Retrieval Without Recollection: an Experimental Analysis of Source Amnesia

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 23, 593-611 (1984)

Retrieval without Recollection: An Experimental Analysis of Source Amnesia

DANIEL L. SCHACTER, JOANNE L. HARBLUK, AND DONALD R. MCLACHLAN

University of Toronto

Clinical observations suggest that patients with organic memory disorders sometimes exhibit the phenomenon of source amnesia: retrieval of experimentally presented information without any recollection of the episode in which it was acquired. To investigate source amnesia experimentally, a paradigm was developed in which either of two experimenters read subjects statements about fictional characteristics of well-known and unknown people; retention of items and sources was tested after varying delays. In Experiment 1, a group of patients with severe memory disorders exhibited source amnesia frequently after retention intervals of just seconds or minutes: On nearly .40 of the trials that they retrieved a target item, patients failed to recollect that either of the sources had imparted it to them. Experiment 2 demonstrated that when normal subjects' level of item recall was equivalent to that of amnesics, they exhibited significantly less source amnesia: Normals rarely failed to recollect that a retrieved item derived from either of the two sources, although they often forgot which of the two experimenters was the correct source. The results are discussed in terms of their implications for theories of normal and abnormal memory. © 1984 by Academic Press. Inc.

In 1911, Clapar6de observed a curious the source of the retrieved information, and phenomenon in a case of organic amnesia instead claimed that it was merely an idea associated with alcoholic Korsakoff's syn- that had passed through her mind by drome. Like other amnesics, Clapar6de's chance. Indeed, the patient flatly denied patient exhibited a profound inability to re- that she had been read a story (1911/1951, call and recognize recently experienced p. 69). events. Clapar6de noted, however, that she We shall refer to the phenomenon exhib- sometimes retained information without ited by Clapar6de's patient as source am- any knowledge of how she had acquired it. nesia (Evans & Thorn, 1966): retrieval of For instance, after he read her a story, the experimentally presented information in patient could sometimes recall details of it the absence of a corresponding recollection when questioned a few minutes later. But of how it was acquired. A number of stu- she failed to recollect that Clapar6de was dents of organic memory disorders have reported observations of source amnesia that are similar to those described by Cla- This research was supported by a Special Research Program Grant from the Connaught Fund, University par6de. MacCurdy (1928), for example, of Toronto. Experiment 2 was conducted as part of a told several Korsakoff patients his name B.A. thesis completed by the second author under the and address. When tested several minutes supervision of the first author. We thank Jason Brandt, later, they chose the correct answers on a Neal Cohen, Morris Moscovitch, Larry Squire, and Endel Tulving for comments and discussion. We also multialternative forced-choice recognition thank Karen Hall and Renata Kirschbaum for exper- task, yet did not remember that MacCurdy imental assistance, and thank Carol A. Macdonald for was the source of the retained information. help with preparation of the manuscript. Joanne Har- Zubin (1948) observed that a patient ren- bluk is now at the Department of Psychology, Univer- dered temporarily amnesic by electrocon- sity of Western Ontario, London, Ontario. Requests vulsive therapy responded with above- for reprints should be addressed to Daniel L. Schacter, Department of Psychology, University of Toronto, To- chance accuracy on a forced-choice test of ronto, Ontario M5S 1A1, Canada. items presented prior to treatment, but

