- Email: [email protected]
Hemifield-specific visual recognition memory impairments in patients with unilateral temporal lobe removals
\ullrOl~vlcholo~;itc i
~
Pergamon
PII: S0028 3932(97)000~C 6
\ ' o l 3"~. No. ~) p p 131 [ 1315, I9LJ7 I¢)q 7 ENc~icr Science I td. All right~ reserved }:'tinted in (}real Blilattl 0(128 3t)32 q 7 5 1 7 0 0 +[L00
Hemifield-speeifie visual recognition memory impairments in patients with unilateral temporal lobe removals J. H O R N A K , * S. O X B U R Y , f J. OXBURY,i" S. D. IVERSEN* and D. GAFFAN*:k *Department of Experimental Psychology, University of Oxford, Oxford. U.K.: and +Department ol'Neurolog~, Radcliffe Infirmary, Oxford, U.K. ( Receit,ed 13 /~,hruar v 1997: accep:ed 22.4pill 1997)
Abstract Recent evidence on visual neglect suggests that each hemisphere maintains a retinotopically organized representation of the visual world contralateral to the current fixation point and that this representation is based not only on analysis of the current retinal input but, equally importantly, on information retrieved from memory. This idea predicts that unilateral damage to memory systems should produce a lateralized impairment of memory for the retinotopically contralateral visual world. To test this prediction we examined visual recognition memory performance in the left and right visual hemifields of patients who had undergone partial unilateral temporal lobe removals for the relief of epilepsy, either in the left hemisphere (n = 5) or the right (n = 5). The patients were given complex artificial scenes to remember, constructed of independent left and right halves, and were then tested for recognition of the left and the right halves separately. Stimuli were exposed tachistoscopically throughout and fixation was maintained on a central position. Patients made significantly more errors with half-scenes in the hemifield contralateral to their removal than in the ipsilateral hemifield, an increase of 50% in the error rate on average. The effect was seen equally in patients with left and right removals. This finding supports the idea that visual memory retrieval is retinotopically organized. ~ 1997 Elsevier Science Ltd,
Key Words: hemifield specific memory; visual neglect; temporal lobe removal.
lect patients show unilateral disorders of retrieval from visual long-term memory [1], a recent investigation of visual neglect in the monkey [7] concluded that neglect is a representational impairment. According to this view, the posterior, retinotopically organized part of the visual cortex of each hemisphere maintains a representation of the halt" of the visual world that is contralateral to the current fixation point, and this representation is normally based not only on analysis of the current retinal input, but also on retrieval of visual memories. This view implies that, in the case where the representation of the visual world is based on retrieval of information from longterm memory, that information is retrieved specifically into either the left or the right posterior visual cortex, according to the left or right position of the remembered visual features in relation to the current visual fixation point at the time of the memory retrieval. Such retrieval requires an interaction between the temporal lobe structures which are known to be important for memory and the more posterior visual cortex into which the visual information is proposed to be retrieved. Since the pathways by which the temporal lobe call influence the pos-
Introduction
Several studies have demonstrated material-specific memory impairments in patients who have undergone unilateral temporal lobe removals for the relief of epilepsy. Patients with left-sided removals are impaired in verbal memory, and patients with right-sided removals are impaired in visual and spatial memory [3, 10, I 1 15]. The present experiment explored the possibility thal, in addition to such material-specific effects, these patients naight show visual memory impairments specifically in the visual hemifield that is contralateral to their removal. The possibility of hemifield-specific visual memory impairments is strongly suggested by some indirect evidence obtained from recent experiments with monkeys. Following upon the well-known demonstration that neg-
Address for correspondence: D. Gaff'an, Department of Experimental Psychology, South Parks Rd, Oxford OXl 3U D. UK.; c-mail: gaffan~opsy.ox.ac.uk: tel.: 01865-271349: fax: 01865-310447. 1311
1312
J. Hornak el al./Hemifield-specific memory
terior visual cortex are p r e d o m i n a n t l y i n t r a h e m i s p h e r i c [2], one should therefore expect that each t e m p o r a l lobe w o u l d p l a y a specific role in the retrieval o f visual m e m ories for objects that are in the r e t i n o t o p i c a l l y cont r a l a t e r a l h a l f o f the visual world. To test this idea, we gave patients a visual recognition m e m o r y test tachistoscopically. Visual scenes consisting o f meaningless c o l o u r e d shapes (Fig. 1) were presented for m e m o r i z a t i o n while the p a t i e n t fixated a central spot. Subsequently, m e m o r y for the left h a l f a n d the right h a l f o f the scenes was tested separately, by presenting just one h a l f o f such a scene, again with t a c h i s t o s c o p i c e x p o s u r e a n d fixation o f the same central spot, a n d asking w h e t h e r the half-scene at the r e t e n t i o n test was half o f one that h a d been p r e s e n t e d for learning o r a new one. We used meaningless shapes in o r d e r to reduce the l i k e l i h o o d that patients c o u l d re-code the visual stimuli into a verbal description.
