a type of long-distance focus movement. My point is not to compare Hungarian and English in this paper, but the fact remains
that Hungarian and English differ in that long-distance focus-movement and question formation have identical syntax in Hungarian,
but not in English.
2Focus-raising may also cross adjectival predicates with clausal complements, however (Kenesei, 1994, p. 317):
(5) Emma´-t fontos hogy megl a´togas-s-ad
Emma-ACC important that visit-SBJ-2SGDEF
‘It is Emma whom it is important that you visit.’
128
Page 4
b. *Ja´nos holnap mondta [ azt, [ hogy e´rkezik ]]
John tomorrow said it-ACC that arrives
‘It was tomorrow that John said it, that he would arrive.’
These parallelisms indicate that Hungarian focus-raising and question formation instantiate a unified phe-
nomenon.
NPs bearing a variety of grammatical functions in an embedded clause may undergo focus raising, as
shown in (8), where a dative argument, an inessive adjunct, and a direct object, respectively, are shown
fronted via focus-raising.
(8) a. Ma´ria Ja´nos-nak akarta, hogy a csomag-ot ke´zbesı´tse´k
Mary John-DAT wanted that the parcel-ACC deliver.3PL
‘As for Mary, it was John that she wanted the parcel to be delivered to.’
b. Legjobban eb-ben a ke´rdes-ben szeretne´m, hogy megegyezz-u¨nk
most this-in the question-in would:like.1SG that agree.SBJ-1PL
‘It is on this question that I would like most that we agree.’
c. Mindenki Ma´ria´-t va´rta, hogy megva´lasztja´k
everyone Mary-ACC waited that elect.3PL
‘It was Mary that everybody expected that they elect.’
Usually, the fronted element bears the morphological case that it would bear if it were expressed in the lower
clause. Thus, the non-fronted version of (8a) is as in (9), where the dative noun phrase Ja´nosnak still bears
dative case.
(9) Ma´ria akarta, hogy a csomagot Ja´nos-nak ke´zbesitse´k.
Mary wanted that the parcel.ACC John-DAT deliver.3PL
‘As for Mary, it was John that she wanted the parcel to be delivered to.’
However, there is one class of exceptions to this general pattern, both in focus-raising and in question
formation. When the subject of the embedded clause raises to the matrix focus position, the fronted phrase
often (optionally for some speakers, obligatorily for others) bears accusative case:
(10) Pe´ter-t mond-t-a, hogy jo¨n
Pe´ter-ACC say-PAST-1SG.DEF that come.3SG
‘It is Peter who he/she said is coming.’
Pe´ter would of course get nominative, not accusative, downstairs:
(11) Ja´nos mond-t-a, hogy Pe´ter jo¨n
I say-PAST-1SG.DEF that Peter.NOM come.3SG
‘John said that Peter is coming’
It is this accusative marking that I seek to explain here.
§1 is devoted to determining whether, and in what sense accusative-marked, focus-raised elements orig-
inating from subject position (henceforth, Pe´tert) are objects, as the case-marking suggests. I will conclude
that while Pe´tert is in fact an object of the matrix clause, it is an athematic object.3 In §2, I will argue for
one dialect of Hungarian that this athematic object position is anaphorically, rather than functionally, iden-
tified with the embedded clause subject. This implies that the phenomenon is a raising construction with
anaphoric control, a possibility that is predicted typologically, but unattested.
3An athematic GF is one that is not semantically selected by the PRED that governs it; notationally, an athematic GF does not
appear inside the angle brackets enclosing the argument list of a PRED.
129
John tomorrow said it-ACC that arrives
‘It was tomorrow that John said it, that he would arrive.’
These parallelisms indicate that Hungarian focus-raising and question formation instantiate a unified phe-
nomenon.
NPs bearing a variety of grammatical functions in an embedded clause may undergo focus raising, as
shown in (8), where a dative argument, an inessive adjunct, and a direct object, respectively, are shown
fronted via focus-raising.
(8) a. Ma´ria Ja´nos-nak akarta, hogy a csomag-ot ke´zbesı´tse´k
Mary John-DAT wanted that the parcel-ACC deliver.3PL
‘As for Mary, it was John that she wanted the parcel to be delivered to.’
b. Legjobban eb-ben a ke´rdes-ben szeretne´m, hogy megegyezz-u¨nk
most this-in the question-in would:like.1SG that agree.SBJ-1PL
‘It is on this question that I would like most that we agree.’
c. Mindenki Ma´ria´-t va´rta, hogy megva´lasztja´k
everyone Mary-ACC waited that elect.3PL
‘It was Mary that everybody expected that they elect.’
Usually, the fronted element bears the morphological case that it would bear if it were expressed in the lower
clause. Thus, the non-fronted version of (8a) is as in (9), where the dative noun phrase Ja´nosnak still bears
dative case.
(9) Ma´ria akarta, hogy a csomagot Ja´nos-nak ke´zbesitse´k.
Mary wanted that the parcel.ACC John-DAT deliver.3PL
‘As for Mary, it was John that she wanted the parcel to be delivered to.’
However, there is one class of exceptions to this general pattern, both in focus-raising and in question
formation. When the subject of the embedded clause raises to the matrix focus position, the fronted phrase
often (optionally for some speakers, obligatorily for others) bears accusative case:
(10) Pe´ter-t mond-t-a, hogy jo¨n
Pe´ter-ACC say-PAST-1SG.DEF that come.3SG
‘It is Peter who he/she said is coming.’
Pe´ter would of course get nominative, not accusative, downstairs:
(11) Ja´nos mond-t-a, hogy Pe´ter jo¨n
I say-PAST-1SG.DEF that Peter.NOM come.3SG
‘John said that Peter is coming’
It is this accusative marking that I seek to explain here.
§1 is devoted to determining whether, and in what sense accusative-marked, focus-raised elements orig-
inating from subject position (henceforth, Pe´tert) are objects, as the case-marking suggests. I will conclude
that while Pe´tert is in fact an object of the matrix clause, it is an athematic object.3 In §2, I will argue for
one dialect of Hungarian that this athematic object position is anaphorically, rather than functionally, iden-
tified with the embedded clause subject. This implies that the phenomenon is a raising construction with
anaphoric control, a possibility that is predicted typologically, but unattested.
3An athematic GF is one that is not semantically selected by the PRED that governs it; notationally, an athematic GF does not
appear inside the angle brackets enclosing the argument list of a PRED.
129
Page 5
1 Objecthood and selection
The accusative case marking on Pe´tert in (10) indicates that it functions as a (direct) object (OBJ), since OBJs
normally go with accusative case in Hungarian. That hypothesis is shown schematically in (i). If that were
true, this construction would exemplify a mismatch of function and meaning; even if Pe´ter is functionally
the OBJ of the matrix clause, intuitively it is semantically interpreted only as the subject of the embedded
clause (Judit Gervain, p.c.). The alternative proposal would be that Pe´ter is not functionally an OBJ, even
though it formally bears accusative case, as shown schematically in (ii).
(12) (i)
FOC
[
CASE ACC
]
OBJ
PRED ‘say. . . OBJ. . . ’
(ii)
FOC
[
CASE ACC
]
PRED ‘say. . . ’
Even if Pe´tert functions syntactically as an argument in the matrix clause, we are not obligated to consider
it a semantic argument of the verb. Assuming that it is functionally an object is consistent with an analysis
conforming to schema (i) below, where the verb syntactically and semantically selects an object (i.e., the
object is thematic). That assumption would also be consistent with an analysis like (ii), where the verb
syntactically but not semantically selects an object (i.e., the object is athematic).
(13) (i)
OBJ
[
CASE ACC
]
PRED ‘say
〈
. . . OBJ. . .
〉
’
(ii)
OBJ
[
CASE ACC
]
PRED ‘say
〈
. . .
〉
OBJ’
I will argue in favor of analysis (ii) in (13), in which the accusative-marked, focus-raised element functions
syntactically as an argument of the matrix clause, but is not a semantic argument of the verb.
1.1 Objecthood
1.1.1 Definiteness agreement
Patterns of definiteness agreement suggest that Pe´tert is indeed an object of the matrix verb. In Hungarian,
verbs agree in definiteness with their direct object.4 Thus, the verb form of la´t ‘see’ is different in I saw a
bird, as in (15) from what it is in I saw the bird, as in (14).
(14) La´t-om a madar-at
see-1.SG.DEF the bird-ACC
‘I see the bird’
(15) La´t-ok egy madar-at
see-1.SG.INDEF an bird-ACC
‘I see an bird’
(The indefinite conjugation is also used when the verb has no object.)
When a definite NP such as the proper name Pe´ter (accusative Pe´ter-t) focus-raises, the matrix verb
appears in the definite conjugation (mond-t-a ‘say-PAST-3SG.DEF’), as in (10). Noun phrases with plural
numeric determiners such as ke´t fiu-t in (16) and question words such as ki-t in (17) are indefinite, so the
matrix verb goes in the indefinite conjugation (mond-ott ‘say-3SG.PAST.INDEF’) in these cases.
4See Bartos (1997) for discussion of the precise semantic characterization of the agreement; semantic definiteness is not obvi-
ously sufficient.
130
The accusative case marking on Pe´tert in (10) indicates that it functions as a (direct) object (OBJ), since OBJs
normally go with accusative case in Hungarian. That hypothesis is shown schematically in (i). If that were
true, this construction would exemplify a mismatch of function and meaning; even if Pe´ter is functionally
the OBJ of the matrix clause, intuitively it is semantically interpreted only as the subject of the embedded
clause (Judit Gervain, p.c.). The alternative proposal would be that Pe´ter is not functionally an OBJ, even
though it formally bears accusative case, as shown schematically in (ii).
(12) (i)
FOC
[
CASE ACC
]
OBJ
PRED ‘say. . . OBJ. . . ’
(ii)
FOC
[
CASE ACC
]
PRED ‘say. . . ’
Even if Pe´tert functions syntactically as an argument in the matrix clause, we are not obligated to consider
it a semantic argument of the verb. Assuming that it is functionally an object is consistent with an analysis
conforming to schema (i) below, where the verb syntactically and semantically selects an object (i.e., the
object is thematic). That assumption would also be consistent with an analysis like (ii), where the verb
syntactically but not semantically selects an object (i.e., the object is athematic).
(13) (i)
OBJ
[
CASE ACC
]
PRED ‘say
〈
. . . OBJ. . .
