Los Siete Pecados de la Memoria: Perspectivas desde la Psicología y la Neurociencia Cognitiva

Y 60 flashcards y 40 preguntas más en el documento.

Texto del documento

The Seven Sins of Memory Insights From Psychology and Cognitive Neuroscience Daniel L. Schacter Harvard University Though often reliable, human memory is also fallible. This article examines how and why memory can get us into trouble. It is suggested that memory's misdeeds can be classified into 7 basic "sins": transience, absent- mindedness, blocking, misattribution, suggestibility, bias, and persistence. The first three sins involve differ- ent types of forgetting, the next three refer to different types of distortions, and the final sin concerns intrusive recollections that are difficult to forget.

Evidence is reviewed concerning each of the 7 sins from relevant sectors of psychology (cognitive, social, and clinical) and from cognitive neuroscience studies that include patients withfocal brain damage or make use of recently developed neuroimaging techniques. Although the 7 sins may appear to reflect flaws in system design, it is argued instead that they are by-products of otherwise adaptive features of memory. Question: If Vernon Jordan has told us that you have an extraordinary memory, one of the greatest memories he has ever seen in a politician, would that be something you would care to dispute? Clinton: No. I do have a good memory. At least I have had a good memory in my life . . .

It's also—if I could say one thing about my memory—I have been blessed and advantaged in my life with a good memory. I have been shocked and so have members of my family and friends of mine at how many things that I have forgotten in the last six years—I think because of the pressure and the pace and the volume of events in a president's life, compounded by the pressure of your four-year inquiry, and all the other things that have happened. When President Clinton testified before Kenneth Starr's grand jury, his numerous lapses of memory prompted in- vestigators to query him about his reputation for prodigious recall.

The logic implicit in their question, later articulated explicitly by Starr in his own testimony to the House committee investigating impeachment charges, seems clear: How could someone with such a seemingly excep- tional memory forget as much as Clinton did about the details of his encounters with Monica Lewinsky? Starr's lawyers were, to put it mildly, suspicious about the self- serving aspects of Clinton's failures to recall potentially damning incidents and statements.

Although their skepti- cism may indeed be warranted, the contrast between Clin- ton's reputation for extraordinary memory on the one hand, and his claims of sketchy recollections for his encounters with Lewinsky on the other, also illustrates a fundamental duality of memory. I have previously referred to this duality as memory's "fragile power" (Schacter, 1996). The power of memory is evident when one contemplates what the various forms of memory make possible in our everyday lives: a sense of personal history, knowledge of facts and concepts, and learning of complex skills.

Because of memory's impor- tance in everyday life, it is easy to see why Vernon Jordan would be struck by Clinton's "extraordinary memory" and how that ability would enhance Clinton's prospects as a politician. But, as Clinton professed to have learned during his term as President, memory also has a darker, more fragile side. People may forget events rapidly or gradually, distort the past in surprising ways, and sometimes experi- ence intrusive recollections of events that they wish they could forget. This darker side of memory has occupied center stage in recent scientific, clinical, and popular discussions.

As most psychologists are acutely aware, a bitter controversy has raged throughout the 1990s concerning the accuracy of recovered memories of childhood sexual abuse (see, for instance, Conway, 1997; Freyd, 1996; Herman, 1992; Kihl- strom, 1995; Lindsay & Read, 1994; Loftus, 1993; Pope, 1996; Poole, Lindsay, Memon, & Bull, 1995; Read & Lindsay, 1997; Schacter, Norman, & Koutstaal, 1997). Some recovered memories have been corroborated and appear to be accurate, but there are also good reasons to believe that many such memories are inaccurate (e.g., Lindsay & Read, 1994; Schacter, 1996; J. W. Schooler, 1994).

False memories of childhood sexual abuse are as- sociated with devastating psychological consequences for accusers and their families (Loftus & Ketcham, 1994; Pendergrast, 1995). As the debate concerning recovered memories has raged, memory researchers have focused increasingly on developing experimental paradigms to ex- plore illusory or false memories in which people confi- dently claim to recollect events that never happened (for Editor's note. Denise C. Park served as action editor for this article. Author's note. Preparation of this article was supported by grants from the National Institute on Aging, National Institute of Mental Health, and Human Frontiers Science Program.

I thank Wilma Koutstaal, Susan McGlynn, and Anthony Wagner for useful comments and discussion, and I thank Sara Greene and Carrie Racine for help with preparation of the article. Correspondence concerning this article should be addressed to Daniel L. Schacter, Department of Psychology, Harvard University, 33 Kirkland Street, Cambridge, MA 02138. Electronic mail may be sent to dls@wjh.harvard.edu. 182 March 1999 American Psychologist Copyright 1999 by the American Psychological Association, Inc. 0003-066X799/S2.00 Vol. 54. No. 3, 182-203 This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Daniel L. Schacter Photo by liza HMS Medio Services review and discussion, see Estes, 1997; Roediger, 1996; Schacter, Norman, & Koutstaal, 1998). Although false memories have been discussed inten- sively in recent years, forgetting is perhaps the most famil- iar of memory's indiscretions. Psychologists and neurosci- entists have studied forgetting ever since Ebbinghaus (1885) applied experimental methods to the study of mem- ory and provided quantitative estimates of forgetting.

The general public, too, has become increasingly concerned with forgetting, even prior to the release of Clinton's forgetting-filled grand jury testimony. As highlighted by a recent cover story concerning memory in Newsweek (Cow- ley & Underwood, 1998), millions of aging babyboomers in addition to Clinton are trying to understand why they now forget more frequently than in the past and what, if anything, they can do about it (e.g.. Crook & Adderly, 1998). We are all affected by memory's shortcomings in our everyday Jives, and scientists have studied them for de- cades.

But there have been few attempts to systematically organize or classify the various ways in which memory can lead us astray and to assess the state of the scientific evidence concerning them. Given the scientific attention paid recently to (he fallibility of memory, and the important real-world consequences that are sometimes associated with forgetting and distortion, such an undertaking would appear to be both timely and potentially useful. 1 suggest that memory's transgressions can be divided into seven basic "sins." I call them transience, absent- mindedness, blocking, misattribution, suggestibility, bias, and persistence. The first three sins reflect different types of forgetting.