Methods Sul~/ects A consecutive series of patients was recruited from the Epilepsy Surgery Service at the Radcliffe Infirmary. Any patient returning postoperatively for assessment or for research investigations was invited to take part in the present study and none refused. The interval between surgery and testing ranged from 3 months to 5 years. In this report individual patients are identified by letters of the alphabet (A through J). There were nine female patients and one male (patient E). Their ages ranged from 20 to 41 years at the time of testing. Language dominance was left-sided in all but case B, who had bilateral language representation. Neuropsychological assessment Pre-operatively IQ scores were in the range 77-I 13, and postoperatively they were in the range 84-130. Most candidates for
Fig. 1. An artificial complex scene of the kind that was exposed taschistoscopically to be remembered. The scene is rendered in shades of grey but in the original was coloured. An algorithm based on a random number generator has produced coloured shapes chosen independently in the left and in the right halves of the scene. At the centre a small letter H is displayed and the patient has to report this letter (which can be either C or H, varied randomly from trial) to ensure central fixation.
J. Hornak et al./Hemifield-specific memory
1313
Control O/ifi.vation and trial onset
left-sided surgery had verbal memory impairments and in some cases these were exacerbated by the operation. Nonverbal memory assessment did not discriminate between patients with leftsided and right-sided pathology pre-operatively; nonverbal memory improved after left-sided removals in some patients, but deteriorated mildly in some patients with right-sided removals. These are typical findings. Some of the postoperative neuropsychological assessments are presented in Table 1. The table presents verbal and performance IQ scores (WAIS) and the recall scores in delayed paragraph recall (WMS-R) and in delayed reproduction of the Rey-Osterrieth figure. None of these patients showed visual agnosia clinically, and in accordance with this they all performed normally in the picture completion subtest of the performance IQ test.
On every trial in both of the tasks described below (preliminary test and memory testing) the patient initiated the trial by touching the computer keyboard. The stimulus for that trial was then displayed for 200 msec. On every such trial there was a letter in the middle of a small box in the centre of the screen (see Fig. 1) and the letter was either H or C, determined randomly at each trial. The patient was required to report the letter first, before any o t h e r j u d g m e n t about the scene. The box alone appeared first, and indicated that the trial was ready to be initiated. Practice at simply naming letters presented tachistoscopically in the box, with no other stimuli on the screen, was given for at least 12 trials before introducing the further requirements described in the following tasks. Since the letter in the box was too small to be identified in peripheral vision, this requirement ensured that the central box was fixated at the beginning of every trial. Further, the 200 msec exposure duration ensured that eye movements could not be made away from the central box during the exposure [16]. The letter was correctly identified by every patient on every trial in the following tasks.
Surgery The unilateral removals were on the left side in patients A, D, E, H and .1 and on the right side in the others. Patient A underwent a small-sized partial left temporal lobectomy; all other left-sided removals were amygdalohippocampectomies with an approach through the Sylvian fissure [17]. A m o n g the right-sided removals the patients B and G had amygdalohippocampectomy, while patients C, F and I had an en bloc" temporal lobectomy including the amygdala and the anterior 2 3 cm of the hippocampus.
Preliminary teet: colour idenlilication At each trial in this preliminary test a coloured patch (quarter of a disk) was presented in one of the four corners of the display area, chosen sit random, and the patient was asked to report (alter first reporting the letter in the central box) the colour of the patch which varied randomly between red, green and blue. The nearest part of the patch was 120mm from the fixation point and the patch extended to the corner of the display area, 165 mm from the fixation point. This task served to familiarize the patients with the requirement to attend to tachistoscopically exposed peripheral stimuli at the same time as reporting the letter in the central box, and also tested whether the patients" visual fields were intact over the area of the stimuhis display.
Apparatus and stimulus material Coloured visual images were generated by a computer and displayed on a monitor screen. The rectangular display area of the monitor screen was 330mm in diagonal and was placed 330mm from the patient's eyes. Figure 1 shows a grey-scale rendering of one of the scenes (whole-screen visual displays) that was presented for learning. Each half of each scene was constructed of a random number of ellipse segments of random size, orientation and colour, with an algorithm similar to that used in a previous experiment [4] except that, unlike the previous experiment, there were no alphanumeric characters in the scenes. The algorithm could generate a large number of unique scenes, and a different set of particular scenes was used for each patient. The left and right halves of each scene were constructed independently of each other. Both at learning trials and in retention tests, the screen was blank grey between trials. To ensure that no part of the left and right halves of the scene should fall into the opposite visual field, every scene had a blank grey funnel shape in the middle, in which no coloured ellipse segments could be displayed. Thus, no part of the half-scene could be less than 30 mm from the fixation point. In addition some other stimulus material was used in the preliminary test (see below) to ensure that the patients' visual fields were intact over the area of the display screen.
