Psycholinguistics (PLE: Psycholinguistics)

Central Topics

Alan Garnham

• Publisher: Psychology Press
• Year: 2013
• ISBN:
  9781135006785
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

Originally published in 1985, this title was an important new teaching text at the time. Alan Garnham focuses on current theories about the central cognitive aspects of language understanding, and attempts to reflect the emergence of cognitive science, an inter-disciplinary approach to the study of language and other cognitive processes. As well as describing psychological studies, the text includes ideas from linguistics, artificial intelligence, the philosophy of language and formal logic.

Some introductory remarks on the study of language understanding precede a discussion of word recognition and the computation of the syntactic structure of sentences. The central part of the book is concerned with questions about meaning, the mental representation of word meanings, and text comprehension. The final two chapters address questions of how the parts of the language processing system operate together, and how language production is related to comprehension. Rather than attempting an exhaustive discussion of empirical research on his chosen topics, the author gives the reader the flavour of linguistic arguments. In particular, Psycholinguistics attempts to indicate the problems and also the possibilities of relating experimental data to theories of language processing.

Psycholinguistics will still be useful reading on courses in psycholinguistics, language and thought, and cognitive psychology.

1

Introduction – how to study language understanding

Overview

Psycholinguistics is the study of the mental mechanisms that make it possible for people to use language. It is a scientific discipline whose goal is a coherent theory of the way in which language is produced and understood. This book describes the most important discoveries that have been made in the attempt to construct such a theory. The present chapter begins by explaining certain biases in psycholinguistic research – why, for example, the emphasis is on language understanding rather than language production. It then outlines the overall goal of psycholinguistics – the identification of what information is conveyed by language, and how. This goal is broken down into a number of subgoals, and the subprocessors that contribute to the interpretation of discourse and text are described. The genesis and evaluation of psycholinguistic theories are then considered with reference to some ideas from the philosophy of science, and the use of statistical methods for analysing psycholinguistic data is briefly discussed. Some problems with the experimental approach to psycholinguistic problems are outlined, and the computer modelling approach of artificial intelligence (AI) is introduced as a possible alternative. Its strengths and weaknesses are discussed, and a synthesis of the two approaches – cognitive science – is suggested.

In the remaining chapters of the book theoretical and empirical issues in psycholinguistics are discussed in detail. The major subprocessors of the language understanding system are introduced starting with low-level processors, and moving on to higher-level ones. When all the processors have been described, the relation between them is considered, as is the relation between understanding and producing language. Theoretical material from related disciplines is discussed as and when it is appropriate.

The topics covered in the individual chapters are as follows. Chapter 2 is about linguistic theory and its relevance to psycholinguistics. The first half of the chapter shows that a simple description of how language is used is simple-minded, and considers the question of how a description of language might contribute to an account of how language is understood and produced. The second half gives an overview of the branch of linguistics that has had most influence on psycholinguistic research – syntax. Chapters 3 and 4 describe two subprocessors of the language understanding system – the word recognizer and the parser. Chapter 5 provides a general introduction to the concept of meaning, in its broadest sense, drawing on work from a number of disciplines – most importantly philosophy – which have contributed to theories of language processing. Chapter 6 discusses the mental representation of word meanings, and chapter 7 describes the way the language understanding system determines the meaning of discourses and texts. Chapter 8 addresses the question of how the subprocessors of the language understanding system operate together, and chapter 9 discusses language production and its relation to language understanding. The final chapter provides an overview of the book, and considers the likely directions that future research will take.

Biases in psycholinguistic research

The contents of this book reflect certain biases in psycholinguistic research. For purely practical reasons attention has focused on understanding language understanding rather than language production. It is easier to study understanding in a systematic way, since the experimenter can select words and sentences with great care. In production the investigator is constrained by what people happen to say, although the situation in which they say it can be controlled. For this reason comprehension has been studied more thoroughly than production. However, it is reasonable to assume that much of our linguistic knowledge is shared by the mechanisms responsible for generation and analysis, and that the study of comprehension sheds light on production. Nevertheless, production needs to be studied in its own right, for two reasons. First, there are a number of questions about production that are independent of questions about comprehension. These questions are briefly considered in chapter 8. Second, it is possible that some mechanisms are not shared between production and comprehension. For example, people might have different lexicons for speaking and listening – we tend to have many more words in our sight vocabularies than in our speaking vocabularies. However, stronger evidence would be required before this unpar-simonious view was adopted.