Transience involves decreasing accessibility of information over time, absent-mindedness entails inatten- tive or shallow processing that contributes to weak mem- ories of ongoing events or forgetting to do things in the future, and blocking refers to the temporary inaccessibility of information that is stored in memory. The next three sins all involve distortion or inaccuracy.

Misattribution in- volves attributing a recollection or idea to the wrong source, suggestibility refers to memories that are implanted as a result of leading questions or comments during at- tempts to recall past experiences, and bias involves retro- spective distortions and unconscious influences that are related to current knowledge and beliefs. The seventh and final sin, persistence, refers to pathological remembrances: information or events that we cannot forget, even though we wish we could. Like the biblical seven deadly sins—pride, anger, envy, greed, gluttony, lust, and sloth—the seven sins of memory occur frequently in human affairs.

The biblical sins, however, can also be seen as exaggerations of human traits that are in many respects useful and even necessary for survival. So, too, is the case for the seven sins of memory. As annoying and occasionally dangerous as they may be, I suggest later in this article that memory's sins should not be viewed as flaws in system design or unfor- tunate errors made by Mother Nature during the course of evolution. Instead, the seven sins are more usefully con- ceptualized as by-products of otherwise desirable features of human memory (cf. J. R. Anderson & Schooler, 1991; Bjork & Bjork, 1988).

Perhaps paradoxically, then, the seven sins can provide insights into the very operations of memory that make it such a valuable resource in numerous aspects of our everyday lives. In the body of this article, I summarize two major types of evidence and ideas concerning each of the seven sins. First, much of what is known about the seven sins comes from work in cognitive, social, and clinical psychol- ogy; I summarize recent research from each of these do- mains. Second, I consider what we have learned aboui the seven sins from the perspective of contemporary cognitive neuroscience.

During the past 20 years, cognitive neuro- science analyses of human memory have become increas- ingly influential (for general summaries, see Gabrieli, 1998; Gazzaniga, 1995; Schacter. 1992, 1996). The cogni- tive neuroscience approach has relied heavily on studies of patients with brain lesions that selectively affect particular forms of memory and, more recently, on studies using functional neuroimaging techniques, such as positron emis- sion tomography (PET) and functional magnetic resonance imaging (fMRI).

PET and fMRI measure local changes in hemodynamic responses that are correlated with changes in neuronal activity: PET is sensitive to changes in blood flow, whereas fMRI is sensitive to oxygenation-level- dependent changes in the magnetic properties of blood, usually referred to as BOLD contrast. Both techniques allow relatively precise localization of the observed changes in hemodynamic response. To make inferences about (he activation of particular brain regions during per- formance of behavioral tasks, investigators generally mea- sure changes in blood flow or blood oxygenation level in one experimental condition relative to another condition.

Estimates of blood flow or oxygenation level can Ihen be subtracted from one another, or assessed with various other March 1999 American Psychologist 183 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. analysis strategies (for further discussion of the nature and logic of neuroimaging approaches, see Posner & Raichle, 1994). The neuroimaging of human memory has pro- gressed rapidly during the past five years (for reviews, see Buckner & Koutstaal, 1998; Cabeza & Nyberg, 1997; Schacter & Wagner, in press; Ungerleider, 1995).

In the sections that follow, I ask what (if anything) cognitive neuroscience approaches have taught us about each of the seven sins. I then conclude by considering the sense in which, and the extent to which, the seven sins reflect the operation of otherwise adaptive features of memory. Understanding the Seven Sins: Evidence From Mind and Brain Transience Memory for facts and events typically becomes less acces- sible over time. Gradual forgetting was first documented in Ebbinghaus's (1885) well-known studies, where he at- tempted to learn and remember nonsense syllables.

He assessed his own memory at various delays after initial learning and observed a rapid drop-off in retention at the early delays, followed by a more gradual drop-off at later delays. Cognitive psychologists continue to investigate and characterize the form of the forgetting function. Although the basic characteristics of the forgetting curve described by Ebbinghaus have been observed in numerous situations, recent evidence indicates that forgetting over time is best described mathematically with a power function: the rate of forgetting is slowed down by the passage of time (e.g., Wixted & Ebbesen, 1997).

Discussions about the cause of long-term forgetting have focused on whether forgetting is attributable to actual loss of information from memory storage, to retrieval failure that can be reversed by provision of appropriate cues, or both (e.g., Loftus & Loftus, 1980; Schacter, 1996; Squire, 1987). There is no doubt that retrieval failure plays an important role in forgetting.

Some experiences may be rendered temporarily inacces- sible because of interference from related experiences (Postman & Underwood, 1973), and it is well-estab- lished that cues and hints can elicit recall of seemingly forgotten memories (e.g., Tulving & Pearlstone, 1966; for a recent review, see Koutstaal & Schacter, 1997b). Nonetheless, such findings need not indicate that all forgetting is attributable to access failure. The view that experiences are recorded permanently, with all forget- ting attributable to access failure, is surprisingly com- mon— even among psychologists (Loftus & Loftus, 1980).

However, it seems likely that information is also lost from storage over time (Schacter, 1996; Squire, 1987). Loss of information over time may be particularly likely to occur when people do not "use" a memory. It is known, for instance, that retrieving and rehearsing ex- periences plays an important role in determining whether those experiences will be remembered or forgotten (e.g., Bjork, 1988; Koutstaal, Schacter, lohnson, Angell, & Gross, 1998) and in determining what aspects of those experiences will be retained (e.g., Suengas & Johnson, 1988). Memories that are not retrieved and rehearsed may slowly dissipate over time.

