This book "asserts that the origin and spread of languages must be examined primarily through the time-tested techniques of linguistic analysis, rather than those of evolutionary biology" and "defends traditional practices in historical linguistics while remaining open to new techniques, including computational methods" and "will appeal to readers interested in world history and world geography."
Date: Tue, 18 May 2004 15:05:23 +1200 From: Frank Lichtenberk Subject: Subordination
Cristofaro, Sonia (2003) Subordination, Oxford University Press.
Frantisek Lichtenberk, Department of Applied Language Studies and Linguistics, University of Auckland, Auckland, New Zealand.
[Another review of this book appears in http://linguistlist.org/issues/15/15-682.html --Eds.]
The book under review is a major cross-linguistic, typological study of subordination (as defined by the author). It is impressive in its scope and some of its findings, but there are also, in my view, some problematic issues.
Broadly, the volume can be divided into three major parts. Chapters 1-4 provide the background and set out the basic premises and assumptions. Chapters 5-7 deal with the three basic types of subordination: complement relations, adverbial relations, and relative relations. It's these chapters that present the major empirical results. Chapters 8-11 deal with general issues arising from the findings, such as comparison of the types of subordination, explanations for the phenomena found, and prospects for future research. There are also several appendices, one of which gives information on the constructions and their properties found in the languages in the sample.
In any large-scale cross-linguistics study, defining the phenomenon to be investigated is an important issue. Cristofaro opts for a strictly non-structural definition of subordination: ''By subordination will be meant a situation whereby a cognitive asymmetry is established between linked SoAs [states of affairs], such that the profile of one of the two (henceforth, the main SoA) overrides that of the other (henceforth, the dependent SoA).'' (p. 33). This property of subordination is referred to by Cristofaro as ''the Asymmetry Assumption''. While a functionally/cognitively based definition of the phenomenon under investigation may be necessary in a cross-linguistic study, a very strict application of such a definition may have as its consequence exclusion of certain syntactic constructions that other linguists would regard as qualifying for inclusion. And this is what happens here. Cristofaro leaves out from her investigation of relative relations what she calls ''non-restrictive relatives'' (including asserting relative clauses), because the profile of the state of affairs that a non- restricting relative expresses is not overridden by that of the other state of affairs (which would normally be thought of as being expressed in the main clause). Conversely, the definition allows for inclusion of constructions that would traditionally not be considered to involve subordination. Cristofaro compares the English sentence (1.3) ''After she drank the wine, she went to sleep'' and a corresponding sentence in Mandarin and says (p. 2): ''... the notion of subordination is independent of the way in which clause linkage is realized across languages. For instance, the English sentence in (1.3) involves a clause that would be identified as subordinate under most traditional criteria. However, the corresponding Mandarin Chinese sentence in (1.4) involves two morphosyntactically independent clauses.'' The translation of the Mandarin sentence is 'After s/he drank the wine, she went to sleep'. While reading the subsequent chapters of the book, the reader may need to remind himself/herself of what Cristofaro means by subordination.
Cristofaro's adoption and strict application of a functionally/cognitively based definition of subordination has further relevance, which has to do with another methodological principle adopted in the study, although the former does not entail the latter. The methodological principle has to do with heavy reliance on translation in determining whether a given construction in a certain language involves subordination or not. In cases where the source of the data does not provide information about the assertional value of a given construction type (and this, according to Cristofaro, is the usual case), ''[t]he solution ... is to assume that the translation used preserves the conceptual organization of the linked SoAs in the original sentence'' (p. 41). This strikes me as a risky assumption to make.
There is one more general premise that Cristofaro assumes that has important methodological implications, and that is the premise that ''all languages are able to express any cognitive situation, and all the ways in which a particular cognitive situation is expressed should be taken into account'' (p. 49). This allows her to include cases where ''the relevant situation is only inferred from the context, if that is the standard means to express that situation in the language. [footnote omitted]'' (ibid.). Inference, then, can be subsumed under subordination, in the absence of a syntactic construction. Since, as Cristofaro assumes, all languages must be able to express the same ''cognitive situations'', then if ''a particular construction is the only means available in a language to express a particular semantic relation between SoAs ... one has to assume that the relevant construction can express all of the cognitive correlates of that semantic relation, including subordination'' (p. 46). Consequently, ''there may be cases where one has to regard a particular construction as an instance of subordination independently of how it is translated'' (ibid.). There may, then, be a conflict between the assumption that all languages may express the same ''cognitive situations'' and how a certain construction is translated, and in such cases the former takes precedence.
