Handbook of Learning and Cognitive Processes (Volume 5)

Human Information Processing

William Estes

• Publisher: Psychology Press
• Year: 2014
• Print ISBN: 9781848723962
• eBook ISBN: 9781317672012
• Topic: Psychologie
• Subtopic: Kognitive Psychologie & Kognition

Originally published in 1978 Volume 5 of this Handbook reflects a single theoretical orientation, that characterized by the term human information processing in the literature at the time, but which ranges over a very broad spectrum of cognitive activities. The first two chapters give some overall picture of the background, goals, method, and limitations of the information-processing approach. The remaining chapters treat in detail some principal areas of application – visual processing, mental chronometry, representation of spatial information in memory, problem solving, and the theory of instruction.

The first three volumes of the Handbook presented an overview of the field, followed by treatments of conditioning, behavior theory, and human learning and retention. With the fourth volume, the focus of attention shifted from the domain of learning theory to that of cognitive psychology.

1

The Information-Processing Approach to Cognition: A Confluence of Metaphors and Methods

W. K. Estes

Rockefeller University

I. Behavioral versus Informational Modes of Description

Doubtless no one would disagree that the first step toward theory construction in any scientific area must be an adequate description of facts, accomplished by means of terms that do not presuppose the theory. What is not so obvious is that there may be alternative frameworks for the description of facts and lower level empirical laws, and the decision made at this choice point may have profound implications for theory.

In psychology it has seemed to many investigators, perhaps a majority, that an adequate descriptive base can be achieved by means of an accurate recording of an organism’s observed behaviors, that is, the relationships between stimulus inputs from the environment and the consequent responses elicited from or emitted by the organism. Indeed, it has been demonstrated that one can go a good way toward theory on this basis in the treatment of conditioning and animal learning. It turns out that one needs to augment descriptions of directly observed behaviors with inferred concepts such as interoceptive stimuli and covert responses, but these can be tied closely to verification procedures relating to classes of observed behaviors.

In the sphere of cognitive activity, difficulties emerge in that nearly everything that goes on of interest is unobservable. When a person is reading, the investigator can observe and record the movements of the reader’s eyes, but under no illusion that he obtains from this record an account of the cognitive activity termed reading. Similarily, in a problem situation one can observe the environmental context and the way in which it is ultimately changed by the individual solving the problem, but this account provides no information as to what the individual did in arriving at a solution. In order to proceed toward theoretical accounts of activities such as these, we evidently need factual accounts of what the individual is doing between the initiation of the observed stimulus context and the ultimate observable responses indicating that the individual has obtained information or solved a problem. To be adequate as a basis for theory, this account must be as firmly anchored in observation as descriptions of the movements of an animal in a maze. How are we to arrive at this objective?

Historically, there have been three principal approaches of consequence. First, it seemed on a commonsense or philosophical basis that one might be able to achieve a direct solution by way of introspection. It might be that an individual has access to his own unobservable behaviors of a kind quite comparable to the observations he can make of the overt behaviors of others. For present day psychologists I need not dwell on the sustained and sometimes brilliant efforts that were made to implement this approach nor on the apparently insuperable difficulties that brought these efforts largely to nought. We now understand that introspection cannot stand alone as a source of factual data for theories but provides at most a supplementary and relatively fallible tool for generating useful descriptions of cognitive activity. The second main approach, that of the behaviorists, proceeded on the supposition that perhaps all of the unobservable behaviors of a cognizing individual involve small scale activations of the same effectors that serve overt behaviors and that, consequently, internal activity can be adequately monitored by cleverly placed electrodes. The insufficiency of this approach has proved essentially the same as that of the first; it seems clear now that electrical recording of the activity of motor units again provides only a supplementary and often not readily interpretable source of information about the activity of an individual engaged in reading or thought.

The third approach to the problem of description, and the one that has become dominant in current cognitive psychology, arises from a conjunction of a method with a set of concepts. The method had its origin in early laboratories of experimental psychology. I refer, of course, to the use of reaction times to trace mental processes. The idea of ascertaining the durations of unobservable mental processes by subtracting reaction times for initiation of earlier from those of later stages of a task, first formulated by Donders in 1868 (see Chase, Chapter 2 of this volume), has evolved into an extremely sophisticated technique. The related concepts are those of information theory, which began to filter into psychology in the 1950s but have only become highly fruitful when conjoined with the reaction-time method in the “information-processing” approach of current cognitive psychology. The critical change in frame of reference associated with this approach was the shift from basic data obtained in the form of descriptions of stimuli and responses to basic data expressed in terms of transformations of information.