Although there is no direct evidence for such dissipation in studies of hu- mans, neurobiological evidence from invertebrate organ- isms has revealed loss of synaptic connectivity over time (e.g., Bailey & Chen, 1989). It has also been established that forgetting can occur quite rapidly—on a time scale of seconds, rather than minutes, hours, or days. Beginning with the classic studies of J. Brown (1958) and Peterson and Peterson (1959), rapid forgetting has been attributed to the operation of a short- term or working memory system (Baddeley, 1986).

Work- ing memory is necessary for holding information "on-line," usually for brief periods of time, while other cognitive activities are carried out. The exact cause of rapid forget- ting within working memory has long been debated (for a discussion of alternative views, see Crowder, 1989) but, as noted below, the evidence suggests that it involves differ- ent mechanisms than long-term forgetting. Cognitive neuroscience analyses have been instru- mental in providing insights into the underlying bases of both gradual (i.e., long-term) and rapid (i.e., short-term) transience.

Beginning with the pioneering studies of Sco- ville and Milner (1957), studies of brain-damaged amnesic patients have shown that damage to the medial temporal lobes, including the hippocampus and related structures, produces profound long-term forgetting. Such patients per- form reasonably well on tasks that tap short-term or work- ing memory (e.g., digit span) and can maintain information in working memory if they are not distracted (Parkin & Leng, 1993; Squire, 1987). However, once distraction or delay are introduced, amnesic patients forget experiences rapidly (anterograde amnesia).

They also exhibit varying degrees of forgetting for experiences that occurred prior to the injury (retrograde amnesia; for alternative perspectives, see Nadel & Moscovitch, 1997; Squire & Alvarez, 1995). However, studies with amnesic patients have not been highly successful in pinpointing the exact source of the pathological forgetting that patients exhibit. Thus, for in- stance, investigations of anterograde amnesia have not yet resolved whether forgetting is attributable specifically to problems in encoding, storage, or retrieval of information, or some combination of these processes (for a recent dis- cussion, see Mayes & Downes, 1997).

Neuroimaging studies are also limited with respect to the kinds of insights they can provide concerning the neural bases of transience. Current neuroimaging techniques do not permit direct study of the storage or consolidation processes that intervene between encoding and retrieval and that are directly related to the occurrence of transience. However, neuroimaging techniques do allow a separation between encoding and retrieval processes.

Thus, studies using fMRI and PET have provided some initial insights into the role of medial temporal lobes in encoding and retrieval processes that has not been possible to obtain in studies of amnesic patients (for review and discussion, see Lepage, Habib, & Tulving, 1998; Schacter & Wagner, 1999). 184 March 1999 American Psychologist This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Two recent fMRI studies illuminate an important source of transience: initial encoding of information into memory.

Both studies used "event-related" fMRI proce- dures, which make it possible to track encoding processes on a trial-by-trial basis. Responses can be sorted later according to whether an item is remembered or forgotten, much in the manner that data from studies using electro- physiological measures, such as event-related potentials, can be selectively sorted and averaged (for discussion of event-related fMRI methods, see Dale & Buckner, 1997). Wagner, Schacter, et al. (1998) used event-related fMRI to determine whether responses to individual words during encoding predict subsequent remembering and for- getting of those words.

During the scanning period, partic- ipants viewed a long list of words and decided whether each word was abstract or concrete. This encoding phase was followed by a nonscanned recognition test in which participants made old-new judgments indicating whether or not they remembered having encountered an item in the study phase and also indicated their degree of confidence in theirjudgment. Analysis of the fMRI data revealed that two regions of the brain were more active during the encoding phase of the experiment for words that were subsequently remembered with high confidence than for words that were subsequently forgotten.

One of these regions was in the posterior (i.e., back) portions of the left temporal lobe (the left parahippocampal gyrus), and the other was in the lower portion of the left frontal lobe (the left inferior frontal gyrus). Thus, level of activity in these two regions at the time of encoding predicted whether an individual word was later remembered or forgotten. Brewer, Zhao, Desmond, Glover, and Gabrieli (1998) carried out a similar event-related fMRI encoding study, using pictures of everyday scenes instead of words. Their results converged nicely with those of Wagner, Schacter, et al. (1998). Brewer et al.

found that degree of activity during encoding in parahippocampal regions of both the left and right hemispheres, and inferior frontal regions in the right hemisphere, predicted subsequent remembering and forget- ting: There was greater parahippocampal and right prefron- tal activity during encoding for subsequently remembered than forgotten pictures.

The fact that right hemisphere regions predicted subsequent remembering and forgetting of pictures, whereas only left hemisphere regions predicted subsequent memory for words, fits with other neuroimag- ing data linking the left hemisphere with verbal encoding and the right hemisphere with nonverbal encoding (Kelley et al., 1998; Wagner, Poldrack, et al., 1998). Gradual forgetting of different types of information thus appears to be attributable, at least in part, to initial levels of activity in regions involved with encoding verbal and nonverbal experiences.

Turning from long-term to short-term transience, neu- ropsychological studies of brain-damaged patients have illuminated the nature of short-term forgetting by demon- strating a pattern opposite to that observed in amnesic patients. Specifically, a number of studies have described patients who show relatively intact long-term memory— they can remember the ongoing experiences of everyday life and perform well on tests of memory that amnesic patients fail—but have severe problems with immediate retention (for a review, see Vallar & Shallice, 1990).

Such patients may exhibit entirely normal performance on such long-term memory tasks as learning pairs of associated words, yet exhibit virtually instant forgetting on such im- mediate memory tasks as digit span. These patients have problems within the working memory system. According to Baddeley (1986), this sys- tem contains a number of interrelated components. In many cases of short-term memory loss, the observed deficit can be attributed to a specific component of working memory known as the phonological loop (see Baddeley, Gathercole, & Papagano, 1998). The phonological loop is necessary for holding small amounts of speech-based information.

It is the type of memory one would rely on when attempting to hold on to a telephone number as one races from the phone book to the telephone. Patients with damage to the phono- logical loop usually have a lesion in the lower part of the left parietal lobe, and they exhibit rapid forgetting of speech-based information. It is important to note that pa- tients with damage to the phonological loop also have difficulty with long-term retention of phonological infor- mation, such as learning new vocabulary (Baddeley et al., 1998).

