Levels of Processing in Human Memory (PLE: Memory)

Laird S. Cermak, Fergus I.M. Craik

• Publisher: Psychology Press
• Year: 2014
• ISBN:
  9781317749790
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

As a conceptual framework for the investigation of human memory, the levels-of-processing paradigm had enjoyed immense popularity since its introduction in the early 1970s. It was the impetus behind literally hundreds of experiments and was used as an "explanation" for a wide range of retention phenomena. Consequently, a wealth of data and theory had emerged, and this title assimilates and evaluates this information. Originally published in 1979, the distinguished contributors to the volume – both proponents and opponents of the levels-of-processing framework – present here their latest data and ideas on a viewpoint that has been a tremendous influence in memory research and related areas.

I

THEORETICAL AND EMPIRICAL DEVELOPMENTS

1	Effects of Elaboration of Processing at Encoding and Retrieval: Trace Distinctiveness and Recovery of Initial Context

Larry L. Jacoby

Fergus I. M. Craik

One central point made by Craik and Lockhart (1972) in describing a levels-of-processing framework for the study of memory was the intimate association between memory and perception. In their view, the memory trace was characterized as the record of operations carried out initially for the purposes of perceiving and interpreting the stimulus array. It was also proposed that deeper, more meaningful analyses of perceived events would be associated with more durable memory traces than would relatively superficial analyses of the sound or appearance of incoming stimuli; and in support of their proposal, Craik and Tulving (1975) showed that words for which meaningful decisions are made show higher levels of retention in an incidental memory task than do the same words after decisions about their sound or appearance. However, although the levels-of-processing view has had some success in describing general features of remembering in many situations, some difficulty has been encountered in specifying precisely what is meant by “deep” and “meaningful.” Further, the view does not readily allow for distinctions to be made within the domain of meaningful analyses; it simply postulates that all events processed in terms of their meaning should be well remembered. In some instances, however, types of meaningful processing have been associated with quite poor retention (e.g., Mandler & Worden, 1973). Given a set of tasks, each requiring subjects to deal meaningfully with presented items, the original levels framework provides no basis for predicting differences, although such differences are found.

In later papers (Craik & Jacoby, 1975; Craik & Tulving, 1975; Lockhart, Craik, & Jacoby, 1976), the original levels formulation has been altered in a number of ways. Differences in retention are no longer explained purely in terms of depth of processing; additional mechanisms have been introduced. First, it has been suggested that processing differs in breadth or degree of elaboration as well as in level, thereby allowing an account of differences in retention that arise within a level of processing. Lockhart et al. also implicated discriminability as an important determinant of retention. By this was meant that the memory of a particular event must be discriminable from those of other events. Discriminability in turn was related to differences in the “depth” and “elaboration” induced by initial processing.

The present paper expands on these and related notions. First, the role of distinctiveness of encoding is examined more fully; the idea that distinctiveness is not an absolute characteristic but is always relative to some particular background or set of items is emphasized. Following from this idea is the second point—that the original context must be recreated at output if the encoded distinctiveness is to be effective in allowing discriminability of the wanted trace from others. Third, the notion is developed that retrieval, like encoding, is a matter of degree; just as an encoded item can be elaborated to a greater or lesser extent, so can retrieval information be processed more or less extensively. For example, the degree to which the initial encoding context is retrieved is largely under task control. Fourth, the point is made that some forms of recognition do not depend on retrieval of the encoding context; the parallels between context-dependent and context-free recognition on one hand and Tulving’s (1972) notions of episodic and semantic memory on the other are explored. These ideas are illustrated by the results from some recent experiments.

Encoding Distinctiveness

One of the major difficulties with the original levels framework comes from the way that meaning was treated. Although not explicitly defined, there was a tendency in that paper to treat meaning as if it were a fixed entity; the meaning of a word was said to be either encoded or not encoded, depending on task demands. By this commonsense approach, each word has a single meaning or at most a few meanings. Such an approach quickly runs into difficulty, as can be seen in the simple situation of naming a given object. As pointed out by several writers (e.g., Brown, 1958; Garner, 1974; Olson, 1970), a concrete object does not have a single name or description. Rather, what an object is called or how it is described depends on the other objects from which it is to be discriminated. For example, a chair is a chair; but it is equally a piece of furniture, a thing, a wooden artifact, and any number of other descriptions, depending on what the chair is to be distinguished from. Similarly, the meaning of a word in a given context depends on distinctions that are to be conveyed by that word in that context.