〉
’
(ii)
OBJ
[
CASE ACC
]
PRED ‘say
〈
. . .
〉
OBJ’
I will argue in favor of analysis (ii) in (13), in which the accusative-marked, focus-raised element functions
syntactically as an argument of the matrix clause, but is not a semantic argument of the verb.
1.1 Objecthood
1.1.1 Definiteness agreement
Patterns of definiteness agreement suggest that Pe´tert is indeed an object of the matrix verb. In Hungarian,
verbs agree in definiteness with their direct object.4 Thus, the verb form of la´t ‘see’ is different in I saw a
bird, as in (15) from what it is in I saw the bird, as in (14).
(14) La´t-om a madar-at
see-1.SG.DEF the bird-ACC
‘I see the bird’
(15) La´t-ok egy madar-at
see-1.SG.INDEF an bird-ACC
‘I see an bird’
(The indefinite conjugation is also used when the verb has no object.)
When a definite NP such as the proper name Pe´ter (accusative Pe´ter-t) focus-raises, the matrix verb
appears in the definite conjugation (mond-t-a ‘say-PAST-3SG.DEF’), as in (10). Noun phrases with plural
numeric determiners such as ke´t fiu-t in (16) and question words such as ki-t in (17) are indefinite, so the
matrix verb goes in the indefinite conjugation (mond-ott ‘say-3SG.PAST.INDEF’) in these cases.
4See Bartos (1997) for discussion of the precise semantic characterization of the agreement; semantic definiteness is not obvi-
ously sufficient.
130
Page 6
(16) Ke´t fiu-t mond-ott hogy jo¨n.
two boy-ACC say-3SG.INDEF.PAST that come.3SG
‘It was two boys that he/she said were coming.’
(17) Ki-t mond-ott, hogy jo¨n?
who-ACC say-3SG.INDEF.PAST that come.3SG
‘Who did he/she say is coming?’
If the choice between the definite and indefinite conjugation for a verb is based on the properties of its object,
as is normally assumed, then these accusative NPs are objects.
1.1.2 Reflexives
Another argument for the objecthood of accusative-case marked, focus-raised objects is the fact that reflexive
pronouns coreferential with the matrix subject may appear in this position, as in (18).
(18) ¨Onmaga´-ti mondta Pe´teri hogy szeret-i Mari-t
themself-ACC say-PAST-3SGDEF Peter.NOM that love-3SG.DEF Mary-ACC
‘It is himselfi that Peteri said loves Mary.’
(This example is well-formed only as an answer to the Hungarian equivalent of the question, “Who did
Peter say loves Mary?”, or as a correction to an assertion that Peter said someone else loves Mary.) As I
will argue, the reflexive pronoun o¨nmaga´(t) must be bound by a syntactic coargument; an argument func-
tion syntactically selected by the same PRED. Therefore, the accusative NP in (18) occupies an argument
function.
Under the non-object analysis of accusative marked focus raised subjects schematized in (ii) of (12),
the reflexive pronoun in (18) is the embedded clause subject, bound by the subject of the matrix clause. If
o¨nmaga´t could be bound by the subject of a higher clause, we would predict ambiguity in (19); the reflexive
anaphor should be able to take either Ja´nos or Mari as an antecedent. (Note that pronouns in Hungarian are
not specified for gender.)
(19) Ja´nosj mond-t-a hogy o¨nmaga´ti,∗j szeret-i Marii
John.NOM mond-PAST-3SG.DEF that themself love-3SG.DEF Mary.NOM
‘John said that Mary loves herself.’
‘*John said that Mary loves himself.’
In fact, in this case o¨nmaga´t may only refer to Mari, the potential binder that is a coargument.
Towards the same point, if the non-object analysis were right, then the matrix subject should be able to
bind a reflexive pronoun in the subject of the embedded clause, even if focus raising has not taken place.
There is a nominative reflexive pronoun in Hungarian (o¨nmaga), and this pronoun cannot be bound by the
subject of the matrix clause either, as in (20).
(20) *Ja´nos mond-t-a hogy o¨nmaga szeret-i Mari-t
John say-PAST-3SG.DEF that themself love-3SG.DEF Mary-ACC
‘It is John who said that himself loves Mary.’
Thus, the reflexive pronoun must be bound within its clause. Example (18) therefore provides evidence that
accusative-marked focus raised subjects have a grammatical function in the matrix clause.5
5Thanks to Amy Dahlstrom for suggesting this avenue of inquiry. Thanks also to Agnes Mihalik for providing these native
speaker judgments.
131
two boy-ACC say-3SG.INDEF.PAST that come.3SG
‘It was two boys that he/she said were coming.’
(17) Ki-t mond-ott, hogy jo¨n?
who-ACC say-3SG.INDEF.PAST that come.3SG
‘Who did he/she say is coming?’
If the choice between the definite and indefinite conjugation for a verb is based on the properties of its object,
as is normally assumed, then these accusative NPs are objects.
1.1.2 Reflexives
Another argument for the objecthood of accusative-case marked, focus-raised objects is the fact that reflexive
pronouns coreferential with the matrix subject may appear in this position, as in (18).
(18) ¨Onmaga´-ti mondta Pe´teri hogy szeret-i Mari-t
themself-ACC say-PAST-3SGDEF Peter.NOM that love-3SG.DEF Mary-ACC
‘It is himselfi that Peteri said loves Mary.’
(This example is well-formed only as an answer to the Hungarian equivalent of the question, “Who did
Peter say loves Mary?”, or as a correction to an assertion that Peter said someone else loves Mary.) As I
will argue, the reflexive pronoun o¨nmaga´(t) must be bound by a syntactic coargument; an argument func-
tion syntactically selected by the same PRED. Therefore, the accusative NP in (18) occupies an argument
function.
Under the non-object analysis of accusative marked focus raised subjects schematized in (ii) of (12),
the reflexive pronoun in (18) is the embedded clause subject, bound by the subject of the matrix clause. If
o¨nmaga´t could be bound by the subject of a higher clause, we would predict ambiguity in (19); the reflexive
anaphor should be able to take either Ja´nos or Mari as an antecedent. (Note that pronouns in Hungarian are
not specified for gender.)
(19) Ja´nosj mond-t-a hogy o¨nmaga´ti,∗j szeret-i Marii
John.NOM mond-PAST-3SG.DEF that themself love-3SG.DEF Mary.NOM
‘John said that Mary loves herself.’
‘*John said that Mary loves himself.’
In fact, in this case o¨nmaga´t may only refer to Mari, the potential binder that is a coargument.
Towards the same point, if the non-object analysis were right, then the matrix subject should be able to
bind a reflexive pronoun in the subject of the embedded clause, even if focus raising has not taken place.
There is a nominative reflexive pronoun in Hungarian (o¨nmaga), and this pronoun cannot be bound by the
subject of the matrix clause either, as in (20).
(20) *Ja´nos mond-t-a hogy o¨nmaga szeret-i Mari-t
John say-PAST-3SG.DEF that themself love-3SG.DEF Mary-ACC
‘It is John who said that himself loves Mary.’
Thus, the reflexive pronoun must be bound within its clause. Example (18) therefore provides evidence that
accusative-marked focus raised subjects have a grammatical function in the matrix clause.5
5Thanks to Amy Dahlstrom for suggesting this avenue of inquiry. Thanks also to Agnes Mihalik for providing these native
speaker judgments.
131
Page 7
1.2 Semantic selection
Again, if accusative-marked focus-raised embedded-clause subjects such as Pe´tert in (10) are objects, they
are not necessarily semantically selected objects. Indeed, I will argue that they are not.6
1.2.1 Dependency is not lexically enforced
First, let us suppose the alternative: accusative-marked fronted subjects are semantically selected. Then we
have a lexical entry for bridge verbs such as mond ‘say’ that includes the following lexical specification:
(21) (↑PRED)=‘say〈SUBJ OBJ COMP〉’
Supposing this were the lexical entry, we would predict (22) to be possible, in the absence of any other
mechanism for ruling it out. (22) is ungrammatical because the matrix object function cannot be identified
functionally or anaphorically with any participant in the embedded clause.
(22) *Pe´ter-t mond-t-am, hogy Ja´nos jo¨n
Pe´ter-ACC say-PAST-1SG that John.NOM come.3SG
‘[roughly] Peter did I say that John is coming’
The putative lexical entry in (21) is like the lexical entry for an equi control verb, which by definition
thematically selects the controller along with a complement clause (normally XCOMP). An equi verb such
as try enforces identity between the controller and the controllee through a stipulation of referential equality
between the matrix controller and the SUBJ of the embedded clause, in the verb’s lexical entry (Dalrymple,
2001). This precise solution will not do for us, because focus-raising is possible over long distances, as
exemplified earlier in (4). Thus, the identity stipulation would have to involve a functional uncertainty
equation allowing the controllee to be embedded within multiple COMPs. However, the possible extraction
paths for focus-raising mirror precisely the possible extraction paths for wh- questions. For example, it is
not possible to focus-raise out of a complex noun phrase, as demonstrated earlier in (7b). This parallelism
shows that the range of dependencies in focus-raising constructions is not determined lexically, by bridge
verbs, but by the aspect of the grammar that regulates long-distance dependencies. Supposing that bridge
verbs semantically select for accusative-marked focus-raised subjects prevents us from making use of the
grammar to derive the constraints on extraction.
In contrast, if we assume that accusative-marked fronted subjects are not semantically selected by the
verb, then we can use the grammar to impose identity between the higher-clause object and an embedded
subject. If the object is athematic, then the representation for (10) will consist at least of (23).
(23)
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
OBJ
[
PRED ‘John’
]
COMP
PRED ‘come
〈
SUBJ
〉
’
If OBJ is not functionally identified with any function that is semantically selected, then the condition in (24)
will be violated.
6I assume with little argument that these accusative-marked elements are indeed subcategorized for by the verb if they are
objects. However, an analysis using nonsubcategorized objects may be correct for the cognate object construction (e.g. Suzy
smiled a pretty smile) as well as certain objectlike temporal adverbials (e.g. I slept two hours), so it may be fruitful to explore that
possibility.