Another bias that recurs throughout the book, though a less marked one, is toward the study of written, as opposed to spoken, language. Again the reasons are largely practical – the problems, both experimental and theoretical, are, or at least can be made, more tractable in the case of written language, so more research has been carried out. This bias is stronger in some areas than others, perhaps being most apparent in the study of word recognition.

Psycholinguistic theory

The primary purpose of language is communication, and hence a psycholinguistic theory must give an account of what information language conveys. Although there may be no single answer to the question of what language expresses, in the most important cases communication is about, in a rather general sense, how the world is, was, will be or might be. Newspapers report recent events. History books describe how things used to be. Textbooks survey current knowledge in the sciences and humanities. This information has been successfully communicated if the reader correctly updates his or her mental representation of the world. However, not all uses of language are about the real world, and not all conversations and texts are descriptive.

Works of fiction describe fictional worlds rather than the real world. Representations of such worlds can be created by analogy with those of the real world. They contain people, things and events, and locations at which events can occur. A mental representation of a fictional world is similar in form to one of the real world. The difference lies in whether the representation corresponds to how things really are.

In the case of spoken language in particular, description is only one use of language among many. Much communication is in the form of commands, questions, requests, promises and so on. In an extended sense these uses of language are also ‘about’ the world, since in order to understand them it is necessary to have an internal representation of the world, and of how it would be if, for example, the command was obeyed, or the request complied with. In each of these cases, therefore, understanding language requires the construction of, or reference to, internal mental representations of how things are or might be. It also requires that the point of the utterance should be recognized, and an appropriate response formulated. Examples of appropriate responses are that a description should be added to an internal representation of the world, and that a question to which the answer is either ‘yes’ or ‘no’ should initiate a search through such a representation to see if the fact that is questioned is true. There are only a very few uses of language – some exclamations, for example – that do not require for their comprehension an internal representation of the real world or a fictional world.

These considerations indicate the outline of a theory of language understanding. The language understanding system extracts the content of incoming sentences, and constructs a representation of the situation to which they refer. It further determines the point of what is being said so that the intended message can be computed, and an appropriate response formulated. The content extracted from a sentence or a set of sentences will be referred to as the mental model of the situation that the sentences are about. This term will be used rather than semantic representation, which was favoured in earlier writings on psycholinguistics and AI. There are two main reasons for rejecting the earlier term. First, to say that something is a semantic representation suggests that it is closely related to a linguistic structure, but a mental model is structurally similar to a part of the world, rather than to any linguistic representation. Second, in line with recent practice in linguistics, the term semantics will be used in this book in a restricted sense to be explained in chapter 5. A mental model contains much more than semantic information in this restricted sense.

Subprocessors of the language understanding system

As in any scientific discipline, it is useful to divide the overall problem that psycholinguists are trying to solve into a number of subproblems. From the point of view of language understanding, psycholinguists would like to know: first, how words are recognized, second, how the structure of a sentence is determined, third, how its meaning is computed, fourth, how its meaning is integrated with what has gone before, and fifth, how the intended message is worked out. When tackling these subproblems it is, of course, important that the overall goal of psycholinguistics should be kept in mind.

Low-level perceptual processing

This book is concerned only with cognitive processes, and not the ‘low-level’ auditory and visual analysis that precedes them. Nevertheless, all language understanding starts with perception, and it will be assumed that the results of perceptual processing are made available to the language understanding system. In the case of written language, low-level processes identify lines and curves, and perhaps the shapes of words. With spoken language there is an additional complication since it has been suggested (for example, Liberman, 1970) that there is a special linguistic mode of perception.

Word recognition

The understanding system receives an input of lines, curves and spaces, or an acoustic waveform, and it must use this input to make contact with a store of information about words – the mental equivalent of a dictionary – in order to decide what words have been presented. The processing of auditory inputs is especially difficult. They almost always occur against a background of noise, which must not be confused with the speech; there are no simple cues to segmentation in the waveform (that is to deciding which parts of the waveform belong to the same word, and which to different words), and no simple cues about how fast words are arriving. Analysing untidy cursive script is almost as difficult, but recognizing typewritten or printed words is comparatively easy. The cues to segmentation – spaces and new lines without hyphenation – are relatively simple to detect, and every occurrence of the same word in a single fount is effectively identical. The question of how we are able to cope with many different founts is another, more difficult, one.

For a number of years word recognition was studied primarily by cognitive psychologists whose main concern was not with language understanding. However, identifying words is an essential stage in comprehension, and psycholinguists (e.g. Forster, 1979; Marslen-Wilson and Tyler, 1980) are becoming increasingly interested in both word recognition itself, and its relation to other parts of the language understanding system.

