Handbook of Learning and Cognitive Processes (Volume 3)

Approaches to Human Learning and Motivation

William Estes

• Publisher: Psychology Press
• Year: 2014
• Print ISBN: 9781848723931
• eBook ISBN: 9781317672289
• Topic: Psychology
• Subtopic: Cognitive Psychology & Cognition
• Index: Psychology

Originally published in 1976, Volume 3 of this Handbook deals primarily with conditions of acquisition, retention and forgetting, and the manner in which acquired information and motivation combine to determine performance. The organization of this volume can be understood in terms of four principal categories. The first category deals with general problems of methodology, the second and third with basic concepts arising from research on human learning and performance and the fourth with applications.

Volume 1 presented an overview of the field and introduced principal theoretical and methodological issues that persistently recurred in the expanded treatment of specific research areas which comprise the later volumes. The areas traditionally associated with conditioning, learning theory and the basic psychology of human learning are treated in Volumes 2 and 3. The last three volumes will range over active lines of research having to do with human cognitive processes, at the time: Volume 4, attention, memory storage and retrieval; Volumes 5 and 6, information processing, reading, semantic memory, and problem solving.

1

Methodology of Human Learning

Leo Postman

University of California, Berkeley

One of the fields conventionally listed in taxonomies of psychological research is “human learning and memory.” This designation implies the existence of separate but related domains of inquiry—acquisition or learning on the one hand and retention or memory on the other. Since the two are invariably linked, both in theory and in experimental practice, why do the separate labels continue to be retained? The primary reason is that one cannot avoid making a distinction between learning and memory on abstract theoretical grounds. The distinction has also remained pragmatically useful in the classification of experimental procedures and the organization of empirical findings. No operations, however, have been devised that permit fully independent measurements of the two processes. Hence, the methodologies and the explanatory concepts of the domains of learning and memory have always been inextricably interwoven; they have become increasingly so in recent years. In order to focus on the problems considered in this chapter it is useful to begin with a brief discussion of the rationale of the distinction between learning and memory.

I. THE DISTINCTION BETWEEN LEARNING AND MEMORY

A. Acquisition and Retention

A differentiation between acquisition and retention, or an equivalent distinction, appears to be logically required. As McGeoch (1942) put it succinctly a long time ago, “An act must be fixated before it can be retained [p. 4].” More recently, Melton (1963) made the same point in more precise terms. Our basic experimental observations are changes in performance from Trial n to Trial n + 1. Whenever such changes occur, they must be taken to be the result of three theoretically separable but functionally confounded events, namely (a) trace formation, (b) trace storage, and (c) trace utilization. (The concept of trace is used here as an hypothetical construct, with no specific physiological connotations, to refer to the representation of input events in the nervous system.) When one speaks of encoding, storage, and retrieval, the postulated sequence of events is, of course, exactly the same.

Within this descriptive framework we can consider the difference between acquisition and retention in three different ways:

1. If one equates trace formation with acquisition, and storage and trace utilization with retention, it follows logically that acquisition must precede retention.

2. One can make an essentially arbitrary operational distinction between studies of learning and of memory in terms of the independent variables to which changes in performance are related. This is the conventional definition articulated by Melton. Changes from Trial n to Trial n + 1 are taken to index learning “when the variable of interest is the ordinal number of Trial n.” On the other hand, the changes are used to assess retention “when the variable of interest is the interval, and the events during the interval, between Trial n and Trial n + 1” (Melton, 1963, p. 3).

3. Given this conventional definition, one can inquire into the relationship between measures of learning and of retention. When that is done, it becomes readily apparent that independent measurement of the logically separable processes is not possible.

Inferences about trace formation must of necessity be based on observations of trace utilization. We cannot know that learning has occurred except by measuring retention after some arbitrarily chosen interval. We typically chart the progress of learning by repeatedly administering tests of retention after a constant interval. The learning curve rises over trials not only because new traces are formed but also because there is an increasing number of traces that persist between input and output. This state of affairs has, of course, been recognized for a long time. To quote McGeoch (1942) again, “A curve of learning represents a progressively greater balance in favor of retention, so that it is, in part a retention curve [p. 4].” On the assumption that a single input is always sufficient for trace formation, one can carry the argument further and equate learning with increased resistance to forgetting (Murdock, 1974, p. 12).

