Memory Retrieval Research Paper

1. Two Views of Memory Retrie al

Two simple views provide a necessary foundation for the presentation of contemporary accounts of memory retrieval. One view claims that retrieval is determined by the state of the memory trace. The other claims that it depends on the presence of an effective retrieval cue.

1.1 The Trace-dependent View of Retrieval

The simplest account of memory retrieval assumes that it depends only on the strength of the memory trace relative to other traces. Retrieval failure occurs because the memory trace has become too weak, or because competing traces have become stronger than the target trace. Such a trace-dependent view of retrieval held sway in much of experimental psychology until well into the 1960s.

According to the trace-dependent view the strength of the memory trace (and thus the likelihood of retrieval) is a function of its initial strength and the length of time between this original experience and the attempted retrieval. The body of evidence supporting both these claims is overwhelming. The second assumption is supported by countless experiments that show a gradual decrease in retrieval success as the retention interval increases. Consider now the first assumption. The initial strength of a trace is determined by a number of factors, most notably by the way in which the experience is encoded, and it is not at all difficult to demonstrate that such encoding processes influence retrieval. For example, in experiments requiring the recall of names of simple objects, recall levels can be greatly enhanced if, during presentation of the names, participants are asked to form a visual image of each object.

The trace-dependent view can account for many of the basic phenomena of memory retrieval. For example, the fact that retrieval becomes more difficult with passing time is a consequence of a weakening of trace strength. The fact that an unrecallable item can nevertheless often be recognized is explained by the claim that recall requires a stronger trace than does recognition. By appealing to fluctuations in trace strength over time it can even account for the common experience of being unable to recall something at one moment but being able to do so at a later time.

In fact, the trace-dependent view of retrieval is not so much false as seriously incomplete. It is incomplete in at least three ways. First, strength is not the only feature of the memory trace that is important to retrieval; so too is its content. A unidimensional concept such as strength fails to capture the qualitative (or multidimensional, semantic) aspects of the encoding process. As will be seen, this is a feature of the memory trace that is of great importance in understanding retrieval. Second, the state of the memory trace, no matter how it is characterized, is not the only factor that influences retrieval. These further influences will be described in the remainder of this research paper. Third, the fact that encoding processes strongly influence retrieval does not constitute an account of the retrieval process itself, any more than the fact that practice increases skill at a game such as chess is an account of the processes involved in skilled chess playing.

1.2 The Cue-dependent View of Retrieval

An alternative, or rather a complement, to the tracedependent view of memory retrieval is that of cue dependency. According to this view, retrieval failure is often a consequence of ineffective cueing rather than of a faded memory trace. From this perspective, the problem is one of trace access. The trace may be quite intact but without an effective retrieval cue it remains inaccessible, rather like a mistakenly shelved library book. The book is quite intact, but the retrieval code is no longer effective. Similarly, information stored in computer memory is of no use without an effective means of accessing it, and a major research area in computer science is the development of efficient systems of information retrieval. Although such systems have had only limited success as models of human memory retrieval, they do embody one important principle. Memory retrieval depends on the relationship between how a memory is stored and the type of information available at the time of retrieval. Successful retrieval therefore depends not only on the state of the memory trace, but also on the conditions at time of retrieval.

It is not difficult to show that material that resists all efforts at retrieval, and thus gives every appearance of having been forgotten, can subsequently be remembered in response to an appropriate cue. Such findings demonstrate that in many cases inability to retrieve information may be a failure to access information that is potentially available, rather than of forgetting in any permanent sense. This fact accounts for the common experience described as the ‘tip of the tongue’ state in which people feel they are on the verge of recall. Such states tend to occur in association with attempts to retrieve simple factual information such as the name of a person or place. Formal studies of the phenomenon confirm its validity. When they report being in a tip-of-the-tongue state, participants in experiments can often accurately describe features of the word they cannot quite recall—its first letter, or the number of syllables. A closely related phenomenon that has also been studied experimentally is called ‘feeling of knowing.’ With better than chance accuracy, participants can predict whether or not they will be able subsequently to recognize an item of information they cannot presently recall. For a more detailed description of these phenomena see Tulving and Craik (2000, Chap. 13).

The success at recovering apparently forgotten memories in response to appropriate cueing has led some to entertain an extreme form of the cuedependent view of forgetting. The claim is that nothing is ever forgotten but, rather, memories simply become inaccessible. This extreme view has dubious scientific status in that it cannot be falsified by empirical methods. Even the failure of the most exhaustive efforts to cue the lost memory will remain unconvincing to the firm believer: it can always be claimed that such failure simply reflects an inability to find an effective cue. For this reason, and because of its general implausibility, this view has never found much support from within experimental psychology, although it is widely held in other circles (for a review of this point see Loftus and Loftus 1980).

