It seems quite unnecessary to be concerned with a definition of “forgetting.” Each of us has had innumerable experiences, often painful, with this phenomenon, and we know that, in general, the longer the period between the point at which we learn something and the point at which we try to remember it, the less likely we are to remember correctly. The growth of living organisms is always accompanied by deterioration processes. Forgetting is the behavioral deterioration process that may be likened to organic deterioration and is in opposition to the growth process called “learning.” We do not, in fact, know much about the organic or physiological correlates of forgetting, and it is not easy to understand why, in the evolutionary process, forgetting should remain so powerful a process in man when, from almost any point of view, it is unadaptive. It is almost as if we have “room” for only so much learning, and when more learning is to be added than the room will hold, some of the old learning must go. Indeed, in a manner of speaking, the theories of forgetting are based on this notion.
Learning must occur before forgetting can occur. This fact, seemingly trivial, is the key to an understanding of the processes causing forgetting and must also be the foundation of any applied program aimed at retarding forgetting.
A review of the experimental studies of forgetting can but lead to the impression that the evidence on which we base our knowledge of forgetting comes from quite a restricted set of materials and procedures, and thus our generalizations about forgetting may be severely restricted. It is a fact that most of our data come from what is classically known as rote-learning tasks, and the subjects for experimentation have been preponderantly college students. The study of the forgetting of motor skills has lagged; the memory for ideas, concepts, and principles has never been systematically investigated. Yet, when probes into these areas have been made, the results have given us no reason to believe that the basic laws are different from those derived from rote-learning tasks. The extensive use of college students—whose use has not been dictated by theoretical concern but has been largely a matter of convenience—may make us wary of generalizing these conclusions to all ages and ability levels. But again, the evidence available does not indicate that any serious distortions in the laws of forgetting are attendant upon the particular subjects that have been used in experimental work.
Nonsense syllables and nonwords . We must look to the tasks and materials from which most of our contemporary knowledge of forgetting is derived. The materials are simple verbal units. These may be words, or they may be a sequence of letters not found in English dictionaries. Calling these latter units “nonsense syllables,” or “nonsense material,” has produced considerable misunderstanding. It has often been assumed that a nonsense syllable (for example, RUH) is an artificial unit which removes the study employing such units from contact with real-life situations. It might be argued that experimentation is always, in fact, artificial (a vacuum does not exist in nature) and that it is only by such procedures that fundamental laws are obtained, because these procedures allow for the control of variables. In fact, the original use of nonsense syllables by Ebbinghaus in his pioneering studies on memory was, to a large extent, based on this premise. But there are three other facts that must be set forth and that may be more critical than an argument based on a philosophy of science.
To say that a nonsense syllable is not related to real life is simply erroneous. A word common to an adult must be, in essence, a nonsense syllable to a child before its meaning is acquired. The use of verbal units not in dictionaries may be tapping the same or similar processes that are involved when the child changes from a nonverbal organism to an organism with a verbal repertoire that can be correctly applied. Or a nonsense syllable may be likened to a word in a foreign language, and thus to the college student it is anything but removed from real life.
The second point is that a derogation of the use of nonsense syllables assumes a dichotomy between what is meaningful and what is nonsense. But measurements of meaning, by any of many techniques, do not support such a schism. There is a continuum between words and nonwords. Many so-called nonsense syllables have greater meaning than some words. Nonsense words, like “real” words, have varying affective connotations. Indeed, it could probably be demonstrated that affective reactions to such nonsense syllables as WOR are much more intense than those to such common words as THE. A nonsense unit, therefore, should mean only a word that does not appear as an entry in a dictionary, but this distinction per se between words and nonwords appears to have little psychological relevance.
The third reason for not viewing nonwords as something significantly apart from words is probably the most critical when reference is made to the use of these materials only in the study of forgetting. This reason is simply that the evidence indicates that words and nonwords “behave” the same as far as forgetting is concerned. This point will be given elaboration at a later point in the discussion.
Manner of presenting verbal units . In the bulk of the studies on forgetting, verbal units are formed into lists, so that in a single task the subject learns several items. The items might be formed into a serial list in which the subject must learn to reproduce not only the items but also a particular order of the items. A second widely used type of list is the paired-associate list. The subject is presented a series of pairs and is required to learn to reproduce the second member of each pair when the first member is shown alone. Acquiring the order of the pairs is not a part of the task; indeed, the investigator usually prevents his subject from learning the order by presenting the pairs in a different order from time to time. The paired-associate task is a counterpart of the task of learning a vocabulary of a foreign language, in which the first member of the pair is a word in the learner’s native language and the second member is its foreign equivalent.
Other tasks that have also been used include prose passages, sentences, and lists of words to be recalled in any order. But for reasons that need not be detailed here, at the present stage in the development of psychological research, the pairedassociate task possesses properties that make it a superior vehicle for studying forgetting, and much of the discussion will be cast around this task. This does not imply that there is evidence that forgetting varies as a function of the task. Such comparisons are difficult to make, although insofar as they have been made there is no convincing evidence that the type of task is a critical variable in the study of forgetting.