132
Again, if accusative-marked focus-raised embedded-clause subjects such as Pe´tert in (10) are objects, they
are not necessarily semantically selected objects. Indeed, I will argue that they are not.6
1.2.1 Dependency is not lexically enforced
First, let us suppose the alternative: accusative-marked fronted subjects are semantically selected. Then we
have a lexical entry for bridge verbs such as mond ‘say’ that includes the following lexical specification:
(21) (↑PRED)=‘say〈SUBJ OBJ COMP〉’
Supposing this were the lexical entry, we would predict (22) to be possible, in the absence of any other
mechanism for ruling it out. (22) is ungrammatical because the matrix object function cannot be identified
functionally or anaphorically with any participant in the embedded clause.
(22) *Pe´ter-t mond-t-am, hogy Ja´nos jo¨n
Pe´ter-ACC say-PAST-1SG that John.NOM come.3SG
‘[roughly] Peter did I say that John is coming’
The putative lexical entry in (21) is like the lexical entry for an equi control verb, which by definition
thematically selects the controller along with a complement clause (normally XCOMP). An equi verb such
as try enforces identity between the controller and the controllee through a stipulation of referential equality
between the matrix controller and the SUBJ of the embedded clause, in the verb’s lexical entry (Dalrymple,
2001). This precise solution will not do for us, because focus-raising is possible over long distances, as
exemplified earlier in (4). Thus, the identity stipulation would have to involve a functional uncertainty
equation allowing the controllee to be embedded within multiple COMPs. However, the possible extraction
paths for focus-raising mirror precisely the possible extraction paths for wh- questions. For example, it is
not possible to focus-raise out of a complex noun phrase, as demonstrated earlier in (7b). This parallelism
shows that the range of dependencies in focus-raising constructions is not determined lexically, by bridge
verbs, but by the aspect of the grammar that regulates long-distance dependencies. Supposing that bridge
verbs semantically select for accusative-marked focus-raised subjects prevents us from making use of the
grammar to derive the constraints on extraction.
In contrast, if we assume that accusative-marked fronted subjects are not semantically selected by the
verb, then we can use the grammar to impose identity between the higher-clause object and an embedded
subject. If the object is athematic, then the representation for (10) will consist at least of (23).
(23)
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
OBJ
[
PRED ‘John’
]
COMP
PRED ‘come
〈
SUBJ
〉
’
If OBJ is not functionally identified with any function that is semantically selected, then the condition in (24)
will be violated.
6I assume with little argument that these accusative-marked elements are indeed subcategorized for by the verb if they are
objects. However, an analysis using nonsubcategorized objects may be correct for the cognate object construction (e.g. Suzy
smiled a pretty smile) as well as certain objectlike temporal adverbials (e.g. I slept two hours), so it may be fruitful to explore that
possibility.
132
Page 8
(24) Semantic Coherence (Dalrymple, 2001, p. 243)
A meaning derivation for an utterance is semantically coherent if the meaning derivation produces a
meaning for the utterance with no additional unused premises remaining.
This principle is independently motivated as a means of ruling out examples like (25).7,8
(25) a. *John rained.
b. *John seems that Mary is nice.
For simplicity of exposition, we may obtain the effect of the semantic coherence condition for our pur-
poses without using glue semantics, if we assume that every instance of a PRED attribute corresponds to a
glue semantic resource, and that semantic government corresponds to the using up of premises. Using this
correspondence, we may state the semantic Coherence condition as follows:
(26) Semantic Coherence (non-glue version):
Every governable function containing a PRED value must be semantically governed.
(27) Semantic government (for our purposes)
f is semantically governed iff f is functionally linked to a position inside the angle brackets in the
semantic form value of a PRED attribute.
In the examples in (25), the SUBJ function contains a PRED attribute, yet is not semantically governed. In
this way, examples (25) fail to satisfy (26).9
We can use this principle to make it necessary that the athematic object in (10) Ja´nos-t ‘John-ACC’ is
linked to some function of the embedded clause in the following way. If, as in (22), the athematic object is
not linked to any participant in the embedded clause, the semantic Coherence condition is violated.
The possibility of identifying the matrix focus with the embedded subject also follows from independent
principles of the grammar. We may assume a rule for regular question formation whereby an element from
a possibly embedded clause raises to the preverbal focus position, such as the following:10
(28) FP → (NP) F′
(↑FOCUS)=↓ ↑=↓
(↑COMP* GF)=↓
Thus, principles of the grammar (the semantic Coherence condition and the regular rule of question forma-
tion) can be made responsible for the identification between the matrix focus and some embedded function.
In fact, these principles alone are insufficient for deriving an f-structure representation of (10) in which
the matrix FOC, the matrix OBJ, and the embedded SUBJ are all functionally identified with one another.
7Under usual assumptions about the lexical entry for seem, the examples in (25) are also ruled out by the requirement that the
PRONFORM attribute have the value IT. So, the semantic Coherence condition technically isn’t needed to rule these out. However,
supposing that a more principled theory of the distribution of it vs. there were to replace this lexical stipulation, the semantic
Coherence condition would be needed in order to rule that sentence out.
8It appeared to me at first that the English “copy raising” construction represented below involves an athematic subject that is
merely anaphorically integrated: Richard seems like he’s in trouble (Potsdam and Runner, 2001). Asudeh (2002) gives an analysis
of the copy-raising construction in LFG, in which the subject is in fact governed by like, so there is no semantic Coherence violation
in this type of example according to that analysis.
9The semantic Coherence condition may be generalizable in such a way that it applies to discourse functions TOPIC and FOCUS.
In fact, such a condition is proposed in Falk (2001), under the name of the Extended Coherence Condition: “All functions in an
f-structure must be incorporated into the semantics. Argument funcitons are subject to the Coherence Condition. Overlay functions
[incl. focus –EEC] must be identified with arguments or adjuncts; adjuncts must be in f-structures containing PREDs.”
10See Szendro˝i (2001) for a good argument for a focus projection, as well as Brody (1990, 1995) for the original suggestion that
Hungarian uses a focus projection.
133
A meaning derivation for an utterance is semantically coherent if the meaning derivation produces a
meaning for the utterance with no additional unused premises remaining.
This principle is independently motivated as a means of ruling out examples like (25).7,8
(25) a. *John rained.
b. *John seems that Mary is nice.
For simplicity of exposition, we may obtain the effect of the semantic coherence condition for our pur-
poses without using glue semantics, if we assume that every instance of a PRED attribute corresponds to a
glue semantic resource, and that semantic government corresponds to the using up of premises. Using this
correspondence, we may state the semantic Coherence condition as follows:
(26) Semantic Coherence (non-glue version):
Every governable function containing a PRED value must be semantically governed.
(27) Semantic government (for our purposes)
f is semantically governed iff f is functionally linked to a position inside the angle brackets in the
semantic form value of a PRED attribute.
In the examples in (25), the SUBJ function contains a PRED attribute, yet is not semantically governed. In
this way, examples (25) fail to satisfy (26).9
We can use this principle to make it necessary that the athematic object in (10) Ja´nos-t ‘John-ACC’ is
linked to some function of the embedded clause in the following way. If, as in (22), the athematic object is
not linked to any participant in the embedded clause, the semantic Coherence condition is violated.
The possibility of identifying the matrix focus with the embedded subject also follows from independent
principles of the grammar. We may assume a rule for regular question formation whereby an element from
a possibly embedded clause raises to the preverbal focus position, such as the following:10
(28) FP → (NP) F′
(↑FOCUS)=↓ ↑=↓
(↑COMP* GF)=↓
Thus, principles of the grammar (the semantic Coherence condition and the regular rule of question forma-
tion) can be made responsible for the identification between the matrix focus and some embedded function.
In fact, these principles alone are insufficient for deriving an f-structure representation of (10) in which
the matrix FOC, the matrix OBJ, and the embedded SUBJ are all functionally identified with one another.
7Under usual assumptions about the lexical entry for seem, the examples in (25) are also ruled out by the requirement that the
PRONFORM attribute have the value IT. So, the semantic Coherence condition technically isn’t needed to rule these out. However,
supposing that a more principled theory of the distribution of it vs. there were to replace this lexical stipulation, the semantic
Coherence condition would be needed in order to rule that sentence out.
8It appeared to me at first that the English “copy raising” construction represented below involves an athematic subject that is
merely anaphorically integrated: Richard seems like he’s in trouble (Potsdam and Runner, 2001). Asudeh (2002) gives an analysis
of the copy-raising construction in LFG, in which the subject is in fact governed by like, so there is no semantic Coherence violation
in this type of example according to that analysis.
9The semantic Coherence condition may be generalizable in such a way that it applies to discourse functions TOPIC and FOCUS.
In fact, such a condition is proposed in Falk (2001), under the name of the Extended Coherence Condition: “All functions in an
f-structure must be incorporated into the semantics. Argument funcitons are subject to the Coherence Condition. Overlay functions
[incl. focus –EEC] must be identified with arguments or adjuncts; adjuncts must be in f-structures containing PREDs.”
10See Szendro˝i (2001) for a good argument for a focus projection, as well as Brody (1990, 1995) for the original suggestion that
Hungarian uses a focus projection.
133
Page 9
The rule given in (28) allows either the embedded subect to be identified with the matrix focus (in which
case the functional uncertainty path is instantiated as: COMP SUBJ), or the matrix object (in which case
the functional uncertainty path is instantiated as: OBJ). If we link the embedded SUBJ to the matrix FOC,
another equation is necessary for linking the matrix OBJ into the functional “chain” if you will, either by
equating it with the matrix FOC or by equating it with the embedded SUBJ. To allow this, I propose the
additional annotation on the focus position in (29).
(29) FP → (NP) F′
(↓CASE)=ACC⇒ (↑OBJ)=↓
In other words, if any constituent is accusative-marked in the focus position of a clause, then it is interpreted
as the object of that clause, whether or not it is also interpreted as the object of another clause. This will link
the matrix OBJ into the functional chain, even if the focus rule is used to link the embedded subject with the
matrix focus, instead of the matrix object.
Independent evidence for this solution comes from the fact that definiteness agreement in focus-raising is
not limited to embedded-clause subjects. Any focus-raised phrase with accusative case triggers definiteness
agreement with the matrix verb. Thus, phrases corresponding to embedded-clause objects, which retain their
definiteness marking from the embedded clause, also trigger definiteness agreement in the matrix clause, as
shown in the contrast between (30a) and (30b).
