Introduction

Phonetics deals with the study of the sounds used in the languages of the world. Its main goals are to catalog the existing sounds, to examine their acoustic properties, and to determine how they are produced and perceived by human beings. Thus, by its very nature, phonetics relates to a wide variety of disciplines: not only general linguistics, but also physics, biology (physiology and neurology), and psychology.

Our own perspective in this chapter will be relatively narrowly circumscribed. First, we shall confine ourselves to what is traditionally known as articulatory phonetics (‘la phonétique articulatoire’), that is, the study of sound production at the level of the vocal cords and the cavities above the vocal cords (the supraglottal cavities). We shall not consider, for instance, acoustic phonetics or sound perception. Second, rather than aiming for a comprehensive review of the articulation of all the sounds, or even all the sound types, that have been documented in natural languages, we shall for the most part focus only on the knowledge of articulatory phonetics that is required in order to tackle the sound systems of French and English and their contrastive study. In sum, this chapter is simply devoted to an overview of the operation of the speech organs and to a presentation of the main articulatory parameters which permit the description, classification, and comparison of French and English sounds.

This chapter is meant as an overview of the fundamental concepts of conversational analysis which will be put to work in Chapter 3. It is by no means a complete guide to this approach; I have included only those notions which will be applied in the analyses in Chapter 3. For a more complete introduction to CA, see Levinson 1983 or Atkinson and Drew 1979. Readers already familiar with CA can skip to Chapter 3.

Notation conventions

Before we examine some of the substantive findings of conversational analysis, I would like to present and discuss some of the notational conventions used in CA-style transcripts. Many of these conventions may be unfamiliar to linguists, but they are all in fact fairly straightforward. (The sources of the transcripts reproduced in this book are given in abbreviated form – AD: 14, SN-4:30, etc. – at the end of the extracts. Further details of these sources are given in section 3.2.)

All talk is transcribed in a pseudo-phonetic system, using the basic orthographic symbols of written English; that is, if the speaker pronounces a word in a way that is not the only possible pronunciation for that word, then special care is taken to transcribe that particular pronunciation (but without using a special alphabet). This practice makes the transcripts somewhat difficult to read (especially for the non-native speaker), but since it brings out useful information I have not normalized the transcripts (except for crucial pronouns, which have been normalized for ease of reading and exposition).

In addition to the issues mentioned explicitly in Chapter 1, the previous chapters have raised several larger theoretical issues which I have not yet addressed. I would like to touch on two of these fundamental issues here, if only superficially, to open these concerns up for discussion. The two issues I would like to address explicitly here are the localness of linguistic patterning and the nature of discourse structure.

Localness

In Chapter 6, I raised the issue of genre-specific conventions of anaphoric patterning. As we have seen, there is no single rule for anaphora that can be specified for all of English (even the limited subset of third-person singular human examined here); instead, we have a variety of specific patterns which obviously share a number of general characteristics, but which nevertheless differ enough to require separate formulation (Fox, forthcoming, b). This is one level at which patterns are local (specific to a particular genre) rather than global (specifiable for all of English). There are other levels at which the idea of localness seems fruitful; it is to these levels that I would like to turn now.

The structure of expository prose has captured the interest of a wide range of disciplines, including rhetoric (Dillon 1981; Young et al. 1970; D'Angelo 1975; Winterowd 1975), cognitive psychology (van Dijk and Kintsch 1983; Meyer and Rice 1982; De Beaugrande 1980; Bransford 1979; Sanford and Garrod 1981; Britton and Black 1985), artificial intelligence (Alvarado 1986; Schank 1982; Brown 1985; McKeown 1982) and linguistics (Hinds 1979; Grimes 1975; Kamp 1981). While several of these studies have designed detailed and insightful notations for representing the hierarchical structure of expository prose (for example, the macrostructures of van Dijk and Kintsch 1983; the argument units of Alvarado 1986; the rhetorical schemas of McKeown 1982; the conceptual graph structures of Graesser and Goodman 1985; the structures of Meyer 1985), none of them provides all that is needed for an in-depth exploration of anaphora. For this study it was necessary to have a notation with the following characteristics:

Introduction

In the preceding chapters I have explored the distribution of pronouns and full NPs in written (expository) and conversational (non-narrative) English. In this chapter I present a comparison of the anaphoric patterns found in the conversational and written texts, including same- and different-gender environments. Before beginning the presentation, however, a note about past research on the differences between spoken and written language is in order.

