In all Discourse Modes and genres, one finds passages that suggest a particular voice. They convey a sense of subjectivity, a point of view toward propositional information. “Point of view” is familiar as a literary term referring to presentation of the mind of a fictional character in narrative. More generally, point of view is “the perceptual or conceptual position in terms of which narrated situations and events are presented” (G. Prince 1987:73). Linguists now use the term for expressions of speech and thought, evidentiality, perspective, and other indications of an authorial or participant voice. “Point of view” is used almost interchangeably with “perspective” and “subjectivity.” I shall use the latter term as more general. Subjectivity is conveyed by grammatical forms and lexical choices.

Three traditions come together in the area of subjectivity. One is deixis and its linguistic expression. Deixis is a general term for the centrality of the here and now in language. The study of deixis takes as basic the canonical speech situation with Speaker and Addressee, and explores its linguistic ramifications. The second tradition involves evidentiality, indications of the source and reliability of information. Evidentiality is a relatively new term for the semantic field of attitude toward knowledge, a kind of modality. Linguistic resources for this vary strikingly across languages. Finally, subjectivity conveys the contents of mind and personal perspective; here linguistic study is complemented by a strong literary tradition.

This work would not have been possible without many kinds of support that I have received from individuals and institutions. I am grateful to them all. The New York Community Trust has supported much of my research on discourse. It also funded three small conferences on discourse at the University of Texas which advanced the work. I received a Faculty Research Award from the University of Texas Research Institute in 1994, which enabled me to pursue this project. During that period I spent some time as a Visiting Scholar at the Maison Suger of the CNRS in Paris. I was the beneficiary of a Dean's Fellow award from the College of Liberal Arts in 1998.

Parts of this work have been presented at conferences and colloquia. I benefited greatly from the discussions that followed, as well as the presentations themselves. They include three conferences on discourse structure at the University of Texas; an International Round Table on The Syntax of Tense and Aspect at Université de Paris, 2000; a symposium on Information Structure in a Cross-Linguistic Perspective at the University of Oslo, 2000; a conference on Linguistics in the Next Decade at the Academia Sinica, Taipei, Taiwan, 2000; a Linguistics Colloquium at the University of Siena in 1998; an invited lecture series at the City University of Hong Kong in 1998.

Sentences that differ in arrangement and sentence accent also differ in meaning, although their propositional content may be the same. As Bolinger puts it, discussing the difference between active and passive (1977:9):

The meanings that Bolinger talks about involve the way a text presents information, its presentational structure.

Surface structure presentation instructs the receiver about how to organize the information in a sentence. The sentences of a text are not undifferentiated wholes, nor simple linear arrangements of words. I adopt the approach to presentation originally put forth by Prague School linguists and further developed in recent years. This approach uses the notions of communicative dynamism, topic–comment, and focus–background, to understand the internal organization of sentences and how they are deployed in texts.

Canonical sentences furnish the background for variation. I consider here structures that play off this background with different word orders and syntactic structures. Non-canonical structures have special force, because of their features and because they depart from the basic case. Writers choose structures. I assume that choice is based on assessment, not necessarily conscious, of how a structure affects interpretation in a specific context.

Sentence-internally, non-canonical structures highlight or downplay the material in certain positions. Syntax may enhance connectedness between sentences by placing information that is familiar to a discourse first in a sentence. A given structure may allow or block a topic relation with the following sentence. Changes in direction may be conveyed by sentences that lack such connection, and by breaks in the syntactic pattern. Thus syntactic patterning affects the organization and progression of discourse passages.

This chapter concentrates on non-canonical structures that affect topic and sentence connectedness, the main factors of presentational progression. I draw on discussions in the literature of a variety of constructions. Together they give a sense of the different tools that the language makes available. I will also look briefly at multi-clause sentences, and will discuss paragraphs as text units.

The interpretations involve inference. Semantic presuppositions are close to the linguistic forms: they are triggered by particular structures, such as cleft sentences and temporal clauses; and by particular forms such as the focus particles “only” and “even.” Pragmatic presuppositions of familiarity status and linking inferences depend on context, world knowledge, and convention.

Text understanding is a constructive process that results in a mental representation. In this it is rather like vision. Seeing a tiger, for instance, is the result of a process of construction. The perceiver's mind/brain converts information from a pattern of light and dark on the retina to a representation, an interpretation of the object – an image of a tiger (Marr 1982). Inference often plays an important role at the final stages. Similarly, to understand a sentence, one goes from a sound wave or set of marks to a conceptual representation which brings together information of different kinds, some of it supplied by inference and world knowledge. The active nature of understanding informs the approach to text structure and its representation that I take here.