Recognition memory. h'arninc, trials Scenes were presented for memorization in sets of t\mr, and each set of four was tested for retention (see below~ before the next set of four began. In each repetition of the learning trials for each set the patient saw the four scenes in random order, one trial for each scene. Each trial was self-initiated and displayed the scene for 200 msec, as described above. The same set of four was then displayed again in random order for a number of repetitions before retention was tested. The number of repetitions was determined on-line by the experimenter for each set for each patient, in the light of the individual patient's overall proficiency and confidence at the task: the median number of repetitions ~as 6.
Table 1. Post-operative results of neurospychological assessments Patients
A
B
C
D
g
F
G
H
I
J
Side of removal Verbal IQ Performance IQ Paragraph recall Rey Osterrieth
L 88 99 7 22
R 99 117 8 20
R 96 98 4 18
L 90 109 5 24
L 101 122 7 23
R 118 118 14 18
R 86 85 7 13
L 80 92 5 15
R 124 130 11 24
L 97 9(11 9 24
J. Hornak et al./Hemifield-specific memory
1314
Recognition memory: retention tests Each set of four scenes was tested immediately after the learning trials with that set. Trials at the retention test displayed only half of a scene, which could be either familiar (half of one of the four scenes that had been learned) or novel. Familiar halves were always presented for test in the same hemifield as during the learning trials. Both halves of each learned scene were tested, generating eight trials with familiar stimuli, and there were in addition eight trials with novel half-scenes, four on the left and four on the right. Thus the retention test for each set consisted of 16 trials in total, and these were administered in random order. At each trial the patient first reported the letter in the centre and then reported a judgment as to whether or not the half-scene was part of one that had been learned. Thus each set of scenes generated a total of 16 recognition memory judgments, half relating to the left and half to the right side of the display. During the retention tests the experimenter did not indicate whether the patient's recognition judgments were correct although general encouragement was given. The number of sets completed was on average 6 (range 4-10).
Results
Preliminary test: colour identification
to the patient's temporal lobe removal (contralateral or ipsilateral to the removal). It can be seen that every patient but one made more errors in the hemifield contralateral to the removal than in the ipsilateral hemifield. On average 9.6 errors were made with half-scenes ipsilateral to the removal, and 14.4 errors with half-scenes contralateral to the removal. This difference was statistically significant [t = 3.273, d.f. 9, P < 0.01 two-tailed]. As can be seen from the individual patients" results as identified by capital letters in Fig. 2, the size o f the difference between the hemifields seemed not to be affected by the side o f the removal or the extent o f the removal (see Subjects for information about the removals in the individual patients). The difference between contralateral and ipsilateral errors was on average 4.2 in the patients with right-sided removals and 5.4 in the patients with left-sided removals ( t < l ) . Additionally, there was no difference between the overall performance o f patients with right and left removals, combining results from the two hemifields; the patients with left hemisphere removals made on average 23.8 errors each in total and the patients with right hemisphere removals made on average 24.2 errors each (t < 1). (The median n u m b e r o f repetitions o f the learning trials was 6 in each group.)
All the patients reported the colour correctly on every trial. Discussion
Recognition memory Figure 2 shows the errors at retention tests for each patient, classified as to the side o f the display in relation
20
15
.o I0
o
I
r
Contra
Ipsi Hemifield
Fig. 2. Recognition memory errors in the hemifields contralateral and ipsilateral to the temporal lobe removal Each line shows results from one patient and the individuals are identified by letters, A-J.
The patients showed a clear impairment in visual recognition m e m o r y for half-scenes that were presented and tested contralateral to the side o f the temporal lobe removal (Fig. 2). This effect supports the idea, put forward in the introduction, that each temporal lobe has a specific role in visual m e m o r y for the half o f the visual world that is contralateral to the current fixation point. The effect appeared to be equally strong in patients with left and right temporal lobe removals. The patients were an unselected sample, and thus there is no reason to d o u b t that the effect observed is a general one in this population. Before the m e m o r y test, each patient was first shown to have intact visual fields in the area that was subsequently used for the recognition m e m o r y test, in that they could accurately report the colour o f a patch presented tachistoscopically in any o f the four corners of the display area. It m a y nevertheless be asked whether the m e m o r y impairments seen in these patients could be seco n d a r y to some more subtle form o f perceptual abnormality or attentional deficit in the hemifield contralateral to their temporal lobe removal. O u r simple test with coloured patches cannot exclude such a possibility. Furthermore, we have argued that there is not a sharp categorical distinction between m e m o r y impairments and perceptual impairments [5, 6]. Clearly, it will be important in future investigations to study a wider range o f m e m o r y tasks and also o f perceptual tasks with tachistoscopic lateralized presentations in patients similar to
J. Hornak et al./Hemifield-specific memory those studied here, in order to determine the relationship, if any, between lateralized perceptual and memory processes in these patients. The hemifield-specific effect in the present experiment outweighed any material-specific effect of the unilateral removals, since the average performance (combining the data from the two hemifields) of the patients with right hemisphere removals was not inferior to that of the patients with left hemisphere removals. It may be that the hemifield specificity of visual memory is a more powerful effect than the specificity of the right hemisphere for visual material. All but one of our patients had partial or complete removal of both the amygdala and the hippocampus in one temporal lobe. The effect seen could be due to the removal of either one of these structures or to the combined removal of both, or to the accompanying damage to cortical structures surrounding the amygdala and hippocampus. One likely possibility, however, is that at least part of the effect should be ascribed to the hippocarnpal damage, because bilateral section of the fornix (one of the main oulput pathways of the hippocampus) in the monkey produced an impairment in memory for artificial scenes similar to those used in the present experiment [41. In the future it will be important to follow up these findings with other patients who have unilateral lesions. such as unilateral tk~rnix damage, in structures which are important l~r memory. The remaining patient tpatient A) had a removal of the cortex in the temporal pole, sparing the hippocampus and amygdala. The removal in this patient is similar to the ablation of cntorhinal and perirhinal cortex in the monkey, which produces severe impairments in visual recognition memory lbr objects [8]. Memory for visual scenes in the monkey requires both the perirhinal cortex and the t\~rnix, and is severely impaired by disconnection of these two structures [91]. Taken together with the evidence on visual neglect [7], the present evidence suggests that widespread areas of cortex in each hemisphere interact with each other m memory retrieval to produce a retinotopically organized representation of the contralateral visual world.
