Brain (1995), 118, 789-800 Knowledge, implicit knowledge and \ in visual agnosia Todd E. Feinberg,1'2 Diana Dyckes-Berke,2 Christian R. Miner2 and David M. Roane2 Departments of ^Neurology and ^Psychiatry, Albert Eistein College of Medicine and Beth Israel Medical Center, New York, USA Correspondence to: Todd E. Feinberg, MD, Neurobehavior and Alzheimer's Disease Center, Beth Israel Medical Center, 317 East 17th Street, New York, NY 10003, USA residual knowledge on forced-choice matching tasks. While the patient could not consistently demonstrate awareness of knowledge prior to stimulus matching ('Do you know what this is?'), she was able to reliably demonstrate awareness of knowledge for response accuracy ('Are you sure?') assessed after stimulus matching. Further, the extent of the patient's metaknowledge corresponded to her degree of preserved knowledge. We propose that this pattern of performance suggests limited or partial access to preserved semantic knowledge which, though degraded, is not 'non-conscious'. Keywords: implicit knowledge; metaknowledge; associative visual agnosia; pure alexia Introduction Associative visual agnosia refers to a disorder of visual recognition not attributable to primary perceptual defects or linguistic impairments (Teuber, 1968; Rubens and Benson, 1971; Taylor and Warrington, 1971; Mack and Boiler, 1977; Ferro and Santos, 1984; Feinberg etal., 1986,1994; McCarthy and Warrington, 1986; Farah, 1990; Iorio et al., 1992; Bauer, 1993). Ideally, these patients have intact perceptual performance (e.g. as demonstrated by good figure copying skills) which distinguishes their disorder from apperceptive agnosia (but see Farah, 1990; Bauer, 1993). These patients cannot demonstrate visual object recognition via nonverbal means such as pantomime, a distinction which may help to differentiate visual associative agnosia from optic aphasia (Freund, 1889; Lhermitte and Beauvois, 1973; Beauvois, 1982; Coslett and Saffran, 1989, 1992). They show intact pantomime to verbal stimuli, which distinguishes their impairment from apraxia and receptive aphasia, while intact naming to verbal description rules out anomic aphasia. In actuality, the distinction between associative visual agnosia and optic aphasia is not always clear (Geschwind, 1965; Bauer and Rubens, 1985; De Renzi et al., 1987; Kertesz, 1987). By definition, pure alexia or alexia without agraphia © Oxford University Press 1995 (Dejerine, 1892; Geschwind and Fusillo, 1966; Albert et al., 1973; Coslett and Saffran, 1989a; Farah, 1990; Coslett etal., 1993) is a disorder of reading not attributable to purely perceptual factors or to aphasia (hence the preservation of writing). The patient with pure alexia cannot read normally, either via phonological or whole word strategy, but may read by means of a letter-by-letter strategy (Kinsbourne and Warrington, 1962; Warrington and Shallice, 1980; Patterson and Kay, 1982; Shallice and Saffran, 1986). Associative visual agnosia and pure alexia frequently co-occur, particularly in the context of dominant occipitotemporal infarction due to posterior cerebral occlusion (Michel et al., 1979; De Renzi etal., 1987; Farah, 1990, 1991; Feinberg etal., 1994). While associative visual agnosics and pure alexics may fail on tests of explicit object or word recognition tasks, it is of interest that both disorders may be characterized by some degree of preserved (residual, implicit, covert, tacit) knowledge. For instance, Taylor and Warrington (1971) reported a visual agnosic with a severe deficit in recognition of objects and pictures, who showed a surprising ability to orient pictures correctly and sort pictures into crude categories. Similar abilities have been demonstrated more Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 Summary Residual or implicit knowledge has been observed in patients with object agnosia, optic aphasia and pure alexia. Previous investigators have considered implicit knowledge in these patients to be dissociated from awareness on the basis of intact semantic capabilities that are consistent with right hemisphere processing. The absence of explicit verbal identification is presumably dependent upon damaged left hemisphere systems. We describe a 72-year-old woman with a left occipital infarction, object agnosia and pure alexia who was unable to explicitly identify visual stimuli (objects and words), but was able to make reliable judgements of her 790 T. E. Feinberg et al. 1 1 Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 recently in an additional case of associative visual agnosia (Jankowiak et al., 1992). Covert face recognition by prosopagnosic patients has also been demonstrated, using both psychophysiological and behavioural measures. For example, skin conductance responses may reveal more accurate recognition than overt verbal responses in some prosopagnosics (Bauer, 1984). Residual reading capabilities in pure alexia have been noted even more frequently (Albert et al., 1973; Bauer, 1984; Caplan and Hedley-Whyte, 1974; Landis et al, 1980; Shallice and Saffran, 1986; Coslett and Saffran, 1989a, b, 1992; Coslett et al., 1993). In these cases, patients who are unable to read aloud are nonetheless able to perform lexical decision tasks, word sorting tasks and/or semantic matching tasks. Several different explanations have been proposed to account for residual recognition abilities in agnosic and alexic patients. The most straightforward is that recognition mechanisms are intact, but have been disconnected from some other brain mechanisms necessary for conscious awareness. Schacter et al. (1988) has proposed this hypothesis as a general way of accounting for implicit/explicit dissociations in neuropsychology, and De Haan et al. (1992) have specifically endorsed such an account of covert recognition in prosopagnosia. Shallice and Saffran (1986) suggested that the preserved comprehension of some pure alexics results from right hemisphere processing of the printed word. According to this account, overt responses require the left hemisphere, but some degree of covert processing is possible in the right hemisphere. Coslett and Saffran (1989a, b) ' ^ T g . ' l ' l C/Cll StCilOilSxVeiil/v_-Vii SUiiii i i phWC'/iiMx^L. J'in, IXUII provided evidence for this hypothesis by showing that the demonstrates infarction of the calcarine cortex and the fusiform, profile of their subjects' reading abilities was similar to the lingual and parahippocampal gyri. There was involvement of the known abilities of the right hemisphere and further suggested temporal white matter; infarction of the splenium of the corpus that the right hemisphere may also mediate preserved object callosum was also evident. recognition in optic aphasia. A final possibility is that the preserved recognition capability of some agnosics and alexics visual difficulties and a change in behaviour. Her past medical reflects the residual processing of damaged, but not history includes hypertension. She had no prior neurological obliterated, recognition systems. Farah etal. (1993) suggested history and was working up until the time of admission. this type of explanation for covert recognition in prosopOn examination in the hospital, the patient was found to agnosia, and used computer simulation to show that with have a complete right homonymous hemianopia. Otherwise, increasing levels of damage to a neural network, covert task her general neurological examination was