Bound-Variable Anaphora and Left Branch
Condition
Marijana Marelj
Abstract. Since the seminal work of Reinhart (1976, 1984), it has been recognized that the
structural configuration for bound-variable anaphora is that of c-command. One way of
deriving this variable-binding condition is to capitalize on the similarities between traces and
bound pronouns and to analyze the latter as spelled-out traces (see Aoun 1982, also Hornstein
2001 for a somewhat different implementation of Aoun�s idea). Taking Hornstein�s (2001)
analysis of bound pronouns in English as a starting point, I account for some puzzling
differences between English and Serbo-Croatian in the domain of the intrasentential
pronominal anaphora. By establishing the correlation between the way English and Serbo-
Croatian behave with respect to the Pronoun Insertion strategy, on the one hand, and the Left
Branch Condition (Ross 1967/1986), on the other, I argue that the differences under
consideration boil down to the presence or absence of a DP-layer in a given language. The
validity of this hypothesis is tested on a broader set of crosslinguistic data, including data from
Dutch, Latin, Russian, and Spanish. The analysis I put forth supports the minimalist view of
binding (Hornstein 2001, 2006; Grohmann 2003) as well as accounts of left branch extraction
languages as D-less (Uriagereka 1988; Corver 1992; Bosˇkovic´ 2005, 2008).
1. Introduction
It is well known that pronouns like his in (1) are ambiguous between the bound and the
referential readings. It is also well known that not all languages pattern like English.1
As seen in its Serbo-Croatian counterpart, the /-complete njegov �his� cannot get
the bound-pronoun reading in (2a). This reading requires the use of the /-deficient
pronoun svoj in (2b).
(1) Everyonei loves hisi/j mother.
(2) a. Svakoi voli njegovu*i/j majku.
everyone loves his-3sg.masc.poss mother
�Everyone loves his mother.�
b. Svako voli svojui/*j majku.
everyone loves self-poss mother
�Everyone loves his own mother.�
I am grateful to J. Ballantyne, Zˇ. Bosˇkovic´, M. Everaert, S. Franks, M. van Koppen, R. Nouwen,
T. Reinhart, and E. Reuland for valuable comments, questions, discussions, or criticism on different parts
and stages of the materials that ended up in the final version of this paper, and to M. Everaert and
R. Nouwen (Dutch), M. Weidner (German), L. Meroni (Italian), T. Apostolovski and V. Bubalova
(Macedonian), K. Migdalski (Polish), N. Slioussar (Russian), Zˇ. Bosˇkovic´ (Serbo-Croatian), and S. Baauw
(Spanish) for judgments. I would also like to thank the anonymous reviewers for Syntax, whose careful
comments and criticism helped to improve this paper. I am thankful to J. C. Ballantyne for proofreading
the manuscript. Any and all errors remain my own. This work was supported by a grant from the
Netherlands Organization for Scientific Research (NWO), which I hereby gratefully acknowledge.
1 See Vikner 1985, Hestvik 1992, Avrutin 1994, Dayal 1994, and Kidwai 2000, among others. See
section 2.3 for elaboration and discussion.
� 2011 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
Syntax 14:3, September 2011, 205–229 DOI: 10.1111/j.1467-9612.2011.00156.x
The goal of this paper is to provide an account of the pattern in (1) and (2).
Trivially, English has one pronoun—for example, his—where Serbo-Croatian (SC)
has two: a /-complete njegov (marked for person, number, and gender) and the /-
deficient svoj (unspecified for person, number, and gender). The question is whether
the pattern in (1) and (2) can be explained by appealing to this trivial fact of numbers.
In other words, can the lack of ambiguity in (2) be accounted for by the simple fact
that SC has two distinct morphological forms for two distinct readings, where English
has only one?
