Language, Memory, and Thought

John R. Anderson

• Éditeur: Psychology Press
• Année: 2013
• ISBN:
  9781134918898
• Édition: 1
• Sujet: Psychology
• Sous-sujet: History & Theory in Psychology

Published in 1976, Language, Memory, and thought is a valuable contribution to the field of Cognitive Psychology. This book presents a theory about human cognitive functioning, a set of experiments testing that theory, and a review of some of the literature relevant to the theory. The theory is embodied in a computer simulation model called ACT.

1

Preliminary Considerations

I believe that parsimony is still inappropriate at this stage, valuable as it may be in later phases of every science. There is room in the anatomy and genetics of the brain for much more mechanism than anyone today is prepared to propose, and we should concentrate for a while more on sufficiency and efficiency rather than on necessity. MARVIN MINSKY

1.1 INTRODUCTION TO THIS BOOK

This century has seen an amazing intellectual blossoming. Psychology at the beginning of this century was largely a parlor room avocation. Today it is approaching a mature science. There have been momentous developments in many of the disciplines adjacent to psychology–mathematics, philosophy, logic, linguistics, and computer science. We in psychology are still trying to digest the implications of these developments for the practice of our own science.

The computer revolution of the past 25 years has been particularly important to me. When researchers in artificial intelligence began to apply themselves seriously to the task of reproducing human intelligence on a machine, it became clear in a very concrete way how remarkable is human cognitive functioning. Despite the impressive advances in artificial intelligence, current computer programs do not approach the power and generality of human intelligence. Today we really have little idea about how humans perceive, remember, process language, or reason. This respect for the power of human cognitive functioning is reinforced by the more abstract analyses coming out of philosophy, logic, and linguistics.

All this poses an intellectual challenge. How is the power of human cognitive functioning possible? There have been two main approaches to this question. Those in artificial intelligence have tried to program machines to be as intelligent or more intelligent than man. There is a range of justifications for the artificial intelligence approach (see Newell, 1970, for a review). One justification asserts mat the abstract logical character of a program which matches feats of human intelligence must be very similar to that of the human mind. This is asserted despite the fact a computer and a human brain are very different. It is argued that feats of human intelligence are so difficult that there can be only one way of accomplishing them on any physical device. Another justification is more modest. It simply asserts that it is interesting to know how a machine could be intelligent. This knowledge would improve our perspective on the nature of human intelligence, and also has obvious practical applications.

The approach of cognitive psychology, on the other hand, has been to build models of human information processing which predict features of human behavior. In some respects this is easier and in others it is harder than the artificial intelligence approach. It is easier because there is no requirement that the model be translatable into an intelligent computer program. The model certainly must be specified well enough so that it can be simulated on a computer, but that simulation can be so inordinately slow and inefficient as to disqualify it as an intelligent program. The cognitive psychology approach is more difficult because its models must predict the specifics of human behavior and not just reproduce the global intelligent aspects of that behavior.

Purpose of the Book

In this book I present a model in the cognitive psychology tradition. Various aspects of the model have received computer implementation, but these implementations are in no way artificial intelligence programs. This model is an attempt to provide an integrated account of certain linguistic activities–acquisition, comprehension, and generation of language, reasoning with linguistic material, and memory for linguistic material. Perhaps this model could be extended to deal with other areas of human cognitive processing–in particular, perception. At this time, I do not know whether such extentions will be possible.

This book is greatly concerned with the construct of memory, which has undergone drastic change in the last decade in Cognitive Psychology. It used to refer to the facility for retaining information over time. Now it more and more refers to the complex structure that organizes all our knowledge. In its concern with memory the book is an extension of the HAM model that I developed with Gordon Bower (Anderson & Bower, 1973). However, the model developed here is remarkably different from the HAM model. Research subsequent to formulation of the HAM model indicated there were some mistakes in that theory, which have been corrected in the new model. Other theoretical changes have been introduced, not because HAM made incorrect predictions, but because it seemed that a more parsimonious theory could be achieved if changes were made.

