Emotional Perception and Memory in Amnesia

Neuropsychology In the public domain 1997, Vol. ll.No.1, 104-113

Stephan B. Hamann Larry Cahill University of California, San Diego University of California, Irvine

Larry R. Squire Veterans Affairs Medical Center and University of California, San Diego

The authors examined whether perception of emotional stimuli is normal in amnesia and whether emotional arousal has the same enhancing effect on memory in amnesic patients as it has in healthy controls. Forty standardized color pictures were presented while participants rated each picture according to emotional intensity (arousal) and pleasantness (valence). An immediate free-recall test was given for the pictures, followed by a yes-no recognition test. Arousal and valence ratings were highly similar among the amnesic patients and controls. Emotional arousal (regardless of valence) enhanced both recall and recognition of the pictures, and this enhancement was proportional for amnesic patients and controls. Results suggest that emotional perception and the enhancing effect of emotional arousal on memory are intact in

Many studies have indicated that emotional material is emotional arousal was selectively abolished by propranolol, more memorable than neutral material (Christiansen & a p-adrenergic antagonist. Yet, participant ratings of the Loftus, 1987; Craik & Blankstein, 1975; Heuer & Reisberg, emotional arousal produced by the story were unaffected by 1990; Matlin & Stang, 1978). The primary factor underlying the drug. the effect of emotion on memory appears to be emotional Several studies of experimental animals indicate that the arousal, independent of whether the arousal is pleasant or amygdala is involved in the enhancing effects of emotion on unpleasant. In studies of nonhuman animals, emotional memory (McGaugh et al., 1993). Findings from a patient arousal has been demonstrated to affect memory through the with bilateral lesions apparently restricted to the amygdala interaction of peripheral adrenergic systems with choliner- (Cahill, Babinsky, Markowitsch, & McGaugh, 1995) are gic, opioid peptidergic, and GABAergic systems in the consistent with evidence from experimental animals. Using amygdala (McGaugh et al., 1993). The amgydala also has the same emotional memory task as in the study with the been implicated in Jear conditioning (Davis, 1994; LeDoux, p-adrenergic antagonist (Cahill et al., 1994), this patient 1993). failed to benefit from emotional arousal, suggesting that the Although emotional arousal also can benefit memory by amygdala plays a crucial role in mediating the memory its effects on cognition (e.g., increased attention, deeper facilitation effect in humans. elaborative processing), the enhanced memory associated Bradley, Green wald, Petry, and Lang (1992) examined the with physiological arousal is independent of these cognitive effect of arousal (emotional intensity) and valence (pleasant- factors and can be dissociated experimentally from them. ness) on memory in college students, systematically examin- Thus, a recent study in humans (Cahill, Prins, Weber, & ing the relative influence of each factor on memory. Stimuli McGaugh, 1994), involving memory for an illustrated story, were drawn from the Internationa] Affective Picture System demonstrated that the memory facilitation associated with (IAPS; Lang, Ohman, & Vaitl, 1988), a standardized set of color slides depicting a wide variety of naturalistic subjects Stephan B. Hamann, Department of Psychiatry, University of that spans a wide range of affective reactions. The primary California, San Diego; Larry Cahill, Center for the Neurobiology finding was that the effect of emotion on memory (both of Learning and Memory, University of California, Irvine; Larry R. recall and recognition) was mediated by arousal rather than Squire, Veterans Affairs Medical Center, San Diego, and Depart- ments of Psychiatry and Neurosciences, University of California, valence. Specifically, after the effect of emotional arousal San Diego. Stephan B. Hamann is now at the Department of had been partialed out, there was virtually no effect of Psychology, Emory University. stimulus valence (pleasantness) on memory. This study and This research was supported by the Medical Research Service of others (Lang, Bradley, & Cuthbert, 1990; Russell, 1980) the Department of Veterans Affairs, National Institute of Mental have validated the dimensional approach to measuring Health Grant MH24600, the McKnight Foundation, and the emotion (arousal and valence). We have adopted the same McDonnell-Pew Center for Cognitive Neuroscience. We thank I. dimensional approach in the current study. McGaugh for his helpful comments and N. Champagne and J. Here, we examined whether the effects of emotion on Zouzounis for research assistance. Correspondence concerning this article should be addressed to memory would be observed in patients with amnesia. Little Larry R. Squire, Department of Psychiatry (116A), Veterans Affairs is known about the effect of emotion on memory in amnesic Medical Center, 3350 La Jolla Village Drive, La Jolla, California patients or even whether the emotional reactions and emo- 92161. Electronic mail may be sent to [email protected]. tional perceptions of amnesic patients are normal (however,