claimed that the materials had never been ploring the extent to which the phenom- shown to her previously. Luria (1976, pp. enon can be produced experimentally in 284-285) described a case of amnesia sec- subjects with intact memory function. ondary to a ruptured anterior communi- Before describing the paradigms that are cating artery aneurysm in which the patient used in the present experiments, it is first could learn a list of picture-word paired necessary to consider exactly what is associates, and could produce the correct meant by the term source amnesia. One of response when given the stimulus 2 days or the central characteristics of the phenom- 1 week after learning. The patient, how- enon is that in a situation in which memory ever, had no knowledge of how he acquired for two attributes of an event is probed, the information, and could not recollect the subjects demonstrate knowledge of one at- learning episode when asked about it even tribute but not the other. The remembered after a brief delay. attribute corresponds to a fact or item that More recently, Schacter, Tulving, and has been presented to the subject; the for- Wang (1981; described in Schacter & gotten one is the source of the recalled in- Tulving, 1982b) provided quantitative doc- formation. It is not uncommon in the study umentation of source amnesia in a case of intact human memory to encounter sit- study of a young man who had developed uations in which subjects are tested for re- a severe memory disorder after closed-head tention of multiple attributes of an event, injury. The patient was asked a series of and recall one or some of the to-be-remem- questions about little-known facts (e.g., bered attributes at the same time that they Who holds the world's record for shaking do not remember others. For example, An- hands? Theodore Roosevelt), and was told derson and Bower (1971) provided evi- the an'swer by the experimenter. Twenty dence that subjects can remember one at- minutes later, the patient recalled or rec- tribute of a sentence (a verb) even though ognized the correct answer for over half of they are unable to recall another (an ob- the items. However, given that he had re- ject). Jones (1976) reported an experiment tained a fact, the probability of remem- in which normal subjects studied photo- bering the source was only .11. On most graphs that depicted an object of a partic- trials, the patient either claimed that he had ular color in a specific location. When guessed the appropriate answer, or had tested by cued-recall methods, subjects oc- read about it in a newspaper or magazine. casionally remembered all or none of the By contrast, a matched control subject re- attributes. Frequently, however, they remembered the source on all questions that called a particular attribute of an object he answered correctly. (e.g., color) and failed to recall another one There is a somewhat dramatic flavor to (e.g., location). Using a similar experi- the cited clinical descriptions and case mental paradigm, Jones (1979) has demon- studies of source amnesia that is probably strated that different attributes of a memory attributable to the fact that patients forget trace can be forgotten at different rates over so quickly the source of the newly acquired the course of a retention interval. information. But these observations of Further evidence for the occurrence of source amnesia have told us little about the partial attribute recall can be found in re- phenomenon other than that it can be ob- cent studies concerned with normal sub- served in some cases. The purpose of the jects' memory for the source of acquired present article is to provide a systematic information. Geiselman and Crawley (1983) experimental analysis of source amnesia by found that presence versus absence of prior examining some of its properties in a group knowledge of a source's personal history of memory-disordered patients, and by ex- did not affect subjects' recognition memory SOURCE AMNESIA 595 of sentences read by the source. Prior mals, they also forget different attributes at knowledge did, however, substantially aid different rates, then it would not be partic- subjects' ability to remember which of two ularly surprising that amnesics sometimes sources had read a particular sentence. The retain an experimentally presented fact and finding that level of source memory forget its source. On closer analysis, how- changed across conditions while level of ever, it is apparent that there may be an sentence memory remained constant sug- important difference between observations gests that there were at least some in- of source amnesia in memory-disordered stances in which subjects recognized a sen- patients and partial attribute recall in nor- tence but did not remember its source. In mals. Consider, for example, the Brown et a study by Rothkopf, Fisher, and Billington al. (1977) study of face recognition. One (1982), subjects watched three sources could claim that subjects in this experiment make a series of statements on either one exhibited "source amnesia": They retained or three television monitors. The number information about an experimentally pre- of monitors did not affect recall of state- sented item, as indicated by accurate face ments, but source recall was substantially recognition, but did not remember how it higher in the three-monitor condition than was acquired in the sense that they could in the one-monitor condition. Again, such not state accurately in which of two rooms a pattern of results implies that subjects a recognized face had been studied. But it sometimes did not remember the source of is probably not unreasonable to assume that recalled information. Similar evidence has these subjects could remember that a rec- also been reported in a study by Brown, ognized face had been encountered in one Deffenbacher, and Sturgill (1977). They of the two contexts. By contrast, the most showed college students photographs of prominent feature of clinical observations faces in two different rooms. Two days of amnesia is that patients seem to have no later, subjects were given a two-alternative recollection at all of the occurrence of a forced-choice face recognition test, and prior learning episode. were also asked to indicate in which of two These observations suggest that it may rooms the face had been presented. Rec- be useful to distinguish between source for- ognition performance was virtually perfect getting and source amnesia. The difference (.96), whereas memory for the presentation between source forgetting and source am- room was only slightly above chance (.58). nesia lies in the type of source errors that Thus, these subjects frequently recognized subjects make when they recall or recog- a particular face without a corresponding nize an experimentally presented item. An memory for the room in which it was en- intraexperimental source error occurs countered. when subjects recall or recognize an item, The foregoing studies establish that when and recollect that it was presented to them normal subjects attempt to remember mul- earlier in the experiment, but attribute it tiattribute events, they often gain access to incorrectly to one of several possible ex- one attribute and do not gain access to an- perimental sources. An extraexperimental other attribute. In view of these findings of source error occurs when subjects re- partial attribute recall in normal memory, it member an item, but fail to recollect that it could be argued that what has been called had been presented by any experimental "source amnesia" represents yet another source, and attribute their knowledge to example of this apparently common phe- guessing or to an extraexperimental source nomenon: Amnesics, like normal subjects, such as radio or a newspaper. We can thus may sometimes retrieve one attribute of a define source forgetting as recall or recog- memory trace but not another. If, like nor- nition of an experimentally acquired item 596 SCHACTER, HARBLUK, AND MC LACHLAN that is accompanied by an intraexperi- ment 1) and in subjects with intact memory mental source error. We define source am- function (Experiment 2). nesia as recall or recognition of an experi- Although the particulars of the experimentally acquired item that is accompanied mental paradigms differ in the two experi- by an extraexperimental source error. ments because of the substantial differ- The literature discussed thus far suggests ences between the subject groups, there are that source forgetting may be characteristic several features that they share in common. of normal subjects, whereas source am- First, to-be-remembered information is im- nesia is observed in memory-disordered pa- parted to subjects by either of two people tients. This suggestion, however, must be who act as experimental sources. Retention regarded cautiously because of the meth- of to-be-remembered items is tested by odological limitations of relevant research. cued recall; retention of the source is as- Studies of normal subjects have not been sessed by asking subjects how they ac- designed to permit examination of the pos- quired the recalled information. Second, sibility that normals do commit extraexper- the experimental materials are constructed imental source errors and hence exhibit in such a way that subjects could not have source amnesia. For example, in the Brown acquired knowledge of a previously pre- et al. experiment, the subjects were explic- sented item from other than an intraexper- itly instructed to indicate the experimental imental source (one of the two experimen- room in which each face had been encoun- ters), yet still could plausibly attribute tered. Under these instructional conditions, knowledge of a recalled item to an extraex- it would not have been possible for subjects perimental source. This was accomplished to make an extraexperimental error, even if by using as to-be-remembered items state- they did not remember seeing the face in ments that describe fictional characteristics either of the two rooms. Similar consider- of people, such as "Bob Hope's father was ations apply to the experiments reported by a fireman." Retention of the item is tested Geiselman and Crawley (1983) and by by presenting the name and asking about Rothkopf et al. (1982): Their subjects were the characteristic (e.g., "What job did Bob instructed to indicate which of the experi- Hope's father have?") A correct answer to mental sources had made a particular state- this question could derive only from an in- ment, and hence they could not commit extraexperimental source. When subjects are traexperimental source errors. Interpreta- tested, however, they also encounter some tion of clinical observations of source previously unpresented lure items con- amnesia is also uncertain because of meth- cerning commonly known characteristics of odological limitations: Clinical demonstra- a person, such as "What country did Win- tions have not used more than one experi- ston Churchill rule" or "What did A1 Ca- mental source, so the only kind of source pone do?" We assume that when subjects errors that patients could make in these sit- are asked to indicate how they know the uations were extraexperimental ones. To answer to these questions, they will re- evaluate the extent to which source am- spond with an extraexperimental source nesia and source forgetting occur in a par- such as "newspapers" or "school." By in- ticular situation, it is necessary to create cluding a number of test lures, we hoped to conditions in which subjects can commit ei- create a situation in which it is plausible for ther intraexperimental or extraexperi- subjects to make extraexperimental as well mental source errors. For the present re- as intraexperimental source attributions. search, we have devised experimental par- A third shared feature of the two exper- adigms that fulfill this requirement. These iments is that subjects are told about char- paradigms make it possible for us to study acteristics of either well-known people both source forgetting and source amnesia (e.g., Bob Hope) or unknown people whose in memory-disordered patients (Experi- names were generated by the experimen- SOURCE AMNESIA 597 ters (e.g., Alice Reznak). This prior knowl- from diverse forms of neurological dysfunc- edge manipulation was included for the pur- tion that are commonly associated with pose of delineating some of the properties memory deficits, including early stages of of source amnesia: We wanted to determine Alzheimer's disease, ruptured anterior whether source amnesia occurs across dif- communicating artery aneurysm, anoxia, ferent types of materials, or whether it is closed-head injury, and encephalitis. The observed only when subjects do or do not mean IQ (88.9) of the amnesics, as mea- have some prior knowledge of to-be-re- sured by the Wechsler Adult Intelligence membered names. Scale (revised), is in the low normal range, whereas their mean MQ (72.8) on the EXPERIMENT 1 Wechsler Memory Scale (WMS) reflects The patients included in this experiment substantial impairments (Table 1). None of are characterized by severe memory dis- the patients is aphasic, anomic, or agnosic. orders (Table 1). For example, most of them The memory task that was used in the are unable to recall more than one or two present experiment is extremely easy for items from a 20-word categorized list on im- people with intact memories. In pilot work, mediate free recall or cued-recall tests. It is we examined the performance of four non- necessary, of course, that patients recall amnesic control subjects, who were some experimentally presented items for matched to the memory-disordered patients there to exist even a possibility of ob- with regard to age and intelligence, and serving source amnesia. Pilot work indi- found that performance in all cases was at cated that presenting patients with lists of ceiling in all conditions (over 90% fact and to-be-remembered items in a standard source recall). In light of this result, and study-test format did not yield levels of because we compared patients' perfor- item recall appreciably above zero. As an mance to that of the normal subjects tested alternative to a study-test procedure, we in Experiment 2, we did not include a sep- developed what we shall refer to as a con- arate set of control subjects in this experi- tinuous recallparadigm. One of the two ex- ment. perimental sources asks the patient a ques- Design and materials. The experimental tion concerning a characteristic of a person, design is a 2 x 2 within-subjects factorial and supplies the correct answer (e.g., What in which the two independent variables are job did Bob Hope's father have? fireman). retention interval and level of prior knowl- The question is asked again by one of the edge of study materials. Retention interval sources after either one or four similar was varied by manipulating the number of questions have been posed; the recurrence items intervening between the first and of the question constitutes the retention second presentations of a test question: In test for both the item and its source. Lure the short-delay condition there was one in- questions concerning well-known charac- tervening item, and in the long-delay con- teristics of well-known people are inter- dition there were four intervening items. spersed among the critical items. Pilot work Two levels of prior knowledge--high indicated that at these extremely short de- knowledge and no knowledge--were used. lays, even severely amnesic patients could Sixteen high-knowledge and sixteen no- retain the answers to some of the questions. knowledge names were used, yielding a total of 32 critical items. High-knowledge Method names were derived individually for each Subjects. The eight patients selected for patient. This was accomplished by asking the experiment are all characterized by se- patients about approximately 30 famous en- vere difficulties in the storing and retrieving tertainers and politicians in a separate ses- of new information. As indicated by Table sion conducted about 6 months prior to the l, the memory disorders in the group result experiment. The patients were required to 598 SCHACTER, HARBLUK, AND MC LACHLAN

