(as described by Oscar-Berman and ZOLA-MoR~AN [1]). The experimenter and the subject sat facing each other across a table and were separated by a wood ...
Neuropsychologia, Vol. 18, pp. 513 to $25. Pergamon Press Ltd. 1980. Printed in Great Britain.
0028--3932/80/1001--0513502.00/0
COMPARATIVE NEUROPSYCHOLOGY AND KORSAKOFF'S SYNDROME. II--TWO-CHOICE VISUAL DISCRIMINATION LEARNING MARLENE OSCAR-BEILMAN Boston Veterans Administration Medical Center and Dopm-tment of Neurology, Boston University School of Medicine and STUART M. ZOLA-MORGAN Veterans Administration Hospital, San Diego, and Department of Psychiatry, University of California at San Diego
(Received 20 December 1979) ~et--Pefformar~ by 12 alcoholic Korsakoff patients was compared to that of 15 patients with Huntington's disease, 11 patients with Broca's aphasia, and 13 alcoholic control subjects on two types of two-choice visual discrimination problems which, in brain-damaged monkeys, successfully dissociate deficits of visual perception from those of visual associative learning. Korsakoff and Huntington patients w~e slow to learn individual discriminations, but eventually performed well on them. Both groups were severely, but differently, impaired on concurrent discrimination tasks in which several pairs of stimuli are presented together within each testing session. It is suggested that the Korsakoffs have a severe deficit of associative learning combined with abnormal sensitivity to interference, whereas the Huntingtons are impaired primarily in visually-based associative learning. Deficits in Korsakoff patients are related to their widespread brain damage which includes limbic-system structures and prefrontal cortex. Huntington deficits are related to atrophy of frontal and temporal cortex, perhaps consequent to atrophy of the caudate.
IN A PREVIOUSpaper [1], we showed a severe impairment by Korsakoff patients in the ability to acquire new stimulus-reinforcement associations as measured by performance on visual and spatial reversal tasks. Since the Korsakoff deficit was at least as severe on visual reversals as it was on spatial reversals, we felt that it would be informative to explore the possibility that Korsakoffs' performance might resemble the abnormal visually-based behavior of monkeys with lesions of the temporal lobes. This is especially relevant in light of (a) Hol~L's [2] suggestion that at least one aspect of human amnesia might be related to damage of visual association cortex or projection fibers in the temporal stem adjacent to the hippocampus, Co) MZSHKIN'SSuggestion [3, 4] that human amnesics may suffer from an inability to form functional links between sensory association areas (for appreciation of stimulus qualities) and the limbic system (for appreciation of reward value), and (c) considerable evidence of direct neuroanatomical connections between regions implicated in Korsakoff memory impairments (e.g., dorsomedial thalamic nuclei, hippoeampus, pulvinar) and cortex of the temporal lobe [2, 5-8]. The present study was aimed at exploring these hypotheses further by using two procedures which have successfully dissociated deficits of selective visual perception from those of visual association learning in monkeys with temporal lobe lesions [9-11 ]. One task involves the learning of two-choice visual discriminations, each (in succession) to some predetermined criterion; these are called individual pair (IP) 513
514
MXI~LL~- OSCAR-Bm~a,~ and STUNT M. ZOLA-MoI~oAN
discriminations. The other task, called concurrent pair (CP) discrimination learning, also involves the learning of pairs of visual stimuli, but a number of different pairs are presented in random order within each session such that several discriminations are learned simultaneously. Iw~a and MxSHg~,~[11] and Cowry and GRoss [9] found severe impairments in acquisition of IP discrimination problems after lesions of foveal prestriate cortex (the portion of the circumstriate belt which receives projections from the cortical representation of the fovea in striate cortex); this was interpreted to reflect a deficit in discriminating or in detecting differences between the stimuli. By contrast, lesions of inferotemporal association cortex (corresponding to VoN BO~N and BAmEY's area TE, [12]) produced selective impairment on performance of the CP problems. Learning CP problems requires that subjects form several associations simultaneously; therefore, performance on the CP task is assumed to be a more sensitive measure 0f associative-learning capacity than IP problems. In addition, since considerable proa~tive interference is likely to be generated while CP discriminations are being performed, we thought that the task might be sensitive to the disruptive proactive interference effects observed for verbal learning by Korsakoff patients [13]. We reasoned that even if the rate of acquisition or the final level of performance by Korsakoffs was normal on IP discriminations, as soon as the problems are presented in a concurrent format, more interference would be generated, and performance would deteriorate. We also wanted to compare performance on CP problems which already had been learned individually to criterion, with acquisition of new CP problems in order to see if interference is greater during initial learning than it is with familiar stimuli. The genera/approach in this paper, comparative neuropsychology, is the same as that described in our previous paper [I]. That is, we are using performance measures on human neurological patients which are similar to those used with brain-damaged experimental animals. It is hoped that when performance requirements are in common for humans and nonhuman animals, assessment of psychological components will be measured more analogously than when vastly different procedures are used, and that the behavioral relevance of homologous brain structures would be clearly comparable. METHOD $~/ects The same 51 patients described in our previous paper [1] were tested on the visual discrimination problems of this experiment. They were selected from inpatient populations at the Boston or Brockton Veterans Administration Medical Centers. All of the patients were n~les, and their conditions were diagnosed by the r~uroiotDr and/or psychology services of at least one of the two participating hospitals. There were four patient groups: (a) 12 Korsakoff subj~ts (K) with a mean a ~ of 53.6 years (range 40 to 72). They all had a hlstory of chronic alcoholism and showed clinically significant memory deficits; (b) 15 patients with Huntington's disease (HD) with a mean age of 46.1 years (range 30-65); (¢) l l aphasies (AP) with a mean age of 51.4 years (range 38-67) and with varying degrees of right-sided hemiplegia and left-anterior frontal danmge causing Broca's aphasia; and (d) 13 alcoholics (AL), mean age 40.1 years ( r a n ~ 25-50), with no clinical evidence of memory impairment. As noted in our previous paper [1], three of the four groups (K, H D and AP) were chosen because they have relatively well-defined areas of brain damage which cause vastly different clinical symptoms. In addition to providing behavioral data which may he helpful in understanding cognitive deficits in Huntington's disease and aphasia, results of tests on the latter two groups provide data on brain-damag~xl controls for Korsakoff's disease. The alcoholic patients were included in order to rule out the effects on behavior of alcoholism per se. Table I, taken from our previous paper [1], summarizes the performance of the patients on three standardized tests of inte||eetual function, the Raven Progressive Matrices [14], the W~hsler Adult Intell i g e r ~ Scale [15] and the Weehsler Memory Scale [16]. Missing scores signify difficulty in obtaining the information due to a patient's symptomatololB~, to early discharge, etc.
Age
53.6 40-72 12
46.1 30-65 15
51.4 38-67 11
40.1 25-50 13
Group
Korsakoffs Mean Range N
Huntingtons Mean Range N
Aphasics Mean Range N
Alcoholics Mean Range N 61 2-95 12
33 3-64 8
14 0-71 14
45 22-80 9
Percentile Raven Matrices
113 107-120 2
? ? 0
84 72-104 10
105 92-115 12
VIQ
104 96-112 2
91 80-105 9
76 47-104 8
103 91-134 12
WAIS scores PIQ
111 109-114 2
? ? 0
81 59-103 8
104 94-124 12
Full-scale
5.8 -13
? -0
4.7 -15
2.8 -12
Info.
13.0 13
13
0
0 4.9
?
15
15 ?
7.8
12
12 4.4
6.4
2.6
13
6.7
0
?
15
5.3
12
6.3
13
5.3
0
?
15
3.6
12
3.9
Wechsler memory scale subtests Digits Orien. P.A. For. Back.
