ASSOCIATIVE VISUAL AGNOSIA RESULTING FROM A DISCONNECTION BETWEEN INTACT VISUAL MEMORY AND SEMANTIC SYSTEMS Giovanni A. Carlesimo1, Paola Casadio1, Maurizio Sabbadini1 and Carlo Caltagirone1,2 (1I.R.C.C.S. S. Lucia, Roma, Italy; 2Clinica Neurologica, Università di Roma “Tor Vergata”, Italy) ABSTRACT We report the case of a patient (RC) who developed a severe visual agnosia, associated to alexia without agraphia, color anomia and amnesia, following an ischemic stroke in the territory supplied by the left posterior cerebral artery. Based on his proficient performance on tests evaluating analysis of elementary visual features, formation of viewer-centered and object-centered representations of visual stimuli and discrimination between drawings representing real and unreal objects, we concluded that the critical locus of deficit was a disconnection between the normally functioning visual memory store and the semantic system. RC’s disturbance in visual processing of human faces paralleled his recognition disorder of other classes of objects. The possible contribution of neurobiological factors in determining RC’s agnosic deficit is discussed. INTRODUCTION Visual agnosia is a relatively rare clinical syndrome characterized by an impairment in recognizing objects by vision. Typically, agnosic patients are free from elementary neurosensorial deficits, do not suffer from a global cognitive impairment and promptly name by touch or definition the same objects that they do not recognize visually. Lissauer (1890) first proposed distinguishing between two forms of visual agnosia. The first one, which he termed apperceptive, is due to a deficit in constructing a coherent visual percept of the object. Thus, the apperceptive agnosic patient is unable to copy simple line drawings, to track figure contours and to make simple visual judgements (e.g., figure matching). The second type of visual agnosia identified by Lissauer (1890) is termed associative and is characterized by an inability to assign the correct meaning to visual stimuli which, however, are adequately perceived (as demonstrated by normal performance in copying, tracking and matching tasks). Following Lissauer’s (1890) first descriptions, a number of cases fitting diagnostic criteria of apperceptive agnosia (e.g., Benson and Greenberg, 1969; Campion and Latto, 1985) or, more frequently, associative agnosia (e.g., Mack and Boller, 1977; Riddoch and Humphreys, 1987; Rubens and Benson, 1971) have been reported in the literature (see for a review, De Haan, Heywood, Young et al., 1995). Cortex, (1998) 34, 563-576 564 Giovanni A. Carlesimo and Others An advance in comprehension of the basic mechanisms of visual agnosia has been made possible by the development of cognitive models that provide a description of the computational stages involved in the processing of visual information. The most influential of these models is, no doubt, that proposed by Marr (1980, 1982). Following preliminary extraction of the elementary visual features to form a visual image of the percept (primal sketch), this author distinguishes a so-called 21/2-D stage, in which a coherent viewer-centered representation of the object is achieved, from a subsequent 3-D stage in which, through a sort of mental rotation of the visual image or based on previous “knowledge of the world”, an object-centered view of the stimulus is obtained. This 3-D reconstruction of the visual image is then compared with the previously-stored representations of the visual structure of objects (visual memory store); when a match occurs between the presently-perceived image and a stored representation, a node in the semantic system for access to meaning is activated. Coming back to Lissauer’s terminology, an apperceptive agnosia can result from a lesion in the 21/2 D and 3 D stages of Marr’s model. In contrast, disruption of the memory store for the visual structure of objects or its disconnection from previous stages of the visuoperceptual analysis or from the semantic system can result in an associative form of visual agnosia. Several attempts have been made to interpret basic mechanisms of visual agnosia in the light of the above mentioned cognitive models. Thus, the patient described by Benson and Greenberg (1969) who, despite good performances on tasks assessing elementary visual functions, was unable to make simple visual judgements of shapes, has been reinterpreted as suffering from an impairment in constructing an adequate viewer-centered representation (Ellis and Young, 1988). In contrast, patient JL (Humphreys and Riddoch, 1984, 1985), who was specifically deficient in recognizing objects observed from unusual points of view, was likely impaired in performing mental computations ordered to generate an adequate 3-D representation. Finally, Riddoch and Humphreys (1987) reported the case of HJA who, despite being deeply