Acknowledqemenl -This research was supported by the Medical Research (~ouncil.
References 1. Bisiach, E., Capitani, E., Luzzatti, C. and Perani, D., Brain and conscious representation of outside reality. Neuropxychologia 19, 543 551, 1981. 2. Felleman. D. J. and Van Essen, D. C., Distributed hierarchical processing in the primate cerebral cortex. Cerehral Cortex 1, 1-47, 1991.
1315
3. Frisk, V. and Milner, B., The role of the left hippocampal region in the acquisition and retention of story content. Neurop,v.rc/toloyia 28, 349 359, [990. 4. Gaff'an, D., Scene-specific memory for objects, a model of episodic memory impairnlent in monkeys with fornix transection. Journal 01 ('ognitire Neuroscience 6. 305 320, 1994. 5. Gaffan, I)., Associative and perceptual learning and the concept of memory systems. ( ' o q l l i l n ' d Brdin Research 5, 69 80, 1996. 6. Gaff'an, D. and Hornak. J., Amncsia and neglect, beyond the Delay Brion system and the Hebb synapse. Philosophical Tran.vacfion,v of dw Roral Societ r London, in press. 7. Gaffan, D. and Hornak, J., Visual neglect m the monkey, representation and disconnection. Brain, in press. 8. Gattim, D. and Murray, E. A.. Monkeys (Macaca 121.wicularis) with rhinal cortex ablations succeed in object discrimination learning despite 24-hr intertrial intervals and fail at matching to sample despite double sample presentations. &,harioral New'oscience 106, 30 38, 1992. 9. Gufl-an, I). and Parker, A., Interaction of perirhinal cortex ~ ith the fornix-lilnbria, melnory lot objects and object-in-place memory..lore'hal ol'Neuro.vcience 16,5864 5869, 1996. 10. Morris, R. G., Abrahams, S. and Polkey, C. E., Recognition memory for words and faces following unilateral temporal lobectomy. British Journal o/ Clinical P.u'cholog 3' 34, 571 576, 1995. 11. Pigott, S. and Milner, B., Memory for different aspects of complex visual scenes after unilateral temporal- or frontal-lobe resection. Neurolzvycholo.qia 31, I 15. 1993. 12. Rains, G. D. and Milncr, B., Right-hippocampal contralateral-hand effect in the recall of spatial location in the tactual modality. :\k'uropu'clu~loqia 32, 1233 1242, 1994. 13. Rams, G. D. and MiMer. B., Verbal recall and recognition as a function o1" depth of encoding in patients with unilateral temporal Iobectomy. Nem'ot~sychol<,.qia 32, 1243 1256, 1994. 14. Smith, M. L. and Milner. B., Right hippocampal ilnpairment in the recall of spatial location, encoding deficit or rapid forgetting? Neuropsycholo.oia 27, 71 81. 1989. 15. Szatko~ska, I.. Szymanska, O.. Bednarek, D., Skowronsku, R. and Grabowska, A.. Disturbances in time lilnited storage of sensory inl\)rmution after right temporal lobectom~.. Acre ,%,urohio/ogica Experimenlalis $6, 259 262. 1996. 16. Treislnan, A., Cavanagh, P., Fischer, B., Ramachundran. V. S., and Heydt. R., Form perception and attention striate cortex and beyond. In Ki,vual Perception- The Neurol#U'siologica[ Foumlalivnx, ed. g. Spilhnann, and J. S. Werner. Academic Press, N e w Y o r k , 1990, pp. 273 316. 17. Yasargil, M. G., Teddy. P. and Roth, P , Selective amygdalohippocampectomy, operative anatomy and surgical technique. ,4dram'e.v am/Technica/Standarrl, v in Neur,~,vm'.qerr 12, ¢)3 123. 1985.