TABLE 1 CHARACTERISTICS OF MEMORY-DISORDERED PATIENTS IN EXPERIMENT 1

Education Patient Diagnosis Age (years) WAIS-R° WMS b

1 Alzheimer 58 13 86 62 2 Alzheimer 61 21 92 79 3 Alzheimer 60 17 90 74 4 Aneurysm 58 13 89 79 5 Anoxia 53 10 83 71 6 Head injury 50 8 88 74 7 Encephalitis 30 12 82 61 8 Undiagno sed 70 11 101 82 Mean 55.0 13.1 88.9 72.8

Wechsler Adult Intelligence Scale (revised). Wechsler Memory Scale.

state as many facts as possible about each and long-delay conditions. Each of the 32 of the names; the 16 names about which name-characteristic combinations appeared they generated the largest numbers of facts twice during the task in the form of a test were used in the present experiment. Pa- question. Half of the questions in each of tients generated an average of 4.56 facts the four experimental conditions were about each high-knowledge name. Sixteen asked initially by Source A, and half were fictitious names of both sexes were gener- asked initially by Source B. When the ques- ated by the experimenters for the no- tion recurred, on half the trials it was asked knowledge condition. Thirty-two fictitious by the same source who asked it initially, characteristics of people were also gener- and on the other half it was asked by the ated by the experimenters, with the con- other source. In all cases, patients were straint that they would not be specific to first told to try to provide the appropriate sex or to known attributes of the high- characteristic in response to the question. knowledge names, and hence could be When they could not do so, they were then paired with any name on the list. In addi- required to state whether or not the question to the critical items, 10 questions were tion had been put to them previously by included as lures. Questions were selected either of the experimental sources. If pa- to be relatively easy for patients to answer tients said "No," the answer to the ques- and to provide a variety of source re- tion was then provided by the experimenter sponses other than the two experimenters. who had just posed it. If patients said The lure questions were interspersed ran- "Yes," they were then required to state domly among the targets and appeared only which of the two sources had asked them once. The combination of the 10 once-presented buffer items and the 32 twice-presented targets yielded a total of 74 ques- TABLE 2 tions on each list. Examples of materials EXAMPLES OF QUESTIONS USED 1N THE CONTINUOUS are presented in Table 2 in the form of a RECALL TASK 1N EXPERIMENT 1 typical experimental sequence of six ques- 1. What job did Bob Hope's father have? (fireman) tions. 2. What did A1 Capone do? (gangster) Four forms of the continuous recall task 3. What is Alice Reznak addicted to? (nicotine) were created so that each characteristic ap- 4. What sport did Babe Ruth play? (baseball) peared equally often with a high-knowledge 5. What is Alice Reznak addicted to? (nicotine) 6. What job did Bob Hope's father have? (fireman) and no-knowledge name in the short-delay SOURCE AMNESIA 599 the question previously, and they were then then repeated it in the form of a complete told the answer to the question. sentence ("Bob Hope's father was a When patients guessed the appropriate fireman."). Questions concerning the characteristic on the first appearance of a source of the response, and the prior oc- question, they were told that they were in- currence of the question in the experiment, correct, and a second, "backup" answer were asked in the manner described in the was then introduced as the correct re- previous section. The entire procedure sponse. This procedure was adopted to took about 30 s per item, although it was counter the possibility that apparent item several seconds shorter if patients provided retention on the second appearance of a a rapid correct response to a buffer or to a question could be attributed to guessing. second appearance of a critical item. When patients provided the correct char- Results acteristic on the second appearance of a question, they were asked to indicate the Item recall. All of the memory-disor- source of their knowledge. If they said that dered patients were able to recall some of one of the experimenters had told them the the experimentally presented items. Com- item, they then indicated which experi- bined across experimental conditions, the menter had done so; if they claimed to be proportion of items recalled by individual guessing, or confabulated a source, they patients ranged from. 16 to .50. Moreover, were asked to state whether or not they had all of the patients recalled at least one item heard the question at any point in the ex- in each of the four experimental conditions, periment. with the exception of one patient who did Procedure. Patients were informed that not recall any in the no-knowledge, long- they would be asked a series of questions delay condition. Incidence of correct about some well-known and some unknown guesses on the first appearance of a ques- people. They were also told that they would tion was relatively low: There were only probably know the answers to some of the eight cases, distributed among five different questions, but that in many cases, the an- patients, across all experimental condi- swers were bits of trivia that most people tions. did not know. It was explained that when Both of the experimental variables had they could not answer the question cor- some influence on level of item recall. As rectly, the experimenter who had asked it indicated by the data in Table 3, recall of would provide the right response. They the fictional characteristic to high-knowl- were instructed that their task was to re- edge names was higher at the short delay member the characteristic of the person (.39) than at the long delay (.30). Recall to and who had told it to them, because many no-knowledge names was slightly lower, of the questions would be asked again a few and in this condition, too, recall at the short minutes after initial presentation. delay (.33) exceeded recall at the long delay All testing was done individually. The (. 17). Analysis of variance revealed a sig- two experimental sources (one male and nificant main effect of delay, F(1,7) = 14, one female) were seated at opposite ends p < .01, MS e = .57; the main effect of prior of a testing table across from the patient, knowledge fell short of statistical signifi- who sat midway between the two sources cance, F(1,7) = 2.52, p > .05, MS e = 1.79. and faced them. Patients were given about There was a nonsignificant interaction be- 15 s to respond to each question, and were tween prior knowledge and delay, F(1,7) = encouraged to guess. When an experi- 1.4, p > .05, MS e = .5. menter provided the answer to a question, Level of item recall was also analyzed he or she first stated the characteristic according to whether the same source (e.g., "The answer is a fireman."), and asked a particular question twice, or 600 SCHACTER, HARBLUK, AND MC LACHLAN whether different sources asked the ques- sented items. The data in Table 4 indicate tion on the two trials. Overall item recall that even when patients did accurately re- was virtually identical when test questions call a characteristic of a person, proportion were posed by the same source (.29) and of correct source recall was still low, when they were posed by different sources ranging from .33 to .45 across individual ex- (.31). perimental conditions. The observation Source recall. Because two experimental that source recall given item recall was sources were used in this experiment, pa- slightly higher at the long delay than at the tients could achieve .50 source recall on a short delay is probably attributable to vari- chance basis if they remembered that a par- ability associated with the small number of ticular question had been asked previously observations per condition, as indicated by by one of the experimenters. However, the values in parentheses in Table 4. Recall overall level of source recall was consid- of the target item accompanied by an ex- erably lower than .50 in all experimental traexperimental source error--that is, conditions (Table 3). In view of the low source amnesia--occurred frequently in all level of performance, it is not surprising conditions of the experiment. On about .35 that the experimental variables did not sig- to .55 of the trials on which they supplied nificantly influence level of source recall: the correct answer to a question, patients There were nonsignificant main effects of did not remember that the information had both retention interval, F(1,7) = 3.94, p > been imparted to them by one of the ex- .05, MS e = 1.53, and prior knowledge, perimenters. Source amnesia occurred F(1,7) = 1.87, p > .05, MS e = .82, as well about as often when a characteristic was as a nonsignificant interaction between recalled concerning a high-knowledge name these variables, F(1,7) = .66, p > .05, MSe as it did when a characteristic was recalled -- .42. The low level of source recall by the about a no-knowledge name. Moreover, in memory-disordered patients is attributable each of the four experimental conditions, to the fact that they made many extraex- three to five patients made extraexperi- perimental source errors. Combined across mental source responses for all of the items levels of the two experimental variables, that they recalled accurately, and every pa- patients made .29 correct source responses, tient exhibited source amnesia on at least .26 intraexperimental source errors, and .45 one question. extraexperimental source errors. Clinical observations of source amnesia The critical data in the experiment con- suggest that patients frequently confabulate cern patients' source responses condition- source responses. Across all experimental alized upon recall of experimentally pre- conditions patients confabulated sources on .30 of the trials on which they recalled an item and committed an extraexperi- TABLE 3 PROPORTIONS OF ITEM RECALL AND SOURCE RECALL mental error. For example, one amnesic AS A FUNCTION OF RETENTION INTERVAL AND PRIOR who recalled the correct answer ("cats") KNOWLEDGE IN EXPERIMENT 1 when asked for the second time "What was Item recall Source recall Henry Fonda allergic to?", claimed that she had acquired this bit of knowledge from One-item delay a newspaper article. On .70 of the extraex- High knowledge .39 .36 No knowledge .33 .28 perimental source responses, patients ei- Mean .36 .32 ther claimed to be guessing, or said that Four-item delay they had "deduced" or "figured out" the High knowledge .30 .28 recalled characteristic on the basis of their No knowledge .17 .25 knowledge of the person. Amnesics did, Mean .24 .27 however, make appropriate extraexperi- SOURCE AMNESIA 601