Rapid single-unit technique. Most of the research on forgetting has been with multiple-unit tasks (for example, several pairs of items in a paired-associate task). What must be considered a most amazing technical breakthrough was made in 1959 by Lloyd R. Peterson and Margaret J. Peterson. These investigators were able to devise a situation in which a single verbal unit was presented for learning and the subsequent course of forgetting could be measured over a period of time. A syllable, such as RZL, was presented for study for two seconds. Then, in order to prevent the subject from rehearsing the syllable, he was given a neutral task (for example, counting) for a few seconds. Finally, he was asked to recall the syllable originally shown to him. The startling fact is that lawful and reliable forgetting curves were obtained in which forgetting was essentially complete (the subject could .iot reproduce the unit) after 20 seconds. The exploitation of this technique has been, and is being, pursued by many investigators. The classical studies have used forgetting intervals of hours or days, and thus the sheer labor and mechanical problems in the performance of these studies were great. Although it is perhaps too early to say with assurance that the laws of forgetting observed under the Peterson-Peterson technique are the same as those for classical studies, the evidence thus far points strongly to this conclusion. If this is true, the study of forgetting may proceed at a pace hitherto impossible with classical procedures. In the Peterson-Peterson technique, seconds are comparable to hours or even days with the classical list method.
We may turn next to an examination of the reliable experimental facts in single-task studies of forgetting. A distinction must be drawn between single-task and multiple-task studies if we are to focus on the processes believed critical for an understanding of forgetting.
The laboratory situation can be described as follows. A subject who has not previously served in a laboratory experiment involving verbal learning is given a single list of verbal units to learn. These are normally presented over and over, each unit being exposed for a short interval of time (for example, two seconds), until the subject is just able to reproduce the list perfectly. Assume, further, that the subject returns 24 hours later and is asked to recall all of the units he can. We ask now about the general quantitative aspects of forgetting in such a situation and, in particular, what variables cause the amount of forgetting to vary.
Amount forgotten. Over the 24-hour period, the subject will forget from 15 to 20 per cent of the units. Thus, if the list contained ten paired associates, we will expect about eight of these to be recalled correctly. Under the conditions specified above, this amount of forgetting remains fairly constant across a wide range of materials and tasks. Some gross measures may indicate a degree of forgetting as great as 30 per cent, but when refined techniques of measurement are used, the 15 to 20 per cent loss will usually be found. With intervals longer than 24 hours, more forgetting will, of course, occur, but we do not know with any precision the rate of fall for many longer intervals with various materials. Because of the relatively large amount of data available on forgetting after 24 hours, we will use this interval as a basic reference interval.
Significance of the degree of learning. We may now ask about variables that influence the amount of forgetting during the 24-hour period. The facts, as viewed at the present time, provide a conclusion that seems not only to be opposite to that expected as research in an area proceeds but also opposite to that often expected according to “common sense.” For we can name only one variable that substantially influences the rate of forgetting in the single-task situation. Obviously, considerable clarification and elaboration are needed.
Characteristics of the verbal units will produce enormous differences in the rate at which the list is learned. As might be expected, a list of common words is learned much more rapidly than a list of difficult nonwords. The differences are so large that even the casual reader will be convinced that this is true. Thus, it is clear that a list of five 3-letter words, such as CAT, PEN, BUS, FAR, and ELK, would be learned more rapidly than a list of five nonwords, such as RZL, DBQ, HFG, BJX, and PCR. Technically, we say the words have a higher level of meaningfulness than do the nonwords. But if both sets are learned to an equal degree, they will be equally remembered 24 hours later. The critical point is the “equal degree of learning,” because degree of learning is the one variable that does influence the amount of forgetting. This can be demonstrated in two ways. First, when the number of practice trials is varied within a given task, it will be found that the greater the number of practice trials, the less the forgetting. Second, when units within the task are examined, one may note that those units which have been given correctly many times during learning will prove to be remembered better after 24 hours than those given correctly only a few times. So we say that if the degree of learning is equal for materials of widely different meaningfulness, they will be forgotten at the same rate.
Meaning related to degree of learning. The foregoing conclusion might seem contradictory to the experience of many, because most of us have long believed that retention of more meaningful material is better than that of less meaningful material. But this belief may have arisen as follows. If we spend an equal amount of time in learning the two kinds of material, retention of the list of words would assuredly be better after 24 hours than the retention of the list of nonwords. Even if a subject spent twice as much time studying the nonword list, retention of the words might still be better. This is because differences in rate of learning such lists vary so markedly that even after a longer study period on the nonwords the degree of learning will still be less than the degree of learning for words. But under such circumstances, the differences in amount remembered after 24 hours must be ascribed to the degree of learning achieved and not to the differences in the material. Differences in the material produce differences in learning, but if the level of learning achieved for materials of different difficulty prior to the introduction of the retention interval is equivalent, we may expect essentially equal amounts of forgetting, which, as noted earlier, will be about 15 to 20 per cent over 24 hours.