~
Pergamon
PII: S0028 3932(97)000~C 6
\ ' o l 3"~. No. ~) p p 131 [ 1315, I9LJ7 I¢)q 7 ENc~icr Science I td. All right~ reserved }:'tinted in (}real Blilattl 0(128 3t)32 q 7 5 1 7 0 0 +[L00
Hemifield-speeifie visual recognition memory impairments in patients with unilateral temporal lobe removals J. H O R N A K , * S. O X B U R Y , f J. OXBURY,i" S. D. IVERSEN* and D. GAFFAN*:k *Department of Experimental Psychology, University of Oxford, Oxford. U.K.: and +Department ol'Neurolog~, Radcliffe Infirmary, Oxford, U.K. ( Receit,ed 13 /~,hruar v 1997: accep:ed 22.4pill 1997)
Abstract Recent evidence on visual neglect suggests that each hemisphere maintains a retinotopically organized representation of the visual world contralateral to the current fixation point and that this representation is based not only on analysis of the current retinal input but, equally importantly, on information retrieved from memory. This idea predicts that unilateral damage to memory systems should produce a lateralized impairment of memory for the retinotopically contralateral visual world. To test this prediction we examined visual recognition memory performance in the left and right visual hemifields of patients who had undergone partial unilateral temporal lobe removals for the relief of epilepsy, either in the left hemisphere (n = 5) or the right (n = 5). The patients were given complex artificial scenes to remember, constructed of independent left and right halves, and were then tested for recognition of the left and the right halves separately. Stimuli were exposed tachistoscopically throughout and fixation was maintained on a central position. Patients made significantly more errors with half-scenes in the hemifield contralateral to their removal than in the ipsilateral hemifield, an increase of 50% in the error rate on average. The effect was seen equally in patients with left and right removals. This finding supports the idea that visual memory retrieval is retinotopically organized. ~ 1997 Elsevier Science Ltd,
Key Words: hemifield specific memory; visual neglect; temporal lobe removal.
lect patients show unilateral disorders of retrieval from visual long-term memory [1], a recent investigation of visual neglect in the monkey [7] concluded that neglect is a representational impairment. According to this view, the posterior, retinotopically organized part of the visual cortex of each hemisphere maintains a representation of the halt" of the visual world that is contralateral to the current fixation point, and this representation is normally based not only on analysis of the current retinal input, but also on retrieval of visual memories. This view implies that, in the case where the representation of the visual world is based on retrieval of information from longterm memory, that information is retrieved specifically into either the left or the right posterior visual cortex, according to the left or right position of the remembered visual features in relation to the current visual fixation point at the time of the memory retrieval. Such retrieval requires an interaction between the temporal lobe structures which are known to be important for memory and the more posterior visual cortex into which the visual information is proposed to be retrieved. Since the pathways by which the temporal lobe call influence the pos-
Introduction
Several studies have demonstrated material-specific memory impairments in patients who have undergone unilateral temporal lobe removals for the relief of epilepsy. Patients with left-sided removals are impaired in verbal memory, and patients with right-sided removals are impaired in visual and spatial memory [3, 10, I 1 15]. The present experiment explored the possibility thal, in addition to such material-specific effects, these patients naight show visual memory impairments specifically in the visual hemifield that is contralateral to their removal. The possibility of hemifield-specific visual memory impairments is strongly suggested by some indirect evidence obtained from recent experiments with monkeys. Following upon the well-known demonstration that neg-
Address for correspondence: D. Gaff'an, Department of Experimental Psychology, South Parks Rd, Oxford OXl 3U D. UK.; c-mail: gaffan~opsy.ox.ac.uk: tel.: 01865-271349: fax: 01865-310447. 1311
1312
J. Hornak el al./Hemifield-specific memory
terior visual cortex are p r e d o m i n a n t l y i n t r a h e m i s p h e r i c [2], one should therefore expect that each t e m p o r a l lobe w o u l d p l a y a specific role in the retrieval o f visual m e m ories for objects that are in the r e t i n o t o p i c a l l y cont r a l a t e r a l h a l f o f the visual world. To test this idea, we gave patients a visual recognition m e m o r y test tachistoscopically. Visual scenes consisting o f meaningless c o l o u r e d shapes (Fig. 1) were presented for m e m o r i z a t i o n while the p a t i e n t fixated a central spot. Subsequently, m e m o r y for the left h a l f a n d the right h a l f o f the scenes was tested separately, by presenting just one h a l f o f such a scene, again with t a c h i s t o s c o p i c e x p o s u r e a n d fixation o f the same central spot, a n d asking w h e t h e r the half-scene at the r e t e n t i o n test was half o f one that h a d been p r e s e n t e d for learning o r a new one. We used meaningless shapes in o r d e r to reduce the l i k e l i h o o d that patients c o u l d re-code the visual stimuli into a verbal description.