Developmental studies have shown that perfor- mance on working memory tasks that require the phono- logical loop is closely associated with long-term vocabu- lary acquisition and related aspects of language learning (Gathercole & Baddeley, 1994). Thus, when the phonolog- ical loop does not operate normally (as in cases of left parietal damage or some children with developmental dis- abilities), problems with short-term forgetting arise that have important consequences for such fundamental abili- ties as language learning.

Neuroimaging studies have also begun to illuminate some of the neural systems that are relevant to short-term transience by revealing the component structures of work- ing memory (for reviews, see Awh & Jonides, 1998; Schacter, Wagner, & Buckner, in press; E. E. Smith & Jonides, 1997). To take just one example, neuropsycholog- ical studies suggest that two key components of the pho- nological loop—a phonological store and a rehearsal mech- anism that maintains the contents of this store—are based on distinct neural substrates, because storage and rehearsal processes can be impaired selectively (e.g., Vallar & Bad- deley, 1984).

Neuroimaging studies support this distinc- tion: A number of studies suggest that regions within the left (posterior) parietal lobe subserve phonological storage, whereas portions of left (inferior) prefrontal cortex (Broca's area) are important for phonological rehearsal (e.g., Paulesu, Frith, & Frackowiak, 1993). Short-term for- getting can result from a failure of rehearsal processes, storage processes, or both.

Absent-Mindedness Transience—forgetting over time—can occur even when an event or fact is initially well-encoded and remembered immediately and can occur even when we deliberately search memory in an attempt to recall a specific event or March 1999 American Psychologist 185 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. fact.

However, a good deal of forgetting likely occurs because insufficient attention is devoted to a stimulus at the time of encoding or retrieval or because attended informa- tion is processed superficially. Such incidents of forgetting associated with lapses of attention during encoding or during attempted retrieval can be described as errors of absent-mindedness (e.g., Reason & Mycielska, 1982). Absent-mindedness during encoding is a likely source of common everyday memory failures, such as forgetting where one recently placed an object (e.g., car keys).

Such absent-minded encoding failures occur when actions are carried out automatically and attention is focused else- where (Reason & Mycielska, 1982). Consistent with this observation, cognitive studies have established that divid- ing attention at the time of encoding results in poor sub- sequent memory for target information (e.g., Craik, Go- voni, Naveh-Benjamin, & Anderson, 1996). Likewise, even when attention is nominally devoted to a target item, subsequent memory suffers when that item is initially en- coded at a shallow level. This effect was firmly established in cognitive studies demonstrating the well-known "depth of processing" effect (Craik & Lockhart, 1972).

When people are induced to carry out "shallow" encoding of target information by making judgments about low-level, nonsemantic features of target information (e.g., Is TABLE printed in upper- or lowercase letters?), they later have considerably worse memory for the target (TABLE) than when they are induced to carry out "deep" encoding by making judgments about semantic features of the target item (e.g., Is TABLE a type of furniture?; Craik & Tulving, 1975). A form of shallow encoding also plays a role in recent demonstrations of an intriguing phenomenon termed "change blindness" (for a review, see Simons & Levin, 1997).

In studies of change blindness, people observe ob- jects or scenes in which various features are changed over time. Change blindness occurs when people fail to detect these changes. For example, Levin and Simons (1997) showed participants a movie in which an actor carried out a simple action. Unknown to the observers, the actor was replaced by a different person during the course of the scene. Only one third of observers noticed the change. Even more striking, Simons and Levin (1998) described a naturalistic study in which one of the experimenters asked a person on a college campus for directions.

While they were talking, two men walked between them holding a door that hid the second experimenter. Behind the door, the two experimenters traded places, so that when the men carrying the door moved on, a different person was asking for directions than the one who had been there just a second or two earlier. Remarkably, only 7 of 15 participants re- ported noticing this change! One explanation for change blindness is that people typically encode features of a scene at an extremely shal- low level, recording the general gist of the scene but few of the specific details (Simons & Levin, 1997).

As Simons and Levin (1998) noted, "successful change detection probably requires effortful encoding of precisely those features or properties that will distinguish the original from the changed object" (p. 648). Support for this explanation is provided by a follow-up to the "door study." Simons and Levin (1998) noted that the people who failed to notice that a different person emerged from behind the door were middle-aged and older adults; college students tended to notice the change.

They hypothesized that the older indi- viduals might have encoded the initial (young) experi- menter categorically as a "college student," whereas the college students (for whom the person asking directions was a peer) encoded the experimenter at a more specific level. To determine whether college students would be more susceptible to change blindness when induced to encode at a categorical or generic level, Simons and Levin repeated the "door study," but now attired as construction workers. They reasoned that college students now might tend to encode them categorically as "construction work- ers" and, hence, show higher levels of change blindness.

Results supported the hypothesis: Only 4 of 12 students noticed when a different construction worker emerged from behind the door to ask instructions. Thus, shallow encoding that does not proceed beyond a categorical level results in poor recollection of the details of a scene and consequent vulnerability to change blindness. Some of the neural bases of absent-minded encoding have been elucidated in neuroimaging studies that have compared "deep" and "shallow" encoding tasks.

Several early studies revealed greater activation in the lower (i.e., inferior) regions of the left frontal cortex during semantic encoding trials than during nonsemantic encoding trials (e.g., Demb et al., 1995; Kapur et al., 1994), thereby indicating that level of left prefrontal activation is related to semantic encoding processes. More recent work has indi- cated that the left parahippocampal region discussed earlier also shows greater activation during semantic than nonse- mantic encoding tasks (Wagner, Schacter, et al., 1998).

Thus, we can tentatively infer that "absent-minded" encod- ing operations (at least within a verbal domain) are those that involve relatively little recruitment of left inferior prefrontal and parahippocampal regions. These observa- tions are consistent with the previously discussed event- related fMRI data showing that level of activation in infe- rior prefrontal and parahippocampal regions predicts sub- sequent remembering and forgetting. Absent-mindedness also occurs at the time of re- trieval, when people may forget to carry out a particular task or function.