Similarity related to degree of learning. Let us consider another variable that produces wide differences in learning. If the number of duplicated letters among nonwords is varied within a list of paired associates, the greater the number of duplicated letters, the more difficult it is to learn the task. Technically, this is called “variation in intralist formal similarity.” But again, if the level of learning achieved for a list with high similarity is equivalent to that for a list with low similarity, the amount of forgetting will not differ measurably over a 24-hour period. The same situation will prevail if we use a list of synonyms and a list of words with very little similarity of meaning. The former will be more difficult to learn, but differences in forgetting cannot be demonstrated if the levels of learning attained are equivalent.
Individual differences in rate of learning. The general principle that degree of learning is the only critical variable in the single-task situation can be pursued one step further. There is evidence leading to the conclusion that there are no individual differences in forgetting. One proposition that has been ubiquitous in psychological research is that any set of measurements will produce a distribution of reliable individual differences. Some individuals are slow learners; others are fast learners. Some have excellent depth perception; others have poor depth perception. Measurements on living organisms of all levels have supported the proposition that there is a consistent range of individual differences. To suggest that this proposition does not hold true for forgetting requires a careful exposition of the extent of the data available for its support.
For one characteristic of individuals, the data are quite clear. The rate at which a person learns a given task is not related to rate of forgetting. A subject who learns a list very slowly will show no more forgetting than a subject learning the same task very rapidly, if the level of learning attained is equivalent. Such a fact is not without some support in incidental observations. It is not unusual for students in the author’s classes to observe (when discussing their performance in the course) that they do not learn rapidly, but once they have learned something well, they remember it well. A slow-learning student, therefore, may take much longer to achieve a level of learning attained by a rapid-learning student, but given the equivalent degree of learning, forgetting does not differ.
Rate of learning is, of course, only one of hundreds of characteristics on which people differ reliably. Some people are more emotional than others; some are more conservative than others. Do none of these many possible characteristics produce differences in rate of forgetting? Here the evidence is less convincing, but it is pointing toward the conclusion that if these characteristics are related to rate of forgetting, the magnitude of their effect is small.
Consider, for example, the following situation. Subjects of widely different learning abilities are given the same period of time to study a given task, and immediately following this study period they are tested. The test scores will show a wide range, indicating that rate of learning differed markedly among the subjects. After 24 hours, a recall test is given. The recall scores are then correlated with the scores taken immediately after learning. Such a procedure will produce very high correlations across a wide variety of materials. The high correlations can be taken to indicate again the strong relationship between degree of learning and retention. If there were wide individual differences in rate of forgetting that vitiate the relationship between learning and retention, such high correlations would not be found. This is to say that a given person’s retention score can be predicted with a high degree of accuracy if only his learning score is known. Under such circumstances, it does not seem possible to attribute to individual differences a role of consequence in forgetting. That such a conclusion runs contrary to the conclusions concerning the role of individual differences in many other situations is indeed a puzzle, and perhaps future research will change the conclusion. But the available data allow at least a tentative conclusion that individual differences in forgetting are, at best, of very small magnitude.
The reason for making the distinction between single-task forgetting and multiple-task situations may now be examined. Very early in the period in which the study of learning and memory was brought into the laboratory, it was discovered that if two tasks were learned in immediate succession and the subject’s retention of the first task was then tested, severe forgetting was observed. That the second task was in some way responsible for this forgetting was shown by control conditions in which only the single task was learned and tested for retention.
Retroactive and proactive inhibition. The amount of forgetting produced by the learning of a second task, interpolated between the learning and recall of a first task, is called “retroactive inhibition.” The amount of retroactive inhibition that can be produced in the laboratory is very great; under appropriate conditions recall of the first task can be reduced to zero.
The discovery of retroactive inhibition had two important consequences. First, it became an object of study per se, and hundreds of experiments have been performed in which such variables as the similarity between tasks, the degree of learning of the interpolated task, and various time relationships have been studied.
The second consequence of the work on retroactive inhibition is that it has largely set the theoretical thinking of investigators. This thinking has preponderantly tended toward the use of interference as a fundamental cause not only of forgetting observed in the retroactive situation but of all forgetting. Certainly the role of interference in retroactive inhibition cannot be disputed: the higher the similarity between the two tasks, the greater the amount of retroactive inhibition. Furthermore, subjects are regularly observed to give items from the interpolated task when they are asked to recall the first task. A theory built around an interference notion seemed quite appropriate.
