The Covert Learning of Affective Valence Does Not Require Structures in Hippocampal System or Amygdala

Daniel Tranel and Antonio R. Damasio University of Iowa College of Medicine Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021

Abstract

Following bilateral damage to the entire medial temporal First, Boswell’s lesions guarantee that the entorhinal and peri- lobe and interconnected cortices in the anterior temporal and rhinal cortices, hippocampus, amygdala, and higher-order neo- medial frontal regions, patient Boswell developed a severe cortices in the anterior temporal region are not required to learning defect for all types and levels of Factual knowledge, support this form of covert learning. Second, this demonstra- including faces. In the experiments described here, however, tion is possible only in a patient such as Boswell, because in we demonstrate that Boswell can acquire a nonconscious bond individuals with normal or only partially impaired factual learn- between entirely new persons and the affective valence they ing, fact memory will contaminate the performance. display. The finding is important on the followng accounts.

INTRODUCTION motor skills (Damasio et al., 1989). His amnesia can be summarized as follows: (1) In the anterograde compart- Patient Boswell is a 63-year-old man who has been se- ment, he has a severe defect for all types of material verely amnesic for 15 years, following herpes simplex (e.g., objects, faces, events, names) at unique and non- encephalitis. He has been investigated in our lalxmtories unique levels. Without rehearsal, new knowledge can be on numerous occasions, and descriptions of his basic retained only up to about 40 secs, beyond which it will neuroaiiatomical and neuropsychological profiles can be disappear, to all manner of observation, without a trace. found in previous publications (Damasio, Damasio, Tra- (2) In the retrograde compartment, he has a severe de- nel, Welsh, & Brandt, 1987; Damasio, Eslinger, Damasio, fect for unique material. He is limited to a few shreds of Van Hoesen, & Cornell, 1985; Damasio, Tranel, & Da- information regarding his former occupation, the fact masio, 1989). He has bilateral lesions that involve vir- that he was married and had two children, and a few tually all the principal cortical and subcortical limbic details about the town where he lived. He cannot pro- system structures (see Fig. 1). The mesial temporal re- duce any other accurate information regarding his past. gion is destroyed bikaterally, including the amygdala, en- Even in the few instances mentioned above, he is unable torhinal cortex, and hippocampus. The temporal pole to situate the facts and events temporally in his auto- (area 38) and the middle, inferior, and fourth temporal biography, or to correlate such pieces of information to nri (areas 20 and 21) are destroyed on both sides, and one another or to other temporal markers. His retrieval area 37 (in the posterior third of the same gyri) is par- of nonunique material is relatively less impaired. tially damaged on both sides. Both insular cortices are With regard to face recognition, Boswell has a severe destroyed, along with both posterior orbital cortices and defect in the retrograde compartment, and he cannot basal forebrain regions. However, the motor and pre- recognize faces of family members, long-time work as- motor cortices, basal ganglia, internal capsule, and cere- sociates, or even photographs of himself. However, we bellum are intact on both sides, as are all primary and have shown that he does demonstrate electrodermal most early sensory association cortices in occipital, tem- (skin conductance) discrimination of highly familiar faces poral, and parietal cortices. Most association cortices in from the retrograde compartment (Tranel & Damasio, the prefrontal region are intact bilaterally, including the 1987). He also produces accurate forced choice perform- entire expanse of dorsolateral frontal cortices. ances for familiar retrograde faces, and because his con- Boswell does not have perceptual defects (except for fidence ratings in these tasks were virtaally zero (i.e., he anosmia), nor does he have any language impairment. was “guessing” on every selection), we have taken the Motor function is normal, and so is the acquisition of accurate forced choice behavior, together with the elec-

0 1993 hfmuchusetts Institute of Technology Journal of Cognitive 5:1,pp. 79-88 (1993)

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021

Boswell

Figure 1. The neuroanatomical analysis of Boswell’s lesions, based on CT and MR, was conducted according to a standard template charting technique (Damasio & Damasio, 1989), and to the new BRAINVOX technique, which performs in viuo three-dimensional brain reconstruction based on high-resolution MK slices (Damasio & Frank, 1992). The bilateral lesions have destroyed the mesial temporal region (entorhinal and perirhinal cortices, aniygdala, hippocampus), the temporal pole (area 3H), the anterior portions of the middle, inferior, and fourth temporal gyri (areas 20, 21, and anterior sector of 37), the insular region, the posterior orbital cortices, and the anterior sector of the basal forebrain. The motor and premotor cortices, however, are intact on both sides, as are all primary and early sensory association cortices (olfaction excepted). Most association cortices in the prefrontal region are also intact bilaterally, including those in the entire dorsolateral region.