normal and no performance is preserved when overt performance is poor or sensory-motor deficits were present. A CAT scan revealed a at chance. left cerebral infarction consistent with a complete posterior In the present investigation, we examined a patient with cerebral artery occlusion (Fig. 1). There was total destruction clear-cut associative visual agnosia and pure alexia who, of the calcarine cortex and extensive infarction of fusiform, despite failure on all measures of explicit object recognition lingual and parahippocampal gyri. Temporal white matter and explicit reading, showed evidence of preserved was extensively involved. On our initial report of this knowledge on object and word category matching and on patient (Feinberg et al., 1994), splenium appeared spared, lexical decision tasks. Through a series of metaknowledge but subsequent CAT scans showed splenium of corpus tasks, we then examined whether or not this patient's residual callosum to be clearly involved (Fig. 1). knowledge could be construed as unconscious or 'out of the In order to evaluate her perceptual capabilities, the patient patient's awareness'. underwent an extensive series of neuropsychological tests The patient gave informed consent to cooperate in the testing. which were reported in a previous publication (Feinberg et al., 1994; patient 3). In summary, during the acute (weeks 1-9) post-stroke period, the patient demonstrated an Case report apperceptive defect involving shape discrimination which R.E. is a 71-year-old married, college-educated executive spared position discrimination. Her deficits subsequently secretary admitted to the hospital with the acute onset of Knowledge, implicit knowledge and metaknowledge Assessment of object recognition In order to establish that the disorder represented associative visual agnosia, we first sought to demonstrate that the impairment was not due to either perceptual or aphasic impairments. The first series of investigations was designed to confirm that this patient's pattern of deficit conformed to the definition of visual associative object agnosia. We then explored to what extent the patient displayed the type of knowledge that previously has been described as residual, tacit or implicit. Table 1 Results of agnosia testing Stimulus mode Response mode Hand tested* Left Visual objects Tactile object Verbal description of object Object name Line drawing Name or describe use Pantomime use Visual match^ Name Pantomime use Tactile match^ Name Pantomime Locate object Locate object 2/20 3/20 20/20 20/20 20/20 20/20 20/20 20/20 20/20 2/20 Right 2/20 20/20 20/20 20/20 20/20 2/20 *On selected tasks, both hands were tested independently as indicated; Correct object selected from among five choices. visually presented objects. The same objects were then represented (in differing order) and the patient was requested to gesture the use of each without verbal identification. Both hands were tested in alternating fashion. This was followed by a test of visual perception and immediate memory. In this test, stimuli objects werefirstpresented visually, then removed from the patient's view and the same object was then placed in a group with four other objects. The foils were randomly selected from the remaining (non-target) objects and not specifically semantically related to targets. With eyes open, the patient then had to retrieve the target object. Retrieval was performed with left (20 trials) or right (20 trials) hands. The patient experienced near total failure on all verbal identification tasks, as she was virtually unable to name or describe objects from visual presentation. She was also unable to pantomime object use with any degree of consistency. In contrast, the patient had a flawless performance on the nonverbal, visual matching task, obtaining 100% accuracy. Tactile object matching, naming and gesturing Using the same objects presented above, the patient was allowed to manipulate the target object with eyes closed. She was then given a series of five objects, including the target, and directed to select the target object using tactile manipulation with eyes closed. Foils were randomly selected from the remaining (non-target) objects and not specifically semantically related to targets. This task was performed with each hand tested separately. The patient had no difficulties with this task. When, using either hand, she manipulated or palpated object stimuli for verbal (name) or gesture, the patient's performance accuracy was 100%. Recognition of verbal stimuli Visual naming, describing and gesturing The results of these tests are summarized in Table 1. The patient was first required to name or describe the use of 20 Target objects were verbally described and the patient was instructed either to name the object (20 trials) or pantomime its use (20 trials each hand). When given a verbal description Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 evolved (weeks 10-15) toward a syndrome of dissociated impairment of visual memory for shapes, at which time the patient showed impaired shape discrimination with interposed responding delays of 5-60 s, again sparing position discrimination with equal delays. During the subsequent phase (weeks 16—48) the patient's perceptual and memory abilities had dramatically improved while object recognition remained quite impaired in everyday situations as well as on specific recognition tasks, in spite of intact naming to verbal definition. At this stage, it was felt that the patient had associative visual agnosia with some residual apperceptive features. On follow-up neuropsychological testing performed during the current assessment phase (2-3 years after the initial infarction), the patient obtained a verbal IQ of 88 (low average) and a performance IQ of 71 (borderline) on the Wechsler Adult Intelligence Scale—Revised (Wechsler, 1981). The discrepancy between verbal and nonverbal scores is not clinically meaningful. Attention and concentration were intact (Attention/Concentration Index of 91 on Wechsler Memory Scale-Revised (WMS-R); Wechsler, 1987). She had no difficulty repeating digits forward (8) or backward (5) (Wechsler, 1981, 1987), and she repeated 13-word sentences without error. Assessment of language showed severe impairment on tests of visual naming, although comprehension, speed, output and repetition were normal (Goodglass and Kaplan, 1972; Kertesz, 1982). Although the patient had shown remarkable improvement since the initial testing (weeks 1^8), she continued to exhibit some visuospatial impairments. Copying simple figures was adequate [see Feinberg et al. (1994) for an example of the patient's copying and writing] but copying of complex figures was distorted (Osterrieth, 1944; Goodglass and Kaplan, 1972) and judgement of line orientation (Benton et al., 1983) remained impaired. Visuospatial logic and reasoning were average (Raven, 1965) and memory for verbal material was in the low average range on the WMS-R (Verbal Memory Index: 81). Immediate recall of WMS-R paragraphs was average, but delayed recall was impaired. 