In more recent years, a second source of interference was discovered that has only served to refine and reinforce the notion that interference must be a major cause of forgetting. This second source of interference is identified as “proactive inhibition.” Again, a subject learns two tasks, but this time he is asked to recall the task subsequently learned. Of course, some time must elapse between the learning of the second task and the request for its recall to expect forgetting to occur, but given this situation, there will be greater forgetting than would occur for a single list. It is as if the originally learned task interferes with the recall or retention of the subsequently learned task. The magnitude of proactive inhibition increases with time between the learning of the second task and its recall. After 24 hours, we may expect the amount of forgetting produced by proaction to be as great as that produced by retroaction. The combined effects of both retroaction and proaction may be studied by having the subject learn three lists and then subsequently requesting the recall of the second.
Interference theory. The refinements in the interference theory have consisted of increased specification of the mechanisms involved. In order to study these mechanisms, certain fairly standard interference paradigms have been devised. The most widely used is known as the “A-B, A-C paradigm.” This may be visualized as representing a paired-associate arrangement in two successive lists in which the left-hand unit (stimulus term) is identical in both lists and the right-hand unit (response term) differs. With the first list the subject learns to give a response to a specific stimulus, but with the second list he must learn a new response to the same stimulus. When the several items in the two lists have this relationship, interference or negative effects are observed in learning the second list.
Extinction and spontaneous recovery. It is an easy step to assume that the negative effect in learning the second list must somehow be related to the forgetting observed in the retention tests for either list at some later point in time. The evidence points to the fact that something akin to an extinction process occurs in learning the second list. More particularly, the associations developed in learning the first list are extinguished or unlearned in the process of learning the second list. Therefore, if immediately after learning the second list, recall of the first list is requested, serious retroactive inhibition should occur. It does; it is as if the response terms of the first list are not available to the subject. But now, assume that 24 hours elapse before recall of the first list is requested. Retroactive inhibition will occur, but it will be less in magnitude than the amount measured immediately after second-list learning. Independent evidence suggests that following extinction a form of spontaneous recovery occurs, just as in the case of an extinguished conditioned response. Such spontaneous recovery would account for the decreased retroactive inhibition observed after the 24-hour retention interval.
Next, let us examine the role of extinction and spontaneous recovery in proactive inhibition. If following learning of the second list there is a recovery over time of the associations in the first list, these recovered associations should interfere with the recall of the second list. The greater the recovery, the greater the interference; hence, such a mechanism would account for the fact that proactive inhibition increases with time.
The use of extinction and recovery mechanisms in an interference theory of forgetting has further implications. Suppose the second list is learned by distributed practice; that is, we carry out a trial or two, let an interval elapse, give another trial or two, another interval, and so on. These intervals, inserted in learning, should allow some spontaneous recovery of the first-list associations. But the next learning trials should produce some extinction of the first-list associations. Thus, over a series of such learning and rest cycles, we would have successive recovery and extinction cycles. Again, generalizing from work in other areas of learning, we find that such extinction-recovery cycles may lead to more permanent extinction of the first-list associations. Distributed practice on the second list, therefore, should decrease the amount of proactive inhibition. Tests of this proposition are positive. Indeed, when the distribution intervals are long (such as 24 hours), enormous reduction in proactive inhibition will occur. In general, it is fair to say that the use of extinctionrecovery notions, as subprocesses under a general interference theory of forgetting, have proved very useful.
Rapid single-unit technique. Earlier we discussed the technique of studying retention of single items over very short intervals and, in addition, noted that near-complete forgetting had been observed in as short a period as 20 seconds. Experimental evaluations of this phenomenon have now made it fairly reasonable to assume that this rapid forgetting is produced by proactive interference. It is common to use a single subject many times in the single-item studies. The subject is given different units at different times, and retention is tested at various intervals following presentation. Thus, in effect, the subject builds up a repertoire of units as the experiment proceeds. If it is assumed that the previously presented units interfered proactively, the rapid forgetting of the item of the moment may be accounted for. The interpretation seems to hold. Given the appropriate arrangement of conditions, it can be shown that the greater the number of previous items presented, the greater the rate of forgetting the item at hand. The first item presented to the subject shows little, if any, forgetting over time. It seems reasonable to conclude, therefore, that we are not confronted with new principles of forgetting when studying retention of singly presented items over short intervals of time.
Pre-experimental associations. We now come to the final step in the analysis of the interference theory of forgetting. As previously discussed, when the subject learns a single list, forgetting will be approximately 15 to 20 per cent over 24 hours. The facts of retroactive and proactive inhibition produced in the laboratory made it reasonable to suspect that interference also accounted for this 15 per cent to 20 per cent loss. More particularly, it would appear that the interference must arise proactively from sources outside the laboratory. The reasoning is as follows. Assume that the subject is 20 years old at the time he learns this single list. If forgetting the list is to be attributed to interference, it is much more probable that he would have learned “something” that would interfere during these previous 20 years than during the 24-hour retention interval. The latter, of course, would be a source of retroactive interference. The longer the retention interval, the more important would interference from retroactive sources become.