Theories of the differences between spoken and written language

A fair amount of attention has been directed recently to the differences in syntactic structure exhibited by the two modalities (see, for example, Keenan and Bennett 1977; Ochs 1979b; Chafe 1982; Biber 1983; Akinnaso 1982; O'Donnell 1974; Tannen 1982; Rubin 1980). Claims have been made that written texts tend to be characterized by greater complexity of syntactic structure (greater use of nominalizations and complex verb structures, for example), more frequent use of subordination, and a predominance of subject–predicate structure rather than topic–comment (or reference–proposition); that is, in general there seems to be a greater degree of what Chafe (1982) calls syntactic integration in written texts than in spoken texts.

It should be noted at this point that the characteristics of spoken and written language that have been studied are surface phenomena that can be counted, and from which contrastive frequencies can be given.

Introduction

In this chapter I examine the distribution of pronouns and full noun phrases in some expository written English texts. The structural analysis technique used will be rhetorical structure analysis.

The anaphoric patterns established in this chapter are presented in the two modes discussed in Chapter 3. These two modes, it will be recalled, are the context-determines-use mode and the use-determines-context mode (see also Chapter 3 for a discussion of these modes). I argued in Chapter 3 that both of these modes are always present for conversationally interacting parties, although in any particular instance one may be more strongly felt than the other. The argument for this view runs as follows:

1. Anaphoric form X is the unmarked form for a context like the one the participant is in now.

2. By using anaphoric form X, then, the participant displays an understanding that the context is of that sort.

3. If the participant displays an understanding that the context is of that sort, then the other parties may change their understandings about the nature of the context to be in accord with the understanding displayed (cf. McHoul 1982).

I would like to propose now that this same cycle of factors lies behind anaphora in writing as well. Even though the parties (writer and reader) are not co-present at either the time of writing or the time of reading and hence cannot directly participate in such a fluid display of understandings, each feels the other's presence in a way that strongly influences their behavior towards the text.

Major themes of the study

The last ten years have seen a tremendous upsurge in work on discourse production and comprehension, correlated with a growing concern in a variety of disciplines with language as it is used in context. Because of its fundamental place in the understanding of memory, discourse structure and semantic interpretation, anaphora has been the focus of much of this research (e.g. Grosz 1977; Reichman 1981; Sidner 1983; Tyler and Marslen-Wilson 1982; Webber 1983; Givón 1983; Halliday and Hasan 1976; Bosch 1983; Linde 1979; Reinhart 1983). Central to this work has been the belief that there is a strong relationship between the flow of information in a text, the structure of the text, and use of anaphora. A recurrent, and intuitively appealing, finding of this work is that referents which are “in focus” or “in the hearer's consciousness” can be pronominalized, where focus or consciousness are operationalized in terms of the discourse structure (see in particular Grosz 1977 and Reichman 1981).

The present study holds to this interpretation of the relationship between discourse structure and anaphora. One of the themes that runs through this study is that any treatment of anaphora must seek its understanding in the hierarchical structure of the text-type being used as a source of data. Texts may be produced and heard/read in a linear fashion, but they are designed and understood hierarchically, and this fact has dramatic consequences for the linguistic coding employed.

Introduction

In this chapter, I discuss the distribution of pronouns and full noun phrases in non-story conversational texts.

In presenting the distributions, I operate in two modes of description, one of which can be thought of as the “context-determines-use” mode, the other the “use-accomplishes-context” mode. In the context-determines-use mode, it is assumed that the hierarchic structure of the talk determines to some large extent the anaphoric form which the speaker is to use. The type of pattern offered in this mode says that in context X the speaker will use anaphoric form Y. In the use-accomplishes-context mode, on the other hand, it is assumed that it is by virtue of using a particular anaphoric form that the structure is created.