791 792 T. E. Feinberg et al. of target objects, the patient had no problem providing names or pantomimes, as demonstrated by her errorless performance (100% correct) with both hands. Picture to object locating Object categorization To assess the patient's residual recognition abilities, the patient was evaluated on a series of object categorization tasks. In the first task, stimuli consisted of 120 pictures presented in pairs. The patient was required to select the correct item according to the categories real versus non-real (40 trials), living versus non-living (40 trials) and edible versus non-edible (40 trials). The non-real objects were composed of possible but not actual objects joined together (e.g. top-half phone; bottom-half scissors). The patient had flawless performance (100% accuracy) on all three tasks, thus demonstrating that some degree of semantic information could be extracted from visual stimuli. Comments R.E.'s near total failure on visual object identification tasks, as assessed by object naming, verbal description or pantomime, fits the pattern of associative visual agnosia in that she displayed a failure of object recognition confined to visual presentation with normal naming and gesturing in response to tactile presentation and/or verbal description. Intact naming under these conditions also made anomic aphasia an inadequate explanation for her performance. Finally, her near complete inability to correctly pantomime an object's use to visual stimuli suggests optic aphasia is not a likely explanation for her impairments. Likewise, her inability to locate objects in the room from picture stimuli could not be explained by optic aphasia. Her performance was better when she was provided with the object name and then directed 'find one like it' in the room. Pointing to named objects is characteristically better than identifying objects through gesture or naming in patients with associative visual agnosia. This may reflect the fact that while naming involves a nearly infinite set of possible choices, pointing often involves a greatly reduced number of choices (Bauer, 1993). In our object locating task, however, the potential number of choices was the same in the verbal stimuli condition as in the visual stimuli condition. It may be that once recognition of the verbal stimulus occurred, the patient could then utilize contextual clues to facilitate visual recognition. Alternatively, a one-way dissociation of recognition abilities is possible. In spite of her failure on all tests of explicit identification, the patient showed a surprising capacity to distinguish real from unreal objects, sort objects according to semantic category and make appropriate semantic associations. Semantic matching task The patient was shown a series of pictures of line drawings containing a target picture on top and asked to select from five choices pictured below, the item most associated with the target (i.e. 'Which one is most like or goes with . . . ?'). The correct choice was semantically related to the target by common class (e.g. both birds) or association (e.g. hammer and nails). Two of the foils were semantically similar to the correct response, but not as closely related to the target as Assessment of reading As noted previously, the patient's lexical impairments did not extend to written language, as there was no evidence of agraphia in either spontaneous writing or writing to dictation. Therefore, we set out to further examine the characteristics of our patient's alexia and explore any residual or implicit reading capabilities. We presented R.E. with a series of reading tasks utilizing a methodology adapted from Coslett Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 The patient was presented with 20 line drawings of common objects and instructed to 'find one like it in the room' (Appendix). No time constraints were imposed on picture viewing and the patient was encouraged to explore the space thoroughly. On separate control trials, the examiner provided 20 object names (auditorally) and then instructed the patient to find its match in the room (i.e. verbal name-object find). Under the visual (picture)-to-object find condition, performance was markedly impaired (two out of 20 correct). In most instances, the patient was observed to wander about aimlessly with little idea as to where to look for or find the exemplar. At best, her search concluded with an approximated semantic match, as when she found her way to a door hinge after extensive exploration for the target doorknob. This particular behaviour suggested that some degree of semanticassociative knowledge of the object was preserved. However, such semantically related errors occurred on fewer than 10% of the total trials, in fact on most trials the patient appeared to have absolutely no knowledge of the object's identity, as when she was observed looking out the window while searching for the match to the target picture of an eye. In contrast, when target names were provided verbally, the patient was able to point to named objects in the room without error (20 out of 20 correct). the correct response (e.g. target: crow; correct response: duck; foil: frog). The other two foils were visually similar to either the target or the correct response. A total of 40 stimuli were presented in 80 trials. No time limit was imposed. On half the trials, the patient responded by pointing with the right hand and on half she pointed with the left hand. In this task, the patient also demonstrated a surprisingly high level of partial recognition, as she was correct on 71% of trials overall. There was no hand difference in the response accuracy (72.5% correct pointing with the right hand; 70% correct pointing with the left hand), and there were roughly as many semantic as visual errors. Error types and number were unrelated to hand responding. Knowledge, implicit knowledge and metaknowledge and Saffran (1989a, b) and Coslett et al. (1993). These investigators developed sensitive testing measures that enable thorough assessment of both verbal and nonverbal reading capabilities of patients with pure alexia, with particular emphasis on the evaluation of implicit reading capabilities. Explicit word identification Categorical decision tasks These tasks required the patient to make semantic judgements on written word stimuli for three separate categories including: (i) animal names, (ii) food names and (iii) proper names. Each category was assessed individually. For animal and food categories, a total of 75 word stimuli consisting of 25 targets (e.g. turtle), 25 visually similar words (e.g. turban) and 25 visually unrelated words (e.g. abolish) were presented. All foils were matched with targets for frequency and length (<20 counts per million; Kucera and Francis, 1967). Visual foils differed from targets by at least one letter (e.g. lambversus-lamp) and were paired by a minimum of two and a maximum of five letter strings. Word stimuli were arranged in pseudorandom order. The subject was instructed not to read each word, but to attempt to make yes/no judgements on the basis of categorical inclusion or exclusion (i.e. 'Is this an animal?'; 'Is this food?'). No clue was provided as to how many words were grouped within each category. In the proper name category, a total of 100 word stimuli consisting of 50 male and 50 female names, were matched for length (three to eight letters) and visual similarity. Of the 50 male-female cognates, 34 shared at least the first three letters (e.g. Bernard versus Bernice), or differed by just one or two letters (e.g. Mark versus Mary). Stimulus presentation followed that described in the previous tasks. The patient was asked whether each name was suitable for a male or a female (e.g. 