agnosic, was normally proficient in matching objects seen from different points of view, thus demonstrating efficient formation of object-centered representations. He was also particularly brilliant in drawing objects from memory, thus suggesting normal access to the visual memory store from the semantic system. Since the patient’s object recognition was better when stimulus figures were poor in detail than when more realistic figures were utilized, Riddoch and Humphreys (1987) suggested that a difficulty in integrating local details into a coherent percept before accessing representations in the visual memory store was at the base of HJA’s visual recognition deficit. In this paper, we report the case of a patient (RC) suffering from a dense visual agnosia, due to an ischemic stroke in the territory supplied by the left posterior cerebral artery. The results of an experimental investigation conducted 5 to 7 months after the ischemic stroke suggested a unidirectional disconnection between normally functioning visual memory and semantic memory stores. An analysis of visual processing of human faces suggested a similar level of functional impairment at the base of RC’s deficient recognition of objects and people. Finally, we describe the results of follow-up investigations, conducted 11 Fig. 1 – RC’s Goldman perimetry perfomed in September, 1996, showing a right lateral hemianopia without sparing of macula vision. Associative visual agnosia 565 566 Giovanni A. Carlesimo and Others and 16 months after the ischemic stroke, demonstrating improved visual recognition and naming of objects. CASE HISTORY RC was a 69-year-old right-handed man, with a high school education. He had a long history of smoking and was treated for long-standing hypertension and chronic renal failure. On September 19, 1995, the patient was confused upon awakening, could not recognize his wife and presented a mild right hemiparesis. Due to the persistence of these symptoms, he was hospitalized one week later. At admission, he was alert, oriented to place but not to date. Motor and somato-sensory examinations were normal. The clinical exam showed a right lateral hemianopia without sparing of macula vision, confirmed by Goldman perimetry (Figure 1). Spontaneous speech was fluent, with sporadic paraphasias. Comprehension was normal. A severe deficit in naming and utilizing objects suggested the presence of visual agnosia. A cerebral MRI performed in May, 1997, showed an extensive infarction in the left hemisphere region corresponding to the distribution of the posterior cerebral artery (Figure 2). The lesion involved the occipital pole and the mesial surfaces of the occipital and temporal lobes extending forward to the posterior thalamus and limbic cortex. A topographical mapping according to Damasio and Damasio (1989) revealed that the cortical Fig. 2 – RC’s cerebral MRI performed in May, 1997. Associative visual agnosia 567 regions involved by the ischemic lesion were the mesial portion of the superior parietal lobule (Broadman’s area 7), the posterior half of the cingulate gyrus and limbic cortex (areas 23 and 31), the calcarine region (area 17), the lingual and fusiform gyri and precuneus (areas 18 and 19) and the parahippocampal gyrus (area 28). The fibers of the splenium of corpus callosum appeared completely destroyed in the lower half but partially spared in the upper half. RC was first seen in our laboratory in February 1996. At that time, his language was fluent and informative, without paraphasias. His general behavior was adequate and his main complaints were complete inability to read and impossibility to drive a car due to visual field limitations. RC’s wife reported that he frequently misidentified objects (e.g., mistaking his glasses for the TV remote control), but the patient appeared scarcely aware of his agnosic disturbance. In face of troubles in recognizing visual objects, RC never demonstrated difficulties in recognizing people by their own faces. General Neuropsychological Examination De Renzi, Zambolin and Crisi (1987) indicated four main deficits characterizing the neuropsychological impairment consequent to infarction of the territory supplied by the left posterior cerebral artery, namely visual agnosia, pure alexia, color anomia and amnesia. Thus, RC’s early neuropsychological examination was aimed toward evaluating these cognitive areas. His visual naming was extremely poor and did not vary appreciably if real objects (2/16 correct), black and white photographs (5/24) or colored figures (2/40) were utilized. RC generally provided his wrong responses rapidly and appeared completely unaware of his recognition errors. Errors were mainly perseverations. Once he had named (correctly or incorrectly) the first item (e.g., a dog), he tended to attribute all the following items to the same semantic category (i.e., animals). The