TABLE 4 PROPORTIONS OF SOURCE RESPONSES CONDITIONALIZED UPON ITEM RECALL AS A FUNCTION OF RETENTION INTERVAL AND PRIOR KNOWLEDGE IN EXPERIMENT 1

Intraexperimental Extraexperimental Correct source error error

One-item delay High knowledge (25) a .36 .28 .36 No knowledge (21) .33 .29 .38 Mean (46) .35 .28 .37 Four-item delay High knowledge (19) .37 .26 .37 No knowledge (11) .45 .00 .55 Mean (30) .40 .17 .43

a The raw numbers of observations per condition are in parentheses. mental source attributions concerning the Analysis of the source recall data as a lure items that probed well-known charac- function of same versus different sources teristics of well-known people. All patients on study and test trials revealed that overall answered all of the ten lure questions cor- proportion of extraexperimental source re- rectly. In every case, they attributed their sponses did not change when sources were knowledge appropriately to sources such as the same (.47) or different (.43). However, school, television, radio, newspapers, and given an intraexperimental source re- so forth. sponse, there was a tendency for amnesics' Although patients frequently exhibited source identification to appear more accu- source amnesia, they were able to re- rate when the source was the same (.74) member that recalled items derived from versus when it was different (.42). In view one of the experimental sources on close to of the finding that overall accuracy of .60 of the questions across all experimental source identification did not exceed the conditions (Table 4). If we exclude those chance level, this trend probably reflects a cases in which patients attributed a recalled response bias to state that the person item to an extraexperimental source, it is asking the question at test also posed it ear- possible to evaluate their level of source re- lier. call with respect to the chance baseline of .50. Conditionalizing the data so that we Discussion consider only intraexperimental source re- The results of Experiment 1 provide ev- sponses, probability of choosing the correct idence that substantial amounts of source source at the short delay was .56 to high- amnesia can be observed in a group of knowledge names and .58 to no-knowledge memory-disordered patients under con- names. At the long delay, the corre- trolled experimental conditions in which it sponding numbers were .54 and 1.0. The is possible to commit both intraexperi- last value should not be considered too se- mental and extraexperimental source er- riously, however, because it is based on rors: After a delay of just seconds or min- only five observations. When the data are utes, the amnesics frequently retained an- collapsed across experimental conditions, swers to questions in the absence of probability of remembering the correct recollection of the occurrence of a prior source of a recalled item, given an intraex- learning episode. These data corroborate perimental source response, is .61. This and extend previous clinical observations proportion does not significantly exceed and case studies. However, contrary to the chance level, Z = 1.47, p > .05. most clinical observations, we found that 602 SCHACTER, HARBLUK, AND MC LACHLAN source confabulation occurred relatively between study and test. The results of such infrequently. Moreover, the data also indi- studies indicate that a variety of phe- cate that not all of patients' poor source nomena observed in amnesic patients can memory is attributable to source amnesia be simulated by testing normals under con- (i.e., extraexperimental errors); they exhib- ditions of degraded memory (Mayes & ited appreciable amounts of source forget- Meudell, 1981a, 1981b; Mayes, Meudell, & ting (i.e., intraexperimental errors). Som, 1981; Squire, Nadel, & Slater, 1981; Although Experiment 1 does permit us to Squire, Wetzel, & Slater, 1978; Tulving, conclude that source amnesia can be ob- Schacter, & Stark, 1982; Woods & Piercy, served under specifiable conditions, it 1974; see Schacter & Tulving, 1982a, and leaves open the question of how to interpret Weiskrantz & Warrington, 1975, for discus- the theoretical significance of the phenom- sion of methodological issues). enon. A basic issue that needs to be ad- In Experiment 2, we investigated the hy- dressed is whether source amnesia consti- pothesis that source amnesia can be attrib- tutes a qualitatively distinct characteristic uted to generally weak or degraded of patients with severe memory disorders, memory in a simulation study in which or whether it can be produced in normal normal subjects were tested for recall of ex- subjects by appropriate experimental ma- perimentally presented items and their nipulations. It was suggested in the intro- sources at short and long delays after study. duction that source forgetting, but not We assume that testing subjects at 1-week source amnesia, may be characteristic of delay will produce degraded memory in the normal subjects. One way to address this sense that level of item recall should be hypothesis is to attempt to produce both much lower after a 1-week retention in- source amnesia and source forgetting in a terval than on an immediate test. Assuming laboratory simulation study with normal that the expected decline in item recall is subjects. The question of interest in a lab- observed, the hypothesis that source am- oratory simulation study is whether pat- nesia can be attributed to generally de- terns of performance observed in memory- graded memory would receive support if disordered patients are unique features of two further experimental outcomes are ob- amnesia, or whether they also can be pro- tained: (1) Source amnesia occurs more duced in normal subjects when memory is often after a long retention interval than "weakened" or "degraded" by appro- after a short one and (2) normal subjects priate manipulations. The logic that moti- exhibit source amnesia as frequently as do vates such an approach is straightforward: memory-disordered patients when level of ff normals can be made to exhibit a phe- item recall is similar in the two groups. nomenon of amnesia under conditions of degraded memory, then it is difficult to EXPERIMENT 2 argue that the simulated phenomenon con- Experiment 2 is similar to Experiment 1 stitutes a qualitatively unique feature of am- in several respects: Two experimental nesia. Rather, such a finding suggests that sources read to-be-remembered informa- the phenomenon may be observed when- tion to subjects, many of the same materials ever a generally weak or degraded memory were used, and subjects were tested for re- trace is produced. Degraded memory can tention of both item and source in the same be defined operationally in terms of a dec- manner as in Experiment 1. The experi- rement in performance on recall or recog- ments differed, however, in four ways. nition tests as a function of an experimental First, we used a study-test paradigm in this manipulation. In most laboratory simula- experiment because the continuous recall tion studies, degraded memory has been paradigm used in Experiment 1 is far too produced by interpolation of a long delay easy for normal subjects. Instead of testing SOURCE AMNESIA 603