I have assumed in this study that both modes of operation are always present for the participants. That is, for the most part knowledge about how anaphora is usually handled in certain contexts leads a participant to pick the anaphoric form that is “unmarked” for the context, and by picking this form the participant displays his/her understanding of what type of context is currently under development: this display of understanding, in turn, can create for the other parties present the same understanding (when by themselves they might have constructed some other sort of understanding of the current structure). There is thus a continuous interaction, even in the simplest cases, of the following three steps of reasoning:

1. Anaphoric form X is the “unmarked” form for a context like the one the participant is in now.

2. […]

It is almost always a mistake to put a single author's name on a piece of research. This is especially true for the present study, since so many people have had more than a passing influence on the ideas and methods used. Most of the real work of this project was done while I was studying with Bill Mann, Manny Schegloff, and Sandy Thompson, who, in spite of their different approaches, all encouraged me, and helped create what is here now. Without their constant support, this study would never have been possible.

Sandy Chung, Jack Du Bois, Talmy Givón, Paul Hopper, and Paul Schachter have all contributed, sometimes unknowingly, to my general philosophy of language and to my appreciation of other languages. Their influence pervades the text in a somewhat less visible way, but it is still there.

My latest thinking about discourse and grammar has been shaped by colleagues at the University of Colorado, most notably Walter Kintsch and Paul Smolensky. They have increased my appreciation for what cognitive psychology and computer science have to offer linguistics, and have been instrumental in making Colorado an exciting place to be.

Many friends, colleagues, and family members have provided support for the efforts realized here. To my parents I owe the greatest debt, for always being understanding about everything. Loralee MacPike provided support and encouragement for many years, and still does, at a distance.

Introduction

In this chapter, I discuss the problem of syntactic, or structural, disambiguation, which was first introduced in section 1.1.3. I will provide the background for the discussion in chapter 7 of the Semantic Enquiry Desk, a structural disambiguator that works with Absity and Polaroid Words.

Types of structural ambiguity

Structural disambiguation is necessary whenever a sentence has more than one possible parse. There are many classes of structurally ambiguous sentence; below I show some of the more common, but do not attempt to produce an exhaustive list. Included in the list are some local ambiguities (see section 1.1.3) that people can be garden-pathed by.

I will use two methods of demonstrating structural ambiguity. In some cases, I will give one sentence and show its several parses; in others, I will give two sentences such that each has a different preferred parse but each could clearly also have the structure of the other. For simplicity, when I show a parse, I will often show it only for the part of the sentence that contains the ambiguity; pseudoprepositions (see section 3.4) are not usually inserted, except where necessary to make a point.

Attachment problems

The first class of structural ambiguity is that of ATTACHMENT AMBIGUITY: there being more than one node to which a particular syntactic constituent may legally be attached. Attachment problems are mostly problems of MODIFIER PLACEMENT. The most common example is that of a prepositional phrase that may either modify a verb (i.e., be a case-slot filler) or an immediately preceding noun phrase. For example:

(6-1) Ross wanted to phone the man with the limp.

(6-2) Ross wanted to wash the dog with Hoary MarmotTM brand pet shampoo.

The research I have described in this book necessarily leaves unanswered many questions, big and small. Many sections, especially 3.8, 5.3.6, 5.4, and 7.4 have discussed things left undone by Absity, Polaroid Words, and the Semantic Enquiry Desk. In this chapter, I list a number of other open questions, partially baked ideas, and wild speculations, sometimes with my thoughts on how they might be answered or developed. Some could be dissertation topics; others may be good subjects for a term paper or course project. Several are psycholinguistic experiments. At the start of each question (if appropriate) I give in brackets the section or sections of this book in which the matter is discussed.

The representation of knowledge

Exercise 1.1 [1.1.2,1.3.1,5.2,5.6.3] Could a non-discrete representation of knowledge be developed for AI? Such a representation would be able to handle close similarities and differences, such as the head of a pin compared with the head of a hammer. Consider the possibility of pseudo-continuous representations that are to discrete representations as floating-point numbers are to integers. Candidates to consider include some kind of network of neuron-like nodes (cf. Feldman and Ballard 1982; Feldman 1985), a value-passing machine such as Fahlman, Hinton and Sejnowski's (1983) Thistle system, and a simulated-annealing or Boltzmann network (Kirkpatrick, Gelatt and Vecchi 1983; Fahlman, Hinton and Sejnowski 1983; Smolensky 1983) (see also exercise 4.8).