trodermal findings, as evidence for “covert” recognition METHODS AND RESULTS of familiar faces in the retrograde compartment. How- We tested the hypothesis that Boswell is capable of ac- ever, none of these outcomes was obtained in the antero- quiring information about new persons, provided the grade compartment. Boswell did not have discriminatory individuals are paired with a strong affective valence (the skin conductance responses to faces that he ought to operationalization of affective valence is explained be- have learned well, nor did he perform differently from low). In Experiment #1, standard “overt-level” recogni- chance in forced choice paradigms (Trdnel & Damasio, tion was tested (recognition of identity, verbal ratings of 1987). Boswell’s face learning in the anterograde com- familiarity). In Experiment #2, skin conductance and partment, which is the focus of the current set of studies, forced choice techniques were used to probe “covert” is discussed at greater length immediately below. face recognition.’ It is important to note that in these In keeping with his severe amnesia, Boswell has not experiments, the familiar faces did not have any partic- learned the faces of any of the persons he has come into ular affective value. They were faces that Boswell should contact with since the onset of his condition. He shows have learned because under normal conditions, robust no hint of recognition for the faces of physicians, psy- learning would have taken place given the abundant chologists, or caregivers, with whom he has had numer- exposure; however, the faces were not subjected to ous and extensive contacts during the past decade. any systematic manipulation involving affective valence. In However, we observed that in his daily environment (a Experiment #3, affective valences were manipulated to care facility), Boswell appeared to demonstrate a consis- determine whether this could influence face learning in tent preference relative to one of his caregivers (see Boswell, as suggested by the field observations men- Tranel & Damasio, 1990, for a preliminary report of this tioned above. Experiments #4 and #5 provided repli- observation). Specifically, there was one nursing aide to cation and follow-up whom Boswell would frequently go when he desired of Experiment #3. special treats, such as cigarettes or gum. For as long as the aide worked at this facility, he obliged Boswell’s Experiment #1: “Overt-Level”Face Learning requests, and Boswell appeared to gravitate toward this person, even if he could not overtly recognize him from Stimuli and Procedure a photograph, or address him by name. This behavior It was noted above that Boswell has not learned any new suggested that Boswell had learned some new informa- faces during the 15-year period of his amnesia, including tion vis-a-vis the aide, and the experiments we report the faces of caregivers, professionals, or other individu- here were designed to explore this phenomenon. als. To establish in systematic fashion his face learning

80 Journal of Cognitive Neuroscience Volume 5,Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 defect, the following experiments were conducted. We 2A. Skin conductance responses: Boswell’s skin selected eight faces from each of two groups: conductance was recorded, following the proce- dures we have utilized previously to measure non- 1. Professionals: physicians, psychologists, and other conscious, covert face recognition (Tranel & professionals Boswell has encountered frequently since Damasio, 1988; Tranel & Damasio, 1989). The two his illness; sets of faces were the same as in Experiment #1B. 2. Caregivers:various nursing aides and other persons Each set comprised a randomly ordered series of who live with Boswell in his care faci1ity.l The faces were 38 faces either containing the eight Professionals photographed against a white background and prepared faces or the eight Caregivers faces, mixed with non- as black-and-white slides. targets. Boswell viewed the faces but did not make Boswell’s “overt” recognition of these faces was tested any verbal response. as follows: 2B. Two-alternative forced choice: The eight Professionals and eight Caregivers faces were Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 1A. Identity recognition: Boswell was shown the paired with nontargets. Each pair was administered eight Professionals faces and the eight Caregivers on side-by-side Caramate 4000 slide projectors, and faces and asked to identify the person pictured. for each, Boswell was requested to choose the face 1B. Verbal ratings: A verbal ratings experiment that looked “most familiar.” The side (right versus was conducted for each of the two groups. For each left) on which the target face appeared was random- group, the eight targets (familiar faces) were mixed ized. The stimulus sets were administered on two with 30 nontargets (unfamiliar faces), and shown in occasions. random order. Boswell was asked to rate the famil- iarity of each face on a six-point rating scale, which ranges from 1 (“definite familiarity”) to 6 (“definite Results unfamiliarity”). 2A. Skin conductance: Boswell failed to discrim- inate the target from nontarget faces, based on his Results skin conductance responses (SCRs). In the Profes- sionals set, he generated an average SCR of 0.014 1A. Identity recognition: Experiment #1A was pS to the target faces, and an average of 0.010 pS conducted twice, on different occasions. Boswell to the nontargets. For the Caregivers set, his average performed at zero-level, obtaining scores of 018 for target SCR was 0.003 pS, and his average nontarget both sets on both occasions. He failed to identify SCR was 0.008 pS. Statistical comparisons of the accurately any of the faces in either of the two sets. target and nontarget averages for each set were both 1H. Verbal ratings: In Experiment #1R, we com- nonsignificant, according to Mdnn-Whitney U tests pared Hoswell’s verbal ratings for the target and (bothps > 0.30). nontarget faces in each set with a t test. In the 2B. Forced choice: In the forced choice test, Bos- Professionals set, he generated average verbal rat- well selected four target faces on both trials involv- ings of 3.15 and 3.87 for the target and nontarget ing the Professionals faces. For the Caregivers faces, faces, respectively; these ratings were not statisti- he selected three targets on the first trial and four cally different (p> 0.10).3 In the Caregivers set, his on the second trial. Taken together, his overall per- average verbal ratings for the targets and nontargets formance is 47% correct (15/32 across sets and were 3.75 and 4.20, respectively (p > 0.10). trials), which does not differ from chance (50%) The results confirm that Boswell cannot identify, or (p > 0.50). report as familiar, the faces of persons with whom he These two covert-level probes indicate that Boswell has had repeated and extensive contact since the onset cannot discriminate faces of persons he has had extensive of his amnesia. contact with since the onset of his amnesia. The findings are in keeping with a previous report in which we in- Experiment #2: “Covert-Level’’Face dicated that Boswell demonstrates covert-level discrimi- Learning nation of familiar faces in the retrograde, but not the anterograde, compartment (Tranel & Damasio, 1987). ’ Stimuli and Procedure Before turning to Experiment #3, it is important to Two procedures were utilized to investigate Boswell’s note that the target faces utilized in Experiments #1 and “covert-level” recognition of familiar faces. The target #2 did not have any special affective valence (as ex- stimuli (familiar faces) were the same as those described plained above, the original preferred Caregiver was not in Experiment #1A, i.e.,eight faces of Professionals and included in these experiments). Boswell had different eight faces of Caregivers. degrees of exposure to those faces, which was in all