Methods Sul~/ects A consecutive series of patients was recruited from the Epilepsy Surgery Service at the Radcliffe Infirmary. Any patient returning postoperatively for assessment or for research investigations was invited to take part in the present study and none refused. The interval between surgery and testing ranged from 3 months to 5 years. In this report individual patients are identified by letters of the alphabet (A through J). There were nine female patients and one male (patient E). Their ages ranged from 20 to 41 years at the time of testing. Language dominance was left-sided in all but case B, who had bilateral language representation. Neuropsychological assessment Pre-operatively IQ scores were in the range 77-I 13, and postoperatively they were in the range 84-130. Most candidates for
Fig. 1. An artificial complex scene of the kind that was exposed taschistoscopically to be remembered. The scene is rendered in shades of grey but in the original was coloured. An algorithm based on a random number generator has produced coloured shapes chosen independently in the left and in the right halves of the scene. At the centre a small letter H is displayed and the patient has to report this letter (which can be either C or H, varied randomly from trial) to ensure central fixation.
J. Hornak et al./Hemifield-specific memory
1313
Control O/ifi.vation and trial onset
left-sided surgery had verbal memory impairments and in some cases these were exacerbated by the operation. Nonverbal memory assessment did not discriminate between patients with leftsided and right-sided pathology pre-operatively; nonverbal memory improved after left-sided removals in some patients, but deteriorated mildly in some patients with right-sided removals. These are typical findings. Some of the postoperative neuropsychological assessments are presented in Table 1. The table presents verbal and performance IQ scores (WAIS) and the recall scores in delayed paragraph recall (WMS-R) and in delayed reproduction of the Rey-Osterrieth figure. None of these patients showed visual agnosia clinically, and in accordance with this they all performed normally in the picture completion subtest of the performance IQ test.
On every trial in both of the tasks described below (preliminary test and memory testing) the patient initiated the trial by touching the computer keyboard. The stimulus for that trial was then displayed for 200 msec. On every such trial there was a letter in the middle of a small box in the centre of the screen (see Fig. 1) and the letter was either H or C, determined randomly at each trial. The patient was required to report the letter first, before any o t h e r j u d g m e n t about the scene. The box alone appeared first, and indicated that the trial was ready to be initiated. Practice at simply naming letters presented tachistoscopically in the box, with no other stimuli on the screen, was given for at least 12 trials before introducing the further requirements described in the following tasks. Since the letter in the box was too small to be identified in peripheral vision, this requirement ensured that the central box was fixated at the beginning of every trial. Further, the 200 msec exposure duration ensured that eye movements could not be made away from the central box during the exposure [16]. The letter was correctly identified by every patient on every trial in the following tasks.
Surgery The unilateral removals were on the left side in patients A, D, E, H and .1 and on the right side in the others. Patient A underwent a small-sized partial left temporal lobectomy; all other left-sided removals were amygdalohippocampectomies with an approach through the Sylvian fissure [17]. A m o n g the right-sided removals the patients B and G had amygdalohippocampectomy, while patients C, F and I had an en bloc" temporal lobectomy including the amygdala and the anterior 2 3 cm of the hippocampus.
Preliminary teet: colour idenlilication At each trial in this preliminary test a coloured patch (quarter of a disk) was presented in one of the four corners of the display area, chosen sit random, and the patient was asked to report (alter first reporting the letter in the central box) the colour of the patch which varied randomly between red, green and blue. The nearest part of the patch was 120mm from the fixation point and the patch extended to the corner of the display area, 165 mm from the fixation point. This task served to familiarize the patients with the requirement to attend to tachistoscopically exposed peripheral stimuli at the same time as reporting the letter in the central box, and also tested whether the patients" visual fields were intact over the area of the stimuhis display.
Apparatus and stimulus material Coloured visual images were generated by a computer and displayed on a monitor screen. The rectangular display area of the monitor screen was 330mm in diagonal and was placed 330mm from the patient's eyes. Figure 1 shows a grey-scale rendering of one of the scenes (whole-screen visual displays) that was presented for learning. Each half of each scene was constructed of a random number of ellipse segments of random size, orientation and colour, with an algorithm similar to that used in a previous experiment [4] except that, unlike the previous experiment, there were no alphanumeric characters in the scenes. The algorithm could generate a large number of unique scenes, and a different set of particular scenes was used for each patient. The left and right halves of each scene were constructed independently of each other. Both at learning trials and in retention tests, the screen was blank grey between trials. To ensure that no part of the left and right halves of the scene should fall into the opposite visual field, every scene had a blank grey funnel shape in the middle, in which no coloured ellipse segments could be displayed. Thus, no part of the half-scene could be less than 30 mm from the fixation point. In addition some other stimulus material was used in the preliminary test (see below) to ensure that the patients' visual fields were intact over the area of the display screen.