categories most frequently represented in his perseverations were animals, musical instruments and tools in this order. Most of the objects not named following visual presentation were promptly named when the patient was allowed to touch them (14/16). In order to discriminate between visual agnosia (deficient recognition of objects by vision) and optic aphasia (deficient naming of visual objects), we investigated alternative ways by which the patient might manifest recognition of visual objects, namely pantomime of their use and semantic categorization. RC was very poor in the gestural demonstration of the use of objects he could see but not manipulate (2/16 correct). This was not due to an apraxic disorder, since he was very accurate in pantomiming the use of named objects (23/24 correct). He was also poor in associating visual objects according to their semantic or functional properties. When given a triplet of objects (e.g., hammer, pliers and comb) and requested to associate the two objects that were semantically related, RC scored 10/16 correct. In an analogous task involving photographs of objects, he scored 11/15. Reading was severely impaired. RC was unable to read a single word, irrespective of its length, frequency of occurrence and script characters. He read 5/16 single letters printed in capitals but improved when allowed to track the contour with his finger (8/16) or when the examiner traced the letters over his left (11/16) or right (12/16) hand. In contrast, writing was preserved. RC wrote single words or short passages on dictation correctly and without hesitation, though being unable to reread his own script. Color perception was normal (10/10 correct in matching colours). However, he was poor at naming visually presented colours (4/10 correct). RC had severe difficulty in retaining day-to-day events. Due to his visuo-perceptual deficits, long-term memory was formally assessed with verbal material only. On Rey’s 15word learning task (Rey, 1958), he recalled 15/75 words across the five immediate recall trials and no single word following a 15’ delay. In an oral Recognition task for the same 15 words plus 30 distractors, he correctly identified 4/15 of the previously studied words and made 6 false alarms. In a short-story recall task (Spinnler and Tognoni, 1987), he recalled just one information unit in the immediate trial and none following a 15’ delay. In summary, RC was affected by a severe visual agnosia, alexia without dysgraphia, color anomia and dense amnesia. 568 Giovanni A. Carlesimo and Others EXPERIMENTAL INVESTIGATION The following stages of the visual recognition process (Marr, 1980, 1982) were investigated: a) analysis of elementary visual features (size, spatial location, line orientation, etc.); b) formation of the stimulus viewer-centered representation; c) formation of the stimulus object-centered representation; d) access to stored visual representations. Further tests were given to evaluate access to the visual memory store from the semantic system and the visual processing of human faces. These tests were given to RC between February and April 1996. Basic Mechanisms of Impaired Recognition of Objects by Vision Analysis of Elementary Visual Features RC was given five forced-choice tasks exploring perception of the size and shape of simple geometrical figures, orientation and length of lines and spatial location of a dot in a bidimensional plane. Size perception. Twenty pairs of circles and 20 pairs of squares of different size (diameter of circles ranging from 2 to 4.5 cm, sides of squares ranging from 2 to 3.5 cm) were presented arranged in a vertical array. The patient had to point to the bigger shape in each pair. RC’s performance was 100% correct. Simple shape perception was assessed with a simplified version of Efron’s test (1968). In each of 34 pairs, a square (same size in all items) was presented with either the same square (half of the items) or a rectangle, whose base-height ratio varied. The patient had to distinguish whether the members of the pairs were identical or not. RC responded correctly to 33 out of 34 stimuli presented (97%). Perception of spatial locations. The stimulus material consisted of 30 pairs of squares with an internal black dot placed in one of 25 possible locations. In half of the items, the dot location was the same in the two squares, in the remaining half it differed. The patient had to decide whether the dots were located in the same or in different positions. RC scored 29 correct (97%). Line orientation perception. RC was presented 48 pairs of 4 cm lines in a vertical array. Each line could be oriented at 0°, 45°, 90°, 135° relative to the horizontal plane and, in each pair, lines were identically (half of the items) or differently (the remaining half) oriented. RC was 100% correct. Line length perception. Stimuli consisted of 20 pairs of lines