'Is this name suited for a male or female?'). In both animal and food categories, patient R.E. was able to identify target word types with a better than chance degree of accuracy (80% or 20 out of 25 correct for animal names; 84% or 21 out of 25 correct for food names). The patient also demonstrated above chance accuracy in distinguishing proper names (76%, 76 out of 100 correct). Errors in discriminating stimulus words were visually based, in that the patient was more inclined to incorrectly accept the visually related foils as category members. This in fact occurred in nearly 60% of trials for the categorical decision tasks (15 out of 25 animal names; 15 out of 25 food names) and in 70% (17 out of 24) of trials for the proper name discrimination task. R.E.'s relatively strong semantic categorization abilities resemble those alexic patients studied by Coslett and Saffran (1989a, b) and Coslett et al. (1993). Lexical decision task: words versus non-words In this task, stimuli were selected from the word lists of Kucera and Francis (1967) and Coltheart et al. (1977). Word types consisted of 40 high frequency words (mean = 445.8 counts per million; Kucera and Francis, 1967), 40 low frequency words (one count per million), 40 'word-like' words (Hi-N; e.g. 'blane'; Coltheart et al, 1977), and 40 non-words (Lo-N; e.g. 'frip') for a total of 160 word stimuli. Word length varied from three to five letter strings. Stimuli for all reading tests were printed on a 3X5 inch index card in lower case letters. The patient was instructed to indicate whether each letter string formed a 'real' or true word (i.e. 'Is this a word?') via yes/no responding. There proved to be a substantial frequency effect in the patient's individual scores on word types. The patient correctly accepted as real words 95% of the high frequency words and 30% of the low frequency words. Similarly, her responses to Hi-N (wordlike) words (where she was required to reject these letter strings as real/true words) were only 42% accurate, whereas her responses to Lo-N (non-word like) words were as high as 75% accurate. The effects of frequency and orthography on lexical decision are comparable to the findings obtained in Coslett and Saffran's (1989a) subjects. Lexical decision with suffixed stimuli The format for this assessment followed that originally outlined by Shallice and Saffran (1986) and repeated by Coslett and Saffran (1989a, 1992). A total of 63 root morphemes including an equal amount of nouns, verbs and adjectives, were presented under three conditions: unsuffixed, suffixed and incorrectly suffixed. In appropriately suffixed conditions, the following bound morphemes were used: -s, for nouns; -s, -ed, -ing, for verbs; -ly, -(i)est, (i)er, -ous for adjectives. These same morphemes Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 Imagery word lists were adapted from Paivio et al. (1968). Frequency and length variability were controlled using words borrowed from Kucera and Francis (1967). In order to assess the effects of parts of speech, the patient was presented with 60-word stimuli including an equal number of nouns and functors matched for frequency (mean 2=200 counts per million) and length (mean = 2-7 letter strings). Each word was printed on a 3X5 inch index card. To evaluate the effects of word imagery, the patient was shown 80 words consisting of 20 high imagery and high frequency nouns (=£6.0; >50), 20 high imagery and low frequency nouns (=£6.0; <20), 20 low imagery and low frequency nouns (=£3.03; <20), and 20 low imagery and high frequency nouns (=£3.03; >60). Processing of affixes was assessed utilizing 40-word pairs; one suffixed (e.g. towed) and one pseudosuffixed (e.g. tower). Both word types were paired by free morphemes (e.g tow) or root word forms. Viewing time was restricted to 3 s on initial trials, but subsequently eliminated since the patient was unable to identify individual letters (i.e. read letter by letter). Words were matched for frequency and length (<150 counts per million; 1 letter variant). In all the above tasks, the subject was totally unable to read word stimuli aloud regardless of type, length or imageability. 793 794 T. E. Feinberg et al. Comments Considering the extent of our patient's complete lack of explicit-verbal reading, R.E.'s nonverbal processing of written word stimuli, as judged by performance on catagorical decision and lexical decision tasks, though not perfect, was relatively good. Moreover, her reading was largely influenced by word frequency or commonality as well as letter string configurations or orthographic familiarity, but was relatively insensitive to suffixing. This pattern of performance was similar to those patients studied by Coslett and Saffran (1989a, b, 1992) Coslett et al. (1993) and earlier by Shallice and Saffran (1986). Assessment of metaknowledge On tasks where the patient showed poor access to explicit names or to the identities of objects and words, yet demonstrated better than chance performance on matching or forced choice decision tasks, the knowledge that she demonstrated appeared to be outside her awareness. In other words, her performance demonstrated access to stored knowledge, yet raised questions concerning the accessibility of this knowledge to consciousness. In order to determine the relationship between knowledge and awareness of knowledge (metaknowledge), we devised two metaknowledge probes, each of which was intended to measure different aspects of 'knowing'. Predictive meta- knowledge, termed judgement of knowledge, was examined prior to response accuracy measures (i.e. 'Do you know what this picture or word is or means?'). Determinant metaknowledge, termed judgment of accuracy, was examined after obtaining accuracy measures (i.e. 'Are you sure or are you guessing?') We first set out to explore whether the patient had a subjective sense of intact or impaired semantic knowledge when first presented with a pictured object prior to responding (judgement of knowledge). If the domain of knowledge accessed during the matching tasks is truly tacit, 'implicit' or dissociated from awareness in a way that is analogous to blindsight (Weiskrantz et al., 1974; Weiskrantz, 1986), then one would expect the patient to deny recognition of the stimulus. At the very least, there should be little relationship between the patient's judgement of knowledge and her actual accuracy on the matching task. Likewise, if the knowledge utilized in the matching task is unconscious, then once she has responded, the patient should also not know whether her response was accurate or not. Therefore, a poor hit rate on judgement of accuracy would also be predicted and should be no better than the hit rate on judgement of knowledge. Since the assessment of judgement of knowledge is done prior to task response per se, it is not dependent upon the actual conditions of the matching task (e.g. number of choices). However, the judgement of accuracy could be dependent upon such variables since judgement of accuracy is assessed after the matching task. We therefore tested the patient on two versions of the matching task: an easier version (two-choice matching) and a more difficult version (five-choice matching), to yield potentially different levels of accuracy. We reasoned that if judgement of accuracy varies with task difficulty and accuracy (for instance, if judgement of accuracy hit rate increased with accuracy), then this would suggest the patient indeed has insight into her knowledge and, therefore, this