What could be the source of interference from outside the laboratory? The subject has a large verbal repertoire that he has built up during the years. This includes not only a great many verbal units but also associations among those units. It is known, furthermore, that there are varying strengths of associations among letters, and a major reason for the difficulty of learning non-words is that the required associations are contrary to the existing stronger habits or associations. In effect, then, when a subject is required to learn a nonword, he must extinguish, or at least suppress, the associations he brings to the laboratory session. If one were to apply the extinction-recovery theory, it would be expected that during the retention interval the older associations would recover and interfere with those built up in the laboratory, thus producing the forgetting observed at recall. Or, if the materials involved words, pre-experimental associations among those words and with other words would have to be extinguished if the task of the laboratory required new associations. But again, the old habits should recover with time and interfere with recall. Thus the interference theory, as refined and shaped by the facts derived from the studies of retroactive and proactive inhibition in the formal laboratory setting, was given a direct counterpart in the single-list situation, with the interference stemming from habits the subject had learned outside the laboratory. This translation seemed quite appropriate, and the magnitude of the interference to be expected from outside sources could surely account for the 15 per cent to 20 per cent forgetting observed.
To test the application of the theory to the single-list situation requires the construction of learning tasks that will be differentially interfered with by associative habits that the subject brings to the laboratory. This, in turn, requires a knowledge of the associative habits that the subject has, in fact, developed over the years. From various word-association tables and letter-association tables, something is known of the associative network of the average college student. For example, it is known that words occurring with great frequency in our language have more and stronger associations than do words occurring with low frequency. It might be expected, therefore, that if a verbal task were constructed of frequently used words in such a way that the already established habits were inappropriate or contrary to the habits that must be acquired to master the task, an effective proactive situation would be devised. Furthermore, proactive inhibition should be greater than for a task constructed of infrequently used words, since the pre-experimental associations among the infrequently used words are minimal. Clearly, it would be predicted that more of the task constructed of frequently used words would be forgotten; forgetting should be greater because proactive interference after a day, a week, or a month should be greater.
Tests of such expectations have been disappointing. Little, if any, difference in the forgetting of the various tasks has been observed. The same has been true when associations among letters have been used to construct nonwords that should be differentially susceptible to previous habits.
We can see that although the interference theory has been quite successful in accounting for observed facts in the formal retroactive-inhibition and proactive-inhibition paradigms, the translation of this theory to account for the forgetting observed for a single list has not. In study after study we find that 15 to 20 per cent of the task is forgotten over a 24-hour period; we cannot get differential forgetting of any appreciable amount for different materials. The success of the interference theory in other contexts will not allow it to be given up easily. Rather, it would appear that the more fruitful approach would be to continue to devise tasks in which the amount of interference from already established habits would be different. This, in turn, may require further extensive assessment of the habit repertoires of the subjects.
Implications of interference theory. There are further untested implications of the interference theory as outlined above. If proactive inhibition from outside sources is responsible for the forgetting of a single list, we would expect more forgetting for a given task for older people than for younger people. For example, a college student should show more forgetting of a verbal task than should a first-grader. This should follow because the number of potential interfering associations should be greater for the college student. We might also predict that a genius would show more forgetting than an imbecile. The genius has, in his lifetime, learned infinitely more than the imbecile; hence, the sources of potential interference should be greater. Other similar propositions can be deduced from the interference theory. It will probably not be abandoned despite its inadequacy in the single-list situation until a variety of tests of the theory have been negative.
BENTON J. UNDERWOOD
McGEOCH, JOHN A. 1932 Forgetting and the Law of Disuse. Psychological Review 39:352-370.
MELTON, ARTHUR W. 1963 Implications of Short-term Memory for a General Theory of Memory. Journal of Verbal Learning and Verbal Behavior 2:1-21.
PETERSON, LLOYD R.; and PETERSON, MARGARET J. 1959 Short-term Retention of Individual Verbal Items. Journal of Experimental Psychology 58:193-198.
POSTMAN, LEO 1961 The Present Status of Interference Theory. Pages 152-179 in Conference on Verbal Learning and Verbal Behavior, New York University, 1959, Verbal Learning and Verbal Behavior: Proceedings. New York: McGraw-Hill.
UNDERWOOD, BENTON J. 1957 Interference and Forgetting. Psychological Review 64:49-60.
UNDERWOOD, BENTON J.; and POSTMAN, LEO 1960 Ex-traexperimental Sources of Interference in Forgetting. Psychological Review 67:73-95.
It is a common experience to forget what one has learned. Usually, forgetfulness increases with with the length of the retention interval, the time elapsed since the material was last studied or thought about. A graph of the amount remembered (as measured by tests of recall or recognition or relearning) as a function of increasing retention intervals produces a forgetting curve, the slope of which represents the overall rate of forgetting. The first forgetting curve was published in 1885 by Hermann Ebbinghaus, the pioneer in the scientific study of memory. His curve showed the now-familiar monotonic and negatively accelerated form, where the momentary rate of forgetting decreases over time.