104 PERCEPTION AND MEMORY IN AMNESIA 105 see Hamann et al., 1996). Reports of emotional responsivity of memory was high or low (e.g., in the control and the in the amnesic syndrome have focused almost exclusively amnesic groups, respectively). on patients with alcoholic Korsakoff s syndrome (Rapaport, The first question of interest was whether the ratings of 1961; Talland, 1967). Although these reports concur in amnesic patients for the emotional stimuli would be normal characterizing the emotional responses of patients with or abnormal. The second question was whether the effect of Korsakoffs syndrome as abnormal, they differ in their emotional arousal on memory in these patients would be descriptions of the patients. Some patients are described as normal or abnormal. Another question concerned whether having reduced reactivity to emotional stimuli (sometimes the patients with Korsakoffs syndrome would exhibit abnor- referred to as flattened or blunted affect), whereas other mal emotional reactions, as previous reports have described, reports characterize other patients as overly irritable and and whether their reactions would differ from those of other hyperarousable. Similarly, the few experimental studies of amnesic patients. This was of interest partly because patients emotional reactivity in amnesic patients with Korsakoffs with Korsakoffs syndrome sometimes have frontal lobe syndrome (Douglas & Wilkinson, 1993; Markowitsch, dysfunction along with their memory deficits, and frontal Kessler, & Denzler, 1986; Oscar-Berman, Hancock, Mild- lobe dysfunction has been linked to emotional deficits worf, Hutner, & Weber, 1990) have yielded inconsistent (Stuss, Gow, & Hetherington, 1992). Finally, the emotional results. Some researchers have found their reactions normal reactions and the effects of arousal on memory were of (Douglas & Wilkinson, 1993), whereas others have found particular interest in the case of the 2 postencephalitic them either less reactive (Markowitsch et al., 1986) or more amnesic patients with amygdala damage because of the links reactive (Oscar-Berman et al., 1990) than controls. The among amygdala, emotion, and emotional memory. interpretation of these studies is complicated because different measures and materials were used to assess reactivity Method (e.g., self-reported ratings, skin conductance response, assessment only of negative emotions). Participants Apart from the issue of emotional perception in amnesia is the issue of whether the effect of emotional arousal on Amnesic patients. Nine amnesic patients (6 men and 3 women) memory is normal or abnormal. Anormal effect of emotional participated in this study. Four patients had Korsakoffs syndrome. All 4 had participated in quantitative magnetic resonance imaging arousal on memory in amnesia would be consistent with two (MRI), which demonstrated reductions in the volume of the notions: First, the brain structures that mediate the effect of mammillary nuclei (for R. C., J. W., and P. N., Squire, Amaral, & emotional arousal on memory (e.g., the amygdala) modulate Press, 1990; for N. F., unpublished observations). A 5th patient (M. declarative memory (McGaugh et al., 1993). Second, these G.) sustained a bilateral media] thalamic infarction, which was modulatory brain structures are distinct from the medial confirmed by MRI. Of the remaining 4 patients, 3 had bilateral temporal lobe structures important for declarative memory reduction in the size of the hippocampal formation confirmed by (Squire & Zola-Morgan, 1991). Alternately, the effect of MRI (for P. H., Squire et al., 1990; for L. J. and H. W., unpublished emotional arousal on memory may be abnormal in amnesia observations). Patient P. H. had a 6-year history of 1- to 2-min because the perception or expression of emotion is itself attacks (of possible epileptic origin) in association with gastric symptoms and transient memory impairment. In 1989 he suffered a impaired in amnesia. series of small attacks that resulted in marked and persisting The purpose of the current study was to examine whether memory impairment. Patient L. J. became amnesic gradually emotional perception is intact or impaired in amnesia and to during 1988 and 1989 without any known precipitating event. Her determine whether amnesic patients, like normal individu- memory impairment has remained stable since that time. Patient als, exhibit better memory for emotionally arousing material H. W. became amnesic gradually beginning in 1987 and has than neutral material. Perception and memory was assessed remained stable for about 7 years. Slight declines in other cognitive using the standardized IAPS picture stimuli (Lang et al., functions were noted in 1994 and 1995, leading to a diagnosis of 1988) in a group of 9 amnesic patients, 2 postencephalitic probable Alzheimer's disease. The 4th patient (A. B.) was unable to patients who were amnesic but also had damage to the participate in MRI studies, but the etiology of his amnesia (anoxia) amygdala, and two groups of controls. Participants rated 40 is consistent with hippocampal damage. All 9 patients were well-characterized neuropsychologically (see Tables 1 and 2). pictures individually for emotional arousal and valence and The patients' mean age was 63.4 years at the beginning of the were given a free-recall test for pictures they had viewed, study, and they had an average of 13.6 years of education. followed by a yes—no recognition test. Immediate and delayed (12-min) recall of a short prose passage The first group of controls (n = 12) was administered the averaged 4.0 and 0 segments, respectively (Gilbert, Levee, & same test procedure as the amnesic patients (n = 9) and the Catalano, 1968; maximum number of segments = 21). Scores on postencephalitic patients (E. P. and G. T). The second group other memory tests are shown in Tables 1 and 2. The mean score on of controls (control-delay; n = 12) was given a longer the Dementia Rating Scale was 131.8 (Mattis, 1976; maximum interval between the end of the study phase and the memory score = 144), with most points being lost from tile Memory subscale (mean points lost = 7.3). The mean score for the Boston tests. The purpose of the delay was to bring the control-delay Naming Test was 56.0 (Kaplan, Goodglass, & Weintraub, 1983; group's level of memory performance into the amnesic maximum score = 60). The number of categories completed on the patients' range. This condition was important because it was Wisconsin Card Sorting Test (Heaton, 1995) for Patients R.C., N.F., possible that the effect of emotional arousal on memory P. N., J. W., M. G., H. W., A. B., P. H., and L. J. were 4,3,3,5,4,2, would differ depending simply on whether the overall level 5, 5, and 6, respectively. Scores for controls on these tests can be 106 HAMANN, CAHUJL, AND SQUIRE

Table 1 Characteristics of Amnesic Patients

WMS-R Age WAIS-R Patient (years) Lesion IQ Attention Verbal Visual General Delay R.C. 76 Dien 106 115 76 97 80 72 N.F. 57 Dien 94 91 62 73 53 <50 P.N. 65 Dien 99 81 77 73 67 53 J.W. 56 Dien 98 104 65 70 57 57 M.G. 60 Dien 111 113 89 84 86 63 H.W. 77 HF 109 97 84 102 89 54 A.B. 55 HF 104 87 62 72 54 <50 P.H. 70 HF 115 117 67 83 70 57 L.J. 55 HF 98 105 83 60 69 <50 M 63.4 103.8 101.1 73.9 79.3 69.4 56.2 G.T. 59 HF+ 92 120 57 <50 <50 <50 E.P. 73 HF+ 103 94 57 82 61 56 M 66.0 97.5 107.0 57.0 66.0 55.5 53.0

Note. The WAIS-R and the WMS-R indexes yield a mean score of 100 in the normal population, with a standard deviation of 15. The WMS-R does not provide scores for those who score below 50. Therefore, the six scores below 50 were scored as 50 for calculating a group mean. WAIS-R = Wechsler Adult Intelligence Scale—Revised; WMS-R = Wechsler Memory Scale-Revised. Dien = diencephalon; HF = hippocampal formation; HF+ = hippocampal formation and additional brain structures. found elsewhere (Janowsky, Shimamura, & Squire, 1989; Squire et included the hippocampus, perirhinal and parahippocampal corti- al., 1990). ces, the amygdaloid complex, and the inferotemporal gyrus. Aside Patients E. P. and G. T. Two additional patients (E. P. and G. T., from his profound amnesia, E. P. also is anomic, scoring 42 of 60 both postencephalitic patients) also were tested. These 2 patients on the Boston Naming Test (normal score > 50; Kaplan, Good- are considered separately from the amnesic group both because glass, & Weintraub, 1983). E. P. also exhibited some behavioral their amnesia was much more severe than that of the other amnesic evidence of frontal lobe dysfunction: On the Wisconsin Card patients and because both patients had sustained bilateral damage Sorting Test, 0 categories were sorted, and there were 51% to the amygdaloid complex, which has been linked to emotion and perseverative errors (Heaton, 1995); on the FAS Test of Word emotional memory in humans and experimental animals (Cahill et Fluency, 18 words were produced, which fell below the 15th al., 1995; LeDoux, 1993). percentile (Benton & Hamsher, 1976). Patient E. P. sustained extensive bilateral medial temporal lobe G. T. sustained extensive bilateral temporal lobe damage con- damage, confirmed by MRI (Squire & Knowlton, 1995), which firmed by MRI (Hamann et al., 1996). The damage extended