subjects at one- and four-item delays, they infer that the item derived from an intraex- were tested either 20 min or 1 week after perimental source. study. Second, to obtain a relatively large number of observations per condition, we Method exposed subjects to a larger set of to-be- Subjects. Twenty-four University of To- remembered materials than was used in Ex- ronto undergraduates participated in the periment 1. Third, we included three, experiment to fulfill class requirements. rather than two levels of the prior knowl- Subjects were tested in small groups of edge variable to gain further insight into the 2-5. consistency of source amnesia and source Design and materials. The design of the forgetting across different types of mate- experiment was a 2 (retention interval) x rials. Fourth, we used slightly different pro- 3 (prior knowledge) mixed factorial. The cedures to create a situation in which it is between-subjects variable was retention in- plausible for subjects to make extraexperi- terval (20-min or l-week delay); the within- mental source errors. As in Experiment 1, subjects variable was level of prior knowl- we included lure questions concerning well- edge of the names that constituted part of known characteristics of well-known the to-be-remembered information (high, people at the time of test to encourage sub- low, or no). jects to make extraexperimental source re- Fifty-four fictitious characteristics of sponses. We also included similar buffeer people were included on the study list. statements on the study list that were not Thirty-two characteristics were the same as subsequently tested (Table 5), so that sub- those used in Experiment 1. The others jects could not infer that the appearance of were generated by the experimenters. a test question probing a well-known fact Level of subjects’ prior knowledge of the about a well-known person signaled an ex- to-be-remembered names was determined traexperimental source response. In addi- in a separate norming study in which 13 tion, we also included test lures that probed University of Toronto undergraduates were fictitious characteristics of both fictitious given a list of 70 well-known entertainers and well-known people (Table 5). These and politicians and were asked to generate lures were included so that subjects could as many facts as they could about each not use the appearance of a test question person. One minute was provided to gen- that probed a fictitious characteristic to erate facts about each of the 70 people.

TABLE 5 EXAMPLES OF STUDY AND TEST MATERIALS USED IN EXPERIMENT 2 Type of item Number Study Number Test High knowledge 18 Jane Fonda refuses to 18 What does Jane Fonda refuse eat chicken. to eat? Low knowledge 18 Leonard Bernstein has never 18 What has Leonard Bernstein owned a television never owned? No knowledge 18 Alice Reznak is addicted to 18 What is Alice Reznak addicted to? nicotine. Buffer 18 Agatha Christie writes mystery - - novels. Lure - - 24 What was Pablo Picasso’s profession? Lure - - 12 What does Frank Sinatra read every night? Lure - - 12 What does Anna Wood grow in her orchard’? 604 SCHACTER, HARBLUtL, AND MC LACHLAN

Names for which one or more people were people. They were instructed to try to re- unable to generate any facts were excluded member the name-item associations, as from consideration. Of the remaining items, well as who told them each item. The input the 18 names about which subjects gener- list was read at the rate of 6 s per item. ated the most facts were designated "high Following list presentation, all subjects knowledge," and the 18 names about which crossed out specified numbers from a nu- subjects generated the fewest facts were merical array for 10 rain, and then gener- designated "low knowledge." Mean ated information about different cities for number of facts generated about high- an additional 10 min. Long-delay subjects knowledge names was 5.50, and that about were then dismissed and told to return to low-knowledge names was 2.37. The 18 no- the laboratory a week later, whereas short- knowledge names included those that were delay subjects proceeded immediately to used in Experiment 1, and several others the retention test. that were generated by the experimenters. The test phase was initiated by distrib- Input lists were constructed by dividing uting response booklets. Both sources were the 54 fictitious characteristics into three present during test. Subjects were encour- sets of 18. Items in each set were randomly aged to produce an answer in response to paired with either high-, low-, or no-knowl- as many questions as possible, even if they edge names in both the short- and long- had to guess. In addition, they were in- delay groups. Across subjects, each char- structed to indicate the source of their re- acteristic was read equally often by each sponse for every question. It was stressed source at high, low, and no levels of prior that the information required to answer the knowledge. Eighteen buffer items were also questions would derive from a variety of included in the input list, consisting of well- sources, not from just the two experimen- known names paired with well-known ters. Examples of possible source re- facts. The order of the input lists was sponses were given (e.g., television, random with the constraint that no more school, friends, and so on), and subjects than three of the same type of item (high were told that if they were not sure of the knowledge, low knowledge, no knowledge, source of their response, they should indi- or buffer) occurred sequentially. cate that they were guessing. In addition, The test comprised 102 questions of the subjects were instructed that if they could form displayed in Table 5. The order of not recall the answer to a question, but did testing the 54 items that had been previ- remember that the item had occurred ously studied was random with respect to during the study session, then one of the input order. Forty-eight lure questions were experimenters should be chosen when they also randomly interspersed throughout the made their source responses. test, for the reasons described earlier. After completion of the test, subjects Twenty-four lures probed well-known facts were debriefed concerning the nature of the about well-known people, 12 probed ficti- materials and the purpose of the experi- tious facts about well-known people, and 12 ment. involved fictitious facts about fictitious people. Results and Discussion Procedure. Two female experimenters Item recall. There are two important fea- sat at opposite ends of a large table; sub- tures of the item recall data that are pre- jects sat facing the two sources about sented in Table 6. First, level of item recall halfway between them. Subjects were told declined substantially as a function of that each of the two experimenters would delay: Subjects tested at the 1-week reten- read some well-known and little-known tion interval recalled fewer of the fictitious characteristics of famous and not so famous characteristics (.31) than did subjects tested SOURCE AMNESIA 605

TABLE 6 levels of prior knowledge, F(2,44) = 3.48, PROPORTIONS OF ITEM RECALL AND SOURCE RECALL p < .05, MS e = .05. There was a nonsig- AS A FUNCTION OF RETENTION INTERVAL AND PRIOR nificant retention interval x prior knowl- KNOWLEDGE IN EXPERIMENT 2 edge interaction, F(2,44) = .73, p > .05, Item recall Source recall MS e = .05. 20-rain delay The most striking feature of the data in High knowledge .73 .79 Table 7 is that normal subjects committed Low knowledge .56 .71 an extremely small proportion of extraex- No knowledge .36 .56 perimental errors when they recalled to-be- Mean .55 .69 remembered items at both the short and l-week delay High knowledge .42 .55 long delays: Subjects made .03 extraexper- Low knowledge .30 .55 imental source responses at the short delay, No knowledge .19 .55 and .06 extraexperimental source re- Mean .31 .55 sponses at the long delay. By contrast, there was a substantial increase in the number of intraexperimental source errors at the 20-minute retention interval (.55). as a function of delay: Subjects committed Analysis of variance revealed a highly sig- • 12 intraexperimental errors on recalled nificant main effect of retention interval, items at the short delay, and .36 at the long F(1,22) = 25.39, p < .001, MS e = 1.13. delay• Statistical evaluation of the differ- There was also a significant maip effect of ences between proportions of source errors prior knowledge, F(2,44) = 48•84, p < made at short and long delays was based .001, MS e = .55, and a nonsignificant re- upon a nonparametric method for compar- tention interval x prior knowledge inter- ison of two proportions (Bennett & action, F(2,44) = 2.93, p > .05, MS e = .03. Franklin, 1964, pp. 611-615), because The second important result is that level of many subjects did not contribute data to all item recall at the 1-week delay (.31) was conditions. This analysis revealed that the equivalent to the overall level of item recall proportion of extraexperimental source er- attained by amnesics in Experiment 1 (•30). rors did not increase significantly as a func- In addition, inspection of individual exper- tion of delay (p > .05), whereas the pro- imental conditions in Tables 3 and 6 indi- portion of intraexperimental errors did (p cates that normals' item recall to high- < .01). knowledge names at i-week delay (.42) was Comparison of the proportion of extraex- similar to patients' item recall in the cor- perimental source responses made by long- responding condition at one-item delay delay normals and by amnesic patients at (.39), and that normals' delayed item recall similar levels of item recall reveals sizable to no-knowledge names (.19) was equiva- and consistent differences. Collapsed lent to amnesics' recall to no-knowledge across levels of the independent variables names at four-item delay (. 17). in Experiment 1, patients committed .39 ex- Source recall. The crucial data in the extraexperimental source errors on recalled periment concern the distribution of sub- items, whereas long-delay normals, whose jects' source responses when they recalled level of item recall was equivalent to that a to-be-remembered item (Table 7). Given of the amnesics, committed only .06 ex- item recall, there were fewer correct source traexperimental source errors. When eval- responses at the long delay than at the short uated by the Bennett and Franklin test, the delay, as indicated by a main effect of re- differences between these two proportions tention interval, F(1,22) = 14.03, p < .01, is highly significant (p < .01). The low pro- MS e = .08. There were also significantly portion of extraexperimental errors made fewer correct responses with decreasing by long-delay normals is observed consis- 606 SCHACTER, HARBLUK, AND MC LACHLAN