ranging from 4.5 to 8 cm in length. In each item, the patient was requested to point to the longer line. RC gave 19/20 correct responses (95%). Formation of Viewer-centered Stimulus Representations In each of the 40 items of this test, two figures from Vanderplas, Sanderson and Vanderplas’ series (1965), approximately 8 × 8 cm in size, were presented in a vertical array. RC had to decide whether they were the same (half of the items) or different. He performed this task twice, at a one month interval. The Associative visual agnosia 569 first time, he scored 23/40 correct (57%), not significantly different from chance (chi-square = .45; p = n.s.) and he gave almost always (37/40) “different” responses. The second time, RC responded correctly to 33/40 items, significantly more than expected by chance (chi-square = 9.45; p < .01). On this task, a group of 5 age-matched controls scored 100% correct. At around the same time RC performed the task with Vanderplas et al.’s (1965) abstract figures, a similar matching task involving 4 × 4 cm line drawings of real objects from Snodgrass and Vanderwart’s series (1980) was also given. Care was taken to select “different” drawings that were visually similar (e.g., a pen and a flute similarly oriented). On this test, RC scored 100% correct without hesitation. Formation of Object-centered Stimulus Representation In order to achieve a mental representation of a visual object centered on its most usual point of view, a sort of mental rotation of the presently perceived stimulus before accessing previously stored memory representations is required. In order to test this ability, two black and white, 18 × 24 cm photographs of 16 common objects were taken, one representing the object from its prototypical perspective [according to Ellis and Young (1988), “in which its distinctive feature and axis of elongation were clearly visible”, p. 41] and the other from an unusual point of view (most of the cases “with its axis of elongation foreshortened”, in a minority of cases “with the saliency of its primary distinctive feature reduced”). In each of 16 items, RC was presented with three photographs in a vertical array. The photograph at the top represented the prototypical view of an object. The two photographs below represented objects seen from unusual perspectives. One of them was the same object as the stimulus target, the other was a different object selected because it shared some degree of visual similarity with the previous one. RC had to indicate which of the two unusual-view photographs represented the same object as the target. He was fast and accurate in performing this task, scoring 15/16 (his only error consisted of mistaking a cup for a bucket, both seen from above). Access to the Visual Memory Store from Visual Perceptual Analysis The visual memory store contains records of the visual aspect of objects. According to Ellis and Young (1988), it can be accessed either directly from the viewer-centered representation of the object (when the object is observed from a typical perspective) or via the object-centered representation (when a transformation is needed to achieve a more typical view of the object). Two tests were given that did not require stimulus meaning recognition. The first test involved discriminating line drawings of real objects from drawings of unreal objects. This test is similar to that developed by Riddoch and Humphreys (1987), on which their patient HJA performed poorly. The 14 drawings of real objects were selected from Snodgrass and Vanderwart’s (1980) series. The 14 unreal objects were made up of by assembling fragments of these and other drawings. RC was requested to decide whether each of the 22 570 Giovanni A. Carlesimo and Others drawings represented a real or an unreal object. He performed accurately (27/28 correct) and quickly. On a second task, RC was requested to decide which of two objects represented by line drawings was larger in the real world. Accord to Forde, Francis, Riddoch et al. (1997), accurate size judgement requires the access to the structural description of the object. Forty drawings from Snodgrass and Vanderwart’s (1980) series were presented in pairs. The difference in size between the two members of the pairs varied from low (e.g., onion and pineapple), to high (e.g., car and truck). Also on this task, RC was very accurate (19/20) and quick. He made only one mistake on a pair formed by a mouse and a frog. In summary, a detailed evaluation of RC’s visuo-perceptual abilities did not reveal any problem in the processing of elementary visual features such as size, line orientation and spatial location. His performance on an abstract shapes matching task suggested a difficulty in the formation of viewer-centered representations of complex visual stimuli. However, it is unlikely that a critical deficit at this stage of visual processing