The paper is organized as follows. In section 2, I show that there is more to the data
than the trivial fact of numbers and argue that the proper treatment of the data in (1)
and (2) provides evidence in favor of the hypothesis that binding should be
reinterpreted in movement terms (see Lidz & Idsardi 1998; Hornstein 2001, 2006;
Kayne 2002; Zwart 2002; Grohmann 2003). Adopting and adapting Hornstein�s
(2001, 2006) account, in section 3, I derive the differences between Serbo-Croatian
and English under consideration here and provide evidence in favor of the competing
derivations view of movement and pronominalization. In section 4, the main
hypothesis of the paper is tested against a broader set of crosslinguistic data. Section 5
ties the relevant crosslinguistic differences to the presence or absence of a DP-layer in
a language. Section 6 concludes the paper.
2. Small Empirical Puzzles and Big Theoretical Issues
2.1 There Is More to the Puzzle than Numbers
Before we embark on the analysis of the data in (1) and (2), let me provide evidence
that corroborates the fact that there is more to the puzzle here than the trivial fact of
the number of pronouns in a given language.
As illustrated in (3), Macedonian has the same inventory of possessive pronouns as
Serbo-Croatian (SC); the third-person singular masculine pronoun negovata corre-
sponds to the /-complete njegov in (2a) and svojata corresponds to the /-deficient
svoj in (2b).
(3) Direktor-oti dojde so negova-tai/j/svoja-ta Macedonian
director-def came-3sg.ao with his-FEM.DEF/own-FEM.DEF
sopruga.
spouse
�The director came with his wife/his own spouse.� (Friedman 2002)
Importantly, however, Macedonian does not pattern with SC with respect to the
pairing between the available readings and forms. Rather, it patterns with English
with respect to the readings available for the intrasentential pronominal anaphora.
‘‘The use of negovata �his� is unremarkable and would not be interpreted as referring
to someone else�s wife. Rather, the use of svojata �his own� would be taken to imply
that it was unusual for him to come with his own wife rather than someone else�s’’
(Friedman 2002:292).
206 Marijana Marelj
� 2011 Blackwell Publishing Ltd.
In other words, Macedonian allows, even prefers, the bound interpretation of
negovata in (3) and the emphatic interpretation of svojata. Unlike in Macedonian (3),
the SC njegov in (4) does not normally allow the bound interpretation, and svoj in (4)
cannot be interpreted emphatically. The emphatic interpretation in SC requires the use
of njegov.
(4) Direktori je dosˇao sa njegovomj/*i/svojomi zˇenom. SC
director is arrived with his-3sg.masc.poss/self-poss wife
�The director arrived with his spouse.�
I conclude, then, that the contrast between the English (1) and SC (2) cannot simply
be accounted for by appealing to the different number of pronouns in these two
languages. The contrast is, furthermore, not restricted to these data but is present in
other relevant environments.
Let us start with VP-ellipsis data. It is well established (see Reinhart 2000 and
references therein) that intrasentential pronominal anaphora have two interpretations:
binding and covaluation:2,3
(5) Felix hates his neighbor and so does Max. strict or sloppy
Under the strict reading in (6a), the pronoun his in (5) is interpreted referentially, so it
covers the cases where Felix and Max hate John�s neighbors and the situation where
Felix and Max both hate Felix�s neighbors. Under the sloppy reading in (6b), it is
interpreted as a variable, which must be linked to a suitable binder, Felix in the first
conjunct and Max in the second.
(6) a. Covaluation: Felix kx (x hates J�s neighbor) & Max kx (x hates J�s neighbor)
b. Binding: Felix kx (x hates x�s neighbor) & Max kx (x hates x�s neighbor)
Unlike in English (5), however, the ambiguity is absent from SC (7). Quite on a par
with the data in (2), the strict reading (covaluation) arises with the use of njegov (7a)
and the sloppy reading (binding) is restricted to the use of svoj (7b).
(7) a. Lorens mrzi njegovog komsˇiju, a i Tristram takodje. strict
Laurence hates his-3sg.masc.poss neighbor and Tristram too
b. Lorens mrzi svoga komsˇiju, a i Tristram takodje. sloppy
Laurence hates self-poss neighbor and Tristram too
The ambiguity present in English in (1) is observed in (8) as well.