If tests of learning are inevitably measurements of retention, it is equally true that an evaluation of retention scores on Trial n + 1 requires an independent assessment of the degree of learning that existed at the end of Trial n. Unless one can specify the condition of the system at the end of Trial n, there is no way of drawing inferences about the effects of the events that presumably occur during the period of storage and at the time of retrieval. The pragmatic resolution of this dilemma again must take the form of an arbitrary convention. Degree of learning at the end of Trial n is defined in terms of performance on an immediate test of retention, and losses after longer intervals are then referred to this baseline. What is required, then, is the determination of a difference between an immediate and a delayed test if appropriate conclusions are to be drawn about processes subsequent to trace formation. While the logic of measurement is straightforward, we shall see that it has proved quite difficult in practice to implement it by procedures that are free of bias.

B. Analysis of Trace Utilization

The definitions of learning changes and of retention changes proposed by Melton include the implicit assumption that the characteristics of the material and the nature of the subject’s task remain constant between Trial n and Trial n + 1. Systematic manipulation of the latter variables between trials constitute the defining operations of a study of transfer, that is, “the availability and utilization of the memorial products of Trial n in a ‘different’ situation” (Melton, 1963, p. 3). This criterion for delimiting the domain of transfer studies requires further examination in the light of current experimental practice. Much of the work on memory today is concerned not only with the examination of progressive changes in storage but also with the composition of the memory trace (see Tulving & Bower, 1974) and the conditions determining access to memorial representations. The investigation of these problems usually involves the deliberate introduction of changes in the characteristics of the task between Trial n and Trial n + 1, for example, between input and output. A variety of probing procedures can be used to obtain information about the composition of the trace; the targets of such probes are not only attributes of the to-be-remembered (TBR) units themselves but also such features of the input events as modality of presentation, frequency and recency of occurrence of individual items, and so on. One may also attempt to infer the composition of the memory trace from the types of cues, for example, semantic and acoustic ones, that are effective in retrieval. This approach is based on the assumption that only those cues can be effective whose informational content overlaps with that of the trace (Tulving & Bower, 1974; Tulving & Thomson, 1973). Manipulation of the conditions of cuing has also been the preferred method for studying the processes by which memory traces become accessible. When apparently forgotten items are recovered in the presence of appropriate cues (for example, the names of the categories of which the TBR items are instances), it can be concluded that these items were temporarily inaccessible rather than unavailable (Tulving & Pearlstone, 1966).

There is a clear continuity, then, between the defining operations of studies of transfer, on the one hand, and the procedures used to investigate the composition of the memory trace and the conditions of access to the trace, on the other. This continuity of methods exists because systematic manipulation of the conditions of testing is required for the evaluation of hypotheses about the contents of storage or about potentially effective modes of retrieval. Transfer designs have, of course, been used traditionally to answer questions about “what has been learned.” Classical examples in the area of animal learning are the latent learning paradigm and the factorial designs for separating the effects on acquisition and performance of such variables as stimulus intensity and drive level (see Kimble, 1961, pp. 118 ff.). In research on verbal learning, transfer procedures have been the primary vehicle for identifying the nature of the functional stimulus, for example, in the acquisition of serial lists (Bewley, 1972; Young, 1968), and for assessing stimulus-component selection (Richardson, 1971, 1972). The methods discussed earlier, which are explicitly directed toward the assessment of the contents and the accessibility of the memory trace, are based on the same logic. The continuity of methods should not be obscured by the fact that in experiments on transfer the emphasis is often on the acquisition of the test task; the observed transfer effects can be understood only to the extent that the products of past learning and their mode of interaction with the requirements of the test task can be specified.