underlay the impaired recognition of visual objects since his performance was good on tasks requiring (a) to match line drawings of real objects, (b) to match photographs of objects observed from different perspectives, (c) to discriminate real from unreal object line drawings, (d) to point to the drawing in a pair that represented the larger object in the real world. RC’s good performance on these tasks also showed his normal or nearnormal ability to form object-centered representations of visual stimuli and to access the memory system storing the visual aspect of objects. In the light of these findings, we hypothesize that the deficit underlying RC’s visual agnosia was the inability of stored visual representations to activate the corresponding nodes in the semantic system. In other words, the locus of RC’s deficit was a disconnection between normally functioning visual and semantic memory stores. Access to the Visual Memory Store from the Semantic System Based on the fact that their agnosic patient was able to point to the object named by the examiner, De Renzi and Saetti (1997) interpreted associative visual agnosia as a unidirectional disconnection preventing visual percepts from accessing the language system but leaving concepts able to activate representations in the visual memory store. However, this observation is not generalizable to all visual agnosic patients. Impaired pointing to objects named by the examiner was found in agnosic patients described by Oxbury, Oxbury and Humphreys (1969), Rubens and Benson (1971) and Ferro and Santos (1984) as well as others. We used three tasks to test RC’s ability to access the visual memory store from the language system. The first was a verbo-visual matching task, in which he was requested to select the object or the color photograph named by the examiner among three alternatives. In both cases, RC was very accurate and fast (13/14 for objects, 23/24 for photographs). The second task was drawing from memory an object named by the Associative visual agnosia 571 examiner. Even though he was very poor in copying drawings, at least 12 out of 19 objects drawn from memory on two separate occasions were well recognizable (Figure 3). In the third test RC was requested to provide a verbal description of the visual characteristics of 34 objects. He was generally very accurate and gave detailed structural descriptions. For example, in describing a spoon, he said “it is metal, it has a handle, the end part put in the mouth is concave” and in describing a chair “it has four legs, a back and a seat; it is made of wood, sometimes covered with leather”. In summary, despite his performance in drawing from memory was unsatisfactory (likely depending on general drawing difficulties), our data document that RC’s ability to retrieve the mental visual image of an object in response to its name was much better than its visual recognition. Processing of Human Faces Human faces represent a distinct category of percepts whose visual recognition can be selectively impaired (e.g., Bodamer, 1947; Carlesimo and Caltagirone, 1995; De Renzi, 1986) or spared (Feinberg, Schindler, Ochoa et al., 1994; Moscovitch, Winocur and Behrmann, 1997) by a cerebral lesion. a b c d Fig. 3 – Some of RC’s drawings from memory: (a) a glass; (b) a hammer; (c) a pencil; (d) a needle. 572 Giovanni A. Carlesimo and Others To evaluate visual processing of human faces, RC was given a Recognition of famous faces test (Carlesimo and Caltagirone, 1995), which assesses the ability to identify a previously known face and to recall relevant semantic data and the proper name of the identified person. Each of the 20 items is composed of four black and white, 9 cm × 6.5 cm photographs of faces in a vertical array. In each item, one face belongs to a celebrity and the other three faces to nonfamous persons. RC was requested to give three types of responses: (a) familiarity judgement (pointing to the face belonging to the famous person), (b) category (main occupation of the famous individual: politics, movies, music, TV), and (c) proper name. RC’s performance on familiarity judgement (15/20 correct) was in the range of scores obtained by a group of 50 normal controls (14 to 20/20 correct). However, when requested to recall the main occupation of the famous individual, he was very poor (only 3/20 correct) and he completely failed on proper name retrieval. The pattern of deficits shown by RC for human faces appeared to be very similar to that found for objects. Though his ability to match unknown faces seen from identical or different points of view was not assessed, his normal performance on the familiarity recognition test suggests that perceptual processing of physiognomic material was substantially preserved. It also suggests that he was able to access the facial recognition units (Bruce and Young, 1986) in his visual memory