When studying an individual who is engaged in some cognitive activity, one has the option either of describing the events that the individual sees and hears and his observed responses to them or of describing the information made available to him and what he accomplishes by way of storing, transforming, retaining, and utilizing the information. Presumably, most psychological investigators have some attachment to a long-term goal of unity of science and are inclined to assume, tacitly or explicitly, that informational can be reduced to behavioral descriptions, just as they tend to assume that the latter can ultimately be reduced to descriptions in the terms of chemistry and physics. But experience in many fields of science, including our own, suggests also that it is rarely if ever necessary, and certainly not often fruitful, to wait unduly upon such reductions. Theory construction can proceed whenever one has at hand a satisfactory descriptive frame of reference and some constructive ideas for organization and interpretation of data. In the present instance both the descriptive frame of reference and the ideas for organization of data that have enabled a fertile new line of psychological theory construction are those of digital computers and information-processing systems.

Analogies between human problem solvers and digital computers have been important in the development of contemporary cognitive theory, and they can also serve a useful function in helping to elucidate the differences between theories evolving in the stimulus–response versus the information-processing traditions. In order to deal effectively with a computer, one needs bodies of fact and theory organized at two distinct levels. On the one hand, one needs to know precisely what happens when electrical impulses are initiated by input operations, what paths they take through the circuitry, and how they lead to the observed outputs.

This information and the associated theory is essential for engineering and maintenance but of little relevance to the user. The programmer needs, rather, to know how the information he enters at the console is encoded, how it is stored, what transformations are produced by various operations at his command, how the state of the information stored in the machine can be queried, and how to decode the output. He needs to know something of what occurs between the moment at which he types an input and the moment at which he can read the result of some procedure on a display screen, but he needs to know this in terms of the sequence of information-processing operations carried out by the computer, not in terms of the passage of impulses along cables and the magnetization and demagnetization of molecules.

Within an information-processing approach to human cognitive activity, one asks much the same kinds of questions as those raised by the user of a computer. One tries to construct a coherent account of what happens between observed stimulus and observed response, just as does the stimulus–response psychologist or physiological psychologist, but one attempts to do so in terms of the sequence of information-processing operations carried out by the individual. That this approach can actually work evidently owes to some rather striking correspondences between the mode of operation of a human information processor and a digital computer. In particular, both operate to an important extent by means of sequences of operations. However, in the case of the human being, the operations are much slower than in the case of the digital computer and it turns out to be possible to estimate their durations from overt response times.

The measurements, to be sure, must proceed on the basis of assumptions and inferences, since for the most part the beginnings and endings of mental operations cannot be directly observed. The strategy that proves effective is to try to assume as little as possible and to pin down one’s account of a particular information-processing operation or stage by means of converging lines of evidence. The remaining chapters of this volume will show how the prosecution of this strategy during the past decade or so has led to substantial advances in the complexity of phenomena that can be dealt with rigorously and quantitatively in nearly every area of cognitive psychology, from the simplest aspects of reaction time or sensory memory to the comprehension of prose and the mental manipulation of spatial relations.

II. SUFFICIENCY OF INFORMATION-PROCESSING MODELS

Those who have been following the literature on information processing in psychology will be aware that in the preceding paragraphs I have rather understated the objectives of the principal devotees of information-processing models. Their purposes seem to go much beyond the building of a descriptive account of certain cognitive activities and to be pointed rather toward the formulation of a self-contained body of theory expressed wholly in information-processing terms. It is of interest to inquire whether, following a decade or so of pursuit of this grander objective, progress seems to be accelerating or whether these efforts are running into definite limits.