2. The Context-sensitive View of Memory Retrieval

This section develops the view that retrieval is a more complicated process than that assumed by either the trace-dependent or the cue-dependent accounts. The claim is that successful retrieval is a constructive process based on an interactive relationship between features of the trace and those of the retrieval cue. For convenience of exposition a distinction is drawn between retrieval of general knowledge (commonly referred to as semantic memory) and retrieval of personally experienced past events (episodic memory).

2.1 Retrie al from Semantic Memory

What makes a retrieval cue effective? In the case of memory for factual material (semantic memory) the answer is fairly straightforward, at least at a general level. Obviously the cue, if it is to be effective, must be knowledge associated with the material that is to be retrieved, or knowledge from which features of the target material can be inferred. The challenge is to embody this general principle into a more precise description and to this end there have been a number of attempts to describe in detail how such retrieval-byassociation might work. The strategy has been to develop a formal description of a person’s knowledge and then explain how this structure can be used to retrieve information. The description usually takes the form of a network, typically embodied in a computer program. The nodes of the network may be words or concepts, and the interconnections between nodes represent various forms of relationships among the nodes. Thus, a node denoting ‘bird’ may be linked to a higher-order node (animal), to an attribute list (has wings, feathers, etc.), and to a list of exemplars (robin, sparrow, etc.). Such networks can retrieve information either directly from the network or by inference. Thus, the fact that a bird has wings may be retrieved directly, but answering the question ‘does a bird breathe?’ may require an inference because the attribute ‘needs to breathe’ is connected, not directly to the node ‘bird,’ but to the higher-order node ‘animal.’ A powerful feature of such networks is that they can explain how intelligent responses can be made to novel questions (Did Isaac Newton have a telephone?) the answers to which cannot have been learned directly.

Connectionism, or parallel-distributed processing (PDP), is a more recent development. In a connectionist network, words and concepts are not represented by nodes, but are themselves networks, hence the term distributed. A typical network consists of a set of input and output units, between which there may be one or more layers of hidden units. Knowledge (say the concept dog) is represented by the pattern of interconnectedness among these units, and the weight of the interconnections. Stimulation of input units sends excitation through the network. Retrieval consists of the resulting firing of appropriate output units. Although PDP models have scored a number of successes, the extent to which such networks accurately model human acquisition of knowledge and its retrieval remains a controversial matter.

2.2 Retrie al from Episodic Memory

What of the retrieval of personally experienced past events, a form of remembering referred to as episodic memory? How can people retrieve the details of the traffic accident they witnessed yesterday, or how they spent New Year’s Eve, 1999? By definition, such experiences represent unique events tied to a specific point in time. General knowledge on the other hand is likely to be used in a variety of contexts that strips its original acquisition of any temporal specificity. Most people know that Paris is the capital of France, for example, without reference to the events through which that knowledge was originally acquired. The major consequence of the difference between these two types of memory is that the retrieval of elements from single events is more strongly influenced by the context within which such elements occurred. To illustrate this point, consider the following example based on an experiment reported by Thomson and Tulving (see Tulving 1983 for a more detailed account). The experiment will be reported in some detail because it embodies a very fundamental principle of memory retrieval.

Suppose the word FLOWER is a sample member of a list of 24 words presented to participants who will be asked subsequently to recall as many of these words as possible. During the initial presentation, some participants see the word FLOWER paired with another word, fruit. The two words of each pair are weakly associated, that is, for example, the word ‘FLOWER’ is occasionally (about 1 percent of the time) given as a response to the word ‘fruit’ in normative free association studies. Participants in a second group see FLOWER without any pairing. For the recall test, each of these groups is divided into two; one subgroup receives the word fruit as a cue, the other subgroup receives no cue. Thus, the experiment has four conditions. In one condition the words are presented and tested in isolation. Consider this the baseline condition. In a second condition words are presented with a paired word and this paired word is re-presented as a cue at the time of recall. For the other two conditions the paired word appears at presentation but not at recall, or vice versa. How does recall for the three conditions involving at least one presentation of the weakly associated word compare with the baseline condition? Not surprisingly, the condition in which the associate appears at both study and test produces substantially better recall than the baseline condition. Even more interesting is that the other two conditions yield equivalent recall levels that are lower than the baseline condition.