store. In contrast, RC’s deficit in recovering semantic information and proper names of familiar people parallels his impairment in assigning a correct meaning to visual stimuli. Follow up When he first attended our laboratory (February-April, 1996), RC appeared completely unaware of his visual recognition deficit. Requested to name an object or a figure, he readily produced wrong response and frequently added unrequested, confabulatory details about the visual appearance of the stimulus (e.g., when presented with a photograph of a stick, RC called it a doll and described its legs and the color of its hair in detail). As from May, 1996, RC attended a cognitive rehabilitation program two times a week. It was primarily aimed at making the patient aware of his visual recognition deficits and developing a compensatory strategy, based on the overt verbalization of the visual characteristics of the object before producing its name. We had the opportunity to retest RC’s naming abilities for the first time in October, 1996, and then in April, 1997. In October 1996, RC’s accuracy in naming both objects (10/12 correct) and line drawings of real objects (56/80 correct) was strikingly improved. Eight of the 26 naming errors RC made were agnosic in nature (e.g., he misidentified an ear for a pair of shoes), the other 18 errors were anomias (e.g., of the drawing of a pineapple, he said: “It is a fruit, it can be peeled, inside it is yellow, juicy, it has a strange name...”). However, he was very slow in performing the task and his response was generally preceded by an overt verbalization that revealed the strategy he applied to recognize the object and to access its verbal label. An Associative visual agnosia 573 analysis of RC’s verbalizations suggested that in more than one-third of the cases (26/66 of correct responses), he utilized a sort of conduite d’approche strategy based on the definition of functional attributes of the object. For example, when requested to name a stick-plaster, he said: “...yes, when someone gets hurt it is placed, ... then the hurt stays covered... yes, it is a stick-plaster”. In a minority of cases (11/66), instead, RC started perseverating a previous response or category, then accessed the category and finally the correct name of the object without any apparent logical strategy. For example: “It is made of wood, it is used to sit on, ... perhaps no... it is food, a sandwich, it might be a fruit, it is a pear!”. In April, 1997, RC’s ability to name 50 objects was tested once on verbal definitions and, in a different day, on visual presentation. The time needed to provide the correct name was not recorded. RC’s accuracy in naming visual stimuli was further improved (41/50 correct in the visual test, 47/50 correct in the verbal test). However, the patient was much slower in producing the name following visual presentation (average: 63 sec., range: 0-150 sec.) than verbal description (average: 5 sec.; range: 0-60 sec.). In the visual task, 5 errors were misidentifications and 4 were anomias. Also in this case, an analysis of the overt verbalizations that preceded correct responses revealed that the correct name was most frequently provided following a conduite d’approche strategy (26/41). In summary, about one year after our first observation, RC was able to name 92% of the visual stimuli. However his responses were extremely slow, while he was generally fast and accurate in retrieving the name of objects described verbally. DISCUSSION At the time of our first observation, RC presented with all the typical neuropsychological symptoms of the left posterior cerebral artery ischemia syndrome (De Renzi et al., 1987). He had great difficulty in recognizing visually presented objects, was alexic but not agraphic, poor in color naming and suffered from verbal anterograde amnesia. The visual recognition deficit exhibited by RC had the characteristics of an associative visual agnosia. He performed normally on visuo-perceptual tasks not requiring meaning identification, while he failed in naming, pantomiming the use and categorizing visually presented objects or figures. On the basis of his normal ability to access stored representations in the visual memory system (correct reality judgement about line drawings of real and unreal objects and size judgement about line drawings of real objects), we conclude that a disconnection between a normally functioning visual memory store and a substantially preserved semantic memory system underlay RC’s visual agnosia. From a neurobiological perspective, RC’s case provides insights about the neural substrate of visual object recognition. As previously pointed out, the ischemic lesion in RC destroyed the primary visual area of the left hemisphere 574 Giovanni A. Carlesimo and Others and a large portion of the mesial surface of the left temporal-occipital lobes and of