Tranel and Dumusio 81

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 cases more than enough for normal learning. In terms like and did not call for actions that Boswell might find of affective value, those faces are similar to the “Neutral especially rewarding or punishing. Guy” described in Experiment #3 below. By the same We also manipulated the exposure time for each of token, they are different from the “Good Guy” and “Bad the stimulus persons. The Good Guy was given a low Guy” because the latter two faces have been paired sys- amount of exposure time (relative to the “high” amounts tematically with a strong affective valence. The finding described below), specifically, 30 midday for the 5-day that Boswell failed to respond to the target faces in period of the experiment (2.5 hr total). The Neutral Guy Experiment #2 (based on SCRS: and forced choice) is was given a high amount of exposure, specifically, 3 hr/ entirely consistent with the findings from Experiment #3 day for 5 days (15 hr total). The Bad Guy was given the below that, as will be described, indicate that affective same degree of exposure as the Neutral Guy, i.e., 3 hr/ valence is the key factor determining whether Boswell day for 5 days (15 hr total). This manipulation was im- can demonstrate some type of learning of new faces. We plemented to avoid possible confounds with mere ex- will return to this issue in Experiment #5 below. posure time (in fact, it tends to bias the experiment against our predictions), and in part for practical reasons. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 Experiment #3: Face Learning after In summary, the Good Guy had positive affective val- Manipulation of Emotional Valence and ence and low exposure time, the Neutral Guy had neu- Degree of Exposure tral affective valence and high exposure time, and the Bad Guy had negative affective valence and high expo- Stimuli and Design sure time. Each stimulus person interacted with Boswell The stimuli were three persons who were previously on a daily basis over a 5-day period, in accordance with unknown to Boswell. The three were designated as fol- their designation in this experiment. On the sixth day, lows: Person # 1-“Good Guy”; Person #2-“Neutral we conducted the following test trials, to determine Bos- Guy”; Person #3-“Bad Guy” (#1 and #2 were male, well’s learning of the three stimulus persons. and #3 was female). 3A. “Overt”learning: We presented Boswell with Boswell was brought to the Clinical Research Center eight faces (one at a time, projected in black-and- (CRC; an outpatient residential unit designated for re- white on a screen), three of which were the stim- search patients, located within the University Hospitals), ulus persons described above, and five of which and the experiment was conducted there and in our were persons he had never seen. This procedure laboratories over a 6-day period. On day 1, Boswell was was repeated on 10 different trials. On 5 trials, we introduced to the three stimulus persons, on separate asked Boswell to identify the individual shown on occasions separated by a few hours. Immediately on the slide (Free Recall), by naming the person or introduction, each stimulus person adopted an affective providing unambiguous, specific identifying infor- attitude toward Boswell that accorded with that person’s mation. On the remaining 5 trials, we made the designation in the experimental design. The affective same request; however, for the three stimulus per- valences were further operationalized as follows: The sons, we provided Boswell with cues (e.g., “You Good Guy was always especially kind to Boswell. He have been spending quite a lot of time with this gave Boswell many compliments regarding his physical person”; “You have been doing many experiments appearance, attire, and personality. He granted all re- with this person”; and so on). These 5 trials were quests for “treats,” and actually offered gum, drinks (cof- designated Cued Recall. fee, pop), and food items to Boswell. He never required 3B. “Covert”learning: Boswell’s “covert” learn- Boswell to participate in experiments, testing, etc., and ing of the stimulus persons was tested in the fol- was never involved in formal neuropsychological or lowing manner. The face of each stimulus person medical procedures. was presented along with an unfamiliar face (via The Bad Guy behaved in a systematically negative man- side-by-side slide projectors), and for each pair, ner toward Boswell, within limits of social and ethical Boswell was asked to “Choose the person that you appropriateness. Thus, the Bad Guy never complimented would go to for a treat.” Three blocks of 18 trials Boswell, and actively refused requests for treats. The Bad each were administered; on each trial block, each Guy was responsible for performing experiments that stimulus person was presented six times as one Boswell found tedious and unrewarding, e.g., executing alternative in a pair. Thus, there were 18 pairings hundreds of trials of the delayed nonmatching-to-sample for each stimulus person. procedure (which Boswell frequently attempted to dis- continue but would not be allowed to). Results The Neutral Guy was neutral in his attitude. He did not compliment Boswell or grant requests, but was al- 3A. “Overt”learning: Boswell’s performance on ways pleasant and friendly. He was not responsible for this task is summarized in Table 1. He failed to executing experiments in which Boswell was disinclined identify any of the stimulus persons, in either Free to participate. The Neutral Guy’s duties were business- Recall or Cued Recall, on any of the trials. This

82 Journal of Cognitive Neuroscience Volume 5, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 Table 1. “Overt” Learning Table 2. “Covert” Learning

