Neuropsychologia 40 (2002) 2167–2177 When writing 0 (zero) is easier than writing O (o): a neuropsychological case study of agraphia M. Delazer a,∗ , A. Lochy a,b , C. Jenner a , F. Domahs a , Th. Benke a a Universitätsklinik für Neurologie, Klinik für Neurologie, Anichstraße 35, 6020 Innsbruck, Austria b Université Catholique de Louvain, Louvain, Austria Received 17 December 2001; received in revised form and accepted 15 March 2002 Abstract Though a few case studies reported a dissociation between intact writing of Arabic and impaired writing of alphabetical script, a detailed experimental analysis of such a dissociation is still lacking. We report a follow-up study of a patient with a parieto-occipital lesion who is affected by severe peripheral agraphia for letters, but not for Arabic digits. While letters in writing to dictation are frequently illegible, distorted, or consist in meaningless strokes, Arabic digits are well-formed and fluently produced. In a series of tasks, including copying of letters with tachistoscopic presentation and handwriting on a digitizing tablet, several processing levels are assessed in order to localize JS’ functional writing impairment and to determine different processing routes for letters and for numbers. Overall, the results of the experimental investigation suggest a notation specific deficit in the activation of graphomotor patterns for letters, but not for digits. The study thus adds evidence to the so far reported dissociations between Arabic and alphabetical scripts. © 2002 Elsevier Science Ltd. All rights reserved. Keywords: Dysgraphia; Numerals; Digits; Digitizer; Kinematic analysis 1. Introduction How is it possible that a subject is able to draw a round shape representing a number without any effort, but struggles to write the same shape representing a letter? Can this happen—even if the subject perfectly recognizes and copies all letters? In this case study, we present a patient affected by peripheral agraphia for alphabetical stimuli, but not for Arabic numerals. Since Déjerine [13] it is well known that in pure alexic patients the elaboration of Arabic and alphabetical script may dissociate with severely impaired reading of words and intact reading of numerals. Less is known about similar patterns in writing. Though a small number of case studies reported a dissociation between intact writing of Arabic and impaired writing of alphabetical script [2,31,64], a detailed experimental analysis of such a dissociation is still lacking. Aim of the present study is to elucidate the processes implicated in handwriting and, in particular, to identify different processing routes for letters and for numbers. The analysis of the performance of patient JS allows to further specify processing levels implicated in the two scripts and adds evidence to the so far reported dissociations ∗ Corresponding author. E-mail address: margarete.delazer@uibk.ac.at (M. Delazer). between the writing processes employed in two cognitive domains, i.e. language in general and number processing. At a very general level, acquired impairments of writing may be classified as central agraphias, affecting both oral and written spelling, and peripheral agraphias, affecting only the written output. Central agraphias arise from damage to one or more components in generating the spellings of familiar or unfamiliar words. The assumption of separate, lexical and sublexical, processing pathways for familiar and unfamiliar words, respectively, received empirical support by selective impairments of writing one or the other stimulus type [4]. Whether lexical or sublexical procedures are used, all pathways access the graphemic buffer, a temporary memory store which holds amodal, spatially arrayed graphemes [51]. These graphemes are thought to be symbolic representations of letter identity, void of name or shape. It is only beyond this central, amodal representation that modality-specific mechanisms, dedicated to oral or written spelling, assign name or shape to the abstract graphemes. These modality specific mechanisms may be disrupted by brain damage leading to selective impairments in oral or written spelling. Several patients with peripheral agraphia have been identified whose oral spelling was intact or only marginally defective [1–3,10,21,23,32,36,43,46,51,55,64]. Once the graphemic buffer holds information about the identity of the required letters, the correct case, font and shape have to 0028-3932/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 8 - 3 9 3 2 ( 0 2 ) 0 0 0 4 4 - 1 2168 M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 be assigned to the letters and finally appropriate motor programs have to be activated. Ellis [17–19] and Margolin [40] proposed two effector-independent levels of representation, the first one consisting in a long-term store of allographs from where the respective allographic forms are retrieved, the second one being a graphomotor pattern which specifies direction, relative size, position and order of strokes, but not their absolute size and duration, or how they will be effected. Finally, the graphic