II. DIFFERENTIATION OF STAGES

As has been noted, it is possible to distinguish among measurements of learning, retention, and transfer on the basis of somewhat arbitrary but pragmatically useful operations. At the same time, it must be recognized that there are no well articulated rules of inference for relating observed experimental effects to one or the other of the theoretically separable stages of encoding, storage, and retrieval. In a recent discussion of the problem of analytic separation of stages, Murdock (1974) emphasized the fact that these processes are sequential and concluded that they should therefore be assumed to be interdependent. He went on to point out that, as a consequence, any variable that influences encoding may affect storage and retrieval as well; similarly, conditions that influence storage may also affect retrieval. Presumably the effects of retrieval variables are unique to that stage. The assumption of interdependence of successive stages is not logically required; it is in principle possible for sequential processes to be independent. However that may be, any attempt at analytic separation must begin with the identification of variables and operations that are assumed to be specific to a given stage. Once that is done, the next step is to resolve the ambiguities entailed by the sequential nature of the hypothetical processes. Many such attempts have been made, but the rules of inference seem to be getting less, rather than more, consistent.

A. Retrieval

Let us begin with what appears on the surface to be the simplest case, namely the identification of effects that are localized in the retrieval stage.

1. Analysis of Cuing Effects

At one time, manipulation of the conditions of cuing during recall appeared to be a rather straightforward method for studying the characteristics of the retrieval process. With subjects treated alike up to the point of the retention test, any differences in performance related to the cueing procedure implicate the retrieval process. The distinction between availability and accessibility of stored information (Tulving & Pearlstone, 1966) is based on this logic. When the inclusion of some cues (namely items from within the list) on tests of free recall proved ineffective or even detrimental, however, conclusions about the characteristics of storage as well as retrieval were drawn from the results. Slamecka (1968, 1969, 1972) argued for the principle of independent trace storage, with interitem relations influencing only the retrieval plan. Other investigators (Roediger, 1973; Rundus, 1973) saw the results as supporting an hierarchical model of the organization of higher-order units and assumed that items presented as cues become a source of output interference. The merits of these arguments need not concern us here. These studies are cited simply as an illustration of the explanatory burden that has been placed on the results of cuing studies.

This burden is increasing. As we noted earlier, the successive administration of different retrieval cues has been proposed as a method for determining the elements of the trace of a previously presented item (Tulving & Bower, 1974; Tulving & Watkins, 1975). The essential steps in the application of the method are as follows. Suppose that an associative cue (X) and a rhyming cue (Y) are used to probe the trace. These cues are administered to different subgroups of subjects in each of the two possible orders (X₁ Y₂ and X₂ Y₁). For each subgroup, the raw probability of recall in the presence of a given cue is a measure of the gross valence (that is, effectiveness) of that cue. Since cues may overlap in informational content, it is also necessary to assess the effectiveness of each cue when such redundancy is taken into account. The measures used for this purpose are the proportions of cases in which a cue proves effective after a failure to recall to the other one (and ). The latter indices are the reduced valences of the two cues. These estimates are taken to be free of any contaminating effects of the preceding test, on the assumption that the administration of an ineffective cue leaves the trace intact. From a set of the observed values (specifically, the gross valences of the first cues and the reduced valences of the second cues) a matrix is finally derived that represents the informational structure of the trace in terms of the probabilities of recall and nonrecall to the nonoverlapping c...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Table of Contents
• Foreword
• Introduction to Volume 3
• 1. Methodology of Human Learning
• 2. An Introduction to Research on Individual and Developmental Differences in Learning
• 3. Cognitive Factors in Classical Conditioning
• 4. Theories of Discrimination Learning and Learning Set
• 5. Probability Learning and Sequence Learning
• 6. Reinforcement and Human Memory
• 7. Motivation and Reward in Human Learning: A Cognitive Approach
• 8. Motivation from the Cognitive Perspective
• 9. Learning Theory and Behavior Therapy
• Author Index
• Subject Index