The categorical status of (anti)-(anti)-agreement
1. The problem: A survey of the agreement patterns that different languages exhibit in the
context of A'-extraction indicates the existence of (at least) three types of agreement patterns.
The first type is instantiated by English where subject agreement is retained in context of
(WH-) subject extraction (1). The second type is exhibited either as a partial or a radical loss
of subject agreement specifications. For example, in Berber (2) and Halkomelem (3), (WH-)
subject extraction obligatorily triggers the loss of subject agreement. Finally in Standard
Arabic (4), (WH-) subject extraction triggers a change from partial agreement (in simple VSO
declarative clauses) to full agreement. The different patterns are summarized below:
Patterns of subject agreement in Terminology used Example language
extraction context
retained agreement English (1)
lost anti-agreement Berber (2); Halkomelem (3)
changed anti-anti-agreement Arabic (4)
The existence of this variation in the agreement specifications associated with a single syntactic
process (subject extraction) is intriguing in many empirical and theoretical respects. We explore these
intriguing aspects through the following main questions: what triggers the occurrence of agreement,
anti-agreement, or anti-anti-agreement in a given language? And how is this variation linked to the
syntactic process of (WH)-subject extraction?
2. The proposal: The generative literature contains individual analyses of the patterns of agreement
described above (Cf. Ouhalla 1993; Chung 1982; Gerdts 1988, a.o.); however, none of these analyses
addresses the variation in (anti)(anti)-agreement patterns in a unified fashion and associate it with a
single underlying syntactic process. In this paper, we argue that the array of agreement patterns we
identify above can be derived straightforwardly from the category of the relevant agreement ending.
To do that, we assume with Déchaine and Wiltschko (2002) that pronouns are not primitives but
rather that their distribution, interpretation and binding-theoretic properties is determined by their
category: D-pronouns are definite R-expressions, -pronouns are variables, and N-pronouns are
nominal constants. Assuming that agreement is pronominal in nature (cf. Taraldsen 1992; Ritter
1995), we expect to find D-agreement, -agreement, and N-agreement. Accordingly, we expect that
the categorical status of agreement will determine its distribution, its interpretation as well as its
binding-theoretic properties.
3. Deriving the problem: We propose that the binding-theoretic properties of D-, - and N-
agreement derive the observed patterns of agreement.
(i) As a nominal constant, N-agreement cannot be A'-bound and is therefore obligatorily lost. Its
presence would induce a strong-cross-over violation as (subject) cleft constructions which constitute
the syntactic environment for the appearance of anti-agreement effects in Berber are not derived by
movement (Cf. Elouazizi, 2005.c). This explains the anti-agreement pattern.
(ii) As a variable, -agreement can occur in the context of A'-binding and is therefore retained. This
explains the pattern of agreement in languages like English and
(iii) As a definite R-expression, D-agreement functions like an operator and as such is analyzed as
operator agreement (cf. Chung 1998). As such, it is expected to change in context of A'-binding
because it marks the discourse role in SpecCP. This explains the pattern of anti-anti-agreement.
4. Independent evidence: The proposal predicts that the different types of agreement involved in
(anti-)(anti-)agreement patterns display different distributional and interpretational properties. On
empirical grounds, we show that these correlations do indeed hold: (i) N-agreement is associated with
and represents agreement with a thematic role. As such, it always appears on the verb (never on an
auxiliary; (5), (6)) and is tied to a thematic role (much like inherent case; (7)), (ii) -agreement is
associated with Infl and represents agreement with a grammatical role. As such it appears on the verb
only if there is no auxiliary and it is tied to a grammatical role (much like structural case) and (iii) D-
agreement is associated with C and represents agreement with a discourse role and as such it will
appear on whatever element occupies C. It is furthermore predicted to interact with other elements in
the C-system (i.e. the realis/irrealis distinction; cf. Chung 1998). This analysis further predicts that:
(a) different agreement endings can be spelled out in a given language (8), and (b) in long-distance
wh-movement, Spec CP marks agreement with a discourse role and it is independent of the arguments
that appear within that clause (9).
5. Data samples
(1) English agreement
a. The boy plays soccer b. Who is the boy who plays soccer?
(2) Berber anti-agreement
tharbathin
a. si-nt lkub
buy.PERF-3PLF girl.3PLF books
`The girls bought the books.'
tharbathin
b. man i g y-si-n lkub
which girl.3PLF RM X PART-buy.PERF-PART books
`Which girls bought the books?'
(3) Halkomelem anti-agreement
q'wel-et(*-es) te
a. slén? e ni sepll
lady DET ASP bake-TRANS-3s DET bread
`A lady is the one who baked the bread.' (Gerdts 1980,303)
(4) Arabic anti-anti-agreement
a. ?ishtara ?attulabu llkutuba
buy.PERF.3SM students.3PLM.NOM books.ACC
`The students bought books.' (VSO=Partial agreement)
b. ?attulabu humu ?laina ?ishtaru llkutuba
students.3PLM. NOM them.3PLM who.3PLM buy.PERF.3PLM books.ACC
`It is the students who bought books.' (A'-extraction=Anti-anti-agreement)
(5) Halkomelem subject (N) agreement is tied to the verb
a. li máy-t-es b. *li-s máy-t
aux help-trans-3s aux-3s help-trans
`He helped someone.' `He helped someone.'
(6) Berber subject (N) agreement is tied to the verb
a. ad y-gar warba lktab b. *ad-y ar warba lktab
fut 3sm-read.aor boy book fut-3sm read.aor boy book
`The boy will read the book.' `The boy will read the book.'
(7) Halkomelem subject agreement is lost in passive voice
máy-t-em te slháli
help-trans-em det woman
`The woman was helped.'
(8) Marking the path on way to CP in Standard Arabic
?arrajul-u huwa ?lladi arrsala ?nnuquda ?ila llwaladi
man.NOM him.3S.M who.3S.M send.PERF.3S.M money to the.boy
`It is the man who sent the money to the boy.'
(9) Long distance wh-movement and the marking of the discourse role in (Chamorro)
Manggi i gima' änai ma-sangan i päli' na pära u-saga ?
where?.is the house CREL WH[obj2].AGR-tell the priest C FUT AGR-live
`Where is the house where they told the priest that he should live?'
(Reintges, LeSourd, P. and Chung, S., 2004:6/eg.11)
6. Selected references
Chung, Sandra. 1998. The design of agreement. Evidence from Chamorro. The University of Chicago Press.
Déchaine, Rose-Marie, and Wiltschko, Martina. 2002. Decomposing Pronouns. Linguistic Inquiry 33:409-442;
Gerdts, Donna. 1988. Object and absolutive in Halkomelem. Garland.
Ouhalla, Jamal. 1993. Subject Extraction, Negation and the Anti-agreement Effect. Natural Language and
Linguistic Theory 11, 477-518.
Ritter, Elizabeth. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic
Theory 13:405-443.
Reintges, C. LeSourd, P. and Chung, S. 2004. Movement, wh-agreement and apparent wh-in-situ. Ms. ULCL,
Leiden University.
Taraldsen, T. 1992. Agreement as pronoun incorporation. Talk delivered at GLOW.