Ident$cation Choice Behatiior Stimulus Free Cued Real-World Stimulus Person Trial I Trial 2 Trial -3 Total PA) Person Recall Recall Behavior ~~~~~~~~ Good Guy 516 516 5/6 83 Good Guy 015 015 No recognition of face, name, etc. Neutral Guy 416 316 316 56 Bad Guy 116 216 116 22 Neutral Guy 015 015 No recognition of face, name, etc. Bad Guy 015 0/5 No recognition of face, name, etc. contacts with during the 4-year period between the orig- inal study and the follow-up. The new persons are des- ignated by their valence (Positive v. Neutral) and Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 failure matched our observations of his behavior exposure (High, Medium, or Low). Person #3 in the throughout the 5-day period of the experiment, follow-up study was “PositivekIigh,”person #4 was “Pos- during which he never gave any indication that he itive/Medium,” and person #5 was “NeutraLMigh.”We knew the stimulus persons, that he had ever seen also included a sixth person (#6), who functioned as a them before, that he knew their names, etc. As in nontarget (“NeutraVLow”). Experiment #1A, it confirms that Boswell does not Neither the original Good Guy (#1) nor the original learn the identity of new faces. Bad Guy (#2) had had further contact with Boswell since 38. “Coveut” learning: The results of Boswell’s the original experiment. The new “Positive/High’ person choice performance are depicted in Table 2, as a (#3) had spent numerous hours with Boswell at various function of Stimulus Person and Trial Block. As can points throughout the 4-year interval since the original be seen, Boswell chose the Good Guy on 83% of experiment, and had always maintained a very positive, the pairings; the Neutral Guy was chosen on 56% reinforcing attitude toward Boswell (this individual had of the pairings, and the Bad Guy was chosen on played the role of “Neutral Guy” in Experiment #3). The 22% of the pairings. Binomial tests confirmed that new “Positive/Medium” person (#4) was similar, except Boswell’s selection of the Good Guy was signifi- that the overall amount of time was less extensive than cantly above chance (p < 0.001), and his selection for #3. The “NeutraVHigh” person (#5) was someone of the Bad Guy was significantly below chance (p< with whom Boswell had frequent encounters during the 0.01). Selection of the Neutral Guy was not signifi- 4-year interval, to about the same extent as the “Positive/ cantly different from chance. These results contrast High” person; however, the “NeutraVHigh” person had with those from Experiment #2, and show that never assumed a particularly positive or negative attitude Boswell can learn to discriminate between newly toward Boswell. Boswell had only minimal contact with met persons according to the positivehegative in- the new “NeutraVLow”person (#6), always of a neutral teractions he has had with them. nature. It is important to note that this follow-up exper- iment did not specifically manipulate the factors of affec- tive valence and exposure time as in the initial Experiment #4: 4-Year Follow-up to Good experiment, but instead, took advantage of the natural Guy-Bad Guy Experiment arrangement of contacts that Boswell had had with var- Four years after the Good Guy-Bad Guy experiment ious individuals over the 4-year interval between the described above, we conducted a follow-up study. The original and follow-up experiments. aim was two-fold: (1) to measure the extent to which 4A. Two-alternativeforced choice: In the 4-year Boswell had retained the learning he demonstrated in follow-up experiment, Boswell was tested with a the initial investigation; and (2) to explore Boswell’s two-alternative forced choice format in which pairs learning of other persons, with whom he had had fre- of faces were presented on side-by-side projectors, quent, systematically valenced contacts during the 4-year and he was asked to “Choose the person that you interval. would go to for a treat.” Each pair contained one of the six stimulus persons, and a nontarget, unfa- miliar face. There were 125 trials, comprising about Stimuli and Procedure 20 trials per stimulus person (see Table 3 for exact The stimulus persons for the follow-up study included number of trials per person). the original Good Guy (#1 in the follow-up study) and 4B. Four-alternative forced choice: In this task, the original Bad Guy (#2 in the follow-up study). There Boswell was asked to pick “the person you like best” were three new targets, selected because they were per- from a set of four on each trial.* There were two sons with whom Boswell had had systematically valenced trial blocks, and on each, 50 trials were conducted.

Tranel and Dammio 83

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 Table 3. 4-Year Follow-up (2-AFC) ever, a new “good guy” (the Positive/High person) could be learned in the appropriate setting. It is of interest that Choice Behavior the new “Good Guy” was the “Neutral Guy” in the pre- Signt@cunce vious experiment, i.e., that although the basic stimulus Target Person Rate 0 remained the same, the manipulation of affective valence #1 (Original Good Guy) 13/20 (65%) 0.132 led to a change in the behavior toward that stimulus. It is also noteworthy that the new NeutraVHigh person was #2 (Original Bad Guy) 11/19 (58%) 0.500 not chosen at rates different from chance. This finding #3 (New: Positive/High) 18/19 (95%) <0.001 corroborates many other sources of evidence in indicat- ing that exposure per se is not sufficient to support the #4 (New: Positive/Medium) 17/21 (81%) <0.005 learning of a new face in patient Boswell. Together with #5 (New: NeutraVHigh) 12/25 (48%) 0.500 the finding that he made reliable selections of the per- #6 (New: NeutralLow) 12/21 (57%) 0.332 sons with positive affective valence, the results point toward the notion that affective valence is the key factor Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 in Boswell’s learning. The target stimulus persons were the same as those utilized in Experiment #& (except that the Neutral/ Experiment #5: Follow-up with Skin Low person was not used). Thus, on each trial, Conductance Responses Boswell was presented four faces, projected in a The results of Experiments #3 and #4 indicate that 2 x 2 array on a large screen. The quadrant in Boswell can learn to discriminate covertly among newly which a target face appeared was randomized, as encountered persons, according to the positivehegative were a variety of other factors, including gender, interactions he has had with them. This outcome was age, facial appendages, and facial expressions of the demonstrated in a forced choice paradigm, and it con- nontarget foils, and the side of the room on which trasts with the results from Experiment #2, which indi- Boswell was placed when he made his selections. cated that Boswell could not discriminate target faces, on the basis of SCRs and forced choice (remember that Results those faces had not been paired with a particular affective valence). A question that arises in this context is whether 4A. Two-alternativeforced choice: As Table 3 Boswell would show discrimination of “good guy” faces shows, Boswell chose the new Positive/High person in an SCR paradigm. The prediction, given the results of (#3) on nearly every opportunity (18/19 trials). the forced choice procedures described in Experiments New new Positive/Medium person (#4) was also #3 and #4, is that he would demonstrate discriminatory chosen at a level well above chance (17121 trials). SCRs to such faces. This prediction was tested in a follow- However, the original Good Guy and Bad Guy (#1) up experiment conducted at the same epoch as Experi- were chosen at chance levels, as was the new (#a) ment #4. Neutralmigh person (#5). As expected, the new Neutrahow person was at chance. The findings indicate that the effects from the original experi- Stimuli and Procedure ment had dissipated over the intervening 4 years, The stimuli were a set of 11 faces, six of which were so that Boswell no longer demonstrated a tendency nontargets (unfamiliar), and five of which were the Pos- to choose for or against the original Good Guy and itivemigh person, Positive/Medium person, NeutraVHigh Bad Guy, respectively. However, he clearly remains person, Positive/Medium person, NeutraVHigh person, capable of forming biases toward newly favored original Good Guy, and original Bad Guy (as described individuals. in Experiment #4). The set was randomly ordered, and 4B. Four-alternative forced choice: The results presented on two occasions. Procedures for SCR record- are presented in Table 4, and it can be seen that ing and stimulus delivery were the same as those de- the outcome confirms the findings obtained in Ex- scribed in Experiment #2A. periment #&. Specifically, Boswell chose the Pos- itivemigh person on nearly every trial (19/20), and the Positive/Medium person at a rate significantly Results above chance (13/20 trials). The NeutraYHigh per- For the first presentation, Boswell generated a large SCR son, and the original Good Guy and Bad Guy, were to the Positive/High person (0.2 pS, averaged across chosen at a chance level (note that chance is 25% hands). There was also a clear response to the Positive/ in this four-alternative forced choice format). Medium person (0.08 pS, averaged across hands). There The results indicate that the covert learning for the was no response to the NeutraVHigh person, or to the original Good Guy and Bad Guy waned with time; how- original Good Guy or Bad Guy, and none of the nontarget