TABLE 7 PROPORTIONS OF SOURCE RESPONSES CONDITIONALIZED UPON ITEM RECALL AS A FUNCTION OF RETENTION INTERVAL AND PRIOR KNOWLEDGE IN EXPERIMENT 2

Intraexperimental Extraexperimental Correct source error error

Short delay High knowledge (158) ~ ,91 .09 .00 Low knowledge (122) ,85 .13 .02 No knowledge (77) ,74 .18 .08 Mean (357) .85 .12 .03 Long delay High knowledge (90) .62 .32 ,06 Low knowledge (64) .55 .40 .05 No knowledge (40) .58 .32 .10 Mean (194) .58 .36 .06

a The raw numbers of observations per condition are in parentheses. tently at each of the three levels of prior The data in Table 7 indicate that when knowledge, and is significantly lower than long-delay normals recall an item, proba- the corresponding proportion made by am- bilities of a correct source response are .62, nesics in any individual experimental con- .55, and .58 at the three levels of prior dition in which item recall of the two groups knowledge. All of these values are consid- is similar. Inspection of individual subject's erably higher than the corresponding pro- data revealed that only 5 of the 12 long- portions of correct source responses made delay normals committed an extraexperi- by memory-disordered patients in any ex- mental error when they recalled a target perimental condition (Table 4). The pre- item. By contrast, all of the memory-dis- vious analyses indicated that the reason for ordered patients committed an extraexper- this difference is the differential incidence imental error on a recalled item in at least of extraexperimental source errors made by one condition. the two groups. To compare level of source Somewhat surprisingly, normal subjects forgetting (i.e., proportion of intraexperi- demonstrated an apparently greater ten- mental errors) in normals and amnesics, we dency to confabulate sources when they must exclude cases in which subjects made made extraexperimental errors on recalled extraexperimental errors so that we can items than did amnesics. At the short delay, evaluate source recall with respect to the five of the extraexperimental source re- chance baseline of .50. When the data are sponses were "guesses" and four were conditionalized in this manner, probability confabulations such as "television" and of source recall is .67 to high-knowledge "newspapers." At the long delay, subjects names, .57 to low-knowledge names, and confabulated sources on each of the 13 ex- .66 to no-knowledge names. The overall traexperimental errors that accompanied level of conditionalized source recall by recalled items. As noted earlier, amnesic long-delay normals (.64) is numerically patients confabulated sources on only .30 quite similar to the level of conditionalized of extraexperimental errors. We cannot be source recall by amnesics (.61). However, too certain of our interpretation of normals' conditionalized source recall of long-delay source confabulations, however, because normals significantly exceeds the chance postexperimental interviews suggested that level (Z = 3.44, p < .01), whereas patients' some subjects may have used confabulating recall in Experiment 1 does not. responses interchangeably with "guessing" One feature of the unconditionalized responses. source recall data (Table 6) merits commen- SOURCE AMNESIA 607

tary. Collapsed across item recall and non- graded memory. It is possible to argue, of recall, source recall decreased as a function course, that normal subjects might have of retention interval, F(1,22) = 11.17, p < demonstrated source amnesia as often as .01, and was also affected by the prior patients did if they were tested at much knowledge manipulation, F(2,44) = 4.47, p longer delays--perhaps months or even < .05 MS e = .08. In addition, however, years after the study episode. It would not there was an unexpected prior knowledge be particularly surprising if such an out- x retention interval interaction, F(2,33) = come were obtained: Everyday experience 4.91, p < .05, MS e = .09. This interaction indicates that it is difficult to recollect the indicates that differences among levels of source of a bit of knowledge that was ac- prior knowledge observed at the short quired during an episode in the distant past delay were absent at the long delay. It is (e.g., do you remember who first told you possible, however, that the interaction is at- that George Washington was the first Pres- tributable to a floor effect on source recall ident of the United States?). The important at the long delay. Although the issue is not point, however, is that normals' level of particularly critical with respect to the item recall would have to be much lower major concerns of the present experiment, than that of memory-disordered patients it may merit exploration in future research. before they would exhibit comparable amounts of source amnesia. Such an out- GENERAL DISCUSSION come would be consistent with the idea that The results of the present experiments level of source amnesia in memory-disor- have yielded information about the nature dered patients cannot be accounted for in of source amnesia in memory-disordered terms of generally degraded memory. patients and normal subjects that clarifies a In contrast to the source amnesia data, number of aspects of the phenomenon. Ex- the present results lend some support to the periment 1 demonstrated that source am- idea that source forgetting may be a prop- nesia can be observed under controlled ex- erty of generally degraded memory: The in- perimental conditions in a group of patients cidence of intraexperimental source errors with severe memory disorders. On about in normal subjects increased significantly .40 of the trials in which they recalled a over the retention interval, and it was sim- target item, patients committed extraexper- ilar in amnesics and long-delay normals. imental source errors. Source amnesia oc- However, some interpretive caution must curled reliably across subjects and different be exercised with respect to this point be- types of to-be-remembered materials. Ex- cause amnesics' level of source recall, periment 2 indicated that when normal sub- given an intraexperimental source re- jects' level of item recall is equated with sponse, did not differ statistically from that of amnesics by interpolation of a long chance, although this may be attributable retention interval, normals exhibit signifi- to the small number of observations in the cantly less source amnesia: Given item re- experiment. In addition, the tests of source call, normals made between .05 and. 10 ex- recall and item recall differed in a signifi- traexperimental source responses at the cant way: The to-be-remembered item three levels of prior knowledge. A second changed on every trial, whereas the two ex- important outcome of Experiment 2 is that perimental sources remained the same. incidence of source amnesia did not in- Thus, poor source recall of long-delay nor- crease when memory was degraded by in- mals and amnesics, given an intraexperi- terpolation of a long retention interval. mental source response, may have oc- The overall pattern of results suggests curred because each of the experimental that the level of source amnesia exhibited sources was related to so many facts that it by memory-disordered patients cannot be was difficult to discriminate accurately be- accounted for in terms of generally de- tween them. 608 SCHACTER, HARBLUK, AND MC LACHLAN