knowledge is not dissociated from awareness. Finally, we investigated the matching trials on which the patient denied knowledge of the stimulus in the assessment of judgement of knowledge. We reasoned that this subset of trials meets the most stringent criterion for implicit knowledge. First, we considered whether the patient had a better than chance accuracy on these trials, a result consistent with implicit knowledge. Then, we considered only the trials when she responded accurately and asked whether her judgement of accuracy remained poor. We reasoned that if the knowledge utilized in the matching tasks was truly dissociated from awareness, her judgement of accuracy should be no better than chance. If judgement of accuracy on these trails exceeded chance performance, than this would suggest the knowledge had some access to conscious awareness. In addition, on this same subset of trials, we compared judgement of accuracy performance on two-choice matching trials with that obtained on five-choice matching trials. We reasoned that if the knowledge used in matching tasks was dissociated from awareness, and the patient was truly 'guessing' her Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 were used for inappropriately suffixed stimuli, except they were mismatched with incorrect roots or free morphemes to form unconventional or unusual orthographies (e.g. ankle— 'anklmg'). Root word length ranged from four to nine letter strings. An additional 63 non-words were formed by changing one to two letters within each noun, verb and adjective root morpheme (e.g. donkey—'norkey'). Non-word stimuli were presented in two instances; unaffixed and correctly affixed (i.e. free morphemes attached to original stem/root word—e.g. 'norkeys'). All word stimuli were presented in pseudorandom order. Testing trials consisted of 21 blocks of 15 word stimuli, including all word and non-word types. For each trial, the patient was required to make a yes-no judgment. Overall, the patient achieved an accuracy of 65%. Breakdown of this performance indicates that R.E. correctly accepted as true words appropriately affixed root forms (66% correct: 80% nouns; 57% verbs; 57% adjectives) and unaffixed root forms (66% correct: 76% nouns; 66% verbs; 57% adjectives), and correctly rejected the unaffixed nonwords (68% correct) and affixed non-words (78% correct). However, R.E. was unable to reliably reject the inappropriately affixed root forms (46% correct). In summary, our patient's ability to identify appropriate word forms was restricted to word/non-word root morphemes, and like Coslett and Saffran's subjects (1989a), R.E. could not reliably discriminate suffixed word forms. Knowledge, implicit knowledge and metaknowledge 795 Table 2 Descriptive statistics and cumulative binomial probabilities (P) for judgement of knowledge (JK), accuracy and judgement of accuracy (JA) across five task domains Task (A) Px-Px (5) (B) Wd-Px (5) (A)+(B) combined (C) Px-Px (2) (D) Wd-Px (2) (C) + (D) combined (E) Wd-Obj locate Accuracy JK JA H/R % P < H/R % P < H/R % P < 51/80 41/80 92/160 36/80 47/80 83/160 54/80 (63.75) (51.25) (57.75) (45.00) (58.75) (51.88) (67.50) 0.01 NS 0.05 NS NS NS 0.0001 50/80 50/80 100/160 80/80 69/80 149/160 30/80 (62.50) (62.50) (62.50) (100.00) (86.25) (93.13) (37.50) 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 54/80 47/80 101/160 63/80 66/80 129/160 59/80 (67.50) (58.75) (63.13) (78.75) (82.50) (80.63) (73.75) 0.001 NS 0.001 0.0001 0.0001 0.0001 0.0001 NS = not significant. Methods A series of five semantic matching tasks were given, comparable to the semantic matching tasks described above. Each required the subject to match a line drawing of a target stimulus with a semantically related response choice. Tasks included: (A) Picture-to-Picture matching (five-choice): (B) Word-to-Picture matching (five-choice); (C) Picture-toPicture (two-choice); (D) Word-to-Picture matching (twochoice); and (E) Word-Object Location (unrestricted choice). All tasks utilized visual and semantic foils {see Semantic matching task, above) except task C which utilized foils generally unrelated to targets to make this task of comparable difficulty to task D. Written word picture tasks also included phonological foils. The foils for task E were the other potential targets randomly present in the room {see Picture to object locating, above). No time limits on responding were imposed on any tasks. Each task consisted of 80 trials utilizing different stimulus items and each trial was comprised of three probes, each yielding a binary response. Upon stimulus presentation, the subject was first asked to judge her knowledge of the stimulus (i.e. was she sure she knew what the stimulus was, or not?). Next, she was asked to match the stimulus with the appropriate response (correct/incorrect?). Finally, she was asked to judge the accuracy of her performance on the matching task (i.e. was she sure her answer was correct, or not?). Data analysis Within each of tasks A-E, judgement of knowledge, performance accuracy and judgement of accuracy were construed as separate questions consisting of 80 independent trials and analysed by means of the cumulative binomial probability distribution (Harvard University Computational Laboratory, 1956; Hays, 1973). For these tests, the underlying assumption of independent trials was evaluated by examination of the autocorrelation function for responses across 80 trials in each probe. Consistent with the assumption of independent trials, no significant autocorrelations obtained. Data analysis tested three main hypotheses. In an initial analysis, we wished to ascertain the probability that the subject's pattern of responses might have arisen by chance. In particular, we tested her accuracy rate for each task and the 'hit rate' for veridical judgement of knowledge and judgement of accuracy responses (i.e. when she accurately predicted or judged her performance). Secondly, in an effort to assess the effect of task complexity on judgement of accuracy, the subject's hit rate for the judgement of accuracy probe were compared across the two-choice tasks, five-choice tasks and the unrestricted-choice task by ANOVA and Tukey HSD multiple comparison procedures. Thirdly, we pooled both two-choice tasks and both five-choice tasks and examined in detail the subset of trials on which the subject judged that she was unsure of her knowledge of the stimulus. For this subset, we assessed her actual performance and the judgements based upon that performance. Results Overall accuracy, judgement of knowledge, judgement of accuracy Although her performance ranges from 37.5% accurate on the Word-Object Locate to 100% on the two-choice PicturePicture matching task, the subject's accuracy levels were significantly above chance expectation for all of the five tasks (Table 2). Table 2 also includes the subject's 'hit rates' for trials where she was certain in her judgements and accurate in her responses along with those where she was uncertain and inaccurate. Taking into account the differing thresholds for chance performance across the five tasks, the subject's judgement of knowledge hit rates were considerably worse than her actual performance. This remains true even Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 responses, then judgement of accuracy should not vary with performance accuracy. On the other hand, if the knowledge utilized in these tasks had access to awareness, then judgement of accuracy performance should vary positively with performance accuracy, and judgement of accuracy on twochoice (high accuracy) tasks would be significantly better than judgement of accuracy obtained on five-choice (low accuracy) tasks. 