(30) a. Csak ez-t akar-om/*-ok hogy el-mond-j-ad (Kenesei, 1994)
only this-ACC want-1SG.DEF/1SG.INDEF that out-say-SBJ-2SG.DEF
‘It’s only this that I want you to say.’
b. Csak ke´t dolg-ot akar-ok/*-om hogy el-mond-j-a´l
only two thing-ACC want-1SG.INDEF/1SG.DEF that out-say-SUBJUNCTIVE-2SG.INDEF
‘There’s only two things that I want you to say.’
In (30a), the focus-raised element ezt ‘this’ is definite, so the verb akarom ‘want’ is as well. In (30b), the
numerically-quantified NP ke´t dolgot ‘two things’ is indefinite, and so is the matrix verb akarok ‘want’. If
definiteness agreement is sufficient evidence for objecthood, then it appears that embedded clause objects
become matrix clause objects when they focus raise as well as embedded clause subjects.
1.3 Expletive construction
If the matrix verb has an athematic object, then that object position should be able to be filled with an
expletive. Indeed, that is what we find in examples like (31), I would like to claim.
(31) Azt la´t-om, hogy Pe´ter jo¨n
it.ACC see-1.SG.DEF that Peter.NOM come.3.SG
‘I see that Peter is coming.’
I will argue that in this construction, the accusative pronoun azt is an expletive, filling the athematic OBJ
position.
Following Berman (2003) in her analysis of similar constructions in German, I will refer to the construc-
tion instantiated in (31) as the “correlative” construction. For such constructions, there are two potential
analyses, corresponding to two different types of lexical entries for bridge verbs.
(32) a. Extraposition analysis
The accusative pronoun is the head of the object selected by the verb, and the clause
134
case the functional uncertainty path is instantiated as: COMP SUBJ), or the matrix object (in which case
the functional uncertainty path is instantiated as: OBJ). If we link the embedded SUBJ to the matrix FOC,
another equation is necessary for linking the matrix OBJ into the functional “chain” if you will, either by
equating it with the matrix FOC or by equating it with the embedded SUBJ. To allow this, I propose the
additional annotation on the focus position in (29).
(29) FP → (NP) F′
(↓CASE)=ACC⇒ (↑OBJ)=↓
In other words, if any constituent is accusative-marked in the focus position of a clause, then it is interpreted
as the object of that clause, whether or not it is also interpreted as the object of another clause. This will link
the matrix OBJ into the functional chain, even if the focus rule is used to link the embedded subject with the
matrix focus, instead of the matrix object.
Independent evidence for this solution comes from the fact that definiteness agreement in focus-raising is
not limited to embedded-clause subjects. Any focus-raised phrase with accusative case triggers definiteness
agreement with the matrix verb. Thus, phrases corresponding to embedded-clause objects, which retain their
definiteness marking from the embedded clause, also trigger definiteness agreement in the matrix clause, as
shown in the contrast between (30a) and (30b).
(30) a. Csak ez-t akar-om/*-ok hogy el-mond-j-ad (Kenesei, 1994)
only this-ACC want-1SG.DEF/1SG.INDEF that out-say-SBJ-2SG.DEF
‘It’s only this that I want you to say.’
b. Csak ke´t dolg-ot akar-ok/*-om hogy el-mond-j-a´l
only two thing-ACC want-1SG.INDEF/1SG.DEF that out-say-SUBJUNCTIVE-2SG.INDEF
‘There’s only two things that I want you to say.’
In (30a), the focus-raised element ezt ‘this’ is definite, so the verb akarom ‘want’ is as well. In (30b), the
numerically-quantified NP ke´t dolgot ‘two things’ is indefinite, and so is the matrix verb akarok ‘want’. If
definiteness agreement is sufficient evidence for objecthood, then it appears that embedded clause objects
become matrix clause objects when they focus raise as well as embedded clause subjects.
1.3 Expletive construction
If the matrix verb has an athematic object, then that object position should be able to be filled with an
expletive. Indeed, that is what we find in examples like (31), I would like to claim.
(31) Azt la´t-om, hogy Pe´ter jo¨n
it.ACC see-1.SG.DEF that Peter.NOM come.3.SG
‘I see that Peter is coming.’
I will argue that in this construction, the accusative pronoun azt is an expletive, filling the athematic OBJ
position.
Following Berman (2003) in her analysis of similar constructions in German, I will refer to the construc-
tion instantiated in (31) as the “correlative” construction. For such constructions, there are two potential
analyses, corresponding to two different types of lexical entries for bridge verbs.
(32) a. Extraposition analysis
The accusative pronoun is the head of the object selected by the verb, and the clause
134
Page 10
is contained within the OBJ as a complement to the pronoun, or related to it through
apposition or adjunct modification. This analysis may be schematized thus:
PRED ‘say
〈
SUBJ OBJ
〉
’
OBJ
“azt”
GF
[
“hogy Pe´ter jo¨n”
]
b. Expletive analysis
The accusative pronoun is an athematic object, and the clause is the direct complement
of the verb. It is schematized thus:
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
OBJ
[
“azt”
]
COMP
[
“hogy Pe´ter jo¨n”
]
(The extraposition analysis is the LFG equivalent of the analysis of correlative constructions given by ´E.
Kiss (1990), wherein the clause is extraposed, leaving a trace within the object NP; the expletive analysis
is more like the analysis given by Kenesei (1994), in which the accusative pronoun is an expletive and the
clause is a direct complement of the matrix verb.) For a summary of the possibilities, either the accusative
pronoun (32a) or the clause (32b) is semantically selected by the matrix verb. I offer two arguments for the
expletive analysis, wherein the clause is semantically selected.
1.3.1 Possibility of extraction
Kenesei (1994) argues for the expletive analysis on the basis of the focus-raising construction itself, which
clearly shows that there must be at least some lexical entry for bridge verbs in which the complement is
directly selected by the verb. His argument is as follows. Extraction is not possible from complex noun
phrases, or adjunct phrases, and yet it is possible in examples like (2) and (10). If extraction is possible
only from argument clauses, then the CP in (10) must be an argument of the verb, not a right-dislocated
subconstituent of a complex noun phrase object.
Further evidence against the analysis involving extraposition from a complex noun phrase comes from
the complementary distribution of focus-raising constructions with the accusative pronoun. Under the ex-
traposition analysis of (31), the CP [hogy e´rkezik] is a subordinate part of a complex noun phrase even in
focus-raising constructions, so an example like (2) has the analysis in (33), with a silent pronoun heading
the noun phrase complement of the verb.
(33) Ja´nos holnap mond-t-a [NP [CP hogy e´rkezik ]]
John tomorrow say-PAST-3.SGDEF that arrives
‘It was tomorrow that John said it that he would arrive.’
But the silent head of the putative complex noun phrase cannot be spelled out:11
11
´E. Kiss (1990) maintains that the extraposition analysis is correct even for focus-raising cases, and proposes to deal with the
ungrammaticality of azt in (34) in the following way.
Owing to the Visibility Condition [Chomsky (1981)], sentential arguments are assigned case. A CP category,
however, cannot bear case – therefore, it must be formally subordinated to an NP. It is the complex NP that functions
as the argument of the V, picking up its case and theta-role. The case will be borne by the pronominal head of the
135
apposition or adjunct modification. This analysis may be schematized thus:
PRED ‘say
〈
SUBJ OBJ
〉
’
OBJ
“azt”
GF
[
“hogy Pe´ter jo¨n”
]
b. Expletive analysis
The accusative pronoun is an athematic object, and the clause is the direct complement
of the verb. It is schematized thus:
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
OBJ
[
“azt”
]
COMP
[
“hogy Pe´ter jo¨n”
]
(The extraposition analysis is the LFG equivalent of the analysis of correlative constructions given by ´E.
Kiss (1990), wherein the clause is extraposed, leaving a trace within the object NP; the expletive analysis
is more like the analysis given by Kenesei (1994), in which the accusative pronoun is an expletive and the
clause is a direct complement of the matrix verb.) For a summary of the possibilities, either the accusative
pronoun (32a) or the clause (32b) is semantically selected by the matrix verb. I offer two arguments for the
expletive analysis, wherein the clause is semantically selected.
1.3.1 Possibility of extraction
Kenesei (1994) argues for the expletive analysis on the basis of the focus-raising construction itself, which
clearly shows that there must be at least some lexical entry for bridge verbs in which the complement is
directly selected by the verb. His argument is as follows. Extraction is not possible from complex noun
phrases, or adjunct phrases, and yet it is possible in examples like (2) and (10). If extraction is possible
only from argument clauses, then the CP in (10) must be an argument of the verb, not a right-dislocated
subconstituent of a complex noun phrase object.
Further evidence against the analysis involving extraposition from a complex noun phrase comes from
the complementary distribution of focus-raising constructions with the accusative pronoun. Under the ex-
traposition analysis of (31), the CP [hogy e´rkezik] is a subordinate part of a complex noun phrase even in
focus-raising constructions, so an example like (2) has the analysis in (33), with a silent pronoun heading
the noun phrase complement of the verb.
(33) Ja´nos holnap mond-t-a [NP [CP hogy e´rkezik ]]
John tomorrow say-PAST-3.SGDEF that arrives
‘It was tomorrow that John said it that he would arrive.’
But the silent head of the putative complex noun phrase cannot be spelled out:11
11
´E. Kiss (1990) maintains that the extraposition analysis is correct even for focus-raising cases, and proposes to deal with the
ungrammaticality of azt in (34) in the following way.
Owing to the Visibility Condition [Chomsky (1981)], sentential arguments are assigned case. A CP category,
however, cannot bear case – therefore, it must be formally subordinated to an NP. It is the complex NP that functions
as the argument of the V, picking up its case and theta-role. The case will be borne by the pronominal head of the
135
Page 11
(34) *Ja´nos holnap mond-t-a azt hogy e´rkezik
John tomorrow say-PAST-3SG.DEF it-ACC that arrives
‘It was tomorrow that John said it that he would arrive.’
This evidence shows that there is a structural difference between extraposed clauses and clauses from which
focus-raising may take place, and therefore suggests that bridge verbs directly select for their clausal com-
plements, rather than being indirectly related to them as in the extraposition analysis.
1.3.2 Lack of direct object NPs
An empirical problem for the extraposition analysis of correlative constructions such as (31) is that it over-
generates direct objects. Because OBJ is a thematic argument of the verb in that analysis, the verb is predicted
to be able to take a simple direct object, but normally, this is impossible:
(35) *A hı´r-t mond-t-a
the news-ACC say-PAST-3SG.DEF
‘He said the news.’