The dependence of meaning on the distinctions to be conveyed is easily illustrated in considering synonymy. Anyone who has ever tried to construct a list of synonyms soon becomes convinced that there are no true synonyms in English. Paradoxically, however, almost any pair of words can carry equivalent meaning in some context. To the driver of a car, the consequences of a statement such as “Look out for the ______” are equivalent if the blank is filled by tree, house, truck, or generally any other concrete object name. Clearly, however, the words would not be synonymous in all other contexts. In contrast, words such as woman and lady seem more truly synonymous, because it is more difficult to imagine a context in which the choice of one word over the other would be meaningful. However, contexts that would distinguish the two words can clearly be found.

In perceptual research, several theorists have come to view perception as the process of describing a stimulus (e.g., Rock, 1975). In many cases, contextual factors such as the alternatives from which a stimulus is to be discriminated influence the description and, consequently, the perception of that stimulus. If we return to the original levels notion that memory for a stimulus is the record of perceptual analysis, we can then claim that the memory trace is functionally a description, or set of contrasts. The resultant view is similar to that advanced by multicomponent (Bower, 1967) or attribute (Underwood, 1969) memory theorists. The major difference between the present view and the previous ones is the claim that description is necessarily relative to a given context. That is, meaning is not simply an attribute that is or is not encoded. Rather, meaning is a set of contrasts resulting from distinctions required when interpreting the item in the context of some task.

It is useful to contrast the notion of distinctiveness with that of elaboration. By elaboration, it is often meant that a change in encoding is largely quantitative; that is, more information is added to the trace. By distinctiveness, however, we mean to emphasize the contrastive value of information. For example, learning that a particular person has a high school diploma adds information but has no contrastive value unless other people in the appropriate set failed to complete high school. Our notion is that perception and, consequently, memory move forward by a set of contrasts. If several events are described with highly overlapping or redundant dimensions, the events are not well-distinguished from each other. Additionally, if the same events are described with dimensions that do not overlap at all, they are also not distinguished because they have never been contrasted.

Distinctiveness requires change against some background of commonality. The phenomenon of proactive inhibition in short-term memory (e.g., Wickens, 1970) can be used to clarify this point. The buildup and release from proactive inhibition is likely to depend on characteristics of the whole series of lists rather than simply on the characteristics of words that appear in adjacent lists. If a set of lists is made up of unrelated words, one would expect no release from proactive inhibition to result when, for example, dog is followed by coat. This is true despite the change in category that is produced by this sequence of words. A change in category will be noted and serve as a basis for distinctiveness only if the commonality among prior words has been sufficient to establish a category that can serve as a background for change. Tversky (1977) has recently made a related point by demonstrating the importance of contrast for scaled similarity. In his experiment, the judged similarity of pairs of European countries (e.g., Italy–Switzerland) was increased when the list also included pairs of American countries (e.g., Brazil-Uruguay) that were to be judged rather than when it included only additional pairs of European countries. In the absence of variation with regard to continents, two countries being parts of the same continent added relatively little to their judged similarity. Tversky used the term diagnosticity to describe the above effect of contrast among other effects found in judged similarity. The work reported by Tversky makes it clear that judged similarity depends on diagnosticity and on the intensity or salience of attributes as well as on the number of attributes that two events potentially share. In considering memory performance, the notion of elaboration is insufficient to the extent that it denotes only a difference in the number of attributes encoded and ignores factors such as the diagnosticity of those attributes. By using the term distinctiveness, we mean to emphasize the importance of relationships among events—particularly the importance of contrast.

It is attractive to consider a memorable encoding as one that is easily discriminable or highly distinctive in the memory system. It seems quite possible that deeper, more elaborate encodings are more distinctive in this sense; a similar view has recently been put forward by Klein and Saltz (1976) and by Wickelgren (1977). The idea is that deeper encodings will be more discriminable from other encoded events and will be more easily retrieved, provided also that an appropriate retrieval cue is given (Tulving, 1974) and that the information is encoded in an organized, “recoverable” form (Norman & Bobrow, 1977). To make this point clearer, again consider the case of describing an object. The details of an object are more fully described when the object must be discriminated from a set of very similar objects; the description of a circle will be more complete if it is to be discriminated from other circles that differ in size and location than if it is to be discriminated from a set of squares. A more complete description results in the utility of the description being less reliant on reinstating the original set of alternatives. That is, the fuller description would also serve to specify the object among any less similar set of alternatives (within the same encoding dimensions, at least). Thus, more complete descriptions confer both greater distinctiveness and greater generality as a basis for discriminating one object from others. Similarly, in the case of memory for words, a more complete encoding or description of a word allows that word to be discriminated from a larger set of alternatives. The memory confusability of words that are similar, such as lady and woman, depends on the distinctions that are compelled by the study task and context. If lady were encountered in a list and lady and woman later appeared as alternatives on a recognition test, we would expect a high number of false recognitions of woman (Underwood, 1965). However, if the initial task required encoding of lady in terms of deportment as well as sex, fewer false recognitions of woman should result.