Because such absent-minded lapses in- volve forgetting to execute a planned action at some point in the future, they are typically referred to as failures of prospective memory (e.g., Brandimonte, Einstein, & Mc- Daniel, 1996; Cohen, 1989). Absent-minded errors of pro- spective memory can have important everyday conse- quences, as when elderly patients forget to take prescribed medications (Park & Kidder, 1996) or when air traffic controllers cannot execute a control action immediately and subsequently forget that they need to take deferred action in the near future (Vortac, Edwards, & Manning, 1995).

Pro- spective memory researchers have found it useful to dis- tinguish between "event-based" and "time-based" prospec- tive memory tasks. Event-based tasks involve remembering 186 March 1999 American Psychologist This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. to perform a future action when a specified event occurs, such as remembering to deliver a message to a friend the next time you see her.

Time-based tasks, in contrast, in- volve remembering to perform an action at a specified time, such as remembering to take one's medicine at 11:00 p.m. or remembering to turn off the burner on the stove five minutes from now. Event-based prospective memory tasks are externally cued, so forgetting tends to occur when a cue is not recognized. Time-based prospective memory tasks, in contrast, depend more on generating appropriate cues at the time an intended action needs to be carried out (Mc- Daniel & Einstein, 1993; Vortac et al., 1995).

Absent- minded forgetting in time-based tasks thus tends to occur because people fail to prospectively generate retrieval cues ahead of time, and then spontaneously fail to do so at the time the intended action needs to be performed. This dis- tinction appears to be particularly relevant to studies of aging memory, which indicate that older adults often per- form well on event-based prospective tasks and more poorly on time-based tasks (for a review, see Einstein & McDaniel, 1996; Maylor, 1996). Cognitive neuroscience has so far contributed rela- tively little to understanding absent-minded errors of pro- spective forgetting.

Shallice and Burgess (1991) have dem- onstrated that failures to carry out future tasks in patients with frontal lobe lesions are associated with level of plan- ning skills and executive functions. Cockburn (1995) in- vestigated a variety of prospective memory tasks in a patient with bilateral frontal lobe damage. The patient performed well on each of five event-based prospective tasks, but performed poorly on several time-based prospec- tive tasks, particularly those tasks that required interrupting one action to perform an unrelated one. In a recent PET study, Okuda et al. (1998) investigated the neural correlates of event-based prospective remembering.

They required participants to perform a routine task (repeating spoken words) during scanning. In a condition that required pro- spective memory, participants were also required to retain a planned action (to tap when prespecified words were presented) while carrying out other activities. Prospective remembering was associated with activation in a number of brain regions, most notably the surface of the right frontal lobe (dorsolateral and ventrolateral regions), the front of the left frontal lobe (frontal pole), and inner parts of the frontal lobe near the midline.

Taken together, the patient studies and neuroimaging experiment implicate prefrontal cortex in aspects of both event-based and time-based pro- spective memory (note, however, that Cockburn's [1995] patient performed well on event-related prospective tasks; it would be important to determine whether this patient performed well on the exact task that produced prefrontal activation in the PET study). This apparent link between prospective memory and frontal lobe function makes sense in view of the role that frontal regions play in allowing "mental time travel" into both the past and the future (Wheeler, Stuss, & Tulving, 1997).

Failure to activate appropriate frontal regions, either at the time of planning a future action or at the moment when it needs to be carried out, may be implicated in absent-minded errors of prospec- tive memory, such as forgetting to keep an appointment or to take a prescribed medicine. Blocking Even when a fact or event has been encoded deeply, and has not been lost over time, it may sometimes be tempo- rarily inaccessible (Koutstaal & Schacter, 1997b; Tulving & Pearlstone, 1966). When people are provided with cues that are related to a sought-after item, but are nonetheless unable to elicit it, a retrieval block has occurred (Roediger & Neely, 1982).

Such blocks occur in both episodic mem- ory (i.e., memory for specific personal experiences) and semantic memory (i.e., general knowledge of the world). Blocking constitutes one of the most subjectively compel- ling of memory's seven sins, in the sense that people are acutely aware of the block at the time it occurs. When blocking occurs at an inopportune moment under high stress—such as with actors who suddenly cannot recall their lines—the accompanying subjective awareness can be overwhelming (Reason & Mycielska, 1982). The most thoroughly investigated example of block- ing is the tip-of-the-tongue (TOT) state.

In a TOT state, people are unable to produce a word or a name, but they have a powerful subjective conviction that the item is available in memory. Further, they can sometimes produce partial phonological or semantic information about the item (R. Brown & McNeil, 1966; for review, see A. S. Brown, 1991). TOT retrieval blocks are often resolved quickly: Several studies have shown that roughly half of the sought- after target items are retrieved within a minute or so after the onset of blocking, although some items may not be retrieved until days later (see A. S. Brown, 1991, pp. 211-213).

TOT blocks appear to be partly attributable to the retrieval of similar but incorrect items that interfere with access to the target (see Harley & Brown, 1998, for other relevant factors). For example, in a diary study, Reason and Lucas (1984) found that over half of naturally occurring TOT states were characterized by the presence of what they termed "ugly sisters," referring to Cinderella's undesirable but dominating older sisters. Ugly sisters are incorrect items that are related to the sought-after target and that recur intrusively during the retrieval attempt.

Consis- tent with this observation, when Jones and Langford (1987) induced TOT states by giving definitions of low frequency words, they found that providing phonologically or seman- tically related cues along with the definitions resulted in more TOT states than when unrelated words were given (see also Smith & Tindell, 1997). Blocking appears to be especially pronounced in old age. The incidence of TOT states increases with aging (A. S. Brown & Nix, 1996; Maylor, 1990), although it remains unclear as to whether older adults are more sus- ceptible to interference from "ugly sisters" than are younger adults (cf. Burke, MacKay, Worthley, & Wade, 1991; A. S. Brown & Nix, 1996).