The only elements that can appear as the sole noun phrase complement to bridge verbs are the pronouns az-t
‘that-ACC’ and mi-t ‘what-ACC’.
(36) a. Azt mondta. ‘He said it.’
b. Mit mondott? ‘What did he say?’
In fact, proper names may appear as the sole overt argument to a bridge verb, as in (37),
(37) Ja´nos-t mond-t-a.
John-ACC say-PAST-3SGDEF
‘He said John.’
but this example has an “elliptical character” and can only be used when the context specifies the content of
the elided clausal complement (Katalin Kiss, p.c.).12
We can explain the grammaticality of the examples in (36) with the following assumptions: the pronouns
az-t ‘that-ACC’ and mi-t ‘what-ACC’ operate as expletives in (36), and their associated clauses undergo CP
deletion. The idea that mit can operate as an expletive is supported by the “scope-marking” construction,
illustrated in (39).
NP. Since the pronominal head is a semantically empty dummy NP, the theta-role will be borne by the clause. The
clause inside the NP is formally an adjunct to the pronominal head; consequently, its CP boundary is a blocking
category and a barrier for its constituents. The NP dominating the CP inherits barrierhood; thus extraction from the
CP is prevented. If the pronominal head is phonologically empty, it becomes transparent, presumably because pro,
recoverable from the verbal suffix, is deleted, and the NP projection, having lost its head, is pruned. This way the
CP is directly governed by the V. [Italics mine]
Kenesei (1994) argues against this analysis on the basis that it violates the Projection Principle: “if the verb is subcategorized for
a noun phrase at one level of structure, it cannot have a clausal complement at another” (Kenesei, 1994, pp. 312–313). In LFG,
the Visibility Condition is built into the very framework; a verb cannot subcategorize for one grammatical function at one level of
representation and a different grammatical function at another, because there is only one level of representation at which grammtical
functions are represented, namely f-structure. Thus, this explanation is out of the question for an LFG analysis.
12There are other genuine examples of noun phrase complements to bridge verbs, such as (38), where the direct object is a
complex noun phrase headed by a pronoun, containing a clause.
(38) Mond-t- a´l [NP az-t [CP hogy J a´nos j o¨n ]]
say-PAST-2SG.DEF it-ACC that John come.3SG
‘I said it that John is coming.’
Assuming that indeed this example contains a complex noun phrase, we may be forced to assume that matrix verbs allow OBJ
complements. I am not able to determine whether or not we are forced to do so at this point.
136
John tomorrow say-PAST-3SG.DEF it-ACC that arrives
‘It was tomorrow that John said it that he would arrive.’
This evidence shows that there is a structural difference between extraposed clauses and clauses from which
focus-raising may take place, and therefore suggests that bridge verbs directly select for their clausal com-
plements, rather than being indirectly related to them as in the extraposition analysis.
1.3.2 Lack of direct object NPs
An empirical problem for the extraposition analysis of correlative constructions such as (31) is that it over-
generates direct objects. Because OBJ is a thematic argument of the verb in that analysis, the verb is predicted
to be able to take a simple direct object, but normally, this is impossible:
(35) *A hı´r-t mond-t-a
the news-ACC say-PAST-3SG.DEF
‘He said the news.’
The only elements that can appear as the sole noun phrase complement to bridge verbs are the pronouns az-t
‘that-ACC’ and mi-t ‘what-ACC’.
(36) a. Azt mondta. ‘He said it.’
b. Mit mondott? ‘What did he say?’
In fact, proper names may appear as the sole overt argument to a bridge verb, as in (37),
(37) Ja´nos-t mond-t-a.
John-ACC say-PAST-3SGDEF
‘He said John.’
but this example has an “elliptical character” and can only be used when the context specifies the content of
the elided clausal complement (Katalin Kiss, p.c.).12
We can explain the grammaticality of the examples in (36) with the following assumptions: the pronouns
az-t ‘that-ACC’ and mi-t ‘what-ACC’ operate as expletives in (36), and their associated clauses undergo CP
deletion. The idea that mit can operate as an expletive is supported by the “scope-marking” construction,
illustrated in (39).
NP. Since the pronominal head is a semantically empty dummy NP, the theta-role will be borne by the clause. The
clause inside the NP is formally an adjunct to the pronominal head; consequently, its CP boundary is a blocking
category and a barrier for its constituents. The NP dominating the CP inherits barrierhood; thus extraction from the
CP is prevented. If the pronominal head is phonologically empty, it becomes transparent, presumably because pro,
recoverable from the verbal suffix, is deleted, and the NP projection, having lost its head, is pruned. This way the
CP is directly governed by the V. [Italics mine]
Kenesei (1994) argues against this analysis on the basis that it violates the Projection Principle: “if the verb is subcategorized for
a noun phrase at one level of structure, it cannot have a clausal complement at another” (Kenesei, 1994, pp. 312–313). In LFG,
the Visibility Condition is built into the very framework; a verb cannot subcategorize for one grammatical function at one level of
representation and a different grammatical function at another, because there is only one level of representation at which grammtical
functions are represented, namely f-structure. Thus, this explanation is out of the question for an LFG analysis.
12There are other genuine examples of noun phrase complements to bridge verbs, such as (38), where the direct object is a
complex noun phrase headed by a pronoun, containing a clause.
(38) Mond-t- a´l [NP az-t [CP hogy J a´nos j o¨n ]]
say-PAST-2SG.DEF it-ACC that John come.3SG
‘I said it that John is coming.’
Assuming that indeed this example contains a complex noun phrase, we may be forced to assume that matrix verbs allow OBJ
complements. I am not able to determine whether or not we are forced to do so at this point.
136
Page 12
(39) Mi-t mond-ott, hogy mikor jo¨n?
what-ACC say-3SG.INDEF that when come.3SG
‘When did he/she say that he/she would come?’
See Horvath (1995) for an extensive set of arguments that mit is behaving as an expletive in the scope
marking construction.
Furthermore, although the pronoun azt cannot be present in the matrix clause when focus-raising or
question formation takes place, it is possible to relativize out of the complement to a bridge verb when the
expletive is present. (These were found using Google, but a native Hungarian speaker informant confirms
the grammaticality of these examples.)
(40) Az egyik dolog, amibeni azt gondolom, hogy jobbak vagyunk ti, mint ma´sok . . .
the primary thing in which it-ACC think-1SG that better be.1PL than others
‘The primary thing in which I think (it) that we are better than others. . . ’
www.hte.hu/kiadvanyok/hirlev2002/ hirlevel 2002jan01.html
(41) egy optima´lis ce´l, amiveli azt hiszem, hogy mindenki egyete´rt ti, . . .
an optimistic goal with which it-ACC think-1SG that everyone agrees
‘an optimistic goal, with which I think (it) that everyone agrees’
www.mkogy.hu/naplo34/143/1430020.html
Similarly, a dative possessor in the embedded clause may also raise past the expletive into the higher
clause, as shown in (42a). This is not be possible with complex noun phrases, for example as in (42b).
(42) a. En-neki is az-t hiszem, hogy [ ti a legfontosabb tulajdonsa´ga] a folytonossa´g.
this-DAT FOC it-ACC think-1SG that the most important property the continutity
‘I think that the most important property of this is continuity.’
b. *Enneki is hallot-t-am az o¨tlet-et, hogy [ ti a legfontosabb tulajdonsa´ga] a folytonossa´g.
this-DAT FOC hear-PAST-1SG the idea-ACC that the most important property the continuity
‘I heard the idea that the most important property of this is continuity.’
If the correlative construction had the same structure as a complex noun phrase, as the extraposition analysis
claims, then (42a) should be just as bad as (42b). The difference in grammaticality suggests a structural
difference between the two constructions. Only the expletive analysis contains such a distinction. Hence,
the construction in (31) reflects the expletive pronoun that we would expect under the athematic object
analysis of Pe´tert in (10).
2 Functional and anaphoric control
In the previous section, I concluded that accusative-marked focus-raised subjects such as Pe´tert in (10) are
athematic objects, based on both the long-distance and island-sensitive nature of focus-raising and the the
expletive construction. In this respect, the focus-raising construction is like raising rather than equi. As
laid out by Bresnan (1982), raising constructions involve functional control, while equi involves anaphoric
control. So it makes sense as a default assumption at this point that the focus-raising construction under
consideration involves functional control. In fact, I have already implicitly proposed that the identification
is functional.
However, Kroeger (1993) documents a type of equi with functional control, in Tagalog, showing that
the distinction between raising and equi is orthogonal to the distinction between functional and anaphoric
control. This carves out the typological possibility of a raising construction with anaphoric control. In this
section, I will argue that the predicted but unattested raising with anaphoric control is in fact attested in one
dialect of Hungarian.
137
what-ACC say-3SG.INDEF that when come.3SG
‘When did he/she say that he/she would come?’
See Horvath (1995) for an extensive set of arguments that mit is behaving as an expletive in the scope
marking construction.
Furthermore, although the pronoun azt cannot be present in the matrix clause when focus-raising or
question formation takes place, it is possible to relativize out of the complement to a bridge verb when the
expletive is present. (These were found using Google, but a native Hungarian speaker informant confirms
the grammaticality of these examples.)
(40) Az egyik dolog, amibeni azt gondolom, hogy jobbak vagyunk ti, mint ma´sok . . .
the primary thing in which it-ACC think-1SG that better be.1PL than others
‘The primary thing in which I think (it) that we are better than others. . . ’
www.hte.hu/kiadvanyok/hirlev2002/ hirlevel 2002jan01.html
(41) egy optima´lis ce´l, amiveli azt hiszem, hogy mindenki egyete´rt ti, . . .
an optimistic goal with which it-ACC think-1SG that everyone agrees
‘an optimistic goal, with which I think (it) that everyone agrees’
www.mkogy.hu/naplo34/143/1430020.html
Similarly, a dative possessor in the embedded clause may also raise past the expletive into the higher
clause, as shown in (42a). This is not be possible with complex noun phrases, for example as in (42b).
(42) a. En-neki is az-t hiszem, hogy [ ti a legfontosabb tulajdonsa´ga] a folytonossa´g.
this-DAT FOC it-ACC think-1SG that the most important property the continutity
‘I think that the most important property of this is continuity.’
b. *Enneki is hallot-t-am az o¨tlet-et, hogy [ ti a legfontosabb tulajdonsa´ga] a folytonossa´g.
this-DAT FOC hear-PAST-1SG the idea-ACC that the most important property the continuity
‘I heard the idea that the most important property of this is continuity.’