84 Journal of Cognitive Neuroscience Volume 5, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 Table 4. 4-Year Follow-up (4-AFC)

Choice Behavior (chance=25%)

Target Person Trial 1 Trial 2 Total @-value) #1 (Original Good Guy) 4/10 4/10 40% (>0.10) #2 (Original Bad Guy) 4/10 4/10 40% (>0.10) #3 (New: PositivehIigh) 10/10 9/10 95% (<0.001) #4 (New: Positive/Medium) 6/10 7/10 65% (<0.01) #5 (New: Neutrdmigh) 3/10 3/10 30% (>0.20)

faces elicited an SCR larger than 0.01 pS. The same mere exposure (Greve & Bauer, 1990; Sergent & Poncet, Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 pattern of results obtained for the second presentation. 1990; Tranel & Damasio, 1985, 1988). However, as Ex- The SCR for the PositivdHigh person was 0.24 pS, the periments #3, #4, and #5 indicate, there is an important SCR for the Positive/Medium person was 0.06 pS, and exception to this summary: For faces that have been SCRs for the remaining faces (including the NeutraVHigh paired systematically with substantial affective valences, and original Good Guy and Bad Guy) did not exceed Boswell develops reliable preferences. The evidence in- 0.01 ps. dicates that he can learn associations between new faces Experiment #5 indicates that Boswell does generate and affective valences, and even though this learning may discriminatory SCRs to faces that have been linked to an be less than normal, it is at least capable of supporting affective valence. The effect appears to be somewhat accurate forced choice behavior and electrodermal dis- stronger when the exposure time is higher; however, crimination. However, the new linkages are not made exposure per se is not sufficient to produce an effect, as available to consciousness. Furthermore, the results indicated by the negative result for the NeutraVHigh foil. show that the associations can be modified over time by This outcome parallels the results of the forced choice changing subsequent contexts of interaction for the faces. procedures in Experiments #3 and #4, and together It is possible that the modifiable nature of these traces with the findings from Experiment #2, which showed is an entirely normal phenomenon, which would reflect that Boswell did not produce discriminatory SCRs (or the fluctuations in affective valence and exposure that accurate forced choice responses) to target faces that are so characteristic of social interactions. In other words, were neutral with respect to affective valence, indicates you would not want the nervous system to learn per- that affective valence is the key factor supporting the manent and rigid associations between given persons covert learning of faces in patient Boswell. In short, and a given affective value, even covertly, since such Boswell appears capable of covert learning of new faces rigidity would preclude appropriate adaptation to evolv- (i.e., in the anterograde compartment), provided those ing social circumstances, i.e.,to all the detestable persons faces are paired with affective valence. Such learning is who in time become marvelous and vice versa. not possible without the affective valence, as demon- Given the design of the experiments, it appears that strated by the negative outcome in Experiment #2 and the critical factor operative in Boswell’s learning was as reported in a previous study (Tranel & Damasio, afective valence. Other factors, such as facial expression, 1987). gender, presence of facial appendages, age, and position on the display, did not influence Boswell’s choices in the experiments. Nor can the results be explained by the DISCUSSION “good guy” and “bad guy” having faces that were inher- Boswell cannot recognize or name any face encountered ently more and less likeable-if this were the case, one since the onset of his amnesia, more than 15 years ago, would not expect the original “good guy” and “bad guy” nor can he recall any information about specific persons faces to yield chance performances in Experiment #4A. given the face or name, no matter how extensive his Also, given the fact that the effect waned over time, it exposure to those persons. Furthermore, provided there appears that holding of the association depends to some is no manipulation of affective valence, he does not even extent on the pairing being re-presented (i.e., recent show “covert” recognition of such faces in skin conduc- exposure may help), and on the valence remaining of tance and forced choice paradigms (in the antwograde the same signal. But exposure per se does not appear to compartment; note that he does show such covert rec- be sufficient, because Boswell failed to respond, even at ognition for retrograde faces). This is clearly different covert level, to target faces that were not paired with from some patients with face agnosia who do not learn affective value. In short, Boswell’s learning cannot be new faces at the conscious level, but who do discriminate explained as an effect of “mere exposure” (cf. Zajonc, them covertly without any special manipulation beyond 1980) or “perceptual fluency” (cf. Jacoby, 1984).