The age-related increase in TOT states may be particularly pronounced when people attempt to retrieve names (Maylor, 1990). This observation is consistent with other evidence indicating that name re- trieval failure is a frequent subjective complaint by older March 1999 American Psychologist 187 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. adults and that older adults exhibit difficulties retrieving proper names.

However, it is unclear whether these prob- lems are disproportionately worse than age-related retrieval problems for other kinds of information (for review and discussion, see Cohen & Burke, 1993; Maylor, 1997). The foregoing observations concerning the role of "ugly sisters" in the TOT state resemble a curious phenom- enon known as the "part-set cueing" effect, which has been documented in laboratory studies of episodic memory in which participants encode and retrieve lists of words.

In part-set cueing, provision of some retrieval cues that are related to a previously studied word can block or inhibit, rather than enhance, retrieval of the target item (e.g., Roe- diger, 1974; Slamecka, 1968; Sloman, Bower, & Rohrer, 1991). In a similar vein, several studies have shown that retrieving and reviewing an item or event is sometimes associated with decreased memory for related but nonre- trieved items, perhaps because the nonretrieved items be- come inhibited as a result of retrieving related items (e.g., M. C. Anderson, Bjork, & Bjork, 1994; M. C. Anderson & Spellman, 1995).

Although most such evidence comes from word-list learning studies, Shaw, Bjork, and Handal (1995) reported inhibitory effects of retrieval in an eyewit- ness memory paradigm involving more complex events. Participants first viewed color slides of a crime scene (a student's room where a theft had occurred). The experi- menters then questioned them repeatedly about certain categories of objects in the scene (e.g., some of the college sweatshirts that were present), resulting in retrieval and review of these objects. No questions were asked about other categories of objects (e.g., college schoolbooks).

Compared with objects about which no questions were asked (schoolbooks), participants recalled fewer of the nonretrieved and nonreviewed objects from the categories that had been repeatedly probed. Thus, access to these nonretrieved items seemed to be blocked by successful retrieval of related items (for similar results in a paradigm involving review of photographs depicting events that par- ticipants themselves had performed, see Koutstaal, Schacter, Johnson, & Galluccio, in press). Cognitive neuroscience has so far provided some rel- evant observations concerning retrieval blocks in semantic memory and relatively little information concerning block- ing in episodic memory.

Within the domain of semantic memory, a variety of studies have examined anomic pa- tients who have difficulties with retrieval of common names of objects or with retrieval of proper names. Some patients have difficulty retrieving proper names but not common names, and still others exhibit a selective deficit in retrieving one type of proper name (e.g., names of people; for recent reviews, see Hanley & Kay, 1998; Semenza, Mondini, & Zettin, 1995). Semenza et al. (1995) observed that proper name retrieval deficits are typically associated with the most anterior regions of the left temporal lobe (temporal pole).

Consistent with the neuropsychological evidence, a PET study by Damasio, Grabowski, Tranel, Hichwa, and Damasio (1996) revealed activation of the left temporal pole during proper name production. A more posterior temporal region showed increased activation dur- ing animal naming, and an even more posterior region showed increased activation during naming of tools. These observations raise the intriguing possibility that retrieval blocks associated with attempts to name different kinds of items (e.g., individual persons, animals, tools) may reflect inhibition of slightly different left temporal regions.

Al- though no neuroimaging studies of TOT states have yet been reported (and may be difficult to carry out because of the relative infrequency of the TOT phenomenon), such studies could provide novel insights into the neural corre- lates of retrieval blocking. Numerous neuroimaging studies of episodic memory retrieval have been reported (for review and discussion, see Buckner & Koutstaal, 1998; Lepage et al., 1998; Schacter & Wagner, 1999), but none have specifically examined episodic retrieval blocks. Perhaps the most relevant data have been reported by Nyberg et al. (1996), who uncovered evidence of what they termed "ensemble inhibition." Ny- berg et al.

(1996) reported that regions that showed in- creased activity during a retrieval task appeared to actively inhibit other regions showing decreased activity. These observations suggest that, consistent with psychological observations (e.g., M. C. Anderson, Bjork, & Bjork, 1994; M. C. Anderson & Spellman, 1995), inhibitory processes may be a normal and perhaps necessary component of episodic retrieval. Whether and to what extent ensemble inhibition processes are related to retrieval blocks remains to be elucidated.

Misattribution Transience, absent-mindedness, and blocking can all be thought of as sins of omission: At a moment when indi- viduals need to remember, the desired information is inac- cessible or unavailable. However, memory is also charac- terized by sins of commission: Situations in which some form of memory is present, but is misattributed to an incorrect time, place, or person (e.g., Jacoby, Kelley, & Dywan, 1989; Johnson, Hashtroudi, & Lindsay, 1993; Roe- diger, 1996; Schacter, Norman, et al., 1998). I find it useful to distinguish among three closely related forms of misattribution.

First, people may remember correctly an item or fact from a past experience but misattribute the fact to an incorrect source. For instance, individuals sometimes recall encountering a bit of trivia in the newspaper that, in fact, they acquired from an experimenter (Schacter, Harbluk, & McLachlan, 1984). Similarly, people may assert that they saw a face in one context when they encountered it in another (e.g., Read, 1994) or that they perceived an event that they only imagined (e.g., Garry, Manning, Loftus, & Sherman, 1996; Goff & Roediger, 1998; Johnson, Raye, Wang, & Taylor, 1979; for review, see Johnson et al., 1993).

Source confusions of this kind can be particularly pronounced in older adults (Mclntyre & Craik, 1987). In a recent study, for example, Schacter, Koutstaal, Johnson, Gross, and Angell (1997) found that older adults often confused whether they had seen an everyday action in a videotape or only in a photograph that they viewed several 188 March 1999 American Psychologist This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. days later, whereas younger adults had little difficulty remembering the correct source.