the splenium of corpus callosum, thus preventing the visual signal elaborated in the right hemisphere to reach the language areas in the left hemisphere. RC’s substantially preserved ability to construct adequate viewer-centered and objectcentered representations of the visual stimulus fits well with the assumption of right hemisphere prevalence for these stages of visual processing (Warrington, 1985). More interestingly, RC’s preserved ability to normally access stored visual images (in agreement with previous reports on visual agnosic patients; e.g., De Renzi and Saetti, 1997) could be explained by assuming right hemisphere lateralization of the neural substrate of the memory system for the visual aspect of objects. However, since associative visual agnosia has never been reported following a unilateral lesion of the right occipito-temporal cortex, it is likely that visual images are bilaterally represented, with each hemisphere able to normally support the visual recognition of objects in the case of a lesion confined to the contralateral hemisphere. A further point of interest in the visual recognition deficit exhibited by RC resides in the way he processed human faces. Farah and coworkers (Farah, 1990, 1991; Feinberg et al., 1994) argued for the existence of two subtypes of visual agnosia, the first one specific for objects and printed words (alexia) but sparing human faces, the second one disrupting recognition of all three classes of visual stimuli. RC’s case was straightforward in that not only did his visual recognition deficit extend to physiognomic material, but his impairment in recognizing people by face closely paralleled his difficulty in recognizing objects by vision. He expressed correct familiarity judgements about most famous faces (just as he expressed correct reality judgements about drawings of objects) but was unable to provide any semantic information about these people (just as with objects). The face recognition deficit exhibited by RC cannot be considered a true prosopagnosia, since he could discriminate familiar from unfamiliar faces. Rather, RC’s inability to recognize people whose faces he felt as familiar suggests a disconnection between a normally functioning store for facial recognition units and the semantic system. The visual-semantic disconnection exhibited by RC for objects appeared to be unidirectional. While he was very poor in providing names or general information about objects he felt as real, he retained the ability to point to objects named by the examiner and to provide recognizable drawings and accurate descriptions from memory. The unidirectionality of the visual recognition deficit in visual agnosic patients is not a new finding (e.g., Albert, Reches and Silverberg, 1975; De Renzi and Saetti, 1997) and can be accounted for by assuming that the fibers connecting the auditory verbal areas of the left hemisphere (where the verbal input is decoded) with the right hemisphere cross the median line anteriorly to the damaged splenium. A final point of interest is represented by the evolution of RC’s visual recognition deficit during the time of our observation and by the qualitative analysis of his naming responses during follow up examinations. As described above, at the time of our first observation, RC appeared completely unaware of his misidentifications. He quickly provided wrong responses, tended to perseverate into one or a few semantic categories of objects and frequently Associative visual agnosia 575 confabulated about the perceptual characteristics of the visual stimulus according to the wrong verbal label he had produced. Following a six-month period, during which he attended a specific cognitive rehabilitation program, RC’s accuracy in naming visual stimuli improved remarkably. Analysis of his overt verbalizations suggested that improvement was mainly the result of a change in the strategy he implemented in the search for the correct verbal label. In particular, at this time RC appeared to consider his verbal responses critically, inhibited confabulations and used a verbal strategy based on the description of perceptual or functional attributes to facilitate access to the correct verbal label. These data seem to suggest that although the cerebral lesion posed very severe limitations on RC’s visual processing abilities and determined the qualitative characteristics of his visuo-perceptual deficit, a preserved cognitive flexibility (likely supported by a specifically aimed rehabilitation program) allowed him to elaborate alternative cognitive strategies that modified his behavioral approach to the visual recognition task and substantially improved the accuracy of his recognition. REFERENCES ALBERT, M.L., RECHES, A., and SILVERBERG, R. Associative visual agnosia without alexia. Neurology, 25: 322-326, 1975. BENSON, D.F., and GREENBERG, J.P. Visual form agnosia. Archives