If the correlative construction had the same structure as a complex noun phrase, as the extraposition analysis
claims, then (42a) should be just as bad as (42b). The difference in grammaticality suggests a structural
difference between the two constructions. Only the expletive analysis contains such a distinction. Hence,
the construction in (31) reflects the expletive pronoun that we would expect under the athematic object
analysis of Pe´tert in (10).
2 Functional and anaphoric control
In the previous section, I concluded that accusative-marked focus-raised subjects such as Pe´tert in (10) are
athematic objects, based on both the long-distance and island-sensitive nature of focus-raising and the the
expletive construction. In this respect, the focus-raising construction is like raising rather than equi. As
laid out by Bresnan (1982), raising constructions involve functional control, while equi involves anaphoric
control. So it makes sense as a default assumption at this point that the focus-raising construction under
consideration involves functional control. In fact, I have already implicitly proposed that the identification
is functional.
However, Kroeger (1993) documents a type of equi with functional control, in Tagalog, showing that
the distinction between raising and equi is orthogonal to the distinction between functional and anaphoric
control. This carves out the typological possibility of a raising construction with anaphoric control. In this
section, I will argue that the predicted but unattested raising with anaphoric control is in fact attested in one
dialect of Hungarian.
137
Page 13
2.1 Gervain’s observation
As noticed by Gervain (2002), there is some disagreement in the literature on focus-raising as to the obli-
gatoriness of the accusative marker -t on the focussed constituent in examples like (10). For some speakers,
Pe´ter can appear in nominative case with the same interpretation as in (10), shown in (43).
(43) Pe´ter mond-t-am, hogy jo¨n
Pe´ter.NOM say-PAST-1SG that come.3SG
‘It is Peter that I said is coming.’
The speakers who allow (43) also allow (10). Yet, the speakers who allow only (10) are more liberal than
the others in a different respect: they don’t require that the embedded verb (jo¨n ‘come’) agree in number
with the focussed constituent in certain cases, whereas the others do. This difference can be observed when
a semantically plural, but syntactically singular phrase such as az o¨sszes la´ny ‘every girl’ in (44) is raised.
That az o¨sszes la´ny is grammatically singular is evident from its nominal morphology, which is singular,
and the fact that it agrees with a singular verb, as shown in (44).
(44) Az o¨sszes la´ny jo¨n/*jo¨n-nek.
the all girl come.3SG/come-3PL
‘All the girls are coming.’
Az o¨sszes la´ny obviously picks out a plurality of girls and is in that sense semantically plural; for further
evidence, the plural personal pronoun o˝k ‘they’ could later be used to refer back to it, as the subject of the
next sentence, for example.
When az o¨sszes la´ny is focus-raised, there are, in principle, four different possible outcomes, depending
on two binary choices. The raised phrase may have nominative case, as in (45), or accusative case, as in
(46). Next, the embedded verb ‘come’ may be either plural, as in the (b) examples of (45) and (46), or
singular, as in the (a) examples. As Gervain (2002) shows, not all possibilities are grammatical, and there
is some disagreement among Hungarian speakers as to which of them are. In her study, Hungarian speakers
fell into two groups (A and B); their judgements for each sentence are represented in the righthand columns
in (45)–(46).
(45) NOMinative-marking
a. Az o¨sszes la´ny mond-t-ad, hogy jo¨n XA *B
the all girl.NOM say-PAST-2SG that come.3SG
b. Az o¨sszes la´ny mond-t-ad, hogy jo¨n-nek *A *B
the all girl.NOM say-PAST-2SG that come-3PL
(46) ACCusative-marking
a. Az o¨sszes la´ny-t mond-t-ad, hogy jo¨n XA XB
the all girl-ACC say-PAST-2SG that come-3SG
b. Az o¨sszes la´ny-t mond-t-ad, hogy jo¨n-nek *A XB
the all girl-ACC say-PAST-2SG that come-3PL
For Group A, both nominative and accusative case are allowed on the focus-raised element, and number
agreement between the focus-raised element and the embedded verb is required. For Group B, only ac-
cusative case is allowed on the focus-raised element, but number agreement is not required; Group B accepts
138
As noticed by Gervain (2002), there is some disagreement in the literature on focus-raising as to the obli-
gatoriness of the accusative marker -t on the focussed constituent in examples like (10). For some speakers,
Pe´ter can appear in nominative case with the same interpretation as in (10), shown in (43).
(43) Pe´ter mond-t-am, hogy jo¨n
Pe´ter.NOM say-PAST-1SG that come.3SG
‘It is Peter that I said is coming.’
The speakers who allow (43) also allow (10). Yet, the speakers who allow only (10) are more liberal than
the others in a different respect: they don’t require that the embedded verb (jo¨n ‘come’) agree in number
with the focussed constituent in certain cases, whereas the others do. This difference can be observed when
a semantically plural, but syntactically singular phrase such as az o¨sszes la´ny ‘every girl’ in (44) is raised.
That az o¨sszes la´ny is grammatically singular is evident from its nominal morphology, which is singular,
and the fact that it agrees with a singular verb, as shown in (44).
(44) Az o¨sszes la´ny jo¨n/*jo¨n-nek.
the all girl come.3SG/come-3PL
‘All the girls are coming.’
Az o¨sszes la´ny obviously picks out a plurality of girls and is in that sense semantically plural; for further
evidence, the plural personal pronoun o˝k ‘they’ could later be used to refer back to it, as the subject of the
next sentence, for example.
When az o¨sszes la´ny is focus-raised, there are, in principle, four different possible outcomes, depending
on two binary choices. The raised phrase may have nominative case, as in (45), or accusative case, as in
(46). Next, the embedded verb ‘come’ may be either plural, as in the (b) examples of (45) and (46), or
singular, as in the (a) examples. As Gervain (2002) shows, not all possibilities are grammatical, and there
is some disagreement among Hungarian speakers as to which of them are. In her study, Hungarian speakers
fell into two groups (A and B); their judgements for each sentence are represented in the righthand columns
in (45)–(46).
(45) NOMinative-marking
a. Az o¨sszes la´ny mond-t-ad, hogy jo¨n XA *B
the all girl.NOM say-PAST-2SG that come.3SG
b. Az o¨sszes la´ny mond-t-ad, hogy jo¨n-nek *A *B
the all girl.NOM say-PAST-2SG that come-3PL
(46) ACCusative-marking
a. Az o¨sszes la´ny-t mond-t-ad, hogy jo¨n XA XB
the all girl-ACC say-PAST-2SG that come-3SG
b. Az o¨sszes la´ny-t mond-t-ad, hogy jo¨n-nek *A XB
the all girl-ACC say-PAST-2SG that come-3PL
For Group A, both nominative and accusative case are allowed on the focus-raised element, and number
agreement between the focus-raised element and the embedded verb is required. For Group B, only ac-
cusative case is allowed on the focus-raised element, but number agreement is not required; Group B accepts
138
Page 14
(46b), in which the embedded verb is plural but the focus-raised element az o¨sszes la´ny is grammatically,
though not semantically, singular.13 The pattern is summarized below.
(47)
accusative-marking agreement (embedded verb is singular)
Group A optional required
Group B required optional
2.2 Functional and anaphoric control
This pattern suggests that while Group A uses functional identification between the matrix focus and the
embedded subject as discussed in the previous section, Group B uses anaphoric identification. The Group
A pattern can be accounted for under the assumption of functional identification as follows. The examples
with number disagreement, (45b) and (46b) are ruled out because the matrix focus is functionally identified
with the embedded subject, so the number features will clash. (46a) has an analysis very similar to the one
given for (10), ignoring the representation of the quantifier.
(48)
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
SUBJ
PERS 2
NUM SG
PRED ‘pro’
OBJ []
FOCUS
PRED ‘every-girl’
CASE ACC
NUM SG
PERS 3
COMP
PRED ‘come
〈
SUBJ
〉
’
SUBJ []
For Group B, the possibility of non-agreement between the raised accusative element and the embedded
subject in (46b) suggests that the identification is, by contrast, anaphoric, for that group.
Under these assumptions, here is an analysis of (46a):
13In fact, the judgments were given on a 5-point scale; I have simplified ‘?’ to unmarked, and ’???’ to ‘*’. The average judgments
for each group were not all stars or “perfect” marks; it remains an open question how to account for the more subtle aspects of the
pattern. See Gervain (2002, 2003) for discussion.
139
though not semantically, singular.13 The pattern is summarized below.
(47)
accusative-marking agreement (embedded verb is singular)
Group A optional required
Group B required optional
2.2 Functional and anaphoric control
This pattern suggests that while Group A uses functional identification between the matrix focus and the
embedded subject as discussed in the previous section, Group B uses anaphoric identification. The Group
A pattern can be accounted for under the assumption of functional identification as follows. The examples
with number disagreement, (45b) and (46b) are ruled out because the matrix focus is functionally identified
with the embedded subject, so the number features will clash. (46a) has an analysis very similar to the one
given for (10), ignoring the representation of the quantifier.
(48)
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
SUBJ
PERS 2
NUM SG
PRED ‘pro’
OBJ []
FOCUS
PRED ‘every-girl’
CASE ACC
NUM SG
PERS 3
COMP
PRED ‘come
〈
SUBJ
〉
’
SUBJ []
For Group B, the possibility of non-agreement between the raised accusative element and the embedded
subject in (46b) suggests that the identification is, by contrast, anaphoric, for that group.
Under these assumptions, here is an analysis of (46a):
13In fact, the judgments were given on a 5-point scale; I have simplified ‘?’ to unmarked, and ’???’ to ‘*’. The average judgments
for each group were not all stars or “perfect” marks; it remains an open question how to account for the more subtle aspects of the
pattern. See Gervain (2002, 2003) for discussion.
139
Page 15
(49)
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
SUBJ
PERS 2
NUM SG
PRED ‘pro’
OBJ []
FOCUS
PRED ‘every-girl’
CASE ACC
NUM SG
PERS 3
INDEX []
COMP
PRED ‘come
〈
SUBJ
〉
’
SUBJ
[
PRED ‘pro’
INDEX []
]
Gervain (2002) proposes an analogous analysis in Minimalist style. She argues that Group B uses a “re-
sumptive strategy”, that is, they use a resumptive pronoun in the embedded clause, whereas Group A uses
regular wh-movement. This explains the fact that both plural and singular agreement are possible for Group
B. Anaphors with semantically plural antecedents may be either plural or singular, giving rise to a col-
lective or distributive interpretation respectively. Thus the singular pronoun o˝t in (50a) has a distributive
interpretation, and the plural pronoun o˝ket in (50b) has a plural interpretation.