One rather obvious limitation is perhaps not critical. If an information-processing model is not to be wholly autistic, it needs to be hooked to bodies of fact and theory outside its sphere, and especially at the input and output sides of cognitive functions. On the input side, we have the critical problem that information presented to an individual may or may not enter his cognitive-processing system. Thus, we are seeing concurrent vigorous development of bodies of visual and auditory information-processing theory bearing on our ability to predict or specify what will be effectively perceived under various circumstances. Then at output, the results of cognitive operations must go through highly limited capacity channels. To deal with this aspect of interface with the environment, we require knowledge and theory of speech production and output interference. The extent to which these bodies of research and theory on sensory and motor processes need to be taken account of by the cognitive psychologist depends, of course, on the problem at issue. When we are trying to understand a person’s ability to carry through difficult mathematical proofs or to comprehend stories, these relatively peripheral processes are of little relevance. But when we seek to deal with the functioning of an individual under circumstances of severe time pressure and information overload, they become critical.

To press our queries further, once the matter of environmental interface has been taken into account, does it seem possible that information-processing concepts can be self-sufficient? Here, on the basis of the actual course of research, we must have grave doubts about the answer. The study of classical varieties of problem solving within the framework of computer-simulation models, as in the extensive work of Newell and Simon (1972), perhaps comes the closest to providing examples of almost autonomous information-processing models. These models, for example, the “general problem solver,” are formulated in terms of operations that can be carried out on a computer; they are, in fact, embodied in programs that are run on computers to simulate the way in which it is assumed a human problem solver operates. But even here we find mixtures of terms from different levels of theory appearing, with, for example, references to concepts of perception and short-term memory adduced to those having to do with transformations of information (see, for example, Simon, Chapter 7 of this volume).

In the currently highly active field of research on semantic memory, which surely should be as apt as any for information-processing theory, we see in the current literature a bewildering mixture of concepts of very different origins. At one extreme the work of Carpenter and Just (1975), Chase and Clark (1972), and Trabasso, Rollins, and Shaughnessy (1971) has remained closest to theorizing strictly within an information-processing framework. However, this work also has stayed closely tied to a few experimental paradigms, with little tendency to move toward broader and more formal models than the specific assumptions needed to handle limited types of data. This last remark is not meant at all to underrate the impressive accomplishments of these investigators, only to suggest that they are not pushing the limits of the theoretical approach.

At another extreme Norman and Rumelhart and their associates (1975) have formulated an extremely broad and elaborate model in information-processing terms. Yet, as they bring their model into contact with data, one of their principal strategies proves to be shaking out special cases of the model, which prove to be formulated in a mixture of terms including concepts of information processing together with concepts common to other bodies of theory on perception and memory (see, for example, Rumelhart, 1970).

Anderson and Bower (1973) formulated almost as ambitious a model for the handling of a wide range of phenomena of human verbal memory. The aspects of Anderson and Bower’s model that are intended to represent the structural aspects of memory are couched in terms of a mixture of ideas drawn from linguistics and from association theory. The hybrid they come out with is a body of directed associations between memory elements labeled in terms of linguistic properties. Inputs to the individual are presumed to come largely by way of sentences, heard or read, which are then encoded in the form of propositions. These in turn are mapped onto a network of labeled associations, and retrieval of information from memory is conceived to be a matter of sending the encoded representations of newly encountered sentences through the network in the search of matches between these and previously stored propositions. However, in the recent developments of this work, we find increasing reliance on ideas having to do with the activation of serial or parallel pathways in the network (Anderson, 1976).

The same idea has become conspicuous in the somewhat related but less formally theoretical approach of Meyer, Schvaneveldt, and Ruddy (1974), where again one sees emerging the same mixture of information and processing and pathway concepts. Again, in the current work of Shepard and his associates (see Shepard & Podgorny, Chapter 5 of this volume) on nonverbal representations in memory, we see one of the individuals who was among the first to employ information-processing concepts in psychology generating an extensive body of research and theory in which essentially none of the interpretive concepts come from the information-processing tradition. On the whole I would say that the trend of developments in a wide range of research suggests strongly that the information-processing analogy has had an exceedingly fertile influence on cognitive psychology but by itself is proving insufficient as a source of explanatory concep...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Table of Contents
• Foreword
• 1. The Information-Processing Approach to Cognition: A Confluence of Metaphors and Methods
• 2. Elementary Information Processes
• 3. Visual Processing and Short-Term Memory
• 4. Chronometrie Analysis of Abstraction and Recognition
• 5. Cognitive Processes that Resemble Perceptual Processes
• 6. Natures of Problem-Solving Abilities
• 7. Information-Processing Theory of Human Problem Solving
• 8. Application of Learning Models and Optimization Theory to Problems of Instruction
• Author Index
• Subject Index