796 T. E. Feinberg et al. Table 3 Descriptive statistics and cumulative binomial probabilities for the subset of trials where the subject reports she is only guessing (G) on judgement of knowledge (JK) Five-choice tasks (A+B pooled) Total items JK = G JK = G and Accuracy = Match n= 160 78/160 (48.75%) 43/78* (55.13%) Two-choice tasks (C+D pooled) 82/160 74/82* . n = 160 (90.24%) (51.25%) Accuracy JK JA Sure 37 JK = G and Accuracy = Match and JA = G Sure 82 I 64.91% Correct 57 1 69.51% | 35.09% Sure 5 5^5%^H 32/43** (74.42%) Incorrect 25 20.00% 30.49% Guess 20 80.00% n= 160 15/74 (20.27%) Sure 11 Items */><0.05; **/ > <0.01. on the tasks where her accuracy levels were quite high. Her judgement of accuracy hit rates were generally better, falling significantly below her actual accuracy only on the fivechoice Word-Picture task. In this analysis, we addressed the question of whether the subject's judgement of accuracy hit rate differed with task complexity. With two-choice items (n = 160) and five-choice items (n = 160) pooled and the unrestricted-choice items from Word-Object Locate (n = 80) included, we obtained a significant task effect overall [F = 6.31 (2,397), P = 0.002]. The Tukey HSD multiple comparisons yield only a single significant result, indicating that the judgement of accuracy hit rate was significantly higher on the two-choice items than the five-choice items (H/R2 = 0.81, H/R5 = 0.63; P = 0.001). Other comparisons from this procedure were nonsignificant. Results of the third analysis are presented in Table 3. Here, we omit the Word-Object Locate task and pool the twochoice and five-choice tasks so that each set comprises 160 items. We restrict the analysis to those trials where the subject states she is uncertain of her knowledge of the target stimulus. A flow chart of these responses is shown in the lower arms of Figs 2 and 3. Of 160 five-choice items, she states she is guessing about the nature of the stimulus for nearly half the tests (78 out of 160, 48.75%). However, on these items where she judged her knowledge of the stimuli to be poor, she accurately solved the task for over half of the tests (43/78, 55.13%). Given a one in five probability of being correct by chance alone her level of accuracy for this subset of items is statistically significant. Moveover, her judgement of accuracy remains poor, accurately judging her performance on only 25.58% 55.13% Guess 32 Incorrect 35 74.42% Sure 2 5.71% 44.87% Guess 33 | 48J5%| 94.29% Fig. 2 Flow chart representing subject's responses to n = 160 items comprising both five-choice matching tasks combined. Reponses are binary for each of three probes: judgement of knowledge (JK), accuracy of match and judgement of accuracy (JA). JK JA Accuracy Sure 63 Sure 78 • 1 n= 160 Analysis of judgement of knowledge 'guessing' responses Correct 43 Items | 84.00% Guess 12 96.15% | 16.00% Sure 2 48.75% 1 1 1 1 | I Correct 75 Incorrect 3 66.67% 3.85% Guess 1 33.33% Sure 59 Correct 74 90.24% Guess 82 | | 79.73% Guess 15 20.27% 51.25% Incorrect 8 9.76% Sure 2 25.00% Guess 6 75.00% Fig. 3 Flow chart representing subject's responses to n = 160 items comprising both two-choice matching tasks combined. Reponses are binary for each of three probes: judgement of knowledge (JK), accuracy of match and judgement of accuracy (JA). 11 out of 43 or 25.58% of trials, and not significantly different from what one might expect from chance. Of 160 pooled two-choice items, the subject again judges Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 I Guess 78 Comparison of judgement of accuracy 'hit rates' two- versus five-choice Guess 20 Knowledge, implicit knowledge and metaknowledge her knowledge of the stimulus to be poor about half the time (82 out of 160, 51.25%). On these items, she nevertheless accurately solves most of the tasks (74 out of 82, 90.24%). However, unlike under the conditions of five-choice responding, once she has accurately solved these items, her judgement of accuracy is comparatively good (59 out of 64, 79.73%), and again far better than predicted by chance. In addition, her degree of sureness on accurate trials (59 out of 74, 79.73%) was significantly higher than her degree of sureness on trials where she performed the match inaccurately (two out of eight, 25%). Comments Discussion Our patient's performance across a number of tasks was consistent with the syndrome of associative visual agnosia. Despite adequate elementary perception, language and praxis abilities, she could not demonstrate object recognition by naming visually presented objects, gesturing to visually presented objects, matching pictures to real objects, or locating an exemplar of a depicted object in the environment. Her agnosia extended to the recognition of printed words, in that she had alexia without agraphia. Nevertheless, on certain tasks she was able to demonstrate considerable knowledge of visually presented objects and words. She could associate pictures that were semantically related, and perform lexical decision and word categorization at levels of performance well above chance. How can the dissociation between impaired and preserved recognition abilities be explained? As mentioned above, three types of explanation have been put forward in the literature on implicit visual recognition. Several aspects of the data collected with our patient are relevant to testing among these alternatives as explanations for the dissociation in this case. The first explanation we shall consider is a disconnection, anatomical or functional, between object and word recognition mechanisms on the one hand and other mechanisms necessary for conscious, overt recognition on the other. The metacognitive task of judging recognition accuracy is relevant to testing this explanation. If anything can be considered a criterion for overt (as opposed to covert) recognition, it is the subject's awareness of the fact that she has recognized. We obtained direct measures of this awareness, and found it to be reasonably accurate and associated with her recognition performance. This is inconsistent with the view that systems required for awareness of recognition are deprived of inputs from recognition systems. The disconnection account also fails to explain the distinctive profile of reading abilities seen in this case and in other cases of preserved reading in pure alexia (Coslett and Saffran, 1989a). The pronounced concreteness and frequency effects and the insensitivity to affixes do not follow in any way from a disconnection from conscious awareness, but are explained by the right hemisphere hypothesis. If our patient's residual abilities were due to right hemisphere processing, we need to explain why unimanual responses were no more accurate with her left hand than with her right. To explain the lack of difference between the hands, we would have to conclude that: (i) the disconnection was only partial, making some semantic information available to the left hemisphere to direct right hand semantic matching, but with visuo-verbal connections damaged which prevented visual naming; (ii) the right hemisphere was completely disconnected from the left but could direct on its own right hand as well as left hand