Table 2 Memory Test Performance

Diagram Paired %Word %Word 50 50 Patient recall associates recall recognition words faces

R.C. 3 0 0 3 19 85 37 30 N.F. 4 0 0 2 36 76 28 27 P.N. 2 1 1 1 29 83 31 31 J.W. 4 0 0 2 28 96 29 34 M.G. 0 0 0 2 33 71 39 41 H.W. 6 0 1 0 31 85 23 22 A.B. 4 1 1 2 33 83 32 33 PH. 3 0 0 1 27 84 33 41 L.J. 3 0 0 0 40 93 33 29 M 3.2 0.2 0.3 1.4 30.7 84.0 31.7 32.0 G.T. 0 0 0 0 20 70 27 27 E.P. 0 0 0 0 24 65 19 18 M 0.0 0.0 0.0 0.0 22.0 67.5 23.0 22.5 Control (« = 8) 20.6 6.0 7.6 8.9 71 98 41.1 38.1

Note. The diagram recall score is based on delayed (12-min) reproduction of the Rey-Osterrieth figure (Osterrieth, 1944; maximum score = 36). The average score for copying the figure was 28.8, a normal score (Kritchevsky et al. 1988). The paired-associate scores are the number of word pairs recalled on three successive trials (maximum score = 10/trial). The word recall score is the percentage of 15 words identified correctly on five successive study-test trials (Rey, 1964). The word recognition score is the percentage of 15 words identified correctly by yes—no recognition across five successive study-test trials. The scores for words and faces are based on 24-hr recognition tests of 50 words or 50 faces (modified from Warrington, 1984; maximum score = 50, chance = 25). The mean scores for healthy controls and alcoholics shown for these tests are from Squire and Shimamura (1986). PERCEPTION AND MEMORY EM AMNESIA 107

through the anterior 7.0 cm of his left temporal lobe and the Materials and Procedure anterior 5.0 cm of his right temporal lobe. The lesion included bilaterally the hippocampus; entorhinal, perirhinal, and parahippo- The experiment consisted of a ratings phase, in which partici- campal cortices; the amygdaloid complex; and the inferior, middle, pants rated the emotionality of 40 color slides; a free-recall test, in and superior temporal gyri. In addition to his profound amnesia, which participants attempted to recall the 40 slides they had seen; G. T. is also anomic, scoring 18 of 60 on the Boston Naming Test and a yes-no recognition test in which 20 of the slides they had (normal score > 50; Kaplan et al., 1983). seen were presented, together with 20 different distractor slides. Patients E. P. and G. T. were 72 and 59 years of age, respectively, Finally, ratings of emotionality were collected for the 20 slides that at the beginning of the study, and each had 12 years of education. had served as distractor items in the recognition test. Immediate and delayed (12-min) recall of a short prose passage Sixty color slides depicting various scenes and subject matter averaged 1,5 and 0 segments, respectively (Gilbert et al., 1968; were selected from the IAPS (Lang et al., 1988) to sample a wide maximum number of segments = 21). Scores on other memory range of emotional content. The majority of these slides also had tests are shown in Tables 1 and 2. The mean score on the Dementia been used as stimuli in the Bradley et al. (1992) study. The Rating Scale was 112.0 (Mattis, 1976; maximum score = 144), procedure for the three groups was identical, except that the with most points being lost from the Memory subscale (mean amnesic and control groups were tested in a single session and the points lost = 16.0). control-delay group was tested with a 2-week delay between the Control groups. The control and control delay groups were ratings phase and the other tests. After the 2-week delay, those in employees or volunteers at the San Diego Veterans Affairs Medical the control-delay group were reminded that they had seen a slide Center or were recruited from the retirement community of the presentation in the previous session 2 weeks earlier. They were also University of California, San Diego. The control group consisted of informed that the memory tests they would be taking would 6 men and 6 women who were selected to match the amnesic concern the slides presented in that session. patients with respect to the mean and range of their ages, years of Ratings phase. Slides were presented on a carousel slide education, and scores on the Information and Vocabulary subtests viewer. Participants were informed that they would be viewing a of the Wechsler Adult Intelligence Scale—Revised (WAIS-R; series of pictures and that they were to rate each picture according Wechsler, 1981). They averaged 64.7 years of age (range = 55- to the feeling they experienced while viewing it. No mention was 73), 13.6 years of education, and 20.4 and 53.0 on the Information made of the memory tests that would follow. and Vocabulary subtests, respectively (amnesic patients = 18.1 and Participants rated each of the 40 slides on two dimensions, 51.4, respectively). Immediate and delayed recall of the short prose arousal (emotional intensity) and valence (degree of pleasant or passage averaged 6.7 and 5.0 segments, respectively. unpleasant affect), using the self-assessment mannequin (SAM) An additional control group (control-delay) was tested after a rating system (Hodes, Cook, & Lang, 1985; Lang, 1980; see Figure longer delay than that of the control group to bring their level of 1). Arousal and valence each were assessed with the help of five memory performance into the performance range of the amnesic mannequin figures. Participants recorded their ratings by placing an patients. This group consisted of 6 men and 6 women selected to X on or between any of the five figures, which resulted in a possible match the amnesic patients with respect to the mean and range of range of responses from 1 to 9. In the case of the arousal dimension, their ages, years of education, and scores on the Information and the appearance of the corresponding SAM figures ranged from Vocabulary subtests of the WAIS-R. They averaged 63.2 years of having an active body and eyes wide open to having an inactive age (range = 53-72), 14.6 years of education, and21.9 and 54.6 on body and closed eyes (see Figure 1). In the case of the valence the Information and Vocabulary subtests, respectively (amnesic dimension, the appearance of the corresponding SAM figures patients = 18.1 and 51.4, respectively). Immediate and delayed ranged from smiling with raised eyebrows to frowning with knitted recall of the short prose passage averaged 7.1 and 6.3 segments, eyebrows (see Figure 1). Participants' ratings of emotion using respectively. these two self-report scales have been demonstrated to covary