The theory has also changed because it has been extended to account for more than just memory. I have been concerned extensively with problems of inference making and language comprehension. In recognition of all these change the theory has been given a new name–ACT. We claimed in the HAM book that we really were not committed to that theory. We had developed that theory because it

provides a concrete realization of a certain theoretical position. It provides something definite for research workers to discuss, examine, criticize, and experiment upon. It is hoped that some resolution will be eventually achieved with respect to the fundamental theoretical questions. We hope that others will be encouraged to provide and motivate other explicit models of fundamentally different theoretical positions [p. 7].

In completing the HAM book we had the feeling that we had more or less defined the HAM theory once and for all and that the major task ahead of us was to test its empirical consequences. I feel less certain that ACT has achieved its final shape as a theory. There remains much exploration to be done as to the potential of the theory and variations on it, largely because ACT is a theory of much broader generality than HAM and consequently has more potential to explore. In this book I have caught the ACT theory in a stage of development and presented that. As this book goes to press new developments are occurring. However, it is unlikely that these new developments will be radical departures from the concepts set forth here. Therefore, this book will provide a comprehensive statement of the ACT theory. It also provides evidence for why this theory constitutes a real advance. Later publications will use this book as a reference for expositions of important ACT concepts.

I would like to emphasize that I do not think that ACT, at any stage in its evolution, will constitute the final shape of a theory of human cognitive functioning, any more than did HAM. In my perception of science it is critical to develop models like ACT. One needs to have explicit models that seem compatible with a wide range of data. Once formulated, such a model can be subject to intensive study and its range of performance understood. Having obtained a firm grasp on the character of the theory, one can then begin to explore its weak points and the potential of variations on the model. One can also explore the differences between this theory and others. All this rumination about the theory serves to increase one's sophistication about the significant issues. It is only after such rumination that one can confidently formulate a different and better model.

Contents of This Book

Despite the fact that a somewhat different theory is being offered in this book, it has a similar structure to the old HAM book. In addition to this chapter there are two general survey chapters in which issues relevant to the theory are considered. In the other three sections of this chapter I discuss some general questions about what the limitations of a cognitive theory are, and what its goals and strategies should be, with special attention to the role of computer simulation in a cognitive theory. The ACT model involves the integration of a memory network with a production system. The memory network is intended to embody one's prepositional knowledge about the world. The production system is intended to embody one's procedural knowledge about how to perform various cognitive tasks. As a form of background, in Chapter 2, I review other prepositional theories of knowledge. In Chapter 3, I consider production system models of procedural knowledge and how they are an advance over the older, but remarkably similar, stimulus-response theories.

The next four chapters present the main body of the ACT theory. In Chapter 4 I present an overview of the model, in Chapter 5 a detailed development of the propositional network, and in Chapter 6 a detailed development of ACT's production system. In Chapter 7 I present a formal semantics for the ACT representational system and an analysis of the expressive power of that system.

The next five chapters attempt to apply the ACT model to various content areas. In Chapter 8 I will consider the ACT model for fact retrieval; in Chapter 9, inference making; in Chapter 10, the learning and retention of prose material; in Chapter 11, language comprehension; and in Chapter 12, induction of production systems. Parts of these five chapters would be incomprehensible without an understanding of the theory set forth in Chapters 4–6. On the other hand, Chapters 8–12 can be read in almost any order as they have relatively little dependence on one another. A final short chapter provides an attempt to evaluate the ACT model. The reader might want to skip ahead and read that chapter to get a sense of how the book “will end.”

Various sections of this book differ greatly in the formal and technical prerequisites they assume of the reader. I have tried to make the general substance of my arguments accessible to all. However, there were occasions when there was no choice but to plunge into formal analysis which I fear will be obscure to many. These sections are starred in the book as a warning to the reader, and are difficult for anyone to comprehend without the appropriate technical background.

1.2 LIMITATIONS OF A COGNITIVE THEORY

The goal of a cognitive theory might be stated as the understanding of the nature of human intelligence. One way to achieve such an understanding would be to identify, at some level of abstraction, the structures and processes inside the human head. Unique identification of mental structures and processes was once my goal and it seems that it is also the goal of other cognitive psychologists. However, I have since come to realize that unique identification is not possible. There undoubtedly exists a very diverse set of ...