Let us stress again that distinctiveness is a context-relative term. A description that is highly distinctive for a particular set of alternatives is not necessarily distinctive for another set. Consequently, the distinctiveness of the description of an event cannot be specified without considering the alternatives from which it has been contrasted. If the set of alternatives is changed drastically, a previously distinctive description may be of very little use. A similar point has been addressed by Tulving and Thomson (1973) as encoding specificity. In agreement with Tulving and Thomson, we feel that it is necessary to focus on interactions between the manner in which initial encoding was carried out and the demands of the particular situation in which memory is assessed. However, it seems too extreme to argue that all aspects of the encoded trace are drastically modified by the context in which an event occurs (that is, that all aspects of the encoded description are relative to the specific context); presumably, some aspects are relatively invariant across contexts. It has been suggested, for example, that physical and structural aspects of words are relatively insensitive to changes in context, whereas semantic aspects are modified by context to a greater degree (Jacoby, 1974; Nelson & Brooks, 1974). The assumption that some aspects of an encoding or description are relatively invariant across contexts proves useful in a later discussion of retrieval processes.

Retrieval Processes

These ideas of distinctiveness and discriminability must be tied to some notions of how retrieval processes operate. We will assume a simple “feature overlap” model of recognition; that is, the probability of recognition increases with an increase in the number of common features activated at encoding and retrieval (or as a function of repetition of the same mental operations; Kolers, 1973). It follows that reinstatement at retrieval of the original encoding context enhances recognition by leading to an increase in the number of activated features in common between the study and test situations. Even out of context, re-presentation of an item may lead to some degree of recognition, and in this situation recognition can be enhanced by active reconstruction of the initial context. In this case the partial recognition may lead to constructions of plausible contexts (“Where might I have seen that person before?”); to the extent that one general line of reconstruction is associated with increased feelings of familiarity (“probably somewhere around the university”), that line is pursued until recognition reaches some acceptable level (“a student in my Introductory Psychology class”). In this sense, episodic and semantic information are thought to interact to enhance recognition (Lockhart, Craik, & Jacoby, 1976). One main point to be made here, and developed later, is that retrieval is not to be thought of as an all–or–none, automatic process. Rather, retrieval operations can be elaborated on or curtailed depending on task demands. The second point is that a distinctive encoding enhances recognition, because in this case when the encoding context is fully reinstated as retrieval information, such retrieval cues specify the prior event more precisely.

A number of memory effects can be described in terms of the interactions between distinctive encodings and adequate retrieval information. First, if the original encoding is difficult to accomplish, later memory of the event will usually be good. It is argued that the initial difficulty is associated with the formation of a more complete “description” of the stimulus and thus with a more distinctive trace of the event. As initial encoding becomes easier (with practice, say), there is a concomitant decrease in the level of recognition (Kolers, 1975). The effects of difficulty of retrieval on subsequent retention of the event can also be described in this way (Gardiner, Craik, & Bleasdale, 1973; Gotz & Jacoby, 1974). The role of distinctiveness in explaining effects of depth of encoding, of difficulty of retrieval, and of interference is described more completely later in the paper.

Our ideas on retrieval start from the assumption that successful recognition involves the activation of some critical number of features or operations that are then matched with the record left by the event on its previous occurrence. If there is a sufficient degree of overlap between the features encoded in the trace and the features presently active, then recognition occurs. We also assume that the trace of the event’s initial occurrence incorporates information about both the event itself and its context. If the event is re-presented in the same context on the second occasion, there will be a greater degree of overlap between trace information and test stimulus information; and the probability of recognition will be increased. Further, due to the more precise description of the original event, fewer new events will be falsely recog...

Table of contents

• Cover Page
• Half Title page
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Contents
• Preface
• I Theoretical and Empirical Developments
• II Extensions and Applications
• III Language Processes
• IV Developmental Issues
• V Perceiving, Acting, and Knowing
• VI Theoretical Alternatives
• Author Index
• Index