Reference in Discourse by Andrej A. Kibrik (review)

This book reflects almost thirty years of research undertaken by Andrej Kibrik on reference in discourse, including numerous publications on anaphora, pronouns, deixis, and other topics in linguistics. It presents a comprehensive summary of this research, its current results, and innumerable suggestions for further investigations. Kibrik calls his book "an exercise in cognitive discourse analysis" (p. 18) because he aims at bringing together two rather separate fields: typology and cognitive linguistics. The title perfectly reflects the topic, though, of course, the author is not able to treat all aspects of it in detail. He takes the production-oriented speaker perspective (as opposed to the perception-oriented hearer's perspective). The book combines a variety of methods of investigation such as corpus analysis, experiments, use of stimuli, analysis of grammatical descriptions, among others.

The book is divided into five parts, containing a total of sixteen chapters. Despite its impressive length, it is structured and written in a reader-friendly manner. Each chapter concludes with a summary, and at the end of each part there is also a summary recapitulating the most important points. The appendix provides a questionnaire and a map of the languages mentioned. The references are followed by a language index and a subject index.

Part 1 contains the introduction and lays the basis for Kibrik's cognitive multifactorial approach. He starts out by stating that reference (including anaphora) is a discourse phenomenon. Syntactic anaphora within a smaller domain such as the clause is simply one instance of this larger domain of reference in discourse. He introduces his typology of referential devices by saying that the basic distinction ("the only true and universal opposition" [p. 42]) is between full referential devices (proper names and common nouns) and reduced referential devices (pronouns and zero forms). Kibrik is largely concerned with the issue of what determines speakers' choice in their usage of full referential devices as opposed to reduced referential devices. Furthermore, he focuses almost exclusively on "specific definite reference" in narrative texts, and on referential devices in the third person.

Parts 2 and 3 deal with the typology of referential devices. Part 2 starts with a discussion of what for Kibrik counts as the three most fundamental types of reduced referential devices: (i) free personal pronouns (including strong pronouns and clitics, i.e., prosodically weak pronouns) (ii) bound personal pronouns (i.e., affixes), and (iii) zero forms. He proceeds by saying that languages can be divided into three types according to their major referential device (e.g., English belongs to the first type, Navajo to the second type, and Japanese to the third type). He states that around two-thirds of languages belong to one of the three pure types. However, this is not so clear. First of all, languages usually have more than one reduced referential device, and may prefer one device under certain circumstances, and another under different circumstances. Furthermore, the boundary between clitics (first type) and affixes (second type) is notoriously difficult to draw. Kibrik only notes this difficulty without discussing it, and says that one has to rely on the judgment of language experts.

Kibrik proceeds by taking a look at the factors that have effects on reduced referential devices, e.g., dependency on clause position, imperatives, clause coordination, and subordination. Such factors are quite important for reference in discourse, since natural texts consist not only-and not even primarily-of indicative main clauses, and lead to considerable language-internal variation that makes it quite difficult to divide languages into the three pure types.

Kibrik devotes an entire chapter to the discussion of bound pronouns and the relationship between bound pronouns and agreement. This discussion is crucial, since agreement is one of the central grammatical notions that also functions as a criterion for other properties such as subjecthood, finiteness, etc. Kibrik tries to show that most of the markers that at least traditionally have been categorized as agreement markers are, in fact, bound pronouns because they are referential...