Studies in cognitive science and psycholinguistics provide background information about text understanding and mental models. Mental models are developed in effect with the structures of Discourse Representation Theory.

Section 3.1 discusses the communicative context of language; 3.2 considers some inferences in language understanding; 3.3 relates language understanding to the notion of mental models; 3.4 discusses the analysis of text passages in Discourse Representation Theory, and introduces the rules and structures of the theory.

The pragmatic background for discourse interpretation

Participants in discourse have in common several kinds of knowledge. They may share a general background and specific knowledge about the particular language activity of a given discourse.

The Discourse Modes introduce situation entities into the universe of discourse, and have different principles of text progression. The information is conveyed by forms of aspect and temporal location. The two are complementary. Temporal location locates a situation in time, while aspect specifies the internal temporal structure of the situation. This chapter is devoted to aspect; temporal location is discussed in Chapter 5.

The understanding of aspect is the key to the analysis of Discourse Modes, because the situation entities are aspectual in nature. The Discourse Modes characteristically introduce different types of situation entities. Passages of the Narrative and Report modes primarily involve Events and States. The Description mode primarily concerns States and ongoing Events. The Information mode primarily has General Statives; the Argument mode primarily has Abstract Entities, Facts and Propositions, and General Statives. After a brief introduction to aspectual systems, this chapter discusses the concepts of situation entities and their linguistic correlates, and how aspectual information is encoded in Discourse Representation Theory.

Section 4.1 introduces the two components of aspectual systems and discusses Event and State situation types, including shifts of one type to another; 4.2 discusses General Statives; 4.3 discusses Abstract Entities; 4.4 discusses the linguistic correlates for the three major classes of situation type; 4.5 covers aspectual information in Discourse Representation Structures.

Aspectual categories

The two components of aspectual systems

Aspect is a sub-system of language that conveys information about the internal temporal structure of situations.

Before turning to the acquisition of particular linguistic devices in different domains, we first examine in this chapter developmental studies concerning general aspects of children's discourse organisation. Some of these studies invoke cognitive macrostructures to account for children's representations of event sequences (Section 4.1), others appeal to different types of social and/or cognitive skills to account for their early or later discourse abilities (Section 4.2). To summarise the discussion of this research (Section 4.3), neither set of studies has sufficiently examined children's uses of particular linguistic devices, relying on revealing but partially anecdotal evidence concerning these uses. One of the main arguments put forth in this chapter is that these devices contribute to two aspects of discourse organisation, coherence and cohesion, which have been implicitly confounded in many studies and which must be related for an adequate account of discourse development.

Macrostructures

A first type of framework has provided a general account of verbal and non-verbal knowledge about event sequences in the form of macrostructures. This research is inspired by a number of traditions across various disciplines such as psychology, linguistics, anthropology, or Artificial Intelligence (e.g. Bartlett 1964; Labov and Waletsky 1967; Propp 1979; Shank and Abelson 1977). For example, in cognitive psychology Bartlett's pioneering work on story recall served his proposal that memory is a reconstructive process that is dependent on the use of cognitive schemata or knowledge structures actively integrating incoming information.

Two questions are central to the psycholinguistic study of first language acquisition. What structural and functional factors determine the acquisition process? What are universal and language-specific aspects of this process? This introductory chapter first presents the general theoretical thrust adopted in this book to address these two questions (Section 1.1). Particular attention is placed on the distinction between the forms and functions of language, the need to relate the sentence and discourse levels of linguistic organisation, and the importance of cross-linguistic comparisons for the study of language acquisition. I then indicate more specific developmental questions that arise in three domains of language to be examined thoroughly in this book: reference to entities, the expression of motion and location, and temporal organisation in discourse (Section 1.2). Finally, this chapter closes with an overview of the contents to be found in subsequent chapters (Section 1.3).

Acquiring language

First language acquisition is a complex process involving two facets: all children acquire the type of semiotic system that is characteristic of our species (human language), while acquiring the particular language that surrounds them (their native language). Providing an adequate account of both facets is perhaps the most difficult puzzle to be solved by theories of language acquisition. The central developmental argument put forth in this book is twofold and can be summarised as follows. First, regardless of their particular native language, children's main task is to relate the forms and functions of language (Section 1.1.1).