(50) a. Ke´t fiu´i hiszi azt, hogy Ma´ria szereti (o˝ti)
two boy.SG think.3SG it that Mary love.3SG him
‘There are two boys each of whom thinks that Mary loves him.’
b. Ke´t fiu´i hiszi azt, hogy Ma´ria szereti o˝keti
two boy.SG think.3SG it that Mary love.3SG them
‘Two boys think that Mary loves them.’
As Gervain points out, the idea that Group B uses a pronoun in the embedded clause is also supported by
the fact that Group B does not rule out focus-raising out of complex noun phrases:
(51) Az elno¨ko¨t mondtad, hogy hallottad a hı´rt, hogy mege´rkezett (ok for Group B only)
the president.ACC said.2SG that heard.2SG the news that arrived.3SG
‘You said you heard the news that the president had arrived.’
whereas Group A does. This is in line with the fact that movement out of complex noun phrases is impossi-
ble, but anaphoric binding into them is possible, as shown in (52).
(52) No mani asked whether or not hei would be fired.
(cf. *Whoi asked whether or not ti would be fired?)
In LFG terms, the movement/resumption distinction may be interpreted using the functional/anaphoric iden-
tification distinction.
However, supposing that Group B uses anaphoric identification does not explain why Group B, unlike
Group A, does not tolerate nominative case on the focus-raised element, as shown in (45a). To explain this,
140
PRED ‘say
〈
SUBJ COMP
〉
OBJ’
SUBJ
PERS 2
NUM SG
PRED ‘pro’
OBJ []
FOCUS
PRED ‘every-girl’
CASE ACC
NUM SG
PERS 3
INDEX []
COMP
PRED ‘come
〈
SUBJ
〉
’
SUBJ
[
PRED ‘pro’
INDEX []
]
Gervain (2002) proposes an analogous analysis in Minimalist style. She argues that Group B uses a “re-
sumptive strategy”, that is, they use a resumptive pronoun in the embedded clause, whereas Group A uses
regular wh-movement. This explains the fact that both plural and singular agreement are possible for Group
B. Anaphors with semantically plural antecedents may be either plural or singular, giving rise to a col-
lective or distributive interpretation respectively. Thus the singular pronoun o˝t in (50a) has a distributive
interpretation, and the plural pronoun o˝ket in (50b) has a plural interpretation.
(50) a. Ke´t fiu´i hiszi azt, hogy Ma´ria szereti (o˝ti)
two boy.SG think.3SG it that Mary love.3SG him
‘There are two boys each of whom thinks that Mary loves him.’
b. Ke´t fiu´i hiszi azt, hogy Ma´ria szereti o˝keti
two boy.SG think.3SG it that Mary love.3SG them
‘Two boys think that Mary loves them.’
As Gervain points out, the idea that Group B uses a pronoun in the embedded clause is also supported by
the fact that Group B does not rule out focus-raising out of complex noun phrases:
(51) Az elno¨ko¨t mondtad, hogy hallottad a hı´rt, hogy mege´rkezett (ok for Group B only)
the president.ACC said.2SG that heard.2SG the news that arrived.3SG
‘You said you heard the news that the president had arrived.’
whereas Group A does. This is in line with the fact that movement out of complex noun phrases is impossi-
ble, but anaphoric binding into them is possible, as shown in (52).
(52) No mani asked whether or not hei would be fired.
(cf. *Whoi asked whether or not ti would be fired?)
In LFG terms, the movement/resumption distinction may be interpreted using the functional/anaphoric iden-
tification distinction.
However, supposing that Group B uses anaphoric identification does not explain why Group B, unlike
Group A, does not tolerate nominative case on the focus-raised element, as shown in (45a). To explain this,
140
Page 16
I would like to propose that the complementizer hogy, for Group B, has a specification similar to that which
Falk (2001) gives of English that, in order to explain the that-trace effect:14
(55) hogy C (↑SUBJ) 6=((GF+↑) GF)
This solution echoes the explanation given by Gervain (2002) for the impossibility of nominative case: she
writes, “I assume that it is the nature of the complementizer hogy ‘that’ that is parametrically different in
the two dialects. In the movement dialect, the complementizer allows the raised operator to properly govern
its trace, thus complying with the ECP. In the other dialect, however, the complementizer is ‘opaque’ and
does not allow proper government. Thus resumption is needed to save the sentence from the resulting ECP
violation” (Gervain, 2002, p. 77). The difference here is that the idea of the “Last Resort” is being modelled
with an OT constraint violation.
This constraint prevents subject extraction, thereby preventing nominative focus-marked constituents.
At the same time, it motivates the possibility of anaphoric identification between the raised element and the
embedded subject.
Notice that in (49), there is an athematic object that is not functionally integrated, so the semantic Co-
herence condition is violated: there is a PRED inside a governable function that is not semantically selected.
This particular violation of the semantic Coherence condition does not make us want to suppose that the
semantic Coherence condition does not hold; it is still important for ruling out the Hungarian equivalent of
examples like *John rained. Rather, the semantic Coherence condition can be violated only in a limited
range of cases. In particular, the semantic coherence condition can be violated when a violation of the con-
straint in (55) would be incurred to satisfy it. Thus, the constraint in (55) outranks the semantic Coherence
condition, in the OT sense.
Let us label the semantic Coherence condition SEMANTIC-COHERENCE, and the constraint in (55),
*SUBJECT-EXTRACTION. With these constraints, it is possible succinctly to describe the pattern associ-
ated with both groups as follows: Group A uses the constraint subhierarchy (56); Group B uses (57).
(56) SEMANTIC-COHERENCE *SUBJECT-EXTRACTION
(57) *SUBJECT-EXTRACTION SEMANTIC-COHERENCE
Hypothesizing these constraints thus allows us to explain both of the differences between Group A and
Group B as consequences of a single difference in constraint ranking.15
14The presence of this constraint in Hungarian is ironic, because as ´E. Kiss (1987) points out, on the surface Hungarian has the
opposite of the English that-trace effect:
(53) a. Kit mondtam, hogy j o¨n?
b. *Kit mondtam, j o¨n?
(54) a. *Who did you say that is coming?
b. Who did you say is coming?
15These are not all the constraints that are relevant. Now that the semantic Coherence condition is violable, a constraint is
needed to rule out (22), which is ungrammatical in both dialects. To account for this, I might propose that the Extended Coherence
condition, a more specific condition than semantic Coherence, is highly-ranked for both Group A and Group B.
(58) Extended Coherence: FOCUS and TOPIC must be linked to the semantic predicate argument structure of the sentence in
which they occur, either by functionally or by anaphorically binding an argument. Bresnan and Mchombo (1987)
Alternatively, the Extended Coherence Condition may be a by-product of GEN; if the OT-LFG Input is semantically well-formed,
it is even difficult to imagine a function producing candidate f-structures from the input that violate the Extended Coherence
Condition.
Also, as was pointed out by Peter Sells at the LFG03 conference, another constraint to the effect, FOCUS-RAISE!, needs to be
among those relevant for evaluation.
141
Falk (2001) gives of English that, in order to explain the that-trace effect:14
(55) hogy C (↑SUBJ) 6=((GF+↑) GF)
This solution echoes the explanation given by Gervain (2002) for the impossibility of nominative case: she
writes, “I assume that it is the nature of the complementizer hogy ‘that’ that is parametrically different in
the two dialects. In the movement dialect, the complementizer allows the raised operator to properly govern
its trace, thus complying with the ECP. In the other dialect, however, the complementizer is ‘opaque’ and
does not allow proper government. Thus resumption is needed to save the sentence from the resulting ECP
violation” (Gervain, 2002, p. 77). The difference here is that the idea of the “Last Resort” is being modelled
with an OT constraint violation.
This constraint prevents subject extraction, thereby preventing nominative focus-marked constituents.
At the same time, it motivates the possibility of anaphoric identification between the raised element and the
embedded subject.
Notice that in (49), there is an athematic object that is not functionally integrated, so the semantic Co-
herence condition is violated: there is a PRED inside a governable function that is not semantically selected.
This particular violation of the semantic Coherence condition does not make us want to suppose that the
semantic Coherence condition does not hold; it is still important for ruling out the Hungarian equivalent of
examples like *John rained. Rather, the semantic Coherence condition can be violated only in a limited
range of cases. In particular, the semantic coherence condition can be violated when a violation of the con-
straint in (55) would be incurred to satisfy it. Thus, the constraint in (55) outranks the semantic Coherence
condition, in the OT sense.
Let us label the semantic Coherence condition SEMANTIC-COHERENCE, and the constraint in (55),
*SUBJECT-EXTRACTION. With these constraints, it is possible succinctly to describe the pattern associ-
ated with both groups as follows: Group A uses the constraint subhierarchy (56); Group B uses (57).
(56) SEMANTIC-COHERENCE *SUBJECT-EXTRACTION
(57) *SUBJECT-EXTRACTION SEMANTIC-COHERENCE
Hypothesizing these constraints thus allows us to explain both of the differences between Group A and
Group B as consequences of a single difference in constraint ranking.15
14The presence of this constraint in Hungarian is ironic, because as ´E. Kiss (1987) points out, on the surface Hungarian has the
opposite of the English that-trace effect:
(53) a. Kit mondtam, hogy j o¨n?
b. *Kit mondtam, j o¨n?
(54) a. *Who did you say that is coming?
b. Who did you say is coming?
15These are not all the constraints that are relevant. Now that the semantic Coherence condition is violable, a constraint is
needed to rule out (22), which is ungrammatical in both dialects. To account for this, I might propose that the Extended Coherence
condition, a more specific condition than semantic Coherence, is highly-ranked for both Group A and Group B.
(58) Extended Coherence: FOCUS and TOPIC must be linked to the semantic predicate argument structure of the sentence in
which they occur, either by functionally or by anaphorically binding an argument. Bresnan and Mchombo (1987)
Alternatively, the Extended Coherence Condition may be a by-product of GEN; if the OT-LFG Input is semantically well-formed,
it is even difficult to imagine a function producing candidate f-structures from the input that violate the Extended Coherence
Condition.