2 Covaluation is understood as in Reinhart 1984, 2000. The free variable is assigned a value from the
discourse storage by being covalued with the entries found in it (e.g., Felix, Max).
3 The known caveat is that ambiguity arises with both referential and nonreferential antecedents (see
Heim 1998; Reinhart 2000).
Bound-Variable Anaphora and Left Branch Condition 207
� 2011 Blackwell Publishing Ltd.
(8) Only Lucie respects her husband.
(9) a. Covaluation: Only Lucie ((kx (x respects her husband) & her = Lucie)
Entails that other women do not respect Lucie�s husband
b. Binding: Only Lucie (kx (x respects x�s husband))
Entails that unlike Lucie, other women do not respect their husbands
(Reinhart 2000)
Not surprisingly—and parallel with (2) and (5)—the ambiguity is absent in SC;
whereas the use of njegov correlates with the strict reading (10a), the use of svoj
correlates with the sloppy reading (10b).
(10) a. Samo Lusi posˇtuje njenog supruga. (covaluation, = (9a))
only Lucie respects her-3sg.fem.poss husband
b. Samo Lusi posˇtuje svog supruga. (binding, = (9b))
only Lucie respects self-poss husband
Thus, the contrast between English and SC is not restricted to (1) and (2). Rather, as
illustrated in (5)–(9), VP-ellipsis data and the focus particle only data in English and
SC are consistent with the pattern in (1) and (2). And, as evident from the discussion
on Macedonian, we can dismiss the possibility of explaining this pattern by the trivial
fact of the number of pronouns.
2.2 Binding and Movement
The analysis I propose for the data in (1) and (2) supports the hypothesis that
anaphors and their antecedents are related via movement (see Lidz & Idsardi 1998,
Hornstein 2001, Zwart 2002, Grohmann 2003). The particular implementation of the
idea that I adopt and adapt here is that of Hornstein (1998, 2000, 2001, 2006).
The distributional similarities between NP-movement and anaphora were observed
as early as Chomsky 1973 (see Chomsky 1973, 1981; Reinhart 1976, 1984; Bouchard
1984; Lebeaux 1985, among others). Ten different paradigms show that anaphor
binding and NP-movement are possible in the same domain. Two of them are given in
(11)–(13).
(11) a. John is expected e to lose the race.
b. John expects himself to lose the race.
(12) a. *John was expected (that) e would lose the race.
b. *John expected (that) himself would lose the race.
(13) a. *was seen e
b. *himself shaved (Lidz & Idsardi 1998)
208 Marijana Marelj
� 2011 Blackwell Publishing Ltd.
As illustrated in (11), both of the relationships are possible in the subject of a
nonfinite clause but not in the subject of a finite clause (12). As illustrated in (13),
both anaphors and NP traces require an antecedent. Crucially, in all the environments,
the antecedent must c-command the anteceded. ‘‘Syntactic movement and pronom-
inal anaphora turn out, then, to obey one general restriction on variable-binding,
namely that the binder must c-command the variable at SS’’ (Reinhart 1984:162).
Given the distributional similarities between the anaphor binding andNP-movement,
it is unsurprising that various scholars have argued that they involve the same syntactic
relation. Although the issue is not uncontroversial in the literature (see Brody 1998,
1999; Lidz & Idsardi 1998; Reuland 2001 and the references in the text for discussions
pertaining to the issues here), in the remainder of this paper I will work under the
assumption that the two phenomena should be unified. Furthermore (again, not
uncontroversially), I will assume here that they are unified through movement, as
implemented in Hornstein 2001, 2006. Before we proceed with the analysis of the
pattern in (1) and (2), allow me to provide the rationale behind the derivational
approaches to binding in general and introduce the basics of Hornstein�s approach.