pointing; or (iii) the information generated by the right hemisphere was made available to both hemispheres but was severely degraded due to limited right hemisphere/lexical abilities. In this case complete Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 When required to perform match to sample tasks, the patient's overall accuracy varied according to task difficulty. The worst performance was achieved in the locate task with a match to as many as 40 possible choices, whereas the best performance occurred in the two-choice match to sample task, with the five-choice falling in between. Thus, performance accuracy varied as a direct function of the number of choices. This is most compatible with a 'partial knowledge' hypothesis, since there is no reason to believe that the stimuli were made 'more implicit' by increasing the number of choices. Judgement of knowledge hit rates were generally poor, such that the patient was as likely to overestimate as underestimate her knowledge. Thus, although she did not deny any knowledge in the sense of being unaware of knowledge (the traditional notion of implicit knowledge) one could argue she was nonetheless a poor observer of her own knowledge, and that the knowledge itself was dissociated from awareness. However, analysis of judgement of accuracy hit rate revealed that under conditions of both high (two-choice) and low (unlimited choice) accuracy, the patient did show significant insight into her knowledge with better than chance performance on judgement of accuracy in both of these tasks. Therefore, she showed maximum insight when she possessed very much or very little knowledge. On five-choice items with middle levels of accuracy, the patient had significant judgement of accuracy on only one of two tasks. Here we witness a clear relationship between knowledge and judgement of accuracy, which again argues for a partial knowledge interpretation. Finally, when we analysed the subset of trials when the patient initially claimed she could not identify the stimulus, we found that even in these trials the patient had better than chance accuracy on both two- and five-choice tasks, which would argue for implicit knowledge. While judgement of accuracy on five-choice trials remained poor, on two-choice trials the patient once again showed significant insight into her abilities. In addition, she was 'sure' of her responses nearly 80% of the trials when she was accurate, as opposed to only 25% when she was inaccurate. Thus, insight into knowledge varied once again with knowledge, and is thus most consistent with a 'partial knowledge' hypothesis. 797 798 T. E. Feinberg et al. Our patient showed an increased rate of accuracy on the matching tasks going from the relatively unbounded set of alternatives in the room to five alternatives and from five alternatives to two. Shallice and Saffran (1986) also found that their alexic patient did worse on a five-choice as compared with a twochoice word categorization task. This pattern of performance is consistent with a degraded object recognition system, which can manifest its residual knowledge when there is sufficient external constraint on possible responses. An analogy would be a nearsighted person who can read a distant sign if he knows that the sign must say one of two words, but cannot decipher it without such a constraint. Although our data do not absolutely distinguish whether the degraded system is a bilateral system that has lost its left hemisphere component (a version of the right hemisphere hypothesis) or is a left hemisphere system that is damaged but not obliterated, the nature of our patient's reading pattern does suggest a disproportionate right hemisphere contribution. Finally, the degraded system hypothesis is consistent with the correlation between metacognitive judgements and recognition performance, because according to this hypothesis a single system underlies both. An additional explanation for the patient's decreased matching abilities with increased choices would be that increasing the number of choices also increases the number of comparisons necessary to make a correct match. The patient's residual object knowledge might be in some way inaccessible to some other cognitive system necessary for such matching capabilities. While this remains possible it does not account for failures of recognition that occurred where matches or comparisons were not required such as spontaneous, real life situations, gesturing to pictures or objects, etc. Some of the differences demonstrated across these tasks could be attributed to chance, as increasing the number of alternatives decreases the probability of a correct response by chance alone. Although awareness of having perceived is clearly the primary criterion by which overt and covert perception can be distinguished, there has been little systematic investigation of subjects' awareness in the literature on covert or implicit perception. To our knowledge, the present study is the first to assess awareness on a trial by trial basis at the same time as recognition itself is assessed. In our patient, we found that her ability to decide prior to responding whether she could identify a stimulus on a given trial, was poor. If this had been our only assessment of her 'awareness', we would have (prematurely) concluded that her task performance was yet another example of implicit responding. However, we found that when her judgement of accuracy was assessed after responding, her metaknowledge was not random, and indeed correlated with her knowledge, both of which varied with task complexity. Our findings suggest that if implicit knowledge is to be considered 'non-conscious', then the absence of metaknowledge (i.e. 'Do you know?', 'Are you guessing?', 'Are you right or wrong?') should be assessed on every trial in a forced-choice paradigm comparable to the assessment of the knowledge itself. We conclude that our understanding of other phenomena, such as covert recognition in prosopagnosia, implicit perception in neglect and blindsight, could be deepened by a more systematic study of patients' awareness of their perceptual abilities. Acknowledgements The authors wish to thank Dr M. J. Farah for valuable suggestions on the earlier version of this manuscript, Michael Grimaldi and Lynne Cooper for illustrative and photographic materials, and Norma Kamen for manuscript preparation. References Albert ML, Yamadori A, Gardner H, Howes D. Comprehension in alexia. Brain 1973; 96: 317-28. Bauer RM. Autonomic recognition of names and faces in prosopagnosia: a neuropsychological application of the Guilty Knowledge Test. Neuropsychologia 1984; 22: 457-69. Bauer RM. Agnosia. In: Heilman KM, Valenstein E, editors. Clinical neuropsychology. 3rd ed. New York: Oxford University Press, 1993: 215-78. Bauer RM, Rubens AB. Agnosia. In: Heilman KM, Valenstein E, editors. Clinical neuropsychology. 