Although most laboratory simulation were also ranked according to the propor- studies have found that phenomena of am- tion of extraexperimental errors, given nesia can be observed in normals with de- item recall, and proportion of items re- graded memories (cf. Mayes & Meudell, called. Mean rank on the two neuropsy- 1981a, 1981b; Mayes et al., 1981; Woods & chological tasks was significantly positively Piercy, 1974), two studies have been re- correlated with proportion of extraexperi- ported in which a phenomenon of amnesia mental errors given item recall (r = + .74, was not simulated in normal subjects. Hirst p < .05), but was not significantly corre- and Volpe (1982) found that amnesics' lated with proportion of items recalled (r = memory for temporal order was signifi- + .34, p > .05). In addition, the proportion cantly lower than that of controls under of extraexperimental source errors was un- conditions in which level of item recogni- correlated with either full scale WAIS-R IQ tion did not differ. Squire (1982) found that (r = +.19) or age (r = -.03). These ob- only some amnesic patients' performance servations must be treated cautiously, of on a temporal discrimination task was course, because they are both correlational poorer than would be expected on the basis and post hoc, but they do suggest that dis- of level of item recognition; others per- proportionately high levels of source am- formed no worse on the temporal discrim- nesia may be associated with poor perfor- ination task than would be expected on the mance on neuropsychological tasks that are basis of item recognition performance. sensitive to frontal lobe dysfunction. Squire found a strong positive correlation The foregoing data imply that patients between patients' performance on the tem- who are characterized by signs of frontal poral discrimination task and their perfor- lobe dysfunction, but not by severe mance on certain neuropsychological tasks memory impairment, might also exhibit that are sensitive to frontal lobe pathology: high levels of source amnesia. To investi- Patients who performed poorly on the neu- gate this possibility, we administered a ropsychological tasks tend to exhibit dis- modified version of the task used in Exper- proportionately poor performance on the iment 2 to four patients (two after left-sided temporal discrimination task. In the present stroke and two after ruptured anterior com- study, every one of our memory-disordered municating artery aneurysms) who were patients did exhibit some source amnesia, matched to the amnesics of Experiment 1 but the proportion of extraexperimental in terms of age, intelligence, and perfor- source responses on recalled items varied mance on the Wisconsin Card Sort. Twelve from .06 to 1.0 across patients. Using a high-knowledge and twelve no-knowledge strategy similar to Squire's, we examined name-characteristic pairs, as well as seven the possibility that this variation is associ- buffer items, were read to the patients by ated with level of performance on neuro- either of two sources; the patientswere psychological tasks. The patients were tested at a 0.5-h delay in the same manner ranked on the basis of their performance on as were the college students in Experiment two of the neuropsychological tasks used 2. Overall level of item recall was .17, by Squire that are sensitive to frontal lobe which is below the level attained by am- pathology: the Wisconsin Card Sort nesics in Experiment 1. But these patients (Milner, 1963), on which subjects sort did not commit any extraexperimental stimuli on the basis of changing dimen- source errors on correctly recalled items. sions, and the Benton Word Fluency test This finding indicates that the presence of (Benton, 1973), in which subjects produce severe memory impairment is a necessary words beginning with the letters F, A, and condition for the occurrence of dispropor- S during successive 1-min periods. Patients tionately high levels of source amnesia. SOURCE AMNESIA 609