Parsing

The words in a sentence are not simply strung together, they form natural groups – phrases and clauses. Working out the appropriate grouping is essential for understanding a sentence. For example, in the sentence:

The little old man walked in the park.

the words the little old man must be grouped together, because they refer to a single thing – in this case a person. Other similar phrases have slightly different interpretations. For example, the little old men refers to a group of people. The words in the park form a group describing a location, and, at a higher level, walked in the park describes an action. Other potential groupings of words, for example man walked in, do not form coherent units. The correct interpretation of the sentence therefore depends on the words being grouped together appropriately. This job is performed by the parser. A parser has access to information about how words can be grouped in a particular language, say, English, and it makes use of this knowledge to determine the structure of particular sentences that it encounters. The information that a parser has access to is usually formulated as a set of rules of the kind that will be described in chapter 2.

Semantic interpretation

Although grouping does not itself amount to interpretation, words are grouped together so that sentences can be correctly interpreted. The meaning of a sentence is determined by a further set of rules specifying the kinds of things that particular groups of words refer to, and what relations hold among those things. The language understanding system uses two kinds of information in interpreting sentences. First, it uses information about the meanings of particular words – the kind of information that is to be found in a dictionary, though this information may be organized very differently in the mind of a language user from the way that it is set out in, say, the Oxford English Dictionary. The other kind of information tells it how word meanings are combined to produce the meanings of phrases and eventually those of sentences. In fact it is wrong to stop at sentence meaning. Units larger than sentences – discourses and texts – have meaning that goes beyond that of their component sentences.

Model construction

In this book the term semantic interpretation is used in a narrow sense. The semantic interpretation of a sentence, for example, specifies the range of situations that that sentence could describe, but does not determine which one it actually describes in a particular use. Neither does it specify which of those situations are more likely than others. For example the sentence:

The man met the woman.

could be used to describe the meeting of any woman and any man. However, on a particular occasion when the sentence is used it will be about one particular meeting, and the sentence has not been fully understood until that meeting has been identified. Some such meetings are much more likely than others, primarily because of where people live and how mobile they are. This fact is irrelevant to semantic interpretation, though it may well affect how readily the sentence is understood in a particular context. From the semantic interpretation of its input, the understanding system must work out what particular situation, in the real or an imaginary world, a particular discourse or text is, or is most probably, about. That is to say it must construct some internal representation of that situation – a mental model of it.

Pragmatic interpretation

Finally the understanding system must decide what to do with the model it has constructed. An utterance might be, for example, a description that should be added to an overall model of the world, a question about how things are in the world, or a request to change them. Psycholinguists aim only to describe how the point of an utterance, and hence the intended response to it, is determined, and not whether a command, say, is actually obeyed. Pragmatic interpretation is made more complicated by the fact that what people mean – the message that they are trying to convey – is often different from what they actually say. To take an extreme case, they may say the opposite of what they mean to create an ironical effect. The language understanding system copes easily with these complications, but constructing an adequate theory of how it does so is difficult.

The major components of the language understanding system have now been mentioned. In the later chapters of this book each of these processors, and the facts that are known about how they work, will be discussed in more detail. However, a theory of language understanding must do more than simply describe each component of the system. It must also describe how the components act together to produce understanding. Although the subprocessors were listed in an order from, roughly speaking, early low-level processors to late high-level processors, there is no implication that they operate serially, with the output of one simply being passed on as input to the next. There may be feedback from ‘later’ processors to ‘earlier’ ones, and some of them may work in parallel. The question of the overall structure of the language processing system will be discussed in chapter 8.

Where do psycholinguistic theories come from?

Now that the scope of psycholinguistic theories has been indicated, the question arises as to where such theories come from. How is a theory of language understanding developed? The answer to this question is that, in a quite straightforward sense, theories can come from anywhere – their origins do not affect their status as scientific theories. This in...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Dedication
• Table of Contents
• Acknowledgements
• Preface
• 1. Introduction – How to Study Language Understanding
• 2. The Contribution of Linguistics
• 3. Recognizing Words
• 4. Parsing – The Computation of Syntactic Structure
• 5. Introduction to the Concept of Meaning
• 6. Word Meaning
• 7. Understanding Discourse and Text
• 8. The Structure of the Language Processor
• 9. Language Production and its Relation to Comprehension
• 10. Overview and Future Directions
• References
• Name Index
• Subject Index