Recognition memory. h'arninc, trials Scenes were presented for memorization in sets of t\mr, and each set of four was tested for retention (see below~ before the next set of four began. In each repetition of the learning trials for each set the patient saw the four scenes in random order, one trial for each scene. Each trial was self-initiated and displayed the scene for 200 msec, as described above. The same set of four was then displayed again in random order for a number of repetitions before retention was tested. The number of repetitions was determined on-line by the experimenter for each set for each patient, in the light of the individual patient's overall proficiency and confidence at the task: the median number of repetitions ~as 6.
Table 1. Post-operative results of neurospychological assessments Patients
A
B
C
D
g
F
G
H
I
J
Side of removal Verbal IQ Performance IQ Paragraph recall Rey Osterrieth
L 88 99 7 22
R 99 117 8 20
R 96 98 4 18
L 90 109 5 24
L 101 122 7 23
R 118 118 14 18
R 86 85 7 13
L 80 92 5 15
R 124 130 11 24
L 97 9(11 9 24
J. Hornak et al./Hemifield-specific memory
1314
Recognition memory: retention tests Each set of four scenes was tested immediately after the learning trials with that set. Trials at the retention test displayed only half of a scene, which could be either familiar (half of one of the four scenes that had been learned) or novel. Familiar halves were always presented for test in the same hemifield as during the learning trials. Both halves of each learned scene were tested, generating eight trials with familiar stimuli, and there were in addition eight trials with novel half-scenes, four on the left and four on the right. Thus the retention test for each set consisted of 16 trials in total, and these were administered in random order. At each trial the patient first reported the letter in the centre and then reported a judgment as to whether or not the half-scene was part of one that had been learned. Thus each set of scenes generated a total of 16 recognition memory judgments, half relating to the left and half to the right side of the display. During the retention tests the experimenter did not indicate whether the patient's recognition judgments were correct although general encouragement was given. The number of sets completed was on average 6 (range 4-10).
Results
Preliminary test: colour identification
to the patient's temporal lobe removal (contralateral or ipsilateral to the removal). It can be seen that every patient but one made more errors in the hemifield contralateral to the removal than in the ipsilateral hemifield. On average 9.6 errors were made with half-scenes ipsilateral to the removal, and 14.4 errors with half-scenes contralateral to the removal. This difference was statistically significant [t = 3.273, d.f. 9, P < 0.01 two-tailed]. As can be seen from the individual patients" results as identified by capital letters in Fig. 2, the size o f the difference between the hemifields seemed not to be affected by the side o f the removal or the extent o f the removal (see Subjects for information about the removals in the individual patients). The difference between contralateral and ipsilateral errors was on average 4.2 in the patients with right-sided removals and 5.4 in the patients with left-sided removals ( t < l ) . Additionally, there was no difference between the overall performance o f patients with right and left removals, combining results from the two hemifields; the patients with left hemisphere removals made on average 23.8 errors each in total and the patients with right hemisphere removals made on average 24.2 errors each (t < 1). (The median n u m b e r o f repetitions o f the learning trials was 6 in each group.)
All the patients reported the colour correctly on every trial. Discussion
Recognition memory Figure 2 shows the errors at retention tests for each patient, classified as to the side o f the display in relation
20
15
.o I0
o
I
r
Contra
Ipsi Hemifield
Fig. 2. Recognition memory errors in the hemifields contralateral and ipsilateral to the temporal lobe removal Each line shows results from one patient and the individuals are identified by letters, A-J.
The patients showed a clear impairment in visual recognition m e m o r y for half-scenes that were presented and tested contralateral to the side o f the temporal lobe removal (Fig. 2). This effect supports the idea, put forward in the introduction, that each temporal lobe has a specific role in visual m e m o r y for the half o f the visual world that is contralateral to the current fixation point. The effect appeared to be equally strong in patients with left and right temporal lobe removals. The patients were an unselected sample, and thus there is no reason to d o u b t that the effect observed is a general one in this population. Before the m e m o r y test, each patient was first shown to have intact visual fields in the area that was subsequently used for the recognition m e m o r y test, in that they could accurately report the colour o f a patch presented tachistoscopically in any o f the four corners of the display area. It m a y nevertheless be asked whether the m e m o r y impairments seen in these patients could be seco n d a r y to some more subtle form o f perceptual abnormality or attentional deficit in the hemifield contralateral to their temporal lobe removal. O u r simple test with coloured patches cannot exclude such a possibility. Furthermore, we have argued that there is not a sharp categorical distinction between m e m o r y impairments and perceptual impairments [5, 6]. Clearly, it will be important in future investigations to study a wider range o f m e m o r y tasks and also o f perceptual tasks with tachistoscopic lateralized presentations in patients similar to
J. Hornak et al./Hemifield-specific memory those studied here, in order to determine the relationship, if any, between lateralized perceptual and memory processes in these patients. The hemifield-specific effect in the present experiment outweighed any material-specific effect of the unilateral removals, since the average performance (combining the data from the two hemifields) of the patients with right hemisphere removals was not inferior to that of the patients with left hemisphere removals. It may be that the hemifield specificity of visual memory is a more powerful effect than the specificity of the right hemisphere for visual material. All but one of our patients had partial or complete removal of both the amygdala and the hippocampus in one temporal lobe. The effect seen could be due to the removal of either one of these structures or to the combined removal of both, or to the accompanying damage to cortical structures surrounding the amygdala and hippocampus. One likely possibility, however, is that at least part of the effect should be ascribed to the hippocarnpal damage, because bilateral section of the fornix (one of the main oulput pathways of the hippocampus) in the monkey produced an impairment in memory for artificial scenes similar to those used in the present experiment [41. In the future it will be important to follow up these findings with other patients who have unilateral lesions. such as unilateral tk~rnix damage, in structures which are important l~r memory. The remaining patient tpatient A) had a removal of the cortex in the temporal pole, sparing the hippocampus and amygdala. The removal in this patient is similar to the ablation of cntorhinal and perirhinal cortex in the monkey, which produces severe impairments in visual recognition memory lbr objects [8]. Memory for visual scenes in the monkey requires both the perirhinal cortex and the t\~rnix, and is severely impaired by disconnection of these two structures [91]. Taken together with the evidence on visual neglect [7], the present evidence suggests that widespread areas of cortex in each hemisphere interact with each other m memory retrieval to produce a retinotopically organized representation of the contralateral visual world.
