Foreground-background information and the syntax of subordination

In general, study of the linguistic coding ofdiscourserelationslikeforegroundbackground Information has been limited by structure-dependent or introspection-dependent methods for Information Identification and by vagueness about what constitutes evidence for syntactic coding. This paper addresses these two persistent problems by considering one specific hypothesis about the syntactic coding of foreground-background Information: independent clauses code foreground and pivotal Information; dependent clauses code background Information. Four types of narrative-discourse production were collected from four independent groups of English-speaking subjects using a videotaped cartoon äs the primary elicitation Stimulus. The resulting data are analyzed using a psychologically grounded, syntax-independent, and introspection-free method for identifying pivotal, foreground, and background propositions. Statistical analysis of the interaction of foreground-background Information and the syntax of clause type shows that the hypothesis is correct, given an explicit deflnition of pragmatic coding and the pattern of evidence required to demonstrate it. l. Introduction The general goal of functional linguistics is to describe and explain empirically Ihe connections possible between syntactic form and semantic or pragmatic functions. This general goal can be pursued by considering one of two questions: (1) what is the function or purpose of a given fragment of syntax, or J2) how is a given semantic/pragmatic function realized in syntax. The 3165-4888/85/0005-0085 $2.00 Text 5(1-2) (1985), pp. 085-122 9 Mouton Publishers, Amsterdam 86 Russell S. Tomlin theoretical apparatus developed to handle such questions must operate under the dual empirical constraints of accounting for the behavior of particular languages and for the behavior of language in general. There are two interrelated and persistent problems which constrain meeting this general goal. One problem is methodological. In order to argue for any particular form-function hypothesis, it is necessary that the pertinent syntactic forms and semantic/pragmatic functions be identified independently of one another. While not so much a problem for syntactic form, the syntax-independent identification of semantic/pragmatic functions, especially those that are discourse-based, remains severely problematic. The second problem is more theoretical. Even though all functionally oriented linguistic descriptions utilize such a coding concept,the term 'coding' or 'correlation' is generally left undefmed and vague. It is unclear, in general, just what sort of relationship holds between syntax and semantic/pragmatic function when a claim of 'coding' is proposed. It is, in addition, unclear what pattern of linguistic evidence would support a particular coding relationship. One way to address these persistent problems in functional linguistics is to consider in detail a specific instance of a proposed form-function correlation. Doing so demands that one addresses specifically both of these problems, developing syntax-independent means for conducting data analysis and establishing clearly what is meant by 'coding' and its relation to pertinent linguistic evidence. The directions taken to deal with these problems for a specific question represent instances of general Solutions that might be taken in functional linguistics. The exploratory experiment described in this paper addresses both of these problems. Following the data elicitation paradigm of Chafe (1980), four independent groups of English-speaking subjects were shown a 108 second Video tape of an animated cartoon and asked to produce narratives of four different types: on-line oral descriptions, post-viewing oral narratives after a short delay, post-viewing written narratives after a short delay, and postviewing edited written narratives after a long delay. The data so produced were analyzed in order t o test the hypothesis that, in English, independent and dependent clauses are used respectively to code foreground and background Information in discourse. The experiment addresses these two persistent problems by testing this hypothesis explicitly in several related discourse domains. The results show that, given a reasonably tough characterization of what 'coding' means, the relationship between clause dependency and foreground-background informaForeground-background Information and syntactic Subordination 87 tion in English is strong enough to demonstrate that a special coding relationship does hold between them. 2. The problem The specific problem examined in this study derives from research into narrative organization conducted by Longacre (1976), Jones and Jones (1979), Hopper (1979), and others. These papers draw a distinction between two levels of Information in discourse concerned with the importance or significance of particular propositions. Foreground Information is Information which is more important, or significant, or central to the n&n&tive.Background Information serves to elaborate or enrich foreground Information. (The prscise characterization of these notions äs utilized in this study is taken up below). Besides introducing an important pragmatic distinction, these papers propose that certain aspects of natural-language syntax have the function of signaling whether Information is foreground or not. Of particular interest for this paper is the possible connection between independent-dependent clause types and foreground-background Information. It appears that in many languages, and perhaps in language in general, independent-dependent clause types correlate in some important way with foreground-background Information. This relationship is described in (1), and represents the specific hypothesuthat will be examined and tested in this paper: (1) Hypothesis: Independent clauses correlate with, or code, foreground Information: dependent clauses correlate with, or code, background Information. Tiis hypothesis intuitively seems correct, for it is, after all, very familiär äs tre kind of hypothesis that has regularly been put forward in school grammar b»oks and writing manuals such äs Glorfeld, Lauerman, and Stageberg (1977: 1*7). Q The general principle is this: Put the important parts of your message in an independent clause and the supporting or collateral parts in subordinate elements. 88 Russell S. Tomlin The problem is that such assertions are not grounded very well in empirical study. Even attempts in functional linguistics to connect clause dependency with Information type fail, because the connections derive primarily from deductive theoretical arguments. To test assertions such äs that in (2) or the hypothesis in (1), a research effort must be designed that satisfies the followingrequirements: 1. Explicit and syntax-independent means of Information Identification. Decisions about the Information value (foreground, background, etc.) of individual propositions in natural-discourse data must be made explicitly, without dependence on introspection, and syntax-independently in order to assure both intrasubjective and intersubjective consistency in analysis. Failure here dooms the final analysis to vagueness or circularity. 2. Conditions for coding. One must establish clear theoretical concepts of linguistic coding or correlation. One must also determine what patterns of evidence in the data will count äs meeting those theoretical Standards. Failure in the former results in vagueness for the Claims; failure in the latter results in empirical vacuousness in any test of the Claims. 3. Multiple-subject data. Analysis must be made of the discourse production of an appropriate number of Speakers in order to distinguish individual Variation from group tendencies. While evidence from single subjects may result in strong hypotheses, such evidence constitutes only the very weakest test of the hypothesis. 4. Variation in discourse setting. One must vary the discourse setting in which data is collected in order to ascertain whether the hypothesis tested holds, or fails to hold, in general or only in some specific type or genre of discourse. The experiment conducted for this study meets all four of these requirements. In doing so, it subjects the hypothesis in (1) to the sort of severe empirical test that all Claims of functional coding of discourse by syntax should undergo. 3. Analytical framework: Category definitions and operationalizations 3.1. Foreground-background Information The .general treatment of foreground-background Information in this paper follows Tomlin (1983), which develops research presented in Hopper (1979), Foreground-background Information and syntactic Subordination 89 Jones and Jones (1979), and Longacre (1976). In these studies, foreground Information is characterized äs Information which is more central or salient or important to the development of the discourse theme. Background Information is that which elaborates or develops foreground Information. In addition, äs discussed primarily by Jones and Jones (1979), the foreground-background distinction is not a simple binary concept, but deflnes a continuum, permitting an individual proposition to be ranked against others in terms of its importance or centrality t o the discourse theme. More explicitly, the foreground-background distinction is here defmed theoretically äs a continuous thematic relation of significance which a given proposition bears with respect to its paragraph-level, superordinate-discourse theme. So, for example, in the text fragment in (3) below, P2 is more significant than P J , because it more directly relates Information developing the general theme of Miss Muffett's encounter with the spider. (3) Little Miss Muffe« sät on a tuffett P i Eating her curds and whey. Along came a spider And sät down beside her P2 And frightened Miss Muffett away. If (4) represents the foreground-background continuum, P2 is higher than P! onthat continuum. (4*