The purely functional/typological definition of subordination and its narrow application, the reliance, up to a point, on translation, and the assumption that all languages can express the same ''cognitive situations'' involving subordination may be controversial to various degrees, but it is impossible to tell whether Cristofaro's findings would be materially affected if the notion of subordination was applied differently, if translation was not such a strong guide, and/or if one did not assume that all languages can equally express relations between states of affairs.
In investigating the three types of subordination relations (complement, adverbial and relative), Cristofaro pays particular attention to two main structural properties of clauses that express subordinate states of affairs. One is the verb forms employed. Following Stassen (1985), Cristofaro distinguishes between two main types of strategy: balancing and deranking. In the balancing strategy, the ''subordinate'' clause contains verb forms that are equivalent to those found in independent clauses. In the deranking strategy, the subordinate clause contains verb forms not found in independent clauses; for example, certain distinctions (such as tense, aspect, and mood) are not expressed, or certain special forms are used not found in independent clauses. The other main structural characteristic of subordinate clauses that Cristofaro pays close attention to is the coding of participants, in particular the non-expression of certain arguments, and the expression of certain arguments as possessors or obliques.
Chapters 5-7 form the empirical heart of the book. Of these, the one on complement relations is the most extensive one. It is in this chapter that Cristofaro discusses two concepts that are crucial to the interpretation of some of the findings concerning two of the three subordination types, complement and adverbial. One of these is predetermination, the fact that ''some of the semantic features of the linked SoAs are predetermined by the nature of the relation itself'' (p. 111). For example, with 'want' as the main verb, the dependent state of affairs cannot be temporally located at a time earlier than the state of wanting. The other concept is that of semantic integration, which has to do with how tightly or how loosely the two states of affairs are interconnected. For example, in a causative relation one state of affairs is brought about by the other (causing) state of affairs. (For an earlier discussion of the notion of semantic integration see Givón 1980, 1990.) In addition to predetermination and semantic integration, what may also be of relevance in some cases is the fact that the dependent state of affairs is unrealized or that there is will or interest on the part of a participant that a dependent state of affairs be realized.
In dealing with complement relations, Cristofaro adopts Noonan's (1985) classification of complement-taking predicates (modal, manipulative, propositional attitude, etc.). The central finding of the study of complement strategies in the languages in the sample is expressed in the form of the Complement Deranking-Argument Hierarchy, formulated in terms of the types of complement-taking predicates:
If, in a given language, deranking is used at a certain position on the hierarchy, it is also used at all the positions to the left on the hierarchy. And, if, in a given language, A and S arguments are not expressed in the subordinate clause at a certain position on the hierarchy, they are not expressed at any of the positions to the left. A number of tables at the end of the chapter provide supporting evidence for the hierarchy as well as other generalizations.
As far as adverbial relations are concerned, Cristofaro considers the following subtypes: purpose, three kinds of temporal relation ('before', 'after', and 'when'), reality conditions (but not counterfactual conditions), and reason. As is the case with complementation, the factors of predetermination, semantic integration, and will/interest are relevant. For example, in an 'after' relation the two states of affairs are, by definition, sequenced in a certain way; and there is a high degree of semantic integration in purpose relations. Cristofaro establishes two major hierarchies for adverbial relations. Tables providing supporting evidence conclude the chapter.
The Adverbial Deranking Hierarchy: Purpose > Before, After, When > Reality Condition, Reason
As far as relative relations are concerned, Cristofaro includes only restrictive relativization in her study, and she includes only the following grammatical functions: A, S, O, Indirect object and Oblique object; not Possessor and not Object of comparison. Cristofaro adopts a methodological principle that some might find surprising. (I do.) In some languages a certain grammatical function cannot be relativized, but the relevant noun phrase may be promoted, which permits relativization. Thus, for example, in a language relativization may be restricted to S and A functions, but noun phrases in other functions may be promoted to S or A, which permits relativization. Cristofaro regards such cases as relativization on the function before the promotion; for example, relativization on O rather than on S. This is because of the assumption that ''any given language should in principle be able to express any given concept'' (p. 200) (see also further above in this review). If a certain thematic role is normally expressed in a certain grammatical function (say O), but promotion is required for the purposes of relativization, this should count as relativization on the non-promoted function ''because the same conceptual situation is being expressed'' (ibid.).