Suppose a further condition is added to the experiment. In this new condition the target word is again paired with a weak associate at presentation (fruit—FLOWER), but in the recall phase, this cue is replaced with a strong associate of flower (bloom) not previously presented. A reasonable prediction for this condition is that the presence of the strong associate bloom when recall is being attempted should increase the likelihood of the recall of FLOWER, perhaps to a level higher than for the condition in which the weak associate fruit was present, even although fruit but not bloom was present at the study phase. In fact, exactly the opposite is true. Not only is bloom a much poorer cue than fruit, its presence reduces recall to well below the baseline level. On the other hand, if FLOWER is presented in isolation then bloom is a better cue than fruit.

What can be concluded from this experiment? The experiment demonstrates a principle that is of wide significance and of fundamental importance to an understanding of memory retrieval. In its most general form, this principle states that successful retrieval depends not only on the state of the memory trace, nor only on the nature of the retrieval cue but rather on the interactive relationship between the two. Retrieval will, of course, be influenced both by the form of the initial processing, and by the nature of the available retrieval cues, but neither of these factors considered separately will be sufficient to explain retrieval.

Consider the Thomson and Tulving results in these terms. The event to be recalled is the presentation of the word FLOWER. If one asks if this event will be better remembered if it occurs in relative isolation or in the context of the word bloom, there is no simple answer. It will depend on what cues are available at the time. Similarly if one asks which cue is better, bloom or flower, for retrieving this event, again there is no simple answer. All that can be said is that cue effectiveness depends on the state of the memory trace, which in turn depends on the context within which the event occurred—in isolation, or in the presence of bloom. Memory retrieval is context-sensitive.

2.3 Some Examples of Context-sensitive Retrieval

This context-sensitive view of memory retrieval has important implications for the understanding of common memory phenomena. In everyday remembering retrieval contexts will differ from one occasion to another, providing a complex and varying set of retrieval cues over which there can be little control. Such contexts may vary greatly in their cue effectiveness depending on their relationship to the conditions under which the original event was experienced. This strong effect of contextual cueing on retrieval offers a more convincing explanation (than the fluctuating trace strength view) as to why a memory that remains stubbornly inaccessible on one occasion readily springs to mind on another.

Even changes in the physical environment are sufficient to influence retrieval. An experiment reported by Godden and Baddeley (see Baddeley 1993) provides an interesting example. These researchers were involved with the training of deep-sea divers. They had divers learn material either while under water or while sitting on land, and then tested them either on land or under water. Their results show a strong context effect in that memory is better if the retrieval and learning environment are the same. Similar results have been obtained by having learning and retrieval occur in different rooms. The practical implications of such result are straightforward and exemplify some general principles. First, retrieval in an environment radically different from that of the original experience may be poorer compared with recall in the same environment. Baddeley describes previous anecdotal reports of the difficulty divers had remembering underwater experiences when being debriefed on the surface. Second, when there is some control over the study environment, that environment should be as similar as possible to the environment in which retrieval will be required.

What is true of changes in the external physical environment is also true of the internal or mental environment, giving rise to a phenomenon known as state dependency. For example, attempting to retrieve material while in a state of marijuana or alcohol intoxication is more likely to succeed if the original event was also experienced in that state, rather than a state of nonintoxication. (It should be noted, however, that memory retrieval is always better when both encoding and retrieval occur in the non-intoxicated state.) Emotionally arousing material shows a similar effect. Thus, people in a depressed mood may find it easier to retrieve unpleasant experiences than pleasant ones, a bias that can only serve to sustain the depressed mood.

Such results can be used to develop techniques that might improve a person’s ability to retrieve past events. The dependency of retrieval on physical and mental context suggests several possibilities. If someone is attempting retrieval in a physical environment different from that of the original experience (which is probably the usual situation), then retrieval can be improved by having the person mentally place themselves in the original environment (Smith 1979). This reinstatement of context is an important aspect of an interview technique developed by Fisher and Geiselman (1992) to facilitate the recall of eyewitnesses. As an aside, it is interesting to note that this interview technique achieves results that are at least as good as those obtained through the use of hypnosis. Contrary to popular opinion, there is nothing magical about hypnosis as a means of eliciting memories; insofar as it succeeds, hypnosis exploits the same principles of context reinstatement.