Also, as was pointed out by Peter Sells at the LFG03 conference, another constraint to the effect, FOCUS-RAISE!, needs to be
among those relevant for evaluation.
141
Page 17
Conclusions
I have argued that accusative-marked focus raised subject such as Pe´tert in (10) occupy an athematic object
function. Further, depending on dialect, this function may host either functional or anaphoric binders for
the embedded subject. If this analysis is correct, then the semantic Coherence condition is violable. This
violability may be exploitable in the explanation of other intrusive pronoun phenomena, such as resumptive
pronouns.16
The analysis has certain typological implications as well. First, if it is true that the accusative-marked
focus-raised phrases are objects, then they represent long-distance (A′) movement to an argument (A) po-
sition, which is not usually assumed to exist. Second, if the analysis is correct, then Hungarian fills out a
typological paradigm slot that is predicted to exist, but not yet attested: raising with anaphoric control.
References
Asudeh, A. (2002). Richard III. In Mary Andronis, Erin Debenport, A. P. and Yoshimura, K., editors,
CLS 38: The main session. Papers from the 38th meeting of the Chicago Linguistic Society, volume 1,
Chicago, IL. Chicago Linguistic Society.
Bartos, H. (1997). On “subjective” and “objective” agreement in Hungarian. Acta Linguistica Hungarica,
44:363–384.
Berman, J. (2003). Clausal Syntax of German. Studies in Constraint-Based Lexicalism. CSLI Publications.
Bresnan, J. (1982). Control and complementation. In Bresnan, J., editor, The Mental Representation of
Grammatical Relations, pages 282–390. MIT Press.
Bresnan, J. and Mchombo, S. (1987). Topic, pronoun, and agreement in Chichewˆa. Language, 63:741–82.
Brody, M. (1990). Remarks on the order of elements in the Hungarian focus field. In Kenesei, I., editor,
Approaches to Hungarian, volume 3, Szeged. JATE.
Brody, M. (1995). Focus and checking theory. In Kenesei, I., editor, Approaches to Hungarian, pages
31–43, Szeged. JATE.
Chomsky, N. (1981). Barriers. MIT Press, Cambridge.
Dalrymple, M. (2001). Lexical Functional Grammar. Academic Press, San Diego, CA.
´E. Kiss, K. (1987). Configurationality in Hungarian. Akade´miai Kiado.
´E. Kiss, K. (1990). Why noun-complement clauses are barriers. In Mascaro´, J. and Nespor, M., editors,
Grammar in Progress, pages 265–277. Foris, Dordrecht.
´E. Kiss, K. (2002). The Syntax of Hungarian. Cambridge University Press.
Falk, Y. (2001). Lexical-Functional Grammar: An introduction to Parallel Constraint-Based Syntax. CSLI
Publications.
16Copy-raising constructions like Richard seems like he’s in trouble (Potsdam and Runner, 2001) seem like good candidates,
because the controller appears to be in an argument slot that is not thematic. However, Asudeh (2002) analyzes these nicely
as involving assignment of a thematic role to the controller by the preposition like. If we take his analysis, then the semantic
Coherence condition is not violated in copy-raising constructions.
142
I have argued that accusative-marked focus raised subject such as Pe´tert in (10) occupy an athematic object
function. Further, depending on dialect, this function may host either functional or anaphoric binders for
the embedded subject. If this analysis is correct, then the semantic Coherence condition is violable. This
violability may be exploitable in the explanation of other intrusive pronoun phenomena, such as resumptive
pronouns.16
The analysis has certain typological implications as well. First, if it is true that the accusative-marked
focus-raised phrases are objects, then they represent long-distance (A′) movement to an argument (A) po-
sition, which is not usually assumed to exist. Second, if the analysis is correct, then Hungarian fills out a
typological paradigm slot that is predicted to exist, but not yet attested: raising with anaphoric control.
References
Asudeh, A. (2002). Richard III. In Mary Andronis, Erin Debenport, A. P. and Yoshimura, K., editors,
CLS 38: The main session. Papers from the 38th meeting of the Chicago Linguistic Society, volume 1,
Chicago, IL. Chicago Linguistic Society.
Bartos, H. (1997). On “subjective” and “objective” agreement in Hungarian. Acta Linguistica Hungarica,
44:363–384.
Berman, J. (2003). Clausal Syntax of German. Studies in Constraint-Based Lexicalism. CSLI Publications.
Bresnan, J. (1982). Control and complementation. In Bresnan, J., editor, The Mental Representation of
Grammatical Relations, pages 282–390. MIT Press.
Bresnan, J. and Mchombo, S. (1987). Topic, pronoun, and agreement in Chichewˆa. Language, 63:741–82.
Brody, M. (1990). Remarks on the order of elements in the Hungarian focus field. In Kenesei, I., editor,
Approaches to Hungarian, volume 3, Szeged. JATE.
Brody, M. (1995). Focus and checking theory. In Kenesei, I., editor, Approaches to Hungarian, pages
31–43, Szeged. JATE.
Chomsky, N. (1981). Barriers. MIT Press, Cambridge.
Dalrymple, M. (2001). Lexical Functional Grammar. Academic Press, San Diego, CA.
´E. Kiss, K. (1987). Configurationality in Hungarian. Akade´miai Kiado.
´E. Kiss, K. (1990). Why noun-complement clauses are barriers. In Mascaro´, J. and Nespor, M., editors,
Grammar in Progress, pages 265–277. Foris, Dordrecht.
´E. Kiss, K. (2002). The Syntax of Hungarian. Cambridge University Press.
Falk, Y. (2001). Lexical-Functional Grammar: An introduction to Parallel Constraint-Based Syntax. CSLI
Publications.
16Copy-raising constructions like Richard seems like he’s in trouble (Potsdam and Runner, 2001) seem like good candidates,
because the controller appears to be in an argument slot that is not thematic. However, Asudeh (2002) analyzes these nicely
as involving assignment of a thematic role to the controller by the preposition like. If we take his analysis, then the semantic
Coherence condition is not violated in copy-raising constructions.
142
Page 18
Gervain, J. (2002). Linguistic methodology and microvariation in language: the case of operator-raising in
Hungarian. Master’s thesis, University of Szeged, Hungary.
Gervain, J. (2003). ´Ujra a focuszemele´sro´l: A re´zumptı´v ne´vma´sok te´rme´sete. Master’s thesis, University
of Szeged.
Horvath, J. (1986). Focus in the theory of grammar and the syntax of Hungarian. Foris, Dordrecht.
Horvath, J. (1995). Partial wh-movement and wh “scope-markers”. In Kenesei, I., editor, Approaches to
Hungarian, volume 5.
Horvath, J. (1998). Multiple wh-phrases and the wh-scope-marker strategy in Hungarian interrogatives.
Acta Linguistica Hungarica, 45:31–60.
Kenesei, I. (1994). Subordinate clauses. In Kiefer, F. and ´E. Kiss, K., editors, The Syntactic Structure of
Hungarian, volume 27 of Syntax and Semantics. Academic Press.
Kroeger, P. (1993). Phrase Structure and Grammatical Relations in Tagalog. PhD thesis, Stanford Univer-
sity.
Lipta´k, A. (2001). On the Syntax of Wh-Items in Hungarian. PhD thesis, Leiden University.
Mara´cz, L. (1989). Asymmetries in Hungarian. PhD thesis, University of Groningen. §7.2–§7.4, pp. 297–
325.
Potsdam, E. and Runner, J. T. (2001). Richard returns: Copy raising and its implications. In Mary Androis,
Chris Ball, H. E. and Neuvel, S., editors, CLS 37: The main session. Papers from the 37th meeting of the
Chicago Lingusitic Society, volume 1, Chicago, IL. Chicago Linguistic Society.
Szendro˝i, K. (2001). Focus and the syntax-phonology interface. PhD thesis, University College London.
Zolnay, G. (1926). Mondata´tszo¨vo˝de´s. ´Ertekeze´sek a Magyar Tudomanyos Akade´mia Nyelv- e´s
Sze´ptudoma´nyi Oszta´lya Ko¨re´bo˝l, 23.
143
Hungarian. Master’s thesis, University of Szeged, Hungary.
Gervain, J. (2003). ´Ujra a focuszemele´sro´l: A re´zumptı´v ne´vma´sok te´rme´sete. Master’s thesis, University
of Szeged.
Horvath, J. (1986). Focus in the theory of grammar and the syntax of Hungarian. Foris, Dordrecht.
Horvath, J. (1995). Partial wh-movement and wh “scope-markers”. In Kenesei, I., editor, Approaches to
Hungarian, volume 5.
Horvath, J. (1998). Multiple wh-phrases and the wh-scope-marker strategy in Hungarian interrogatives.
Acta Linguistica Hungarica, 45:31–60.
Kenesei, I. (1994). Subordinate clauses. In Kiefer, F. and ´E. Kiss, K., editors, The Syntactic Structure of
Hungarian, volume 27 of Syntax and Semantics. Academic Press.
Kroeger, P. (1993). Phrase Structure and Grammatical Relations in Tagalog. PhD thesis, Stanford Univer-
sity.
Lipta´k, A. (2001). On the Syntax of Wh-Items in Hungarian. PhD thesis, Leiden University.
Mara´cz, L. (1989). Asymmetries in Hungarian. PhD thesis, University of Groningen. §7.2–§7.4, pp. 297–
325.
Potsdam, E. and Runner, J. T. (2001). Richard returns: Copy raising and its implications. In Mary Androis,
Chris Ball, H. E. and Neuvel, S., editors, CLS 37: The main session. Papers from the 37th meeting of the
Chicago Lingusitic Society, volume 1, Chicago, IL. Chicago Linguistic Society.
Szendro˝i, K. (2001). Focus and the syntax-phonology interface. PhD thesis, University College London.
Zolnay, G. (1926). Mondata´tszo¨vo˝de´s. ´Ertekeze´sek a Magyar Tudomanyos Akade´mia Nyelv- e´s
Sze´ptudoma´nyi Oszta´lya Ko¨re´bo˝l, 23.
143
Sign up today - FREE
Mendeley saves you time finding and organizing research. Learn more
- All your research in one place
- Add and import papers easily
- Access it anywhere, anytime
Start using Mendeley in seconds!
Readership Statistics
2 Readers on Mendeley
by Discipline
100% Linguistics
by Academic Status
50% Post Doc
50% Researcher (at a non-Academic Institution)
by Country
50% United States