Figure 1. Scales used for arousal (top) and valence (bottom) ratings. For arousal, the scale assessed the degree of emotional intensity. For valence, the scale assessed the degree of pleasant or unpleasant affect. 108 HAMANN, CAHILL, AND SQUIRE reliably with physiological reactions (Bradley et al., 1992; Green- were rated for arousal and valence, using the same procedure as in wald, Cook, & Lang, 1989). Thus, arousal ratings covary reliably the initial ratings phase. with skin conductance responses, and valence ratings reliably covary with facial electromyographic responses. Results The two ratings scales were displayed on each page of a ratings booklet (one page for each of 40 slides). For half the participants, Arousal and Valence Ratings for the 60 Slides the upper scale assessed the arousal dimension and the lower scale assessed the valence dimension; for the other half, the scales were The arousal and valence ratings for the 60 slides were reversed. Because few ratings were placed between the mannequin evaluated for the amnesic group, Patients E. P. and G. T., the figures, the 9-point scale was transformed into a 5-point scale for control group, the control-delay group, and a normative scoring, as described by Bradley et al. (1992). sample of 87 undergraduates reported previously (Lang et Practice was given in the ratings procedure using two slides that al., 1988). First, correlations were calculated between groups did not otherwise appear in the study (a slide of an umbrella and a for both arousal and valence ratings to determine whether slide of a chocolate bar). Participants were informed that for each the groups gave the 60 slides the same pattern of ratings. All slide they would have 12 s to make both ratings and that they should attempt to record their initial emotional reaction to each correlations were significant at the .0001 level (df = 58). picture. It was emphasized further that they should try to use the Correlations were obtained by first calculating the mean whole range of the ratings scale and that each rating should be rating given each slide by all participants in each group and made on the basis of their true feelings. then calculating for each pair of groups the correlation To begin, the first practice slide was presented. After 12 s had between the mean ratings given the 60 slides. The Pearson elapsed, the projector was advanced to a blank slide until the product-moment correlations for mean arousal ratings among participant was ready to rate the second practice slide. After the two the amnesic, control, control-delay, and normative sample practice slides were rated, and additional instruction in the ratings groups ranged from .63 to .74, and the correlations for the procedure was given if needed, the ratings phase proceeded in the mean valence ratings ranged from .81 to .93. Patients E. P. same manner as the practice phase. A card explaining the arousal and G. T. gave arousal and valence ratings similar to the and valence ratings scales was always in view. In addition, the other groups (the correlations for mean arousal ratings amnesic patients were reminded periodically during the ratings phase of the ratings procedure about the meaning of the ratings between E; P. and G. T. and the other groups ranged from .35 scales. to .49, ps < .01; for valence, the correlations ranged from For the ratings phase, the 40 slides were presented in a different .78to.84,/is <.0001). random order for each participant. Slides always were presented for Although the groups assigned similar patterns of ratings the full 12s, even when both ratings were completed quickly. In the to the 60 slides, these groups (except for the control-delay few cases in which the ratings were not completed within 12 s, the group) rated the 60 slides higher on the arousal dimension projector was advanced to the blank slide and the participant than did the normative sample (see Tables 3 and 4), f(59) = completed the ratings. 8.87, p < .0001, for the amnesic group; t(59) = 7.91, p < Recall test. When the ratings phase ended, participants were .0001, for Patients E. P. and G. T.; and t(59) = 2.11, p < .04, tested for recall of the slides. They were instructed to write down in any order a word or phrase describing each picture they remem- for the control group. These differences may be because the bered. They were told that they would be given 5 min to complete participants in the current study were older than the under- this test. graduates in the normative sample. In addition, the amnesic Recognition test. After the free-recall test, a recognition test group rated the 60 slides slightly higher on the arousal was given in which 20 of the pictures presented in the ratings phase dimension than did the control group or the control-delay were presented again as recognition targets together (in a pseudo- group (see Tables 3 and 4), ft(59) > 6.84, ps < .0001. For random order) with 20 new distractor pictures that matched the the valence dimension, the amnesic group rated the 60 slides targets -with respect to arousal, valence, and subject matter (using similarly to the normative sample group, t(59) = 0.05, and the values for arousal and valence reported by Lang et al., 1988). the control-delay group, f(59) = 0.84, but higher than the All 40 pictures in the recognition phase were digitized and control group, f(59) = 2.47, p < .02. Patients E. P. and G. T. presented on a Macintosh Color Classic II computer. Each picture rated the 60 slides similarly to the ratings of the amnesic measured 12x8 cm. Participants were told that pictures would appear one at a time on group, ts(59) < 1.53, for arousal and valence. the computer screen, some of which had been presented in the earlier ratings phase. If the picture had appeared in the ratings Recall of 40 Slides phase, they were to press the button labeled yes (the z key for half the participants and the / key for the other participants). Otherwise, Free-recall responses were scored by comparing each they were to press the button labeled no (the / key for half of the participant's written descriptions with the characteristics of participants and the z key for the other participants). Participants the 40 target recall slides. Two independent judges rated a were instructed to attempt to be as accurate as possible but to response correct if it could be linked clearly to a particular respond as soon as they had decided because response times were slide that had been shown. Responses were not scored as being recorded. The recognition test began approximately 7 min correct if they matched more than one slide (i.e., if the after the end of the ratings phase. Each picture remained on the response was not sufficiently specific to identify one particu- screen until the participant pressed the yes or no key. Pressing either key replaced the picture with a blank screen, which, after 2 s, lar slide). The majority of responses (87%) were scored as was followed by the next picture. correct or incorrect by both judges. Scoring conflicts be- Final ratings phase. After the end of the recognition test, the tween the two judges (13% of the responses) were resolved 20 slides that had served as distractor iisms in the recognition test by a third independent judge. PERCEPTION AND MEMORY IN AMNESIA 109

Table 3 Arousal Ratings for 40 to-Be-Recalled Slides Arousal quartile Group First Second Third Fourth Control 1.57 ± 0.18 2.70 ± 0.21 3.45 ± 0.22 4.42 ±0.15 Control-delay 1.74 ± 0.29 2.72 ± 0.31 3.29 ± 0.30 4.14 ± 0.25 Amnesic 1.92 ± 0.19 3.00 ± 0.24 4.16 ± 0.20 4.83 ±0.11 Patients E.P. and G.T. 2.75 ± 0.45 3.50 ± 0.50 4.20 ± 0.20 4.80 ± 0.20 Note. Numbers are means and the standard errors of the means. Arousal quartiles were constructed by rank ordering slides according to each participant's arousal ratings. Each quartile contained 10 slides.