2.4 Failures of Memory

It may seem that human memory retrieval as it has been thus far described is not radically different from information retrieval in computer systems. In both cases success depends on a compatible relation between a system of storage and a method of retrieval. However, there are important differences. Computer retrieval is typically an all-or-none affair; when successful, the information is retrieved with perfect fidelity. In the absence of physical damage or explicit programming instructions, the stored information remains unchanged and is uninfluenced by other input. The downside of this fidelity is the relative inflexibility of what will function as a successful retrieval cue. Human memory retrieval is different. The highly interrelated structure of human memory allows for great flexibility, enabling it to answer all kinds of unanticipated questions. Rather than locating a complete and intact memory trace, a great deal of human memory retrieval (especially episodic memory) is better thought of as a form of pattern completion, rather like the paleontologist’s reconstruction of an entire animal from a few fossil bones.

The cost of this flexibility and constructive processing is reduced accuracy. With computer retrieval, the function of the retrieval cue (file name, key word, etc.) is to locate the stored information; neither the retrieval process nor the cue itself modifies this information. With human memory the situation is different. There is extensive evidence that, in the case of human memory, the constructive, pattern-completion process, along with the retrieval cue itself, play a role in forming the content of the reported memory.

Errors attributable to the constructive process usually reflect the misleading influence of existing knowledge. Schacter (1996, p.103) describes the following simple demonstration based on experimental work by Roediger and McDermott (1995). Participants studied a list of 15 words all associated with the word sweet (candy, sour, sugar, etc.) although the word sweet itself did not appear. A short time later, they were asked whether the word sweet was in the list. Schacter reports that in demonstrations with very large audiences, 80–90 percent erroneously claim that sweet was part of the list. The compelling aspect of this demonstration is not merely the false recognition of sweet, but that people claim to have a quite vivid memory of the word and are highly confident that it was on the list.

The content of retrieval cues, in the form of questions and hints, can also help determine the content of retrieved memories. In a classic study (Loftus and Palmer 1974), participants saw a film of two cars colliding. Some participants were asked to estimate the speed of the cars when they hit each other. For other participants the word hit was replaced with other words such as smashed, or contacted. This variation produced large differences in estimates, ranging from 40.8 mph for smashed, 34 mph for hit, and 31.8 mph for contacted. The differences in cues produced other effects. For example, participants whose question had used smashed were more likely to report that there had been broken glass although, in fact, there had been none.

The potential distorting effects associated with the constructive nature of retrieval have obvious implications for a wide range of activities ranging from legal eyewitness testimony to psychotherapies that place a strong emphasis on the recollection of past experiences. These implications are often lost to interviewers or therapists who regard memories as akin to books in a library or files in computer memory—a passive entity that once located and activated will emerge fully intact. The reality of memory retrieval is different. On the one hand the flexible constructive nature of memory retrieval does make it possible to recall experiences that at first appear to have been totally forgotten. On the other hand, however, this same property makes such recollections vulnerable to the influences of the particular wording of questions, and the general context (both physical and mental) within which retrieval is attempted.

3. Conclusion

Our understanding of retrieval has advanced considerably beyond the notion that the only determining factor is the strength of the memory trace. It has also become clear that a simple cue-dependency account is inadequate. The memory trace is not a static entity waiting for an appropriate cue to pluck it from some mental shelf. Rather, retrieved memories are the result of a complex interplay of stored information and retrieval cue. The principles governing this interaction are likely to be the major focus of future research. Such research will provide greater understanding as to why human memory, usually so dependable, can sometimes be inaccurate or even totally false. A particular challenge is to understand, not simply why mistaken memories occur, but why such memories are often indistinguishable from those that are genuine, and are believed to be veridical with a high degree of confidence. In this matter behavioral research is being supplemented by brain imaging techniques that promise to identify the brain mechanisms that distinguish false from real memories.

 Bibliography:

1. Baddeley A D 1993 Your Memory: A User’s Guide. Macmillan, New York
2. Fisher R P, Geiselman R E 1992 Memory-enhancing Techniques for In estigative Inter iewing: The Cognitive Inter iew. Thomas, Springfield, IL
3. Loftus E F, Loftus G R 1980 On the permanence of stored information in the human brain. American Psychologist 35: 409–20
4. Loftus E F, Palmer J C 1974 Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior 13: 585–89
5. Roediger H L, McDermott K B 1995 Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition 21: 803–14
6. Schacter D L 1996 Searching for Memory. Basic Books, New York
7. Smith S M 1979 Remembering in and out of context. Journal of Experimental Psychology: Human Learning and Memory 5: 460–71
8. Tulving E 1983 Elements of Episodic Memory. Oxford University Press, New York
9. Tulving E, Craik F I M (eds.) 2000 The Oxford Handbook of Memory. Oxford University Press, New York