Tranel and Dammio 85

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 The question could be raised as to whether Boswell’s states, good or bad, that corresponds to the valence learning of affective valence (i.e., learning to tell a “good accompanying the face. The acquisition might require guy” from a “bad guy”) could be attributed to the task the following steps. First, Boswell perceives a face, say, merely being easier than learning person identification the face of the Good Guy. Second, directly thereafter, the (it., learning who persons are). However, there is ample Good Guy does something pleasant. This causes a posi- evidence that this explanation is not viable. In the literally tive somatic state, i.e., a state that involves a particular hundreds of learning experiments in which Boswell has set of changes in parameters of visceral and musculos- been involved over the past decade (Damasio et al., 1985, keletal activity, which are, in turn, mapped at varied 1987,1989),he has never given any indication of learning neural levels. Third, a bond develops between compo- even the simplest declarative-type information (eg, the nents of the neural processing of the face and of the name of an object or person). Also, in Experiment #2B, somatic state. Given the re-presentation of the appropri- in which Boswell chose between highly familiar faces ate face, this bond regenerates some part of the original and faces of strangers, his performance was at chance. processing set. (It would not be necessary to reactivate Given that the target faces in Experiment #2B were of the entire somatic state, but only part of the process Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 persons with whom Boswell has had extensive contact related to it.) It would be on the basis of such a reacti- over many years, it could be argued that this experiment vated set of signals that Boswell can make the choice is actually easier than the “Good Guy” experiment (#3B); between faces. nonetheless, only in the latter did Boswell demonstrate Where does this limited learning take place in Bos- a discriminatory performance. Juxtaposing the negative well’s brain, considering that neither the medial tem- result of Experiment #2R with the positive result of poral (including the hippocampus proper and the Experiment #3B indicates that affective valence 13 the amygdala) nor basal forebrain regions can possibly sup- critical factor, and that familiarity per se, and exposure port the process? One possibility invokes a neural sub- per se, are not sufficient. This conclusion is supported strate proposed by Mishkin for habit learning (Mishkin, by the negative results obtained for the foil with high Malamut, & Bachevalier, 1984; see also Squire & Zola- exposure but neutral valence, in Experiments #4 and Morgan, 1991). Intact face processing areas in occipito- #5. temporal and parietal cortices project to widespread From the behavioral perspective, similar findings have areas of the dorsal striatum (Selemon & Goldman-Rakic, been described for Korsakoff patients (Johnson, Kim, & 1985; Van Hoesen, Yeterian, & Lavizzo-Mourey, 1981; Yet- Risse, 1985), and for experimental animals in standard erian & Van Hoesen, 1978), including the caudate (see conditioning paradigms. Consider the following phe- Fig. 2). The effect of somatic changes caused by reward nomenon, summarized in a recent paper by Garcia can also reach the caudate, e.g., via dopaminergic pro- (1990). A thirsty rat is placed in a compartment where it jections from the substantia nigra (Nauta & Mehler, 1966). drinks water from a spout. Then, at a particular point, In turn, the caudate can signal by reentrant circuitry to the water is sweetened. The animal tastes the sweetened a variety of frontal cortices (Alexander, Delong, & Strick, water, hesitates at the unfamiliar taste, but finally drinks 1986; Ilinsky, Jouandet, & Goldman-Rakic, 1985), in par- as much as possible in the allotted time. A few minutes ticular the prefrontal, onto which both signals about a later, the animal is anesthetized. Then, while under deep face and about the results of a closely associated reward anesthesia, the animal is given a drug that causes nausea can also converge, and be held in a temporary “working in conscious animals. The animal is kept under anes- memory” (Goldman-Rakic, 1987). Re-presentation of the thesia for a few more hours, then allowed to recover for “signal face” would reactivate the caudate-to-frontal cir- a few days. The thirsty animal is returned to the drinking cuitry, and in so doing, generate a “marker” on the basis place with sweet water, where it goes to the drinking of which the choice would be bia~ed.~The marker might spout, licks the water, and then suddenly recoils and come from the partial regeneration of a positive or neg- attempts to get rid of the water from its mouth. The ative somatic state triggered by the sight of the Good or animal’s brain has learned that the sweet water, once a Bad Guy, followed by an overt feeling caused by its good thing, is now bad. And yet the animal could never sensory read-out, or by the processing of some covert have acquired a conscious memory of the association signal related to a somatic state with a particular valence, between sweet water and nausea, because it was not but in the absence of actual reenactment of a somatic conscious when the physiologic correlates of nausea state. All the structures required by this account are intact were produced. The link that established sweet water as bilaterally in Boswell, namely, the neostriatum, the oc- an aversive stimulus was generated in a nonconscious cipital, parietal, and posterior temporal cortices, and the process. vast majority of prefrontal cortices. Returning to our experiments, we have chosen to in- The foregoing interpretation is our preferred account terpret the findings as indicating that Boswell acquired for the results, although other interpretations may be a bond between some aspect of the processing of a face, possible. Our account is based on a theory developed to and some aspect of the processing of the class of somatic address the altered behavior of some patients with frontal

86 Journal of Cognitive Neuroscience Volume 5, Number 1

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 (8b) t I Decision Somatosensory Prefrontal -- (’b) (parietal) premotor (occlplto- (2) Cortices parietal) ____1 _____. (6a)- Cortices -- Visual Cortices Prefronlal 4 (5a) (occlpllo- Cortices parietal) * Face * (1) Cortices (1) Reward (5b)