Table 1. Performance of subjects on standardized tests of intellectual function
t~
t~
i
516
MARLENE OSCAR-BERMAN and STUART M. ZOLA-MORGAN
We also tested 21 patients (normal controls) from the medical wards of the Boston VA Medical Center on the concurrent task of the present experiment. These patients were hospitalized for non-neurological problems such as broken limbs, urinary infections, etc., and had no known CNS pathology. They were males matched as closely as possible to our other patient groups for age (mean 40.8 years, range 22-60), education (mean 12.4 years of schooling; range 8--14 years), and socioeconomic background. More detail on the rationale for including normal controls, and the testing methods used with these subjects follows the Procedure section.
Apparatus A modified version of the Wisconsin General Test Apparatus was adapted for use with human subjects (as described by Oscar-Berman and ZOLA-MoR~AN [1]). The experimenter and the subject sat facing each other across a table and were separated by a wood frame approximately 61 cm wide and 53 cm high. A black curtain was anchored to the frame in such a way that the experimenter, by pulling a drawstring, could raise the curtain enough to reveal the stimulus board which was then slid forward toward the subject. The board on which the stimuli rested measured approximately 53 cm × 28 cm. The stimuli (described below) covered two reinforcement wells about 24 cm apart from center to center. The entire stimulus board was covered with soft terrycloth to prevent auditory cues during baiting. While the curtain was in the lowered position the subject could see neither the stimuli nor the experimenter. When the curtain was raised for each trial, the subject could see the stimuli and the hand of the experimenter, but not the experimenter's face. Hence, the possibility of inadvertent cueing by the experimenter was minimized. General Procedure A subject was seated opposite the experimenter with the curtain lowered between them. The experimenter then explained to the subject in very general terms the requirements of the task: "Mr. J., I'm going to show you two things on black plaques. Underneath one of them is a penny. I want you to try to get the penny every time the curtain goes up. When you find a penny, put it in the box next to you, and at the end of the session you may keep all the money you've made. If you want to stop at any time, we can. All right ? Remember your task is to try to get the penny every time the curtain goes up. There'll always be a penny under one of the things I show you. Any questions ?" The first trial then was initiated by raising the curtain and sliding the stimulus board forward while the experimenter reminded the subject again: "Remember, you want to get a penny every time." When the subject made his response, the stimulus board was withdrawn, the curtain was lowered and the stimuli and reinforcer were set up for the next trial. The inter-trial interval was approximately 10 sec. Subjects in the four patient groups were tested twice on each of the two types of two-choice visual discrimination talks, individual and concurrent. The first test on each task type employed novel stimuli with which the subjects had no previous experience. By the time the second test on each task was presented to the subjects, the stimuli were now familiar (because they had been used previously with identical pairings and reward values). Order of presentation of tasks is presented in Table 2. Note that prior to the start of this experiment (within a day or two) all subjects had had previous experience with spatial and visual reversals
(see [1]). Typically) testing would last for about two hours in the morning, and for another two hours later in the day. Subjects could stop or take a break whenever they wished, but most subjects completed the experiment within two days. I. Individual Pairs. The groups were first tested successively on six individual discrimination problems (IP-1). These discriminations consisted of unfamiliar pairs of stimuli made from 2-dimensional patterns or 3-dimensional junk objects. The second set of individual discrimination problems (IP-2) consisted of six pairs of 3-dimensional objects to which the subjects had been exposed on a preceding (concurrent) discrimination task (CP-I to be described below). Subjects were reminded that their task was to find a penny under one of the two stimuli on every trial. Then they were tested on the first pattern to a learning criterion of nine correct responses in a block of 10 trials, or to a failure criterion of 300 trials. Upon reaching criterion on the first pair, testing was begun on the next pair. If a subject failed to reach learning criterion at the end of 300 trials he was shown the solution and given an additional few trials to ensure that he understood the solution before going on to the next pair. Testing proceeded in the same manner for all six pairs of IP-I stimuli and for the six pairs of IP-2 stimuli. 2. Concurrent Pairs. The subjects were tested on a novel and a familiar form of the concurrent pairs discrimination [7]. On each of these tasks, six different stimulus pairs were presented intermingled during the testing session so that a subject was required to learn several visual problems simultaneously. The first concurrent task consisted of six pairs of novel stimuli which the subject had not seen before (CP-1). The second concurrent task (CP-2) consisted of six pairs of stimuli drawn from previous discrimination problems on which the subjects had already been tested (i.e., IP-I). Table 2 indicates the relationship among the stimuli and between the two types of tasks. The concurrent tasks were administered in a manner described by C o w r y and GROSS [9]. On each trial a single pair was presented with the spatial position of the positive and negative
13 12 7.