Source confusions can have important implications in everyday life, as exemplified by cases of erroneous eye- witness identifications in which a person seen in one con- text is mistakenly "transferred" to another (D. F. Ross, Ceci, Dunning, & Toglia, 1994). A particularly dramatic example involved the psychologist Donald Thomson, a respected memory researcher who was accused of rape on the basis of the victim's detailed recollection of the rapist (Thomson, 1988). Fortunately for Thomson, he had an airtight alibi: He was giving a live television interview (ironically, concerning memory distortion) at the moment that the rape occurred.

The victim had been watching that interview just prior to being raped. She had confused the source of her vivid memory of Thomson, misattributing the television image to the rapist. Thomson's alibi led to his immediate vindication, but others have not been so fortu- nate. Recent investigations into cases of wrongful impris- onment, where innocence was established by DNA evi- dence, provide sobering evidence. In a sample of 40 such cases, 36 (90%) involved false identification of the perpe- trator by one or more eyewitnesses (Wells et al., in press; this finding is also relevant to the discussion of suggest- ibility in the next section).

In the foregoing examples of misattribution, recall of the item or fact is accompanied by a subjective experience of remembering a past event. A second type of misattribu- tion is characterized by an absence of any subjective ex- perience of remembering. People sometimes misattribute a spontaneous thought or idea to their own imagination, when in fact they are retrieving it—without awareness of doing so—from a specific prior experience (e.g., Schacter, 1987). This phenomenon of cryptomnesia is exemplified in everyday life by instances of unintentional plagiarism and has been studied recently in the laboratory (e.g., A. S. Brown & Murphy, 1989; Ceci, 1995; Marsh & Landau, 1995).

A related type of misattribution has been dubbed the "false fame effect" by Jacoby and colleagues (e.g., Jacoby, Kelley, Brown, & Jasechko, 1989). In Jacoby et al.'s ex- periments, participants first studied lists including famous names (e.g., Ronald Reagan) and nonfamous names (e.g., Sebastian Weisdorf). Either immediately or one day after studying the names, participants were a given a "fame judgment" task in which they made famous-nonfamous judgments about previously studied names and new names. On the one-day delayed test, participants frequently clas- sified nonfamous names they had studied a day earlier as "famous," but they hardly ever made the false fame error on an immediate test.

During the delayed test, having forgotten that they studied a nonfamous name such as "Sebastian Weisdorf," participants misattributed the famil- iarity of the name to the "fame" of the nonfamous individ- ual. Older adults are sometimes more susceptible to the false fame effect than are younger adults (cf. Dywan & Jacoby, 1990; Multhaup, 1995). A third type of misattribution occurs when individuals falsely recall or recognize items or events that never hap- pened.

During the 1960s and 1970s, a number of research- ers produced laboratory demonstrations of false recall and recognition in which people claimed to have seen or heard sentences (Bransford & Franks, 1971), words (Underwood, 1965), or dot patterns (e.g., Posner & Keele, 1968) that had not been previously presented. Although a number of re- searchers have since produced impressive demonstrations of similar kinds of misattributions (e.g., Hintzman, 1988), the most striking such finding was reported recently by Roediger and McDermott (1995). They revived and mod- ified a paradigm that was originally devised by Deese (1959).

Deese had reported that when people studied lists of semantically associated words and later tried to recall them, they frequently intruded or falsely recalled a strongly associated word that had not been previously presented. Roediger and McDermott (1995) replicated this false- recall effect and extended it to recognition. In Roediger and McDermott's extension of the Deese task, participants ini- tially studied 15 semantic associates that were all related to a nonpresented "theme word." After studying a number of such 15-word sets, they were then asked to recall words from the list or to recognize them.

For instance, participants might study a list containing the words candy, sour, sugar, bitter, good, taste, tooth, and other related words. They would later receive a recognition test that includes studied words (e.g., taste), nonstudied words that are unrelated to words that had appeared on the study list (e.g., point) and, most important, new theme words (e.g., sweet) that are associatively related to words from the study list.

Roediger and McDermott (1995) reported that participants made false alarms to approximately 65%-80% of the nonpre- sented theme words in various conditions; indeed, false alarm rates to the theme words were indistinguishable from hit rates to words that were actually presented. Participants expressed as much confidence in these false memories as they did in accurate recollections of previously studied words. Roediger and McDermott also asked participants whether they possessed a specific, detailed recollection of having encountered a word on the list (a "remember" response; cf.

Gardiner & Java, 1993; Tulving, 1985), or whether they thought it was on the list because it just seemed familiar to them (a "know" response). Participants provided as many "remember" responses to nonstudied theme words as they did to studied words. This striking false-recognition effect has been replicated and explored in a number of other laboratories (e.g., Mather, Henkel, & Johnson, 1997; Norman & Schacter, 1997; Payne, Elie, Blackwell, & Neuschatz, 1996; Seamon, Luo, & Gallo, 1998). One explanation for this kind of misattribution is that participants are relying on their memory for the general semantic features or "gist" of the items that they studied (cf.

Payne et al., 1996; Reyna & Brainerd, 1995; Schacter, Norman, et al., 1998). Participants may bind together stud- ied items and generated associates, thereby forming and retaining a well-organized representation of the semantic gist of the study list (for discussion of alternative theoret- ical accounts of false recognition, see Roediger, McDer- mott, & Robinson, 1998; Schacter, Norman, et al., 1998). Theme words that match this semantic representation, such March 1999 American Psychologist 189 This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. as "sweet," are likely to be falsely recognized; unrelated words that do not match it are likely to be correctly rejected (for additional evidence on this point, see Mather et al., 1997; Norman & Schacter, 1997). Consistent with this explanation, recent studies have shown that it is possible to reduce or suppress the false- recognition effect when study conditions are created that encourage participants to focus on distinctive properties of individual items.

For example, Israel and Schacter (1997) reported that when participants study Roediger-McDer- mott (1995) lists along with pictures that represent each word, the false-recognition effect is reduced significantly compared with a word-only study condition. Schacter, Is- rael, and Racine (1999) have provided evidence that this reduction of false recognition involves the use of a "dis- tinctiveness heuristic"—a mode of responding based on participants' expectation that recognition of studied items should be accompanied by recollection of distinctive de- tails (i.e., pictorial information).