Perhaps the earliest and simplest attempt to account for forgetting was the idea of trace decay, which postulated that memorizing something lays down a neurochemical imprint or record in the brain, called a memory trace or engram, whose later reactivation is responsible for remembering. This trace was assumed to fade away spontaneously over time if it was not refreshed by some reacquaintance with the learned material; hence forgetting would ensue. With very few exceptions (noted below), trace-decay theory has been abandoned. This is not only because the hypothetical memory trace has never been identified but also because of subsequent research that undermined the thesis: for example, the findings that forgetting is influenced by other activities taking place both before and after the original learning and that forgetting is also greatly affected by the kinds of cues given at the time of test. Providing different retrieval cues in situations where the trace was supposed to have decayed often results in successful recall of the supposedly forgotten material (Capaldi and Neath, 1995); in many instances memory performance can improve with the passage of time (Bjork, 2001). Thus, although forgetting takes place over time, it is probably not because of some inexorable, autonomous fading of a memory trace.
Forgetting is attributed to decay in only two areas of memory research: pro forma decay parameters in abstract mathematical models of memory and the invocation of an unspecified decay process in certain accounts of short-term memory. The former can be seen as mathematical conveniences rather than as strong theoretical statements. The latter suffer from two major problems: ambiguity in the explication of the decay process (i.e., it is unclear exactly what aspect of memory is decaying, what parts of the memory remain) and the reevaluation of the data supporting such assertions. In principle, decay can be empirically demonstrated if one rules out alternative, better-established causes of forgetting, such as interference. When this is done, the evidence for decay is not appreciable; in the words of Cowan, Saults, and Nugent (2001), "no clear evidence of decay has emerged in, lo, these many years."
One theoretical approach to forgetting that has inspired an enormous amount of experimental effort over several decades is interference theory. As the name implies, it focuses upon forgetting caused by interference. Among its beginnings was the influential demonstration by Jenkins and Dallenbach (1924) that the forgetting of a list of verbal items was markedly reduced if subjects passed the retention interval in sleep rather than in their usual waking activities. This result suggested that the experiences of daily life somehow interfered with the memory of the original material. Eventually, interest narrowed upon other learning experiences as the sources of that interference.
The empirical cornerstones of interference theory are to be found in two kinds of laboratory-produced forgetting. These are schematized in Table 1, and each defines a source of interference by comparing the memory performance of one experimental group, which acquires two lists in succession, with that of a control group, which acquires only one. Retroactive inhibition is the forgetting of the first set of materials, which is caused by the subsequent learning of a second set during the retention interval. Proactive inhibition is the forgetting of the second set, which is caused by the prior learning of a first set. It may seem strange that a preceding list would reduce the memory for a subsequently acquired one, but it can, especially when a retention interval precedes testing. Interference theory construes retroactive and proactive inhibition to be the basic models for its approach to forgetting, inside or outside of the laboratory. It is the task of the theory to devise an experimentally testable description of the processes of retroactive and proactive inhibition.
Whenever two sets of materials are acquired in succession, two processes can impair memory of them. One is the dynamic competition of responses for emergence at the time of the test. To the extent that both sets of materials are activated, a response from the first list may be blocked from consciousness by a stronger competing response from the second list, or vice versa. Two competing responses of equal strength may even block each other. This process is set into motion by the demands of the memory test, and it reduces performances on both lists compared with the single-list control condition. The other process, which takes place during the learning of the second list, consists of the temporary suppression of the contents of the first list, to the extent that they conflict with new response requirements. For example, if the first set of materials contains an A-B association and the second set requires an A-C association, then A-B may be suppressed. With both factors taken into consideration, the memory situation immediately after second-list acquisition is that of a suppressed first list competing against a dominant second list. This correctly predicts that retroactive inhibition will be much stronger than proactive inhibition. Another important observation to be explained is that with a delayed memory test, pro-active inhibition increases in magnitude. This fact is incorporated by postulating a gradual dissipation of first-list suppression, wherein the list regains its strength and competes more effectively, producing increased interference. This implies that a suppressed set of materials should be better recalled after some time has passed, which is the opposite of forgetting. Such an effect has been experimentally verified.
Cue-dependency theory is an alternative approach to forgetting. Although quite different from interference theory, it supplements rather supplants it. Cue-dependency stresses the importance of the reminders, or retrieval cues, that operate at the time of test. It emphasizes that the act of remembering requires not only the stored products of original learning but also an appropriate testing environment in which to make contact with that learning. The successful interaction of stored information and retrieval information produces recollection. Forgetting is proportional to the inadquate accessibility of retrieval cues. Remembering depends on the interaction between the conditions at encoding and the conditions at retrieval.