11 5
2
11
11
11 11
!1
5 (6) I1
6 (5)
Groups Aphasics
1
11
11
13 13
13
7 (6) 13
6 (7)
Alcoholics
--
--
21
---
21
---
--
Normals
*These tasks are described in an earlier paper by Oscar-Berman and Zola-Morgan [I]. has indicated by the arrows in the left hand margin, the stimuli used in IP-I were subsequently used in CP-2, and the stimuli used in CP-I were subsequently used in IP-2 (with identical pairings and reward contingencies). ©Ten of the 21 patients in the Normal group received the CP-I task followed by the CP-2 task; the remaining II patients in the Normal group were tested on these tasks in the reverse order. SThese tasks are described in a forthcoming paper b)/Oscar-Berman, Oherg and Zola-Morgan [29].
Concurrent visual discriminations, familiar pairs (CP-2) b Individual pairs of visual discriminations, familiar (IP-2) 9. Retest of failed visual reversals"
12
15 15
12 12
'o:
15
12
7 (8) 15
7 (7)
4. Concurrent visual discriminations, novel pairs (CP-I) © Delayed alterationd Delayed response e
5 (7)
Huntingtons
7 (4) 12
Spatial (or visual)reversals*
Korsakoffs
2. Visual (or spatial)reversals' 3. Individual pairs of visual discriminations, novel (IP-I) b
1.
Order of tasks
Table 2. Order of tests and number of subjects participating in each group
-,,.I
518
M~Lm~ Oscxit-~l~
and STUARTM. ZOLA-MoRGAN
stimuli randomized across trials. The sequence of presentation of the pairs also was random, with the constraint that each pair appear five times in a block of 30 trials. Whenever a subject displaced the positive member of any pair, he found a penny in the well. Training continued to a learning criterion of 27 correct respomes in a block of 30 trials or to a failure criterion of 600 trials2
Procedure for normal control subjects Since order of testing was one of our independent variables in the visual discrimination problems, it was important that the difficulty level of stimulus pairs presented first (i.e., the IP-1/CP-2 set) be equivalent to that of the stimulus pairs presented second (the IP-2/CP-I set). To evaluate stimulus difficulty, we ran 21 normal control subjects (divided arbitrarily into two groups) on the concurrent version of the discriminations. The CP paradigm was employed here because we wanted to assess difficulty level among stimulus sets without the ceiling effects which would be likely with normal subjects in a straightforward IP paradigm. Both of the normal groups received both sets of stimuli, but in different orders. Ten subjects, N e t , were given the CP-I problem first; this was the task that our four main groups got prior to IP training with the same stimuli. After reaching criterion on CP-1, these same normais got the CP-2 problems. The second group of 11 subjects, NC=, got their concurrent training in the opposite order, i.e., CP-2 followed by CP-1. Results indicated that although the two NC groups were tested initially with different sets of stimuli, their performance levels on first-le.arned problems were indistinguishable; similarly, when performance on secondlearned problems was compared, the two groups did not differ. Only order of testing, not stimulus sets, revealed differences between the groups. Thus, the NCz group made more errors on CP-2 (their first set of concurrent stimuli) than did the NCt group (who got that same set second; t (19) == 2.70,P