The recall results were analyzed according to the arousal recall in the amnesic group fell short of significance score given each slide. The 40 target recall slides were rank (p < .09). The effect of arousal on recall was significant for ordered for each participant according to each participant's both the control group, F(3, 33) = 8.03, MSE = 0.012, p < arousal ratings. (The ranks for slides given identical arousal .001, and the control-delay group, F(3, 33) = 7.05, MSE = ratings by the participant were decided on the basis of the 0.009, p<. 001. arousal means for the participant's group.) Figure 2 shows the results for the amnesic, control, and control-delay Recognition of 20 Slides groups. Neither E. P. nor G. T. was able to recall any studied slides. Table 3 shows the mean arousal ratings that corre- The recognition data were analyzed according to whether sponded to each arousal quartile in Figure 2. a slide belonged to the low- or high-arousal category. The 20 A mixed-factorial analysis of variance (ANOVA) was first target and 20 distractor recognition test slides were rank conducted with group (amnesic vs. control) as the between- ordered for each participant according to each participant's subjects variable and arousal quartile as the within-subjects arousal ratings. (The ranks for slides given identical arousal variable. The proportion of correctly recalled slides was the ratings were decided on the basis of the arousal means for dependent variable. There was an effect of group, F(l, 19) = the participant's group.) The 10 slides in each set (target or 64.29, MSB = 0.022, p < .0001, and of arousal, F(3, 57) = distractor) given the highest ratings were designated as 8.10, MSE = 0.012, p < .0001, but no interaction between high-arousal slides; the other 10 were designated as low- group and arousal, F(3, 57) < 1. Thus, the amnesic group arousal slides. The recognition data were partitioned into was impaired relative to the control group, and both groups two arousal categories (low and high arousal) rather than showed a similar beneficial effect of arousal on recall. into quartiles (as with the recall data), so that the same Next, the same analysis was conducted between the number (10) of slides were in each arousal category for both amnesic group and the control-delay group. There was an the recall and recognition data. effect of arousal, F(3, 57) = 7.09, MSE = 0.010, p < .001, Figure 3 shows die proportion of recognition hits and but no effect of group, F(l, 19) = 1.52, MSE = 0.015, p > false alarms (FAs) for the amnesic, control, and control- .23, and no interaction between group and arousal, F(3, delay groups as a function of arousal score. Table 4 shows 57) < 1. The failure to find a Group X Arousal interaction the mean arousal ratings dial corresponded to the low- and argues against the possibility that the absence of the same high-arousal categories in Figure 3. interaction in the amnesic versus control group comparison Three mixed-factorial ANOVAs were conducted with was attributable to the markedly different levels of recall. group (amnesic vs. control) as the between-subjects vari- Next, the effect of arousal on recall was examined able, arousal (low vs. high) as the within-subjects variable, separately in each group across arousal quartiles. For the and proportion of hits, FAs, and corrected recognition scores amnesic group, the effect of arousal on recall was not as the dependent measures, respectively. The amnesic group significant, F(3, 24) = 1.84, MSE = 0.013, p = .17. A test was impaired relative to the control group on all three for the presence of a linear trend across arousal quartiles for measures. For all three measures, there was an effect of

Table 4 Arousal Ratings for 20 Target Slides and 20 Distractor Slides in the Recognition Test Target Distractor Group Low arousal High arousal Low arousal High arousal Control 2.02 ± 0.17 3.96 ± 0.15 2.37 ± 0.17 3.98 ± 0.16 Control-delay 2.42 ± 0.31 3.87 ± 0.27 2.48 ± 0.24 3.84 ± 0.24 Amnesic 2.51 ±0.19 4.69 ±0.12 2.88 ± 0.17 4.72 ± 0.11 Patients E.P. and G.T. 3.00 ± 0.40 4.65 ± 0.15 2.80 ± 0.30 4.65 ± 0.25 Note. Numbers are means and the standard errors of the means. The low- and high-arousal categories were constructed by rank ordering slides according to each participant's arousal ratings. Each arousal category contained 10 slides. 110 HAMANN, CAHILL, AND SQUIRE

possibility that the absence of the same interaction in the amnesic versus control group comparison was attributable to differing levels of recognition performance or to ceiling effects in the control group. After these group comparisons, the effect of low versus high arousal on hits, FAs, and corrected recognition was examined separately in each group. For the amnesic group, there was a marginal effect of arousal on hits, f(8) = 1.94, p < .09, no effect on FAs, /(8) = \M,p > .14, and an effect on corrected recognition scores, f(8) = 2.64, p < .05. For the control group, there was a marginal effect of arousal on hits, r(ll) = 2.16,/j < .054, no effect on FAs, f(ll) = l.6Q,p> .14, and a marginal effect on corrected recognition scores, t(ll) = 2.06, p < .06. For the control-delay group, there was 234 an effect of arousal on hits, t(ll) = 2.20, p < .05, no effect AROUSAL QUARTILE on FAs (p > .1), and an effect on corrected recognition scores, t(U) = 336,p < .01. Figure 2. Recall scores for 40 slides obtained by the control (CON; n = 12), amnesic (AMN; n = 9), and control-delay (CON- Recognition performance for Patients E. P. and G. T. was DELAY; n = 12) groups as a function of the arousal score given near floor, at an overall level only marginally above chance each slide (first, second, third, or fourth quartile). Error bars (hit rates and FAs were converted to d'): mean d' = .54, indicate the standard error of the mean. t(l) = 3.92 against 0, p < .08 (one-tailed test). The mean hit rate was .40 ± .00 and .55 ± .05 for the low- and high-arousal slides, respectively; the corresponding FA rates group, Fs(l, 19) - 5-55-12.09, ps < .05; an effect of were . 10 ± .00 and .45 ± .05, respectively. In contrast to the arousal, Fs(l, 19) = 5.39-12.37, ps < .05; and no Group X amnesic, control, and control-delay groups, who showed a Arousal interaction, Fs(l, 19) = 0.64-2.11, ps > .16. memory benefit from arousal, the corrected recognition Next, the same analysis was conducted between the scores of Patients E. P. and G. T. were (nonsignificantly) amnesic group and the control-delay group. Memory perfor- lower for high-arousal slides than for low-arousal slides mance was similar in both groups. There was no effect of (.10 ± .10 and .30 ± .00, respectively). group for hits, FAs, or corrected recognition scores, F(l, 19) < 1.60. The effect of arousal was significant for hits Effects of Valence and corrected recognition scores, F(l, 19) > 8.3, and there was no Group X Arousal interaction (Fs < 1.02). The Valence had virtually no effect on memory, independent absence of a Group X Arousal interaction argues against the of arousal scores. This result is understandable given the

