Date: Fri, 29 Oct 2004 17:32:04 +0200 From: Petra Burkhardt Subject: The Syntax of Anaphora
AUTHOR: Safir, Ken TITLE: The Syntax of Anaphora SERIES: Oxford Studies in Comparative Syntax PUBLISHER: Oxford University Press YEAR: 2004
Petra Burkhardt, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig
In this monograph, Safir proposes competitive principles to account for different anaphoric dependencies. The core notion of his approach is the existence of principles that act in competition with each other until the best available anaphoric form wins. As a consequence, complementary distribution between forms is derived. This view stands in contrast to theories that advocate that distributional patterns are accidental, and Safir illustrates at length how distribution can be derived from competition between forms. He further suggests that all principles governing anaphora are universal and that the distribution of anaphora follows from morphological and lexical properties of an anaphoric form. The presentation of the theoretical framework is enriched by numerous examples from different languages.
In chapter 1, as a departure for his proposal, Safir discusses previous approaches to anaphora resolution and points out their strengths and shortcomings. In particular, he focuses on the accounts that have evolved around Binding Theory (Chomsky, 1980, et seq.), as well as on Reinhart & Reuland's (1993) Reflexivity account. Some of the shortcomings of the former family of accounts that are pointed out are that coindexation relations are too symmetric, that the characterization of "domain" for Principle A and B is too English-specific (cf. e.g. instances of long-distance binding across languages); that complete complementary distribution between reflexives and pronouns cannot be attested crosslinguistically (cf. e.g. picture-NPs or logophors); that the notion of noncoreference must be rephrased; and others. Some of the challenges of the Reinhart & Reuland (1993) approach that Safir discusses are related to the formulation of A- chains and the distinction between [+/-R], among others. These considerations set the stage for Safir's proposal, which maintains a version of Principle A, but abandons Principle B and C (or their equivalents).
Chapter 2 illustrates the notions "dependent identity" and "coreference", which are crucial distinctions for Safir's theory. He claims that dependent identity is represented in syntax, while coreference is not. Furthermore, dependent identity describes the assignment of referential value to a (dependent) term as a function of its antecedent, while coreference reflects an independent covaluation. On the basis of this distinction, he argues that indices should be eliminated from representation and that alternatively, asymmetric dependent identity should be encoded (as in Higginbotham's (1983) arrow notation). He further discusses issues of disjoint reference and the notion of c-command, which he does not view as a critical condition for dependent identity.
In Chapter 3, Safir elaborates on the core principles of his theory. The theory is built around the notion of complementary distribution, where the best available form- to-interpretation fit wins the competition of anaphoric expressions. Crucially, complementarity is not defined on the basis of domains, but on the basis of a competition of forms. Consequently, the theory abandons what is traditionally known as Principle B. Safir points out that Principle B is stipulated and complementarity therefore accidental in frameworks that assume some version of Principle B. In contrast, complementarity should be derived. At the center of his theory is the Form to Interpretation Principle (FTIP), which is sensitive to a (language- specific) dependency scale. It serves to evaluate anaphoric forms with respect to a given antecedent. When dependence is blocked by the FTIP, Pragmatic Obviation assigns an obviative pair relation. However, what this chapter shows is that additional principles are required so-called "scalar intervention factors" that help to eliminate competitor forms. These constraints deal for instance with antecedent agreement, local antecedent licensing, (anti)subject orientation, distributivity. In the last section of chapter 3, Safir presents a nice overview of different scenarios in which complementarity does not rule - and in which traditional approaches have struggled to account for the distribution of anaphora. He also provides an overview and assessment of previous accounts of both derived and accidental complementarity in this chapter.
Chapter 4 focuses mainly on aspects of reflexive interpretation (i.e. coargument coconstrual). Safir couches the dependent reading required for the interpretation of reflexives within his general account of dependent identity reading controlled by the FTIP (vs. independent coreference reading) and attempts to illustrate that a separate formulation of a semantic notion of reflexivity (cf. Reinhart & Reuland, 1993) is not necessary. However, to accomplish this, he posits three additional principles the Coargument Dependency Constraint that controls distributed interpretation, the Locally Reflexive Principle that requires an identity- specific anaphor to be dependent on its coargument (if available), and the Principle of Thematic Indistinctness that controls coconstrual. This move allows the theory to account for the distribution of SEG- and SELF- anaphors (in e.g. Dutch or Norwegian), proxy readings (e.g. wax figure examples), the interpretation of guises, and the availability of inherently reflexive predicates on the basis of the dependency scale paired with the FTIP and the scalar intervention factors.