1 Somatic slate Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 I’ I(3)

Figure 2. Diagram of the neural structures possibly involved in the acquisition (A) and retrieval (B) of the face-affective valence bond. (A) Step 1 stands for presentation of face and associated reward, step 2 depicts -related signals arriving at occipitoparietal cortices, and reward generating a somatic state; step 3 represents activation of the dopaminergic system (DA), and monitoring of somatic state by parietal cortices; step 4 represents convergence of signals from occipitoparietal cortices (face-related) and DA and parietal cortices (somatic state-related), onto tail of caudate; step 5 represents reentrant signaling from caudate to prefrontal and premotor cortices. (B) Representation of the face (1) and its processing in visual cortices (2) leads to activation of caudate (3) and prefrontal cortices (4).Two mechanisms are proposed beyond prefrontal xctivation. One (5a-(la) is covert and biases the decision making by reactivation in somatosensory and prefronral cortices, without the obligate reenactment of a somatic state. An alternative mechanism (5b6b7b8b) calls for reenactment of a somatic state (Gb), read-out of that state by somatosensory cortices (7h), and subsequent biasing of other prefrontal cortices by parietofrontal projections.

lobe damage (Damasio, Tranel, & Damasio, 1990, 1991). even if it would normally be the preferred mediator of The current data of course do not speak directly to the such learning. notion of somatic states, but we find this a parsimonious and intriguing explanation. In any event, the discovery that Boswell can acquire a nonconscious bond between Acknowledgment new persons and affective valences is important, because Supported by NINDS Program Project Grant 19632. it exposes a form of learning that is possible even when the entire mesial temporal region is bilaterally destroyed. Reprint requests should be sent to Daniel Tranel, Department The current findings are important because they dem- of Neurology, University of Iowa Hospitals & Clinics, Iowa City, onstrate a form of learning that can occur without the Iowa 52242. support of the hippocampus, related cortices (e.g., en- torhinal and perirhinal), and amygdala. Our findings are consistent with recent evidence obtained in rats, which Notes indicates a limited role for the hippocampus in learning 1. The use of forced choice as a measure of “covert” recogni- associations between a sensory stimulus and an “emo- tion is based on the fact that Boswell’s confidence ratings in tional” state (Phillips & LeDoux, 1992; Kim & Fanselow, forced choice paradigms are zero, i.e., his answers are always given as “guesses.” This behavior, which is not surprising in 1992). Another intriguing result, however, regards the light of the severity of his amnesia, has remained unchanged amygdala. There is a considerable body of work showing in the hundreds of experiments that we have conducted with that the amygdala plays a critical role in emotional con- Boswell over more than a decade. In fact, on every trial Boswell ditioning (e.g., Phillips & LeDoux, 1992; Hatfield, Gra- has to be reminded of the instructions and then prompted to ham, & Gallagher, 1992; see LeDoux, 1990, for a review). choose an answer-his selections are truly “forced.” 2. Note that the caregiver for whom Boswell had appeared to Also, the amygdala has been implicated as an important develop a special preference, according to our field observa- structure in the linking of polymodal information, es- tions, was not included in the Caregivers set. This was done to pecially when there is an affective/emotional component avoid confounding the factor of affective valence in Experi- (e.g.,LeDoux, Romanski, & Xagoraris, 1989; Nishijo, Ono, ments #1 and #2, i.e., the target faces in these experiments & Nishino, 1988). Patient Boswell, however, has no amyg- were basically neutral with respect to affective valence. 3. Normal controls generate ratings in the range of 1.0-1.5 for dala on either side (Tranel & Damasio, 1989), and thus target and 5.0-6.0 for nontarget faces. the current findings indicate that the amygdala is not 4. Other forms of the prompt question were tried, e.g., necessary for the covert learning of affective valence, ‘Choose your favorite person,” and “Choose the person you

Tranel and Damasio 87

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021 would go to for help.” Boswell’s choices were not influenced membering and awareness as separate issues. In L. R. by these variations in question format. Squire & N. Butters (Eds.), Neuropsychologp of human 5. There are some important differences between our experi- memory. New York: Oxford University Press. ment and the “habit” experiment. In the latter, animals learn Johnson, M. K., Kim, J. K., & Risse, G. (1985). Do alcoholic to connect the appearance of a given stimulus, which they do Korsakoff ’s syndrome patients acquire affective reactions? not need to learn, with a standard motor response (e.g., reach- Journal of Eqerimental : Learning, Memory, ing for food reward). Our experiment differs, in that (1) Bos- and Cognition, 1 I, 22-36. well’s responses during testing (indicating which face he Kim, J. J., & Fanselow, M. S. (1992). Modality-specific retro- preferred) never occurred during acquisition of the “habit,” grade amnesia of fear. Science, 256, 675-677. and (2) Boswell’s response is deliberated consciously, even eDoux, J. E. (1990). Information flow from sensation to though the conscious decision making is based on a bond he emotion: Plasticity in the neural computation of stimulus acquired nonconsciously. value. In M. Gabriel & J. Moore (Eds.), Learning and com- putational neuroscience:Foundations of dpative net- works (pp. 3-52). Cambridge, MA. MIT Press. REFERENCES ,eDoux,J. E., Romanski, L. M., & Xagoraris, A. E. (1989). In-