Acknowledqemenl -This research was supported by the Medical Research (~ouncil.
References 1. Bisiach, E., Capitani, E., Luzzatti, C. and Perani, D., Brain and conscious representation of outside reality. Neuropxychologia 19, 543 551, 1981. 2. Felleman. D. J. and Van Essen, D. C., Distributed hierarchical processing in the primate cerebral cortex. Cerehral Cortex 1, 1-47, 1991.
1315
3. Frisk, V. and Milner, B., The role of the left hippocampal region in the acquisition and retention of story content. Neurop,v.rc/toloyia 28, 349 359, [990. 4. Gaff'an, D., Scene-specific memory for objects, a model of episodic memory impairnlent in monkeys with fornix transection. Journal 01 ('ognitire Neuroscience 6. 305 320, 1994. 5. Gaffan, I)., Associative and perceptual learning and the concept of memory systems. ( ' o q l l i l n ' d Brdin Research 5, 69 80, 1996. 6. Gaff'an, D. and Hornak. J., Amncsia and neglect, beyond the Delay Brion system and the Hebb synapse. Philosophical Tran.vacfion,v of dw Roral Societ r London, in press. 7. Gaffan, D. and Hornak, J., Visual neglect m the monkey, representation and disconnection. Brain, in press. 8. Gattim, D. and Murray, E. A.. Monkeys (Macaca 121.wicularis) with rhinal cortex ablations succeed in object discrimination learning despite 24-hr intertrial intervals and fail at matching to sample despite double sample presentations. &,harioral New'oscience 106, 30 38, 1992. 9. Gufl-an, I). and Parker, A., Interaction of perirhinal cortex ~ ith the fornix-lilnbria, melnory lot objects and object-in-place memory..lore'hal ol'Neuro.vcience 16,5864 5869, 1996. 10. Morris, R. G., Abrahams, S. and Polkey, C. E., Recognition memory for words and faces following unilateral temporal lobectomy. British Journal o/ Clinical P.u'cholog 3' 34, 571 576, 1995. 11. Pigott, S. and Milner, B., Memory for different aspects of complex visual scenes after unilateral temporal- or frontal-lobe resection. Neurolzvycholo.qia 31, I 15. 1993. 12. Rains, G. D. and Milncr, B., Right-hippocampal contralateral-hand effect in the recall of spatial location in the tactual modality. :\k'uropu'clu~loqia 32, 1233 1242, 1994. 13. Rams, G. D. and MiMer. B., Verbal recall and recognition as a function o1" depth of encoding in patients with unilateral temporal Iobectomy. Nem'ot~sychol<,.qia 32, 1243 1256, 1994. 14. Smith, M. L. and Milner. B., Right hippocampal ilnpairment in the recall of spatial location, encoding deficit or rapid forgetting? Neuropsycholo.oia 27, 71 81. 1989. 15. Szatko~ska, I.. Szymanska, O.. Bednarek, D., Skowronsku, R. and Grabowska, A.. Disturbances in time lilnited storage of sensory inl\)rmution after right temporal lobectom~.. Acre ,%,urohio/ogica Experimenlalis $6, 259 262. 1996. 16. Treislnan, A., Cavanagh, P., Fischer, B., Ramachundran. V. S., and Heydt. R., Form perception and attention striate cortex and beyond. In Ki,vual Perception- The Neurol#U'siologica[ Foumlalivnx, ed. g. Spilhnann, and J. S. Werner. Academic Press, N e w Y o r k , 1990, pp. 273 316. 17. Yasargil, M. G., Teddy. P. and Roth, P , Selective amygdalohippocampectomy, operative anatomy and surgical technique. ,4dram'e.v am/Technica/Standarrl, v in Neur,~,vm'.qerr 12, ¢)3 123. 1985.