After studying lists of words that are all accompanied by pictures, participants are less likely to rely solely on semantic gist when making a recognition decision. Instead, they demand access to dis- tinctive perceptual information before they are willing to call an item "old" (for other studies on reducing false recognition of semantic associates, see Gallo, Roberts, & Seamon, 1997; McDermott & Roediger, 1.998). As with previous examples of source misattributions, older adults appear to be especially vulnerable to the kinds of misattributions involved in false recall and recognition.

For example, several studies have demonstrated that older adults show disproportionately high levels of false recall and recognition in the Deese/Roediger-McDermott para- digm (Balota et al., in press; Kensinger & Schacter, in press; Norman & Schacter, 1997; Tun, Wingfield, Rosen, & Blanchard, 1998). Koutstaal and Schacter (1997a) pro- vided a particularly striking demonstration of age differ- ences in false recognition. In their paradigm, younger and older adults studied detailed colored pictures from various categories.

When given a recognition test after a three-day delay, older adults showed considerably higher levels of false recognition to nonpresented pictures from studied categories than did younger adults. The age differences were most pronounced when participants studied large numbers of pictures (18) from a given category, with older adults showing approximately twice as many false alarms (60-70%) as younger adults (e.g., 25-35%). These obser- vations contrast with other studies in which older adults exhibit high levels of picture recognition accuracy that do not differ substantially from that of younger adults (e.g., Park, Puglisi, & Smith, 1986).

In Koutstaal and Schacter's (1997a) paradigm, presentation of numerous perceptually and conceptually similar pictures likely increased reliance on memory for the general features or gist of target items in the older adults compared with younger adults (for further relevant evidence and discussion, see Balota et al., in press; Kensinger & Schacter, in press; Norman & Schacter, 1997; Tun et al., 1998). Cognitive neuroscience analyses have begun to ex- plore aspects of misattribution.

Studies of brain-injured patients have revealed that damage to the frontal lobes is often associated with a selective increase in source memory errors (e.g., Janowsky, Shimamura, & Squire, 1989; Schacter et al., 1984). More recent experiments have also linked frontal lobe damage with increased susceptibility to false recognition (Parkin, Bindschaedler, Harsent, & Metz- ler, 1996; Rapcsak, Reminger, Glisky, Kaszniak, & Comer, in press; Schacter, Curran, Galluccio, Milberg, & Bates, 1996). By contrast, several studies of amnesic patients with damage to the medial temporal lobes and related structures have revealed decreased susceptibility to certain forms of false recognition.

For instance, several experiments using the Deese/Roediger-McDermott paradigm and similar pro- cedures have shown that amnesic patients exhibit reduced levels of false recognition compared with nonamnesic con- trols (e.g., Koutstaal, Schacter, Verfaellie, Brenner, & Jackson, in press; Schacter, Verfaellie, & Pradere, 1996; Schacter, Verfaellie, Anes, & Racine, 1998). Balota et al. (in press) reported similar findings in patients with memory deficits attributable to Alzheimer's disease. Thus, false recognition in these paradigms depends on the same or similar brain regions that are usually associated with veridi- cal recollection.

Medial temporal regions appear to be involved in encoding and retrieving the kinds of semantic gist or similarity information that can support both true and false memories. Several neuroimaging investigations of false recogni- tion have also been reported. Using a modified version of the Deese/Roediger-McDermott paradigm, Schacter, Reiman, et al. (1996) investigated true and false recogni- tion with PET, and Schacter, Buckner, Koutstaal, Dale, and Rosen (1997) did so with fMRI.

The main finding from the two studies is that patterns of brain activity were highly similar during the two forms of recognition, including some evidence of medial temporal lobe activation during both true and false recognition. Differences in brain activ- ity during true and false recognition were relatively small and appeared to depend on specific characteristics of rec- ognition testing procedures (for a discussion, see Schacter, Buckner, et al., 1997; see also see Johnson et al., 1997). Frontal lobe activation was quite prominent in each of the PET and fMRI studies of false recognition.

Indeed, both studies reported some evidence suggesting that frontal re- gions may be involved in strategic monitoring processes that are invoked as participants struggle to determine whether a related lure word was actually presented earlier in a study list (for elaboration and further relevant evi- dence, see Johnson et al., 1997). Such findings fit well with the previously mentioned observation that damage to re- gions within the frontal lobe is sometimes associated with heightened false recognition (Parkin et al., 1996; Rapcsak et al., in press; Schacter, Curran, et al., 1996).

Thus, activ- ity in the frontal lobes may signal an attempt to monitor or scrutinize the output of medial temporal lobe structures. As noted earlier, although medial temporal activity often pro- vides a basis for accurate remembering, it is also implicated in the encoding and retrieval of the gist or general similar- 190 March 1999 American Psychologist This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ity information that sometimes produces false recollections (for further discussion, see Schacter, Norman, et al., 1998).

Suggestibility The foregoing material indicates that false memories can occur spontaneously when a current situation or test item is conceptually or perceptually similar to a previous one. But such illusory memories may also occur in response to suggestions that are made when one is attempting to recall an experience that may or may not have occurred.

Suggest- ibility in memory refers to the tendency to incorporate information provided by others, such as misleading ques- tions (Loftus, Miller, & Burns, 1978), into one's own recollections (note that this definition corresponds closely to Gudjonsson's [1992] notion of "interrogative suggest- ibility," which he distinguishes from various other forms of suggestibility). Suggestibility is closely related to misattri- bution in the sense that the conversion of suggestions into false recollections must involve misattribution. However, misattribution can occur in the absence of overt suggestion. Thus, suggestibility seems appropriately viewed as a dis- tinct sin of memory.

Suggestion can influence memory in several different ways. Perhaps the most familiar example to experimental psychologists comes from the work Of Loftus and col- leagues concerning memory distortions produced by mis- leading postevent information (e.g.,…