Cristofaro establishes the following Relative Deranking-Argument Hierarchy:
A, S > O > Indirect object, Oblique
(A slightly different hierarchy holds specifically for the expression and non-expression of aspect and mood, where there is a binary split between A, S, and O at the left end and Indirect object and Oblique at the right end.) The Relative Deranking-Argument Hierarchy is clearly not all that different from Keenan & Comrie's (1977) original Noun Phrase Accessibility Hierarchy, the distinction between indirect and oblique objects not being always clear. As is the case in the preceding two chapters, tables providing evidence for the hierarchy are given at the end of the chapter.
The relativization hierarchy is significantly different from those established for complementation and adverbial relations: it is stated in term of grammatical functions rather than in terms of semantic types (such as modal as opposed to desiderative, etc. predicates; and purpose as opposed to temporal, etc. relations.) There is another important difference between relative relations on the one hand and complement and adverbial relations on the other: the semantic factors of predetermination and semantic integration (and will/interest) are not relevant to the hierarchy. This raises a more general question of to what extent relative relations are comparable to the other two. I will return to this later.
The individual hierarchies established for the three different types of subordination are amalgamated into several global hierarchies in Chapter 8. Two of them are given here:
The Subordination Deranking Hierarchy Phasals, Modals > Desideratives, Manipulatives, Purpose > Perception > Before, After, When, A relativization, S relativization > Reality condition, Reason, O relativization > Knowledge, Propositional attitude, Utterance, Indirect object relativization, Oblique Relativization
The Subordination Argument Hierarchy Modals, Phasals, A relativization, S relativization > Desideratives, Manipulatives, Purpose > Perception > Before, When, After, Reason, Utterance, Propositional attitude, Knowledge, Reality condition
O, Indirect object and Oblique relativization do not appear in the Argument Hierarchy because only A and S arguments are relevant to complement and adverbial subordination.
Phasals and Modals occur at the left end of both hierarchies, while the Knowledge, Propositional attitude, and Utterance categories occur at the right end of both. In almost all cases there is grouping of subordination subtypes at the various positions on the hierarchies. Only the Perception category stands by itself, on both hierarchies.
The global hierarchies reflect, to some extent, two sets of factors, neither of which applies to all three types of subordination: predetermination, semantic integration, and preference apply to complementation and adverbial subordination, but not to relativization, while accessibility to relativization applies only to relativization, not to complementation and to adverbial subordination. One could, then, wonder about the rationale for placing relativization on the same hierarchies with the other two types. Interestingly, however, on the Deranking Hierarchy, not all the relativization subtypes group together: A and S vs. O vs. Indirect object and Oblique.
Building on the work of others (Givón 1980, 1990; Haiman 1983, 1985; Langacker 1987a, b, 1991), Cristofaro identifies some cognitive factors that underlie the morphosyntactic properties of the subordination constructions under investigation. One of these is syntagmatic economy, specifically non-expression of predictable information. Information in a subordinate clause may be predictable because of predetermination or because of obligatory argument-sharing between the clause expressing the main state of affairs and the one expressing the dependent state of affairs. Information that is recoverable may be omitted. This is what Cristofaro names the Principle of Information Recoverability. Another factor that is relevant is iconicity, which has to do with semantic integration: ''there is an iconic correspondence between semantic integration and morphosyntactic integration'' (p. 251). Verbal deranking and non-expression of arguments are manifestations of relatively high morphosyntactic integration.