In Chapter 5, Safir addresses the issue of domain. He argues that the domain of anaphora is smaller than the domain of movement and that the notion of a single domain for both anaphora and movement operations must be abandoned. The view that he advocates is that the typology of movement only plays an indirect role in the distribution of anaphora, where movement is the mechanism that transports an anaphor into its local domain. Safir postulates an independent principle as working hypothesis that states that there cannot be a form of covert movement that is not attested for overt movement. Then he proceeds to look at the different types of movement that are available to the language system. For anaphors, he reports covert A-movement (e.g. in ECM constructions), covert clitic movement(e.g. Dutch SE- anaphors), and covert tense-sensitive operator movement (A'- movement) (e.g. Russian SE-anaphors). But he rejects covert head movement and A'-movement to Spec-CP on the basis that these covert movement types bear no similarity to possible overt equivalents. He further shows that covert movement of anaphors is motivated by independent requirements, such as Case assignment or subject orientation. He concludes that movement of SELF is hence no different from other forms of movement and a separate movement account for anaphora is therefore not required. Finally, chapter 5 also touches upon discourse-sensitive Unbounded Dependent Forms (Safir's term for what is often referred to as logophors, even though he supports a narrow view of logophoricity). He claims that these forms are not derived by movement (as e.g. evidenced by their distribution with extra-sentential and split antecedents) and are therefore not considered anaphors.
Chapter 6 focuses on the morphological and lexical properties of anaphora and attempts to illustrate how the internal structure of a given form determines its distribution. As with other factors that have an impact on interpretation, the internal structure is considered to predict whether a form is subjected to competition or not and then the FTIP takes over again. Safir distinguishes between pronominals which are exclusively bundles of features - and relational anaphors which also carry semantic content. He emphasizes again that there is no need for Principle B (or a grammatical feature [+/-pronominal]), because the internal structure of a form determines whether a form is an anaphor or a pronoun. Furthermore, while the FTIP regulates competition among dependent forms, another principle - Weak Pronoun Competition (cf. Cardinaletti & Starke's (1994) typology of pronoun classes) facilitates selection of the weakest pronoun form among those forms that tie under FTIP. Concerning relational anaphors, Safir demonstrates that these entities form a small class of lexical semantic relations, largely related to body part terms (e.g. "head", "bone", "self", "same", "other", etc.), which are also found separately from anaphora. This indicates that this particular property is not a property of anaphora per se. It is also shown that specific interpretive properties are conveyed by these forms (with respect to identity-specific dependencies, distributivity, etc.). Moreover, a number of interesting case studies are presented, which serve to illustrate that crosslinguistic differences are due to the available competitor set in a given language. In this chapter, Safir also analyzes null elements and discusses how the competitive principles can account for their distribution. He further addresses the distribution of reciprocal elements and mentions some preliminary solutions based on semantic properties but (as he points out) more research is needed to fully account for reciprocals within the present approach.
In chapter 7, Safir highlights that the study of anaphora can be used as a window into the architecture of the language system. The theory of anaphora presented in this monograph depicts the language system as a modular system, where the principle of Local Antecedent Licensing (Safir's version of Principle A) regulates processes of convergence and the other principles proposed throughout the book limit processes of interpretation. Crucially, the majority of principles are regarded to be part of a syntax- sensitive semantic module.