delibility of subcortical emotional memories. Journal of Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/5/1/79/1755039/jocn.1993.5.1.79.pdf by guest on 18 May 2021 Alexander, G. E., Delong, M. R., & Strick, P. L. (1986). Parallel Cognitive Neuroscience, I, 238-243. organization of functionally segregated circuits linking Mishkin, M., Malamut, B., & Bachevalier, J. (1984). Memories basal ganglia and cortex. Annual Review of Neuroscience, and habits: Two neural systems. In G. Lynch, J. L. McGaugh, 9,357-381. & N. M. Weinberger (Eds.), Neurobiology of learning and Damasio, A. R., Damasio, H., Tranel, D., Welsh, K., & Brandt, memory (pp. 65-77). New York: Guilford Press. J. P. (1987). Additional neural and cognitive evidence in Nauta, W. J. H., & Mehler, W. R. (1966). Projections to the patient DKB. Society for Neuroscience Abstracts, 13, 1452. lentiform nucleus in the monkey. Brain Research, I, 3-42. Damasio, A. R., Eslinger, P. J., Damasio, H., Van Hoesen, G. W., Nishijo, H., Ono, T., & Nishino, H. (1988). Single neuron re- & Cornell, S. (1985). Multimodal amnesic syndrome follow- sponses in amygdala of alert monkey during complex sen- ing bilateral temporal and basal forebrain damage. hchives sory stimulation with affective significance. Journal of OfNeurologv, 42, 252-259. Neuroscience, 8, 3570-3583. Damasio, A. R., Tranel, D., & Damasio, H. (1989). Amnesia Phillips, R. G., & LeDoux, J. E. (1992). Differential contribu- caused by herpes simplex encephalitis, infarctions in basal tion of amygdala and hippocampus to cued and contextual forebrain, Alzheimer’s disease, and anoxia. In F. Boller and fear conditioning. Behavioral Neuroscience, 106, 274-285. J. Grafman (Eds.), Handbook of (Vol. 3, Selemon, L. D., & Goldman-Rakic, P. S. (1985). Longitudinal pp. 149-166). Amsterdam: Elsevier. topography and interdigitation of corticostriatal projections Yamasio, A. R., Tranel, D., & Damasio, H. (1990). Individuals in the rhesus monkey. Journal of Neuroscience, 5, 776- with sociopathic behavior caused by frontal damage fail to 794. respond autonomically to social stimuli. Behavioural Brain Sergent, J., & Poncet, M. (1990). From covert to overt recog- Re.cearcb, 41, 81-94. nition of faces in a prosopagnosic patient. Brain, 113, 989- )amasio, A. R., Tranel, D., & Damasio, H. (1991). Somatic 1004. markers and the guidance of behavior: Theory and prelimi- Squire, L. R., & Zola-Morgan, S. (1991). The medial temporal nary testing. In H. S. Levin, H. M. Eisenberg, & A. L. Benton lobe memory system. Science, 253, 1380-1386. (Eds.), Frontal lobe function and dysfunction (pp. 217- Tranel, D., & Damasio, A. R. (1985). Knowledge without 229). New York: Oxford University Press. awareness: an autonomic index of facial recognition by Damasio, H., & Damasio, A. R. (1989). Lesion analysis in neu- prosopagnosics. Science, 228, 1453-1454. rop.vcho1o.g. New York: Oxford University Press. Tranel, D., & Damasio, A. R. (1987). Evidence for covert rec- Damasio, H., & Frank, R. (1992). Three-dimensional in vivo ognition of faces in a global amnesic patient. Journal of mapping of brain lesions in humans. Archives of Neurology, Clinical and Eqerimental Neuropsychology, 9, 15. 49, 137-143. Tranel, D., & Damasio, A. R. (1988). Nonconscious face recog- Garcia, J. (1990). Learning without memory. Journal of Cog- nition in patients with face agnosia. Behavioural Brain Re- nitiue Neuroscience, 2, 287-305. search, 30, 235-249. Goldman-Rakic, P. S. (1987). Circuitry of primate prefrontal Tranel, D., & Damasio, A. R. (1990). Covert learning of emo- cortex and regulation of behavior by representational tional valence in patient Boswell. Journal of Clinical and memory. In F. Plum (Ed.), Handbook ofphyszology: ner- Eqmimental Neuropsyhology, 12, 27. vous system, Vol. V. Higher functions of the brain (pp. 373- Tranel, D., & Damasio, H. (1989). Intact electrodermal skin 417). Bethesda, MD: American Physiological Society. conductance responses after bilateral amygdala damage. Greve, K. W., & Bauer, R. M. (1990). Implicit learning of new Neuropsychologia, 27,381-390. faces in : An application of the mere-expo- Van Hoesen, G. W., Yeterian, E. H., & Lavizzo-Mourey, R. sure paradigm. Neuropsychologia,28, 1035-1041. (1981). Widespread corticostriate projections from tem- Hatfield, T., Graham, P. W., & Gallagher, M. (1992). Taste-po- poral cortex of rhesus monkey. Journal of Comparative tentiated odor aversion learning: Role of the amygdaloid Neurology, 199, 205-219. basolateral complex and central nucleus. Behavioral Neu- Yeterian, E. H., & Van Hoesen, G. W. (1978). Cortico-striate roscience, 106, 286-293. projections in the rhesus monkey: The organization of cer- Ilinsky, I. A,, Jouandet, M. L., & Goldman-Rakic, P. S. (1985). tain cortico-caudate connections. Brain Research, 139, 43- Organization of the nigrothalamocortical system in the rhe- 63. sus monkey. Journal of Comparative Neurology, 236,315- Zajonc, R. B. (1980). Feeling and thinking: Preferences need 330. no inferences. American Psychologist, 35, 151-175. Jacoby, L. L. (1984). Incidental vs. intentional retrieval: Re-

88 Journal of Cognitive Neuroscience Volume 5,Number I

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1993.5.1.79 by guest on 25 September 2021