LOW AROUSAL HIGH AROUSAL

CON AMN CON-DELAY CON AMN CON-DELAY HITS FALSE ALARMS

Figure 3. Recognition hits and false alarms (FAs) for the control (CON; n — 12), amnesic (AMN; n = 9), and control-delay (CON-DELAY; n = 12) groups for low-arousal (open bars) and high-arousal (dark bars) slides. Error bars indicate the standard error of the mean. Corrected recognition scores (hits minus FAs) for the control group were .83 ± .05 and .92 ± .04 for the low- and high-arousal slides, respectively; the corresponding values for the amnesic group were .47 ± .10 and .66 ± .09, respectively, and the corresponding values for the control-delay group were .38 ± .06 and .50 ± .05, respectively. PERCEPTION AND MEMORY IN AMNESIA 111

Table 5 Reaction Times for the 20 Target Slides and 20 Distractor Slides in the Recognition Test

Targets Distractors

Group Low arousal High arousal Low arousal High arousal

Control 1,119 ± 206 963 ± 84 1,052 ± 86 1,178 ± 100 Control-delay 1,427 ± 110 1,663 ± 145 1,843 ± 115 2,269 ± 178 Amnesic 1,360 ± 149 1,292 ± 122 1,435 ± 143 1,722 ± 192 Patients E.P. and G.T. 3,282 ± 1,277 2,256 ± 176 3,101 ± 1,888 2,739 ± 1,511

Note. Numbers are means and the standard errors of the means. The low- and high-arousal categories were constructed by rank ordering slides according to each participant's arousal ratings. Each arousal category contained 10 slides. Only reaction times for correct responses (hits and false alarms) were analyzed.

marked curvilinear (quadratic) relationship between valence Comparison of Patients With and Without and arousal ratings in both the current study (quadratic r Korsakoff's Syndrome values between valence and arousal ratings for the 60 slides ranged from .53 to .74, p < .0001, for the amnesic, control, The performance of the 5 patients with Korsakoff's and control-delay groups and for Patients E. P. and G. T.) and syndrome on emotional ratings of the 60 slides and on the in previous studies (Bradley et al., 1992; Lang et al., 1988). recall and recognition tests was compared with the perfor- Specifically, both highly unpleasant (low-valence) and highly mance of the 4 non-Korsakoff amnesic patients. The two pleasant (high-valence) slides were rated as being more groups rated the 60 slides similarly overall for both arousal arousing than neutral slides. (3.53 ± .11 vs. 3.66 ± .08 for the Korsakoff and non- As a result, when the slides in each arousal category (four Korsakoff patients, respectively), r(59) = 1.60, p > .11, and arousal categories in the case of the recall test and two in the for valence (3.16 ± .16vs.3.15 ± .15 for the Korsakoff and case of recognition) were partitioned for each participant non-Korsakoff patients, respectively), r(59) < 1. Further- into equal numbers of high- and low-valence slides, the more, the pattern of ratings across the 60 slides was highly arousal ratings for the high- and low-valence slides were the similar between the two groups (rs = .71 for between same. Moreover, when valence was included as a variable in groups on arousal ratings, p < .0001, and .83 for valence the ANOVAs that had been carried out to analyze the recall ratings, p< . 0001). and recognition test results, no main effects of valence were To compare memory performance between the two groups, obtained. In addition, there were no interactions involving we conducted four mixed-factorial ANOVAs with propor- valence and the group variable. tion correct recall, recognition hits, FAs, and corrected recognition score as the dependent variables. The between- subjects variable for each ANOVA was subject group, and Reaction Time Analysis the within-subjects variables were arousal and valence. The Only correct trials (hits and correct rejections) were used Korsakoff amnesic patients performed similarly to the to analyze the recognition reaction time data. For each non-Korsakoff patients on each measure, with no effects of group, a within-subjects ANOVA was conducted with slide group or interactions involving group and the arousal or status (target vs. distractor), arousal, and valence as vari- valence variables. The exception was a three-way interac- ables. In addition, reaction times greater than 5 s were tion of group, arousal, and valence for recognition hits, F(1, omitted from analysis (3% of responses). Table 5 shows 7) = 28.64, MSE = 0.004, p < .01. This interaction reflected mean reaction times for each group as a function of slide the fact that the amnesic patients with Korsakoff s syndrome status and arousal. The reaction time measure, in contrast to in general had a slightly lower proportion of hits than the accuracy, was relatively insensitive to arousal. The amnesic non-Korsakoff patients, except for highly arousing, pleasant group, F(l, 8) = 5.63, MSE = 203, 721, p < .05, and the slides. For these slides, the patients with Korsakoff's syn- control-delay group, F(l, 11) = 14.79, MSE = 423, 417, drome outperformed the non-Korsakoff patients (.84 ± .08 p < .01, both responded faster to old (target) slides than to vs. .65 ± .17 correct, respectively). new (distractor) slides (reflecting a savings in encoding time for old slides; 1,326 ± 95 vs. 1,578 ± 121 and 1,545 ± 92 General Discussion vs. 2,056 ± 109 for the amnesic and control-delay groups, respectively). The control group also responded (numeri- We examined whether perception of emotional stimuli is cally)fastertooldslides,butnotsignificantlyso(l,041 ± 111 normal in amnesia and whether emotional arousal has the vs. 1,115 ± 66, F < 1). There were no other main effects or same enhancing effect on memory in amnesic patients as it interactions, with the single exception of a main effect of has in healthy controls. With respect to perception of arousal in the control-delay group, F(l, 11) = 17.61, MSE = emotional stimuli, the amnesic group, as well as amnesic 149, 515, p < .01. High-arousal slides (both targets and Patients E. P. and G. T., rated the 60 slides similarly to our distractors) were responded to more slowly than low-arousal controls and the normative sample originally reported by slides (1,634 ± 85 and 1,966 ± 122, respectively). Lang et al. (1988). The amnesic group and Patients E. P. and 112 HAMANN, CAHILL, AND SQUIRE