code provides the specific neuromuscular instructions to execute writing movements [40]. Alternatively (or additional) to visuospatial letter representations, Rapp and Caramazza [51] suggest a relatively abstract, effector-independent representational level that specifies the features of the component strokes of letters. Over the last decades a number of case studies on peripheral agraphia, traditionally labeled as apractic agraphia, has been published which provide a heterogeneous picture of peripheral writing impairments and associated deficits. Some of these cases were related to limb apraxia [1,8,21,30,61] while others were not [2,3,8–10,12,32,34,36,46,54]. Baxter and Warrington [3] described a patient with severe deficits in writing letters and words to dictation, letters being distorted and illegible. In contrast, he showed no limb apraxia, his language and oral spelling were intact, as well as his copying and constructional skills. Baxter and Warrington [3] introduced the term ‘ideatonial agraphia’ which they differentiated from agraphia due to ‘motor apraxia’. A motor apraxic would have general difficulties in carrying out writing movements, movements being slow and laborious in writing to dictation as well as in copying. In ideational agraphia, on the other hand, access to the appropriate motor programs would be impaired. Crucially, there would be disproportionate difficulty in writing to dictation as compared to preserved copying of letters and words. Indeed, both types of impairment have been observed. While some patients improve their handwriting when copying [3,9,12,21,34,36,54,64] others do not improve with a model [1,32,41,48,61]. Patients affected by peripheral agraphia also vary with respect to other associated deficits. Some agraphic patients show a deficit in letter imagery [9,36], others have good imagery [1] and several patients were not tested on letter imagery. In some patients peripheral agraphia is associated with aphasia [9,21,54], while other patients show intact spoken language [61]. Overall, the term peripheral agraphia denotes a heterogeneous group of writing impairments and a review of the literature proves that different functional lesions may cause peripheral deficits in handwriting. In line with the heterogeneous picture of associated deficits, the analysis of anatomical lesion sites in patients with peripheral agraphia discloses a similarly heterogeneous pattern. The two most frequent sites of lesions are the left posterior region, in particular the left parietal lobe [1,3,8,9,12,27,36,43,48,64] and the left frontal lobe [32] including the left premotor cortex [2]. Also left-sided thalamic lesions have been reported [46], which, however, lead to reduced metabolism in the left dorsolateral premotor area. Though most studies reported left-sided lesions, there are also cases of right-sided lesions causing peripheral agraphia in left-handed subjects [61], as well as in right-handed subjects [54]. Deficits in writing alphabetical script are often associated with deficits in writing Arabic script, thus pointing to common stages of the cognitive processes involved in writing the two scripts or to common resources shared by them [15]. However, there are also important differences in the processing demands of the two scripts which may lead to either distinct error patterns in the two scripts or to the selective impairment of one of the two scripts. As pointed out by Grewel [25,26], Arabic numerals constitute a semiotic system which differs from alphabetic notation in several respects. Each single element of Arabic script has a meaning in itself, for example, the symbol ‘4’ denotes the concept expressed by the word “four” (e.g. four dots). In contrast, elements of alphabetical script, e.g. the letter ‘A’, do not have meaning by themselves. The system of Arabic notation is strictly defined, i.e. the symbols for numbers, such as 2, and for operations, such as + or ×, allow only one interpretation [26]. Moreover, the Arabic notation has a very restricted lexicon using only 10 different symbols. Both the identity of single elements (digits) and the position of the single elements within the number have meaning. In alphabetic script, on the other hand, the single elements have no meaning by themselves and their position within the word cannot be inferred or generated with reference to the semantic representation of the word. The differences between scripts may contribute to selective deficits in writing either letters or numerals and to the typical error patterns observed in the two notations. In 1954, Zangwill [64] reported a left-handed man with a left-sided glioma which caused agraphia as the main symptom. His handwriting of alphabetical stimuli was characterized by numerous malformations and substitutions, while he was seemingly able to write Arabic numerals. Unfortunately the assessment of number writing was sparse and consisted only in writing the series from 1 to 20, whereas, no systematic investigation of Arabic script was performed. The most impressive and best studied case of preserved numeral writing in