Another major factor that Cristofaro identifies as relevant has to do with the cognitive status of dependent states of affairs. Under the Asymmetry Assumption (see further above), a dependent state of affairs lacks an autonomous profile: its profile is overridden by that of the main state of affairs. And while main states of affairs receive sequential scanning, dependent states of affairs do not. This absence of sequential scanning of dependent states of affairs motivates (but obviously does not require) the absence of tense/aspect/mood distinctions in clauses. And suspension of sequential scanning may (but does not have to) result in a dependent state of affairs being conceptualized as a ''thing'' rather than a process. This, in turn, may result in the presence of nominal characteristics in clauses encoding dependent states of affairs, such as case marking on, or with, the verb and possessor forms, and perhaps also in the presence of special tense/aspect/mood forms not found in independent clauses.
Cristofaro acknowledges that relativization is different from the other two types of subordination: ''[a] counterexample to the analysis just outlined is provided by relative relations'' (p. 287). Syntagmatic economy, which may be relevant to complementation and adverbial relations with respect both to a lack of tense/aspect/mood distinctions and to non-expression of arguments, is not relevant to relativization. In relativization, the lack of tense/aspect/mood distinctions results from the suspension of sequential scanning of dependent states of affairs, which may also be the case in the other two types of subordination. However, non-expression of arguments (gapping) is not the result of a cognitive principle; rather, it serves to identify the role of the relativized element. It again makes one wonder how, if at all, relativization fits in with the other two types of subordination in ''global'' hierarchies.
Cristofaro proposes a number of implicational correlations concerning various morphosyntactic properties of dependent clauses. For example: presence of case marking or adpositions implies lack of expression of person agreement; and non-expression of arguments implies non- expression of tense/aspect/mood. A correlation is established if the number of exceptions is not greater than one third of all the significant cases.
In one of the appendices Cristofaro lists the information on the languages in her sample on the basis of which she has formulated the various hierarchies. Nevertheless, one would like to see sets of data (sentences) from some of the languages to show how they follow the hierarchies, the global hierarchies in particular.
In the concluding chapter Cristofaro makes some further general points. Languages that contain only balancing strategies, languages that contain only deranking strategies, and languages that contain both balancing and deranking patterns for all types of subordination are disfavored compared to languages that contain split balancing-deranking systems where the two types are distributed in accordance with the hierarchies. There is a functional/cognitive explanation for that. The use of balancing strategies across the board would result in some non- economic patterns (expression of predictable information), while the use of deranking strategies across the board would result in some loss of information (non-expression in the absence of predetermination). Diachronically, the prediction is that languages are more likely to undergo changes that lead to economic and/or iconic patterns rather than vice versa.
At the very end Cristofaro points out that since her model of subordination is a cognitively based one, then if it is valid, one would expect there to be non-linguistic evidence for the concepts she postulates, such as semantic integration between states of affairs and the distinction between processes and things. At the moment, she is not aware of any such evidence, but this should only serve as an incentive for research.
There are aspects of Cristofaro's approach to subordination that many linguists may find themselves in disagreement with, such as the definition of subordination itself; nevertheless, this is an important study which is bound to stimulate further research on subordination, whether in individual languages or cross-linguistically.
Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in Language 4.333-377.
__________. 1990. Syntax: A functional-typological introduction, vol. 2. Amsterdam/Philadelphia: John Benjamins.
Haiman, John. 1983. Iconic and economic motivation. Language 59.781- 819.
__________. 1985. Natural syntax: Iconicity and erosion. Cambridge: Cambridge University Press.
Keenan, Edward L., and Bernard Comrie. 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry 8.63-99.
Langacker, Ronald W. 1987a. Foundations of cognitive grammar, vol. 1: Theoretical prerequisites. Stanford: Stanford University Press.
__________. 1987b. Nouns and verbs. Language 63.53-94.
__________.1991. Foundations of cognitive grammar, vol. 2: Descriptive application. Stanford: Stanford University Press.
Noonan, Michael. 1985. Complementation. Language typology and syntactic description, vol. 2: Complex constructions, ed. by Timothy Shopen, 42-140. Cambridge: Cambridge University Press.
Stassen, Leon. 1985. Comparison in universal grammar. Oxford: Basil Blackwell.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Frantisek Lichtenberk is a Senior Lecturer in Linguistics at the University of Auckland. His research interests include cognitive and functional linguistics, grammaticalization, and the Oceanic languages.