2.3 Binding in a Minimalist Setting
The fact that an anaphor needs to be bound locally establishes that there is a
dependency between the position of the anaphor and that of its antecedent, just as in
the case of NP traces. Typically, however, construal rather than movement is assumed
to account for the dependency. In what follows, I present evidence that this position is
suspect on both conceptual and empirical grounds.
A movement approach to binding (as proposed by Grohmann, Hornstein, Lidz &
Idsardi, and Zwart) is less radically different from the more traditional take on
binding than it might at first seem. Even a cursory comparison to Chomsky�s
approach to binding as LF cliticization (see Chomsky 1986, 1993) confirms this.
Following Lebeaux (1985), Chomsky�s (1986) revised binding theory links the
antecedent and the reflexive derivationally, albeit through the covert (rather than the
overt) movement of the reflexive (see also Pica 1987; Heim, Lasnik & May 1991;
Chomsky & Lasnik 1993, among others). Further, as in theories that argue for the
overt movement of reflexives, this movement is local. If, under either type of account,
the dependency between the anaphor and its antecedent is captured by local
movement, one might wonder what the role of binding condition A is. Although they
differ in details, the crucial concern of ‘‘methodological economy’’ (i.e., Occam�s
razor) underlies the derivational analyses of Lidz & Idsardi (1998), Hornstein (2001),
Zwart (2002), and Grohmann (2003). As further argued by Hornstein (2001, 2006), a
consequence of eliminating condition A is the elimination of condition B for
empirical reasons: bound pronouns and reflexives are in complementary distribution.
If, however, the distribution and the interpretative properties of the former are
regulated by the theory of movement and the distribution and the interpretative
properties of the latter via binding theory, it is not obvious why they should be in
complementary distribution.
Bound-Variable Anaphora and Left Branch Condition 209
� 2011 Blackwell Publishing Ltd.
Reviving Lees & Klima�s (1963) approach to binding, Hornstein (2001, 2006)
proposes a derivational analysis of reflexivization in minimalist terms. By also
reducing the distribution of pronouns to the theory of movement, he accounts for their
complementary distribution. The treatment of bound pronouns is discussed in section
3. Let us briefly review the relevant portions of his treatment of local reflexives here.
Under Hornstein�s account, the locality conditions specified by condition A are
reduced to the locality conditions on movement (Shortest Move/Minimal Link
Condition). The structure for reflexives is roughly as in (14).
(14) Johni admires [ti himself ].
The derivation of John admires himself proceeds by John raising from the Theme
position to the Agent position. This clearly violates the Chain Condition. Although
Chomsky (1995) maintains the position of Chomsky 1986 and bans the acquisition of
h-roles through movement, a number of researchers have provided evidence against
this position and in favor of treating h-roles as formal features (see Bosˇkovic´ 1994;
Bosˇkovic´ & Takahashi�s [1998] analysis of scrambling in Japanese; Lasnik�s [1999]
analysis of pseudo-gapping; Manzini & Roussou 2000; Rodrigues 2004, among
others). h-roles are features on verbs; they are just like Case, agreement, or focus
features in that they drive movement.
If a well-formed chain can have any number of h-roles, examples like *John
admires cannot be ruled out by theta-theoretic reasons. Hornstein argues that such
derivations are ruled out instead by Case reasons. Under his account, reflexives like
himself are grammatical formatives whose role is to check Case. *John admires is
ungrammatical because John cannot check both nominative and accusative. In John
admires himself there is no Case clash; John checks nominative and self checks
accusative (15).
(15) a. [V¢ admires [[ John] -self]]
b. [vP John v0 [VP admires [[John] -self ]]]
c. [vP John v0 [VP admires [[him]self ]]]
Under Hornstein�s account, the interpretative properties of the dependency between
the reflexive and its antecedent are derived syntactically. To account for the sloppy
reading of reflexive constructions under VP ellipsis as in (16a), as well as for the
unavailability of a strict reading in (16a), Hornstein argues that it is sufficient to treat
reflexives as the residue of movement. That is, if reflexivization is movement, then it
follows that himself (elided in the second conjunct) is the reflex of the movement of
Bill. As a consequence, the only available reading is predicted to be the sloppy
reading. In light of this analysis, one can also account for the fact that the sloppy
reading is the only available reading in both the reflexive (16a) and the raising
construction (16b).