of Neurology, 20: 82-89, 1969. BODAMER, J. Die Prosopagnosie. Archiv für Psychiatrie und Nervenkrankheiten, 179: 6-53, 1947. BRUCE, V., and YOUNG, A. Understanding face recognition. British Journal of Psychology, 77: 305-327, 1986. CAMPION, J., and LATTO, R. Apperceptive agnosia due to carbon monoxide poisoning. An interpretation based on critical band masking from disseminated lesions. Behavioral Brain Research, 15: 227-240, 1985. CARLESIMO, G.A., and CALTAGIRONE, C. Components in the visual processing of known and unknown faces. Journal of Clinical and Experimental Neuropsychology, 17: 691-705, 1995. DAMASIO, H., and DAMASIO, A.R. (Eds.) Lesion Analysis in Neuropsychology. New York, Oxford: Oxford University Press, 1989. DE HAAN, E.H., HEYWOOD, C.A., YOUNG, A.W., EDELSTYN, N., and NEWCOMBE, F. Ettlinger revisited: The relation between agnosia and sensory impairment. Journal of Neurology, Neurosurgery and Psychiatry, 58: 350-356, 1995. DE RENZI, E. Prosopagnosia in two patients with CT scan evidence of damage confined to the right hemisphere. Neuropsychologia, 24: 385-389, 1986. DE RENZI, E., and SAETTI, M.C. Associative agnosia and optic aphasia: Qualitative or quantitative difference? Cortex, 33: 115-130, 1997. DE RENZI, E., ZAMBOLIN, A., and CRISI, G. The pattern of neuropsychological impairment associated with left posterior cerebral artery infarcts. Brain, 110: 1099-1116, 1987. EFRON, R. What is perception? In R.S. Cohen and M.W. Wartofsky (Eds.), Boston Studies in the Philosophy of Science, 4. Dordrecht: Reidel, 1968. ELLIS, A.W., and YOUNG, A.W. Human Cognitive Neuropsychology. Hove and London: Lawrence Erlbaum Associates, 1988. FARAH, M.J. Visual Agnosia. Disorders of Object Recognition and What They Tell Us about Normal Vision. Cambridge: MIT Press, 1990. FARAH, M.J. Patterns of co-occurrence among the associative agnosias: Implications for visual object representation. Cognitive Neuropsychology, 8: 1-19, 1991. FEINBERG, T.E., SCHINDLER, R.J., OCHOA, E., KWAN, P.C., and FARAH, M.J. Associative visual agnosia and alexia without prosopagnosia. Cortex, 30: 395-412, 1994. FERRO, J.M., and SANTOS, M.E. Associative visual agnosia: A case study. Cortex, 20: 121-134, 1984. FORDE, E.M.E., FRANCIS, D., RIDDOCH, M.J., RUMIATI, R.I., and HUMPHREYS, G.W. On the links between visual knowledge and naming: A single case study of a patient with a category-specific impairment for living things. Cognitive Neuropsychology, 14: 403-458, 1997. HUMPHREYS, G.W., and RIDDOCH, M.J. Routes to object constancy: Implications from neurological impairments of object constancy. Quarterly Journal of Experimental Psychology, 36a: 385-415, 1984. HUMPHREYS, G.W., and RIDDOCH, M.J. Authors’ correction to “Routes to object constancy”. Quarterly 576 Giovanni A. Carlesimo and Others Journal of Experimental Psychology, 37a: 493-495, 1985. LISSAUER, H. Ein Fall von Seelenblindheit nebst einem Beitrag zur Theorie derselben. Archiv für Psychiatrie und Nervenkrankheiten, 21: 222-270, 1890. MACK, J.L., and BOLLER, F. Associative visual agnosia and its related deficits: The role of the minor hemisphere in assigning meaning to visual perceptions. Neuropsychologia, 15: 345-349, 1977. MARR, D. Visual information processing: The structure and creation of visual representations. Philosophical Transactions of the Royal Society, B200: 269-294, 1980. MARR, D. (Ed.). Vision. San Francisco: W.H. Freeman, 1982. MOSCOVITCH, M., WINOCUR, G., and BEHRMANN, M. What is special about face recognition? Nineteen experiments on a person with visual object agnosia and dyslexia but normal face recognition. Journal of Cognitive Neuroscience, 9: 555-604, 1997. OXBURY, J.M., OXBURY, S.M., and HUMPHREYS, N.K. Varieties of colour anomia. Brain, 92: 847-860, 1969. REY, A. Memorisation d’une serie de 15 mots en 5 repetitions. In A. Rey (Ed.), L’examen clinique en psycologie. Paris: Presse Universitaire de France, 1958. RIDDOCH, M.J., and HUMPHREYS, G.W. A case of integrative visual agnosia. Brain, 110: 1431-1462, 1987. RUBENS, A.B., and BENSON, D.F. Associative visual agnosia. Archives of Neurology, 24: 305-316, 1971. SNODGRASS, J.G., and VANDERWART, M. A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6: 174-215, 1980. SPINNLER, H., and TOGNONI, P. Standardizzazione e taratura italiana di test neuropsicologici. Italian Journal of Neurological Sciences, suppl. 6: 44-46, 1987. VANDERPLAS, J.M., SANDERSON, W.A., and VANDERPLAS, J.N. Statistical and associational characteristics of 1100 random shapes. Perceptual and Motor Skills, 21: 414, 1965. WARRINGTON, E.K. Agnosia: The impairment of object recognition. In P.J. Vinken, G.W. Bruyn and H.L. Klawans (Eds.), Handbook of Clinical Neurology, vol. 45. Amsterdam: Elsevier, 1985, pp. 333-349. Giovanni A. Carlesimo, I.R.C.C.S. S. Lucia, Via Ardeatina, 306, 00179 Roma, Italy. (Received 19 December 1997; accepted 15 April 1998)