G. T. did tend to give higher arousal and valence ratings to tion was about what it should have been given the deficit in the slides than did controls and the normative sample group, free recall. but these differences were small (see Tables 3 and 4) and the The effect of valence was minor in the current study once overall pattern of slide ratings was remarkably similar across the effect of arousal on memory had been accounted for. all groups. The finding that Patients E. P. and G. T., who This result is understandable, given that valence and arousal have bilateral damage to the amygdala, gave normal emotion- ratings for the 60 slides were correlated highly. The effect of ality ratings is consistent with the previous finding that valence was limited to slightly higher corrected recognition Patient B. P., a patient with bilateral amygdala damage performance for low-valence items in the control and resulting from Urbach-Wiethe disease, was unimpaired in control-delay groups (there was no effect in the amnesic rating the arousal associated with an emotional story (Cahill group), and an interaction between arousal and valence for etal, 1995). FAs in the amnesic versus control group comparison. The Arousal improved recall and recognition performance in minor effect of valence on recall and recognition in the all groups except Patients E. P. and G. T., who exhibited little current study accords well with similar findings by Bradley or no detectable memory for the slides. Although the et al. (1992) for recall and recognition in college students. amnesic patients were severely impaired on both recall and Reaction time to recognition items was relatively insensi- recognition compared with the control group, their memory tive to the effects of emotional arousal. Bradley et al. (1992) was enhanced by high arousal in proportion to the enhance- did report an effect of emotional arousal on reaction time in ment exhibited by the control group. In addition, when the picture recognition. Specifically, hit responses to high- amnesic group was compared with the control-delay group, arousal slides were faster than hit responses to low-arousal which had been matched to the performance of the amnesic slides, and correct rejections were slower for high-arousal group by testing after a long delay, the two groups exhibited slides than for low-arousal slides. Perhaps in the current the same memory enhancement as a function of arousal. study participants emphasized accuracy over speed. Indeed, for recognition, the amnesic group exhibited numeri- The results of the current study suggest that emotional cally more benefit from arousal than the control-delay judgments are intact in amnesia, even in patients with group. For recall, the groups also benefited similarly from bilateral damage to the amygdala. Previous clinical and arousal, and there was no interaction with arousal in any anecdotal reports have suggested that patients with Korsa- group comparison. For the recall measure, the test for a koffs syndrome often are abnormal in their emotional linear trend for an arousal effect did not reach significance responses (Rapaport, 1961; Talland, 1965). However, our 5 when the amnesic group was examined by itself (p < .09), patients with Korsakoff s syndrome rated the emotional likely because of floor effects in the amnesic group. Our arousal and valence of slides highly similarly to the non- results for free recall are consistent with previous findings Korsakoff patients, and the beneficial effect of emotional that patients with left unilateral temporal lobectomy exhibit arousal on memory was similar in both groups. As consid- the same benefit in free recall for emotional versus neutral ered previously (Shimamura, Jernigan, & Squire, 1988), words as controls, despite an overall recall deficit for the alcoholic Korasakoff s syndrome is variable with respect to patients (Phelps, Ziv, & Labar, 1995). Thus, the memory the severity of memory impairment, frontal lobe signs, and enhancement effect associated with arousal appears to be other neuropsychological signs of cortical dysfunction. Our intact in amnesia, and this effect extends to both recall and patients were selected because they had a memory impair- recognition memory. ment disproportionate to any other intellectual impairment. The lack of a memory enhancement effect in Patients E. P. In summary, amnesic patients experienced the same and G. T. cannot easily be interpreted. Both E. P. and G. T. memory enhancement from emotional arousal as healthy have bilateral damage to the amygdala, and when bilateral controls. The results suggest that the emotional arousal amygdala damage is present emotional arousal may not effect is modulated by brain structures that are intact in enhance memory (Cahill et al., 1995). However, the recogni- amnesia. The amygdala, which is spared in all the amnesic tion performance of these patients was also near floor, which patients except Patients G. T. and E. P., is a prime candidate would have made it difficult to detect an enhancement effect. for the source of the emotional arousal effect, based on Either or both of these variables could account for the lack of evidence from both animals and humans (Davidson & a memory enhancement effect in these 2 patients. Sutton, 1995; McGaugh, 1989). The comparison between recall and recognition performance in the amnesic and control-delay groups also is notable. To our knowledge, this is the first demonstration References that free recall and recognition of pictorial information are impaired in amnesia proportional to their difficulty for Benton, A. L., & Hamsher, K. deS. (1976). Multilingual Aphasia controls (see Figures 2 and 3). Previous work (Haist. Examination. Iowa City: University of Iowa Press. Bradley, M. M., Grcenwald, M. K., Petty, M. C, & Lang, P. J. Shimamura, & Squire, 1992) established that free recall and (1992). Remembering pictures: Pleasure and arousal in memory. recognition are impaired equivalently in amnesia when word Journal of Experimental Psychology: Learning, Memory, and lists were used. The current finding provides further evi- Cognition, 18, 379-390. dence that recognition in amnesia is not being supported by Cahill, L., Babinsky, R., Markowitsch, H. J., & McGaugh, J. L. memory systems (such as priming) that are intact in amnesia (1995). The amygdala and emotional memory. Nature, 377, (also see Squire & Knowlton, 1995). The deficit in recogni- 295-296. PERCEPTION AND MEMORY IN AMNESIA 113