2nd ed. New York: Oxford University Press, 1985: 187-241. Beauvois MF. Optic aphasia: a process of interaction between vision and language. Philos Trans R Soc Lond Biol 1982; 298: 35-47. Benton AL, Hamsher K deS, Varney NR, Spreen O. Facial recognition test. New York: Oxford University Press, 1983. Caplan LR, Hedley-Whyte T. Cueing and memory dysfunction in alexia without agraphia. Brain 1974; 97: 251-62. Coltheart M, Davelaar E, Jonasson JT, Besner D. Access to the Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 specification of words or pictures would not be achieved to the point that naming were possible, but information would be processed to the point where relatively simple matches could be made. The patient's performance on the picture/word locate tasks, where, given target pictures, she had to locate objects in the room, helps us to mediate between these possibilities. If pictured objects were indeed recognized, or stimulus words were somehow read by the right hemisphere, it remains to be explained why the patient could not point with the left hand to the pictured object or word in the environment during picture to object locate tasks. The severe recognition defect equally demonstrated in both hands on these pointing tasks stands in apparent bold contrast to the patient's residual matching abilities. This constellation of findings is best explained by the third alternative, that is, degraded information generated by the right hemisphere yet accessible to both hemispheres, was responsible for the patient's performance. Knowledge, implicit knowledge and metaknowledge internal lexicon. In: Dornic S, editor. Attention and performance VI. Hillsdale (NJ): Lawrence Erlbaum 1977: 535-55. Coslett HB, Saffran EM. Evidence for preserved reading in 'pure alexia'. Brain 1989a; 112: 327-59. Coslett HB, Saffran EM. Preserved object recognition and reading comprehension in optic aphasia. Brain 1989b; 112: 1091-110. 799 Jankowiak J, Kinsbourne M, Shalev RS, Bachman DL. Preserved visual imagery and categorization in a case of associative visual agnosia. J Cogn Neurosci 1992; 4: 119-31. Kertesz A. The Western Aphasia Battery. New York: Harcourt Brace Jovanovich, 1982. Coslett HB, Saffran EM. Optic aphasia and the right hemisphere: a replication extension. Brain Lang 1992; 43: 148-61. Kertesz A. The clinical spectrum and localisation of visual agnosia. In: Humphreys GW, Riddoch MJ, editors. Visual object processing: a cognitive neuropsychological approach. Hove: Lawrence Erlbaum, 1987: 175-96. Coslett HB, Saffran EM, Greenbaum S, Schwartz H. Reading in pure alexia. The effect of strategy. Brain 1993; 116: 21-37. Kinsbourne M, Warrington EK. A disorder of simultaneous form perception. Brain 1962; 85: 461-86. De Haan EHF, Bauer RM, Greve KW. Behavioural and physiological evidence for covert face recognition in a prosopagnosic patient. Cortex 1992; 28: 77-95. Kucera H, Francis WN. Computational analysis of present-day American English. Providence (RI): Brown University Press, 1967. De Renzi E, Zambolin A, Crisi G. The pattern of neuropsychological impairment associated with left posterior cerebral artery infarcts. Brain 1987; 110: 1099-116. Landis T, Regard M, Serrat A. Iconic reading in a case of alexia without agraphia caused by a brain tumor: a tachistoscopic study. Brain Lang 1980; 11: 45-53. Lhermitte F, Beauvois MF. A visual-speech disconnexion syndrome. Report of a case with optic aphasia, agnosic alexia and colour agnosia. Brain 1973; 96: 695-714. Farah MJ. Visual agnosia: disorders of object recognition and what they tell us about normal vision. Cambridge (MA): MIT Press, 1990. Mack JL, Boiler F. Associative visual agnosia and its related deficits: the role of the minor hemisphere in assigning meaning to visual perceptions. Neuropsychologia 1977; 15: 345-9. Farah MJ. Patterns of co-occurrence among the associative agnosias: implications for visual object representation. Cogn Neuropsychol 1991; 8: 1-19. McCarthy RA, Warrington EK. Visual associative agnosia: a clinicoanatomical study of a single case. J Neurol Neurosurg Psychiatry 1986; 49: 1233-40. Farah MJ, O'Reilly RC, Vercera SP. Dissociated overt and covert recognition as an emergent property of a lesioned neural network. Psychol Rev 1993; 100: 571-88. Michel F, Schott B, Boucher M, Kopp N. Alexie sans agraphie chez un malade ayant un hemisphere gauche deafferente> (hemianopsie, hemianesthesie, hemianacousie). Rev Neurol (Paris) 1979; 135: 347-64. Feinberg TE, Rothi LJ, Heilman KM. Multimodal agnosia after unilateral left hemisphere lesion. Neurology 1986; 36: 864—7. Feinberg TE, Schindler RJ, Ochoa E, Kwan PC, Farah MJ. Associative visual agnosia and alexia without prosopagnosia. Cortex 1994; 30: 395-411. Ferro JM, Santos ME. Associative visual agnosia: a case study. Cortex 1984; 20: 121-34. Freund CS. Ueber optische Aphasie und Seelenblindheit. Arch Psychiat Nervkrankh 1889; 20: 276-97, 371-416. Geschwind N. Disconnexion syndromes in animals and man. Brain 1965; 88: 237-94, 585-644. Geschwind N, Fusillo M. Color-naming defects in association with alexia. Arch Neurol 1966; 15: 137-46. Goodglass H, Kaplan E. The assessment of aphasia and related disorders. Philadelphia: Lea and Febiger, 1972. Harvard University Computational Laboratory. Harvard tables of the cumulative binomial probability distribution. Cambridge (MA): Harvard University Press, 1956. Hays WL. Statistics for the social sciences. 2nd ed. New York: Holt, Rinehart and Winston, 1973: 190-3. Iorio L, Falanga A, Fragassi NA, Grossi D. Visual associative agnosia and optic aphasia: a single case study and a review of the syndromes. [Review]. Cortex 1992; 28: 23-37. Osterrieth R. Le test de copie d'une figure complexe. Arch Psychol 1944; 30: 206. Paivio A, Yuille JC, Madigan SA. Concreteness, imagery, and meaningfulness values for 925 nouns. J Exp Psychol 1968; 76 (I Suppl): 1-25. Patterson K, Kay J. Letter-by-letter reading: psychological descriptions of a neurological syndrome. Q J Exp Psychol: Hum Exp Psycho! 1982; 34: 411-41. Raven JC. Guide to the standard progressive matrices. New York: Psychological Corporation, 1965. Rubens AB, Benson DF. Associative visual agnosia. Arch Neurol 1971; 24: 305-16. Schacter DL, McAndrews MP, Moscovitch M. Access to consciousness: dissociations between implicit and explicit knowledge in neuropsychological syndromes. In: Weiskrantz L, editor. Thought without language. Oxford: Clarendon Press, 1988: 242-78. Shallice T, Saffran E. Lexical processing in the absence of explicit word identification; evidence from a letter-by-letter reader. Cogn Neuropsychol 1986; 3: 429-58. Taylor A, Warrington EK. Visual agnosia: a single case report. Cortex 1971; 7: 152-61. Teuber HL. Alteration of perception and memory in man. In: Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 Dejerine J. Contribution a l'etude anatomo-pathologique et clinique des differentes varietes de cecite verbale. Compt Rend Biol Par 1892; 4: 61-90. 800 T. E. Feinberg et al. Weiskrantz L, editor. Analysis of behavioral change. New York: Harper and Row, 1968: 268-375. Weiskrantz L. Blindsight: a case study and implications. Oxford: Clarendon Press, 1986. Warrington EK, Shallice T. Word-form dyslexia. Brain 1980; 103: 99-112. Weiskrantz L, Warrington EK, Sanders MD, Marshall J. Visual capacity in the hemianopic field following a restricted occipital ablation. Brain 1974; 97: 709-28. Wechsler D. WAIS-R Corporation, 1981. manual. New York: Psychological Wechsler D. Wechsler Memory Psychological Corporation, 1987. Scale-Revised. New York: Received September 22, 1994. Revised December I, 1994. Accepted February 10, 1995 Appendix Stimuli used in Picture-to-Object Locating Task book scissors window shoe pencil desk picture door telephone stapler 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. doorknob tape dispenser cup shirt chair pen eye computer ear flower Downloaded from http://brain.oxfordjournals.org/ at Athabasca University on June 9, 2016 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.