The overall pattern of results suggests Exploration of these and other hypotheses that some memory-disordered patients-- could provide insight into the cognitive those with signs of frontal lobe dysfunc- functions that are necessary for the occur- tion-have special difficulties remem- rence of episodic remembering. bering the episodes in which information The results of the present research also has been acquired. This observation has have implications for the way that we think implications for theories of amnesia. A about episodic remembering in normal number of investigators have suggested that human memory. It is frequently assumed the fundamental deficit in amnesia is the that subjects' ability to reproduce the fac- inability to remember the contextual fea- tual contents of an episode, such as a word tures that define the occurrence of a partic- on a list, constitutes evidence that they re- ular episode in one's past (e.g., Clapar6de, member the occurrence of the episode 1911/1951; Kinsbourne & Wood, 1975, during which the word was studied. Indeed, 1982; Korsakoff, 1889; Rozin, 1976; this idea was endorsed by Tulving (1972) in Schacter& Tulving, 1982a, 1982b; Stern, his initial description of the characteristics 1981; Winocur, 1982). By contrast, others of episodic memory. Schacter and Tulving have suggested that amnesics' retention of (1982b), on the basis of clinical observa- contextual or episodic information is no tions of source amnesia, suggested that more impaired than is their retention of any there may be a need to distinguish between kind of declarative information, including retention of the factual contents of a items, facts, and associations (e.g., Cohen, learning episode and memory for the oc- 1984; Squire, Cohen, & Nadel, in press). currence of the episode itself (see also, We would suggest that each of these ideas Tulving, 1983). The results of both Experi- may be correct, but only when they are ap- ments 1 and 2 indicate the necessity of such plied to specific types of amnesic patients: a distinction: Amnesic patients, and normal Amnesics who perform well on frontal-sen- subjects to a lesser extent, demonstrated sitive tasks may exhibit approximately retention of the factual contents of a equal mnemonic impairments for all forms learning episode in the absence of recollec- of declarative information, whereas those tion of the prior occurrence of the episode. who exhibit signs of frontal lobe dysfunc- These results indicate that it may be unwise tion may have special deficits remembering to infer episodic memory purely on the the episodes in which items and facts have basis of subjects' ability to reproduce cor- been acquired. Accordingly, an important rectly an item or fact that was acquired task for future research is to clarify the na- during a particular episode. A similar theme ture of the deficits that are picked up by has emerged from recent research con- frontal-sensitive tasks, and to provide a de- cerning priming effects, in which normals' tailed understanding of how the cognitive and amnesics' performance on tasks such processes that are tapped by these tasks as fragment completion and perceptual map onto different domains of mnemonic identification is facilitated by previously function. One possibility is that amnesics studied information, whether or not they with signs of frontal lobe dysfunction have recollect the episode in which they encoun- deficits of processing resource that restrict tered the information (e.g., Graf, Squire, & their ability to encode simultaneously an Mandler, 1984; Jacoby & Dallas, 1981; item and its context (see Rabinowitz, Moscovitch, 1982; Scarborough, Gerard, & Craik, & Ackerman, 1982, for a resource- Cortese, 1979; Tulving et al., 1982; War- based account of memory deficit). Another rington & Weiskrantz, 1974). Although the possibility is that such patients have tem- relation between source amnesia and the poral discrimination deficits (Squire, 1982). aforementioned priming effects is not yet 610 SCHACTER, HARBLUK, AND MC LACHLAN clear, both phenomena illustrate that the JONES, G. V. (1976). A fragmentation hypothesis of multiple consequences of a learning epi- memory: Cued recall of pictures and of sequential position. Journal of Experimental Psychology: sode can be expressed in a dissociated General, 105, 277-293. manner. Understanding the nature of this JONES, G. V. (1979). Multirate forgetting. Journal of dissociation is a task that will no doubt oc- Experimental Psychology: Human Learning and cupy memory researchers for a good many Memory, 5, 98-114. years to come. KINSBOURNE, M., ~; WOOD, E (1975). Short term memory and the amnesic syndrome. In D. D. Deutsch & J. A. Deutsch (Eds.), Short-term REFERENCES memory. New York: Academic Press. ANDERSON, J. R., & BOWER, G. H. (1971). On an as- KINSBOURNE, M., & WOOD, E (1982). Theoretical con- sociative trace for sentence memory. Journal of siderations regarding the episodic-semantic Verbal Learning and Verbal Behavior, 10, 673- memory distinction. In L. S. Cermak (Ed.), 680. Human memory and amnesia. Hillsdale, NJ: Erl- BENNETT, C. A., & FRANKLIN, N. L. (1964). Statis- baum. tical analyses in chemistry and the chemical in- KORSAKOFF, S. S. (1889). ]~tude medico-psycholo- dustry. New York: Wiley. gique sur une forme des maladies de la memoire. BENTON, A, L. (1973). The measurement of aphasic Revue Philosophique, 5, 501-530. disorders. In A. C. Velasquez (Ed.), Aspectos pa- LURIA, A. R. (1976). The neuropsychology of memory. tologicos del lengage. Lima: Centro Neuropsicol- Washington, DC: Winston. ogico. MACCURDY, J. T. (1928). Common principles in psy- BROWN, E., DEFFENBACHER, K., & STURGILL, W. chology and physiology. Cambridge: Univ. Press, (1977). Memory for faces and the circumstances MAYES, A. R., & MEUDELL, P. R. (1981a). How sim- of encounter. Journal of Applied Psychology, 62, ilar is immediate memory in amnesic patients to 311-318. delayed memory in normal subjects? A replica- CLAPARI~DE, E. (1911/1951). Reconnaissance et moiti6 tion, extension, and reassessment of the amnesic [Recognition and "me-ness."] Archives de Psy- cueing effect. Neuropsychologia, 19, 647-654. chologie, 1911, 11, 79-90. In D. Rapaport (Ed.), MAYES, A., • MEUDELL, E (1981b). How similar is Organization and pathology of thought. New the effect of cueing in amnesics and in normal sub- York: Columbia Univ. Press, 1951. jects following forgetting? Cortex, 17, 113-124. COHEN, N. J. (1984). Preserved learning capacity MAYES, A. R., MEUDELL, E R., & SOM, S. (1981). in amnesia: Evidence for multiple memory sys- Further similarities between amnesic and normal tems. In N. Butters & L. R. Squire (Eds.), The attenuated memory: Effects with paired-associate neuropsychology of memory. New York: Guil- learning and contextual shifts. Neuropsychologia, ford. 19, 655-664. EVANS, E J., & THORN, W. A. E (1966). Two types of MILNER, B. (1963). Effects of different brain lesions posthypnotic amnesia: Recall amnesia and source on card sorting. Archives of Neurology, 9, 100- amnesia. International Journal of Clinical and Ex- 110. perimental Hypnosis, 14, 162-179. MOSCOVITCH, M. (1982). Multiple dissociations of GEISELMAN, R. E., & CRAWLEr, J. M. (1983). Inci- function in amnesia. In L. S. Cermak (Ed.), dental processing of speaker characteristics: Human memory and amnesia. Hillsdale, NJ: Erl- Voice as connotative information. Journal of baum, Verbal Learning and Verbal Behavior, 22, 15-23. RABINOWITZ, J. C., CRAIK, E I. M., & ACKERMAN, GRAF, P., SQUIRE, L. R., & MANDLER, G. (1984). B. E (1982). A processing resource account of age The information that amnesic patients do not differences in recall. Canadian Journal of Psy- forget. Journal of Experimental Psychology: chology, 36, 325-344. Learning, Memory, & Cognition, 10, 164-178. ROTHKOPF, E. Z., FISHER, D. G., & BILLINGTON, HIRST, W. (1982). The amnesic syndrome: Descrip- M. J. (1982). Effects of spatial context during ac- tions and explanations. Psychological Bulletin, quisition on the recall of attributive information, 91, 435-460. Journal of Experimental Psychology: Learning, HIRST, W., & VOLPE, B. T. (1982). Temporal order Memory, & Cognition, 8, 126-138. judgments with amnesia. Brain and Cognition, 1, ROZIN, P. (1976). The psychobiological approach to 294-306. human memory. In M. R. Rosenzweig & E. L. JACOBY, L. L., & DALLAS, M. (1981). On the relation- Bennett (Eds.), Neural mechanisms of learning ship between autobiographical memory and per- and memory. Cambridge, MA: MIT Press. ceptual learning. Journal of Experimental Psy- SCARBOROUGH, D. L., GERARD, L., & CORTESE, C. chology: General, 110, 306-340. (1979). Accessing lexical memory: The transfer of SOURCE AMNESIA 611

word repetition effects across task and modality. TULVING, E. (1972). Episodic and semantic memory. Memory & Cognition, 7, 3-12. In E. Tulving and W. Donaldson (Eds.), Organi- SCHACTER, D. L., & TULVIN6, E. (1982a). Amnesia zation of Memory. New York: Academic Press. and memory research. In L. S. Cermak (Ed.), TULVING, E. (1983). Elements of episodic memory. Haman memory and amnesia. Hillsdale, NJ: Erl- Oxford: Oxford Univ. Press. baum. TULVING, E., SCHACTER, D. L., & STARK, H. A. SCHACTER, D. L., & TULVING, E. (1982b). Memory, (1982). Priming effects in word-fragment comple- amnesia, and the episodic/semantic distinction. In tion are independent of recognition memory. R. L. Isaacson & N. E. Spear (Eds.), The expres- Journal of Experimental Psychology: Learning, sion of knowledge. New York: Plenum. Memory, & Cognition, 8, 336-342. SCHACTER, D. L., TULVING, E., & WANG, P. L. (1981). WARRIN6TON, E. K., & WEISKRANTZ, L. (1974). The Source amnesia: New methods and illustrative effect of prior learning on subsequent retention in data. Paper presented to the International Neu- amnesic patients. Neuropsychologia, 12, 419- ropsychological Society, Atlanta. 428. SQUIRE, L. R. (1982). Comparisons between forms of WEISKRANTZ, L., & WARRINGTON,E. K. (1975). Some amnesia: Some deficits are unique to Korsakoff's comments on Woods' and Piercy's claim of a sim- syndrome. Journal of Experimental Psychology: ilarity between amnesic memory and normal for- Learning, Memory, & Cognition, 8, 560-571. getting. Neuropsychologia, 13, 365-368. SQUIRE, L. R., COHEN, N. J., & NADEL, L. (in press). W1NOCUR, G. (1982). The amnesic syndrome: A deficit The medial temporal region and memory consol- in cue utilization. In L. S. Cermak (Ed.), Human idation: A new hypothesis. In H. Weingartner & memory and amnesia. Hillsdale, NJ: Erlbaum. E. Parker (Eds.), Memory consolidation. Hills- WooDs, R. T., & PIERCe, M. (1974). A similarity be- dale, NJ: Erlbaum. tween amnesic memory and normal forgetting. SQUIRE, L. R., NADEL, L., & SEATER, P. C. (1981). Neuropsychologia, 12~ 437-445. Anterograde amnesia and memory for temporal ZUmN, J. (1948). Memory functioning in patients order. Neuropsychologia, 19, 141 - 144. treated with electric shock therapy. Journal of SQUIRE, L. R., WETZEL, C. D., &SLATER, P. C. Personality, 17, 33-41. (1978). Anterograde amnesia following ECT: An analysis of the beneficial effects of partial information. Neuropsychologia, 16, 339-348. STERN 0 L. D. (1981). A review of theories of human (Received June 7, 1983) amnesia. Memory & Cognition, 9, 247-262. (Revision received October 18, 1983)