This chapter highlights some universal vs. language-specific properties of linguistic systems which bear on the developmental issues to be raised in subsequent chapters. I first briefly describe some of the general typological dimensions along which languages vary (Section 3.1). I then present in more detail invariant and variable properties of linguistic organisation at the sentence and discourse levels in each of the three domains to be examined subsequently from a developmental point of view. First, denoting entities (Section 3.2) requires the marking of universal distinctions concerning the discourse status of information as a function of mutual knowledge. However, languages rely to different extents on two types of markings: local markings affecting the nominal system and global markings affecting the entire clause. Similarly, space in language (Section 3.3) involves universal distinctions in the representation of motion and location, as well as general principles governing spatial anchoring anchoring in discourse. However, languages vary a great deal in the particular systems of spatial devices they provide and in how they distribute spatially relevant information in the clause. Finally, systems of temporal-aspectual markings (Section 3.4) universally allow speakers to represent various types of situations from different perspectives, to locate these situations temporally, and to ground information as a function of discourse focus. However, they vary in important ways, such as the extent to which markings are grammaticalised, rich, symmetric, and transparent.

We now examine the questions raised by the preceding chapters in the light of the available research pertaining to children's acquisition of a variety of linguistic devices across several domains. The present chapter focuses on children's uses of referring expressions and of clause structure when denoting entities in discourse. Studies concerning referring expressions (Section 5.1) have focused on different syntactic, semantic, and pragmatic aspects of these devices, reaching strikingly divergent conclusions about the rhythm, course, and determinants of acquisition in this domain of child language. A review of this evidence suggests that, despite some early uses of different types of referring expressions, the discourse-internal cohesive functions of these devices constitute a late development. In addition, some scattered developmental evidence shows variable uses of clause structure across languages (Section 5.2), but little is still known about the functions of such linguistic devices in children's discourse development. In summary (Section 5.3), the proposed synthesis of this literature argues for the need to bring together different lines of research in order to determine the impact of sentence and discourse factors on development within a cross-linguistic perspective.

Referring expressions

A great number of studies have examined children's comprehension and production of referring expressions (nominal determiners, overt pronouns, null elements). A review of this literature shows a striking divergence in the conclusions reached by different authors with respect to the rhythm of development.

This chapter synthesises the results of the study just presented (Chapters 8 to 10) and compares them with those of previous studies discussed in the developmental literature (Chapters 4 to 6). After a summary of the main findings within each domain (Section 11.1), four general conclusions are drawn across all three domains (Section 11.2). First, the findings show that the development of discourse-internal functions is late and gradual. Second, acquisition is determined by two types of interrelated factors: sentence-internal syntactico-semantic factors and functional determinants governing the regulation of information flow in discourse. Third, although these general aspects of the developmental process can be observed in all languages, others are language-specific, showing the impact of the particular systemic organisation of each language. Finally, a comparison of the findings across domains highlights particular properties of each domain, as well as interrelations among them suggesting that conjoined developments occur in all aspects of discourse cohesion. More general implications are then drawn in the context of available models of language acquisition (Section 11.3). It is argued that language acquisition is a gradual process, which requires that the child relate two levels of linguistic organisation, the sentence and discourse, accounting for invariant as well as variable patterns. This type of multifunctionality is universal and central in explaining how language becomes its own context, despite cross-linguistic variations due to the fact that children must confront the particular ways in which their native language maps sentence and discourse functions onto forms.

In this chapter we examine how the animate characters were denoted in the narratives, while analyses concerning inanimate referents are presented subsequently in relation to spatial anchoring (Chapter 9). Recall that the compared languages vary with respect to several properties of the devices that are necessary for the marking of information status in discourse: their obligatory vs. optional nature, their global vs. local nature, and the parametric properties of the languages in relation to morphological complexity (see Chapter 3). In English, French, or German local markings (determiners) are obligatory to mark newness and global markings (clause structure) are optional for this purpose, whereas the reverse is true in Chinese. Furthermore, subjects are obligatory in English and French [–pro-drop languages], while Chinese and German allow null elements to various degrees [zero-topic languages], despite other differences between them in this respect (topic vs. subject orientation). A first set of results (Section 8.1) concerns how the animate referents are introduced in the narratives, with particular attention to speakers' reliance on local vs. global markings of newness and on the relations among these markings. A second set of results (Section 8.2) concerns the different means used for subsequent reference maintenance, with particular attention to the different syntactic, semantic, and pragmatic factors determining the forms and positions of subsequent mentions in different clause structures.