peripheral agraphia was reported by Anderson et al. [2]. After a circumscribed surgical lesion in the left premotor cortex, more specifically in Exner’s area, their patient showed severe and stable alexia and agraphia. While other cognitive functions were found to be intact, including oral spelling and drawing, writing letters and words was virtually impossible, writing attempts resulting in spatially distorted and illegible letters. Most interestingly, this patient was able to write and read Arabic numerals without difficulty and could also easily perform written calculations. Anderson et al. [2] explain the dissociation with the different motor sequences employed in the assembling of letters and numerals out of movement components. As they point out, the precise sequence of movement components as well as the context where they appear are different for the alphabetic M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 script and the Arabic notation. Thus, they interpret the severe agraphia for alphabetical stimuli as, “a highly specific impairment of visuoconstructional ability, which pertained only to the domain of graphemes and words” ([2, p. 761]). As a tentative explanation for the separate neuroanatomical structures involved in writing alphabetical and Arabic stimuli they propose that in childhood number concepts are typically acquired in strong association with the digits of the hand, possibly leading to a separate neural implementation [5]. The separate processing of the two scripts is not only supported by preserved production of Arabic script in cases of peripheral agraphia [2,8,12,64], but also by reports of qualitatively different error patterns appearing in writing numbers or words. Gerstmann [22] described a patient affected by severe agraphia for alphabetical script who produced repeated strokes without any similarity to the intended letters. The writing of Arabic script was, though also affected, remarkably better. She was well able to produce single elements of three-digit numerals correctly, but frequently inverted the order of the last two digits, thus following the verbal structure of German numerals ignoring the inversion rule (e.g. fünf hundert drei und siebzig (literally: five hundred three and seventy) was written as 537). Thus, this patient showed a peripheral agraphia for letters, but not for digits which were well-formed, but written in the wrong order. Delazer and Denes [15] reported number writing in an agraphic patient whose alphabetical script was restricted to graphemic jargon [56] and whose writing of Arabic script was restricted to single digits. While writing alphabetic script remained severely impaired over months, writing Arabic script improved considerably. Interestingly, errors were almost exclusively of the syntactic type (cinquecentoventi (five hundred twenty) was written as 50020) and the changing error pattern over time reflected some difficulties observed in developmental studies on the acquisition of Arabic numeral writing [50,57]. One explanation for the different outcome in writing the two scripts may be found in the compensation mechanisms they offer. Writing Arabic script may be facilitated not only by the lower number of elements, but also by the rule-based relation between semantic representation and syntactic frame of the numeral. Graná et al. [24] reported the gradual recovery of a patient affected by agraphia whose processing of Arabic numerals was initially better than his processing of alphabetical script. However, despite this initial advantage for Arabic script, a follow-up examination evidenced correct writing of letters and words, but still syntactic difficulties in writing complex Arabic numerals. Graná et al. [24] emphasize the different processing levels employed in writing numerals and letters, respectively, leading to differential improvements of the two notations. The initial advantage for Arabic script was explained by a letter-form selection deficit for alphabetical script, but not for Arabic script. The later found disadvantage for Arabic script as compared to alphabetical script was accounted for by the specific 2169 processing requirements of transcoding numerals, implying syntactic rules and algorithms. 2. Outline of the study In the present study we report JS, a patient affected by severe peripheral agraphia. In a series of tasks, including oral spelling, letter identification, recognition of masked letters, letter imagery, allograph letter matching, allograph conversion, writing to dictation and copying, several processing levels were assessed in order to localize JS’ functional writing impairment. In addition to the quantitative and qualitative analysis of written production, writing movements were recorded by means of a digitizing tablet allowing the analysis of the kinematic characteristics of the written production. This technique has only rarely been used in research on peripheral agraphia [65], though it provides a powerful insight into the peripheral processes involved in handwriting which may not be gained by pure observation or error analysis. Another technique of