Cahill, L., Prins, B., Weber, M., & McGaugh, J. L. (1994). neutral or emotional content: A study of Korsakoff patients and f3-Adrenergic activation and memory for emotional events. recently detoxified and longterm abstinent alcoholics. Interna- Nature, 371, 702-704. tional Journal of Neuroscience, 29, 1—35. Christiansen, S.-A., & Loftus, E. F. (1987). Memory for traumatic Matlin, M., & Stang, D. (1978). The Pollyannaprinciple: Selectiv- events. Applied Cognitive Psychology, 1, 225-239. ity in language, memory, and thought. Cambridge, MA: Schenk- Craik, F. I. M., & Blankstein, K. R. (1975). Psychophysiology and man. human memory. In P. H. Venables & M. J. Christie (Eds.), Mattis, S. (1976). Dementia Rating Scale. In R. Bellack & B. Research in psychophysiology (pp. 388-417). New York: Wiley. Keraso (Eds.), Geriatric psychiatry (pp. 77-121). New York: Davidson, R. J., & Sutton, S. K. (1995). Affective neuroscience: Grune & Stratton. The emergence of a discipline. Current Opinion in Neurobiol- McGaugh, J. L. (1989). Involvement of hormonal and neuromodu- ogy, 5, 217-224. latory systems in the regulation of memory storage. Annual Davis, M. (1994). The role of the amygdala in emotional learning. Review of Neuroscience, 12, 255-287. International Review of Neurobiology, 36, 225—266. McGaugh, J. L., Introini-Collison, I. B., Cahill, L. F., Castellano, Douglas, J. J., & Wilkinson, D. A. (1993). Evidence of normal C., Dalmaz, C., Parent, M. B., & Williams, C. L. (1993). emotional responsiveness in alcoholic Korsakoff s syndrome in Neuromodulatory systems and memory storage: Role of the the presence of profound memory impairment. Addiction, 88, amygdala. Behavioural Brain Research, 58, 81-90. 1637-1645. Oscar-Berman, M., Hancock, M., Mildworf, M. S., Hutner, N., & Gilbert, J., Levee, R., & Catalano, K. (1968). A preliminary report Weber, D. A. (1990). Emotional perception and memory in on a new memory scale. Perceptual and Motor Skills, 27, alcoholism and aging. Alcoholism: Clinical and Experimental 277-278. Research, 14, 383-393. Greenwald, M. K., Cook, E. W., & Lang, P. J. (1989). Affective Osterrieth, P. A. (1944). Le test de copie d'une figure complexe judgment and psychophysical response: Dimensional covaria- [The test of copying a complex figure]. Archives de Psychologic, tion in the evaluation of pictorial stimuli. Journal of Psychophysi- 30, 206-356. ology, 3, 51-64. Phelps, E. A., Ziv, O., & Labar, K. S. (1995). Recall and affect Haist, F., Shimamura, A. P., & Squire, L. R. (1992). On the following unilateral temporal lobectomy. Society for Neurosci- relationship between recall and recognition memory. Journal of ence Abstracts, 21, 755. Experimental Psychology: Learning, Memory, and Cognition, Rapaport, D. (1961). Emotions and memory. New York: Science 18, 691-702. Editions. Hamann, S. B., Stefanacci, L., Squire, L. R., Adolphs, R., Tranel, Rey, A. (1964). L'examen clinique psychologic [The clinical exam D., Damasio, H., & Damasio, A. (1996). Recognizing facial in psychology]. Paris: Presses Universitaires de France. emotion. Nature, 379, 497. Russell, J. (1980). A circumplex model of affect. Journal of Heaton, R. K. (1995). Wisconsin Card Sorting Test manual. Personality and Social Psychology, 39, 1161-1178. Odessa, FL: Psychology Assessment Resources. Shimamura, A. P., Jernigan, T. L., & Squire, L. R. (1988). Heuer, F., & Reisberg, D. (1990). Vivid memories of emotional Korsakoff s syndrome: Radiological (CT) findings and neuropsy- events: The accuracy of remembered minutiae. Memory & chological correlates. Journal of Neuroscience, 8, 4400-4410. Cognition, 18, 496-506. Squire, L. R., Amaral, D. G., & Press, G. A. (1990). Magnetic Hodes, R. L., Cook, E. W., & Lang, P. J. (1985). Individual resonance measurements of hippocampal formation and mammil- differences in autonomic response: Conditioned association or lary nuclei distinguish medial temporal lobe and diencephalic conditioned fear? Psychophysiology, 22, 545-560. amnesia. Journal of Neuroscience, 10, 3106-3117. Janowsky, J. S., Shimamura, A. P., & Squire, L. R. (1989). Source Squire, L. R., & Knowlton, B. J. (1995). Learning about categories memory: Impairment in patients with frontal lobe lesions. in the absence of memory. Proceedings of the National Academy Neuropsychologia, 27, 1043-1056. of Sciences, 92, 12470-12474. Kaplan, E. F., Goodglass, H., & Weintraub, S. (1983). The Boston Squire, L. R., & Shimamura, A. P. (1986). Characterizing amnesic Naming Test. Philadelphia: Lea & Febiger. patients for neurobehavioral study. Behavioral Neuroscience, Kritchevsky, M., Squire, L. R., & Zouzounis, J. A. (1988). 100, 866-877. Transient global amnesia: Characterization of anterograde and Squire, L. R., & Zola-Morgan, S. (1991). The medial temporal lobe retrograde amnesia. Neurology, 38, 213-219. memory system. Science, 253, 1380-1386. Lang, P. J. (1980). Behavioral treatment and bio-behavioral assess- Stuss, D. T., Gow, C. A., & Hetherington, C. R. (1992). "No longer ment: Computer applications. In J. B. Sidowski, J. H. Johnson, & Gage": Frontal lobe dysfunction and emotional changes. Jour- T. A. Williams (Eds.), Technology in mental health delivery nal of Consulting and Clinical Psychology, 60, 549—559. systems (pp. 119-137). Norwood, NJ: Ablex. Talland, G. A. (1967). Deranged memory. San Diego, CA: Aca- Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion, demic Press. attention, and the startle reflex. Psychological Review, 97, Warrington, E. K. (1984). Recognition Memory Test. Windsor, 377-398. England: FER-Nelson. Lang, P. J., Ohman, A., & Vaitl, D. (1988). The International Wechsler, D. (1981). The Wechsler Mult Intelligence Scale- Affective Picture System [Photographic slides]. Gainesville: Revised. New York: Psychological Corporation. University of Florida, Center for Research in Psychophysiology. LeDoux, J. E. (1993). Emotional memory systems in the brain. Behavioural Brain Research, 58, 69-79. Received March 1, 1996 Markowitsch, H., Kessler, J., & Denzler, P. (1986). Recognition Revision received May 28, 1996 memory and psychophysiological responses to stimuli with Accepted May 31,1996 •