An experiment reported by Thomson and Tulving (1970) emphasizes the importance of the encoding/retrieval. The researchers asked subjects to recall the target words from a list they were shown. At encoding the target word was sometimes presented by itself, and sometimes with a weak cue. At retrieval there were three cue conditions: Sometimes, no retrieval cue was given, sometimes a weak cue was given, and sometimes a strong cue was given. The terms strong cue and weak cue refer to a cue's ability to elicit a target. For example, if you were asked to respond with the first word that popped into your head when you heard bloom, you are very likely to respond with flower. So bloom is a strong cue for flower. In contrast, you are likely to respond with flower only about 1 percent of the time to the weak cue bloom. The results, shown as the probability of recalling the correct target word, are displayed in Table 2.
When there was no cue at encoding, the weak and strong cues at retrieval functioned as expected: More target items were reported when given a strong cue (0.68) than when given a weak cue (0.43). However, when there was a weak cue at encoding, the weak cue at retrieval elicited the target word more than 80 percent of the time, whereas the strong cue elicited the target word barely 20 percent of the time.
What matters is the extent to which the conditions at retrieval uniquely specify the target information. Thus, if you study in the presence of a particular odor (say, freshly baked chocolate-chip cookies) but are tested in the absence of that odor, your ability to remember is impaired. A change in mood, environment, or pharmacological state can result in similar decrements. Most of the things you have memorized were not accompanied by the smell of chocolate-chip cookies; having the odor at retrieval provides a cue that greatly narrows down the possible items to be recalled. According to this view, the way to overcome forgetting is to envision what cues will be uniquely available at the time the information is going to be needed. Once those cues are identified, then the material can be encoded with reference to those cues.
Are there any conditions of original learning that influence the rate of forgetting of a single set of materials learned in the laboratory? Curiously, normal forgetting rates appear to be independent of the materials memorized. They also seem impervious to the levels of processing (superficial versus meaningful) employed at study. There is little evidence that they yield to deliberate mnemonic strategies. There is even serious doubt about whether the degree of original learning has any effect upon the rate of loss. The present picture suggests a remarkable resistance to experimental manipulation. Surely no theory or model of memory can be expected to account satisfactorily for normal forgetting short of resolution of this question.
Bjork, R. A. (2001). Recency and recovery in human memory. In H. L. Roediger III, J. S. Nairne, I. Neath, and A. M. Surprenant, eds., The nature of remembering: Essays in honor of Robert G. Crowder. Washington, DC: American Psychological Association.
Capaldi, E. J., and Neath, I. (1995). Remembering and forgetting as context discrimination. Learning and Memory 2, 107-132.
Cowan, N., Saults, S., and Nugent, L. (2001). The ravages of absolute and relative amounts of time on memory. In H. L. Roediger III, J. S. Nairne, I. Neath, and A. M. Surprenant, eds., The nature of remembering: Essays in honor of Robert G. Crowder. Washington, DC: American Psychological Association.
Ebbinghaus, H. (1885/1913; 1964). Memory: A contribution to experimental psychology, trans. H. A. Ruger and C. E. Bussenius. New York: Dover.
Jenkins, J. G., and Dallenbach, K. M. (1924). Obliviscence during sleep and waking. American Journal of Psychology 35, 605-612.
McGeoch, J. A. (1932). Forgetting and the law of disuse. Psychological Review 39, 352-370.
Neath, I. (1998). Human memory: An introduction to research, data, and theory. Pacific Grove, CA: Brooks/Cole.
Postman, L., and Underwood, B. J. (1973). Critical issues in interference theory. Memory & Cognition 1, 19-40.
Slamecka, N. J. (1985). Ebbinghaus: Some associations. Journal of Experimental Psychology: Learning, Memory, and Cognition 11, 414-435.
Slamecka, N. J., and McElree, B. (1983). Normal forgetting of verbal lists as a function of their degree of learning. Journal of Experimental Psychology: Learning, Memory, and Cognition 9, 384-397.
Thomson, D. M., and Tulving, E. (1970). Associative encoding and retrieval: Weak and strong cues. Journal of Experimental Psychology 86, 255-262.
Underwood, B. J. (1957). Interference and forgetting. Psychological Review 64, 49-60.
In The Psychopathology of Everyday Life (1901b), Freud discussed forgetting under the rubric of psychosis. The typical example is awareness of having forgotten a proper noun (a name, for example). Like amnesia (where one is unaware that one has forgotten), forgetting is the result of repression. The forgotten name inhabits the preconscious and quickly returns to consciousness. It is attracted by an unconscious mental complex that primarily operates by displacement.
The concept of forgetting in general is present in Freud's earliest works on the theory of neuroses (1894a, 1895b, 1896a). But in "The Psychical Mechanism of Forgetfulness" (1898b) and The Psychopathology of Everyday Life, Freud considered forgetting, like slips of the tongue, to be a parapraxis symptomatic of ongoing repression. To demonstrate the existence of the unconscious, Freud uses the example of forgetting because it was one way of talking about repression before 1900. Forgetting appears in his first theory of neuroses, which explains hysteria as a traumatic infantile sexual seduction that has to be rejected and repressed because the child finds it unacceptable.