(16) a. John admires himself and Bill does too. (Hornstein 2001)
b. John seems to like bagels and Bill does too.
210 Marijana Marelj
� 2011 Blackwell Publishing Ltd.
Not all self-elements are the result of movement. As shown by Reinhart &
Reuland (1991, 1993), when it occurs as a logophor, a SELF-anaphor does not show
any of the trademark properties of local reflexives. For instance, unlike local
reflexives, logophors are permitted in environments where they are not c-commanded
by an antecedent, as in (17). Moreover, in some instances of the logophoric use of
SELF-anaphors, the antecedent is not even present, as in (18). Keeping the local
anaphors and logophors apart is reasonable for any theory of binding. The separation
is also maintained under Hornstein�s account.
(17) a. Max�s eyes watched eagerly a new picture of himself in the paper.
b. Bismarck�s impulsiveness had, as so often, rebounded against himself.
(Reinhart and Reuland 1993)
(18) a. A picture of myself would be nice on that wall.
b. Physicists like yourself are a godsend.
(Reinhart & Reuland 1991, after Ross 1970)
To sum up, I have presented the basics of movement approaches to binding, how
such approaches compare with more ‘‘traditional’’ approaches to the issue, as well as
the main tenets of Hornstein�s (2001, 2006) account. Against this background we now
proceed to discuss the pattern in (1) and (2).
3. Deriving the Contrast between English and Serbo-Croatian
3.1 Pronouns as Last Resort Expressions
Let us begin this section with an overview of the tools and theoretical considerations
relevant for the discussion of the pattern in (1) and (2).
Hornstein treats pronouns as Last Resort expressions. ‘‘Pronouns are (i) costly to
use, (ii) introduced by the computational system, and (iii) not items of lexical array’’
(Hornstein 2001:177). The prediction is that every time there is a comparison
between derivations with and without a bound pronoun, the one without the bound
pronoun is favored. ‘‘A pronoun can get a bound interpretation if and only if it goes
surrogate for a (syntactic) variable that could not licitly be formed by movement’’
(Hornstein 2001:176).4
Under his account, the reason that (1), repeated here as (19a), is acceptable with
the bound reading lies in the fact that (19b) is not a viable derivation. Hence, a
pronoun establishes a relation between two positions that cannot be established
through movement.
(19) a. Everyonei loves hisi mother.
b. Everyonei loves [ti mother]. (Hornstein 2001)
4 Again, the reader is reminded that this remark pertains only to bound pronouns. Referential pronouns
are not grammatical formatives but are selected from the lexicon. The issues are discussed in section 3.2.
Bound-Variable Anaphora and Left Branch Condition 211
� 2011 Blackwell Publishing Ltd.
Hornstein suggests that what blocks the more economical derivation in (19b) in favor
of the less economical derivation in (19a) is a violation of the Left Branch Condition
(Ross 1967/1986). Under the movement derivation, everyon
本文档为【Bound_variable anaphora and left branch condition】,请使用软件OFFICE或WPS软件打开。作品中的文字与图均可以修改和编辑,
图片更改请在作品中右键图片并更换,文字修改请直接点击文字进行修改,也可以新增和删除文档中的内容。
该文档来自用户分享,如有侵权行为请发邮件ishare@vip.sina.com联系网站客服,我们会及时删除。
[版权声明] 本站所有资料为用户分享产生,若发现您的权利被侵害,请联系客服邮件isharekefu@iask.cn,我们尽快处理。
本作品所展示的图片、画像、字体、音乐的版权可能需版权方额外授权,请谨慎使用。
网站提供的党政主题相关内容(国旗、国徽、党徽..)目的在于配合国家政策宣传,仅限个人学习分享使用,禁止用于任何广告和商用目的。