This monograph is a valuable contribution to the study of anaphora since it attempts to address a broad spectrum of phenomena relevant for the study of anaphoric dependencies. Safir uses data from numerous different languages to support his claims, and he discusses examples from various dependent elements, including short- and long-distance reflexives, reciprocals, pronominals, epithets, and others. He takes up the ongoing debate of how to account for the varied patterns of anaphora distribution across languages. As early as Chomsky's influential Binding Theory was published, apparent drawbacks and counterexamples have been discussed in the literature. And even after reading this monograph, it is evident that there are still many puzzles to be solved. However, Safir addresses one of the core issues that has been criticized in the past, which is the invalidity of Principle B (and C). His solution is to discard any version of Principle B and C; instead, distribution is governed by competition of forms. This approach fits well with current views of economy, and whether one wants to commit to the particular principles proposed in this theory or not, this book supports the idea that a theory of anaphora or any theory of language for that matter - should incorporate economy- or competition- based principles (cf. e.g. Reuland (2001) for a different approach). In addition, what is clear form Safir's work as well as previous accounts by other authors is that some version of Principle A must be preserved, since there is strong empirical evidence for a restriction on locality between anaphors and their antecedents, and Safir achieves this by introducing the principle of Local Antecedent Licensing.
At first glance, one might think that the theory proposed in this book accounts for dependency effects on the basis of just two principles and an underlying algorithm: the Form to Interpretation Principle (FTIP) and the Pragmatic Obviation Principle. But then it becomes apparent that many other factors play a role in the selection of the best anaphoric form in a given environment. For instance, matching of morphological information is included as a factor, so is the effect of distributive vs. collective interpretation, locality and thematic requirements, etc. This collection of constraints ("scalar intervention factors") could be considered both a strength and a weakness of the proposal. The availability of all of these scalar intervention factors leaves us with a complex system and more needs to be said about when and how these factors interact with each other - because even if they are unary principles, it must be determined at what point they impact dependency relations. Nonetheless, I generally regard it as one of the strengths of this proposal that it acknowledges a variety of different factors that influence the selection of anaphoric forms across languages. It seems that previous proposals have not spelled out this broad range of factors and incorporated them into their framework. Moreover, another nice effect of the competitive approach coupled with scalar intervention effects is that it allows for interpretations that are supported by contextual factors, but would otherwise be unacceptable. A further strength of the book is that Safir draws on some striking examples and attempts to solve them within his competitive approach. Here, one of the phenomenon he does not touch upon are backward anaphora, and it would be interesting to see how his theory can account for these dependency relations (as it appears to me that the Independence Principle would block dependent reading in these cases).
Generally speaking, the monograph also provides a good, yet obviously selected overview of previous approaches to anaphora and their respective strengths and challenges. Furthermore, the theory stresses some pertinent generalizations (e.g. the discussion of movement operations and domains), and a convincing argument is presented for derived complementarity. These generalizations provide a good starting point for future research, which could test the predictions associated with the dependency scale and the competitive principles. Since the theory is based on principles of economy, psycholinguistic investigations might be particularly fruitful to evaluate the predictive power of underlying principles.
Finally, it should also be noted that the reader is repeatedly referred to two other publications by Safir (2004, in prep.) at many interesting points throughout the book. This is unfortunate because some issues remain unresolved but one can look forward to reading more about these fascinating issues in the future.
Cardinaletti, A., and M. Starke. (1994). The typology of structural deficiency: On the three grammatical classes. University of Venice Working Papers in Linguistics 4:41-109.
Chomsky, N. (1980). On binding. Linguistic Inquiry 11:1-46.
Higginbotham, J. (1983). Logical form, binding and nominals. Linguistic Inquiry 14:395-420.
Reinhart, T., and E. Reuland. (1993). Reflexivity. Linguistic Inquiry 24:657- 720.
Reuland, E. (2001). Primitives of binding. Linguistic Inquiry 32:439-492.
Safir, K. (2004). The syntax of (in)dependence. Cambridge: MIT Press.
Safir, K. (in preparation). Person, perspective and anaphora. Unpublished Ms, Rutgers University.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Petra Burkhardt is a guest researcher at the Max Planck Institute for Human Cognitive and Brain Sciences in Leipzig. Her current research interests include anaphor resolution, the syntax-discourse interface, and sentence processing. In her 2004 dissertation, she investigated pronoun-antecedent relations as an interface phenomenon and showed how syntactic and discourse-based information work together during pronoun resolution. To this end, she utilized psycho- and neurolinguistic studies to examine pronoun interpretation in English and Dutch. Currently, she is working on issues of the integration of different kinds of nominal anaphors in German.