Forgetting is associated with a painful sense of awareness (the "name on the tip of the tongue"), while repression is most often unconscious. Forgetting is associated with the psychology of consciousness and the preconscious, while repression is associated with the metapsychology of the unconscious, like memory traces. As a form of parapraxis, forgetting combines partial failure with partial success and must be distinguished from the customary psychological form of forgetting, a successful act of repression.
The dreamer who has forgotten his dream tries to reconstruct it, but in doing so, constructs it anew: "It is indeed possible that while trying to retell it, we fill in the blanks created by forgetting using new material arbitrarily chosen" (Freud, 1900a). We cannot completely remember what is forgotten, and so we prefer to construct likely hypotheses, capable of introducing conviction about what was forgotten (Freud, 1937d).
The person who has forgotten a name, by concentrating on it, only reinforces the ongoing repression. To remember, Freud tells us, we need to abandon the willful attempt to control what initially appears to be a cognitive disturbance, a shortcoming, and give in to the associations that come to mind.
Freud provides an autobiographical example: Forgetting the name of the painter Signorelli during a conversation, he seeks memories, ideas, and words similar to the name. These bring to his mind other paintings with the sensory acuity typical of a screen memory (an early memory used as a screen for a later event), along with the names of other Italian painters (Botticelli, Boltraffio). The value that Freud attributed to the forgotten name had been transferred to neighboring elements, through displacement, as is the case with a mnemonic symbol, which is also a form of metonymy. "Botticelli" is a metonym of Signorelli, "Botticelli" and "Boltraffio" are metonyms of Bosnia-Herzegovina, which Freud was visiting when he forgot Signorelli's name and which is related to the castration complex involved in this forgetting, since Freud attributes to the Turks in Bosnia-Herzegovina a strong attraction to sexuality and a considerable castration anxiety. "Boltraffio" was a metonym of Trafoi, an Italian city where Freud learned of the suicide of one of his patients, which triggered his thoughts on "death and sexuality." The sentence "Herr, was ist da zu sagen? (Sir, what is there to say)" reported to Freud by his interlocutor as reflecting the Turks' attitude toward the inevitability of death, evoked their attitude toward sexuality ("You know very well, Lord, if that fails, then life has no value"), the source of psychic conflict and repression behind his act of forgetting. The representation of death Freud associated with that of castration (the Turkish sentences imply that a life without sexuality is worth no more than death). Moreover, Herr, pre-sent in Herzegovina, refers to Signor (Lord), to the father figure, and to Herz, the heart, an organ likely to grow sick and cause death. Forgetting the name of Signorelli is thus associated with an oedipal dimension that Freud had discovered through his self-analysis: his repression of sexuality, his attraction for his mother, his rivalry with his father, and his ambivalent identification with his father caught up in a desire for parricide and a fear of losing his father.
Freud analyzed two levels at the same time, the psychology of consciousness and the preconscious and the metapsychology of the unconscious. He thus provided an example of the psychoanalytic method, although repression is not associated with the name "Signorelli" so much as the unconscious complex he represents. The names substituted for the forgotten name are composed of verbal memory traces and other proper nouns. They are substituted for the forgotten name through a process that acts on the phonemic material of words (the signifier) through association, metonymy, homology, as well as translation from one language to another, metaphor, and polysemy (Herr has multiple meanings, as does Herz ). In the process of forgetting the name, displacement is metonymy, and condensation is metaphor.
Forgetting, like remembering, belongs more to the phenomenology of consciousness than to the metapsychology of the unconscious. As a specific form of parapraxis, it also signifies repression according to popular convention. Because it occurs in the preconscious and is attracted by the unconscious, forgetting and the rediscovery of the forgotten are similar to what occurs when the subject clearly formulates for himself something he had always known. There have been few developments in psychoanalysis concerning the pre-conscious ego. As a result, it is easier to formulate psychoanalytic approaches that emphasize the cognitive causality of forgetting.
See also: Amnesia; Cryptomnesia; Déjà-vu; Delusions and Dreams in Jensen's "Gradiva" ; Formations of the unconscious; Memory; Moses and Monotheism ; Mythology and psychoanalysis; Slips of the tongue; "Remembering, Repeating, and Working-Through"; Reminiscence; Repression; Psychopathology of Everyday Life The .
Freud, Sigmund. (1894a). The neuro-psychoses of defence. SE, 3: 41-61.
——. (1895b). On the grounds for detaching a particular syndrome from neurasthenia under the description "anxiety neurosis." SE, 3: 85-115.
——. (1896a). Heredity and the aetiology of the neuroses. SE, 3: 141-156.
——. (1898b). The psychical mechanism of forgetfulness. SE, 3: 287-297.
——. (1900a). The interpretation of dreams. SE, 4: 1-338; 5: 339-625.
——. (1901b). The psychopathology of everyday life. SE,6.
——. (1937d). Constructions in analysis. SE,23:255-269.
Bach, Sheldon. (2001). On being forgotten and forgetting one's self. Psychoanalytic Quarterly, 70, 739-756.