investigation, so far not used in the investigation of agraphia, was a copy task with tachistoscopic presentation of stimuli on a computer screen. Tasks were selected in order to assess two aspects of JS’ writing impairment. First, we aimed to document the dissociation between preserved writing of Arabic script and impaired writing of alphabetical script and to define at least partially different pathways in peripheral writing for letters and for numbers. The comparison between scripts is based on the analysis of error rates, error types and writing velocity in the two performances. Second, a detailed analysis of JS’ deficit in writing alphabetical script was undertaken to provide deeper insight into the peripheral processes involved in handwriting letters and words. In order to localize his functional impairment, the following hypotheses, based on current literature, were drawn. Importantly, intact oral spelling would exclude a central form of agraphia. A deficit at the allographic level should be excluded by intact performance in tasks assessing mental imagery of letters and by intact performance in tasks requiring the recognition of incomplete or degraded letters, the identification of overlapping letters and the matching of different allographs.1 In the output, there should be no errors in selecting the correct case (upper, lower) and font (cursive, print). A deficit in the execution of graphic motor patterns would be excluded by intact and fluent copying of visually presented stimuli. The kinematic analysis of copying with the digitizing tablet allows to differentiate between slow, stroke-by-stroke, copying and fluent production of letters indicating intact graphomotor patterns. Of particular interest is the comparison between writing to dictation and copying. Comparable difficulties in both conditions would point to a deficit in the execution of graphomotor patterns. 1 Let us note, however, that the necessity of accessing the visuospatial shape of letters is an assumption challenged by some authors [51]. 2170 M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 On the other hand, a selective deficit in spontaneous writing, but not in fluent copying, would point to difficulties in activating, but not in executing, graphomotor patterns. In a second copy task, letters were flashed on the computer screen for 200, 100 or 50 ms to be copied by the patient. Presentation times were too short to permit a slow copying or even a precise and detailed recognition of the stimulus. Thus, intact copying in this task would suggest that the patient has access to stored knowledge about letters (shapes or graphic motor patterns) which he is able to activate in this task. Moreover, intact performance in this task would indicate that JS does not depend on a visual model in executing the activated graphic motor code. 3. Case description JS, a 69-year-old, right-handed retired police officer, suffered an infarct of the leptomeningeal branches of the anterior cerebral artery in the left parietal region, causing a right-sided hemiparesis, severe ideomotor apraxia and agraphia. Subsequently he recovered from hemiparesis and apraxia, but remained severely agraphic. Four years after disease onset he was referred to the neuropsychological laboratory for neuropsychological assessment and rehabilitation. A CT scan performed 2 years and 6 months after his CVA showed two old lesions, an ischemic lesion in the left parieto-occipital area and a contusion in the right frontal pole (Fig. 1a and b). Results of the neuropsychological background tests are listed in Tables 1 and 2. JS scored in the average in a short form of the WAIS [11] as well as in a vocabulary test estimating premorbid intelligence [35]. Memory tests evidenced reduced verbal and visual learning and recognition [45,47], but preserved verbal working memory [29]. JS showed no impairment of language; his spontaneous speech was fluent and well articulated, speech comprehension was intact. He performed well on a naming task [45], the token test [33] and a verbal fluency task [44]. He read letters and words fluently, but was unable to write a single word to dictation. Attempts to write letters or words frequently resulted in illegible and distorted shapes or in letter substitutions, which were produced hesitantly. Writing upper and lower case letters seemed equally impaired. In writing JS did not differentiate between print and cursive lower case letters, claiming that he also mixed both types before his infarct.2 There were no signs of residual ideomotor apraxia, neither for the left nor the right hand and no signs of right/left disorientation. In contrast, JS had some difficulties identifying fingers touched by the examiner, in particular on the right hand. In two series of tasks no impairments of visual perception and object recognition were found [63,53]. 2 In the present investigation lower case letters were counted as correct when they were produced correctly in either of the two forms, print or cursive. Fig. 1. Brain lesion standard CT study (left brain on left side) 30 months after disease onset showing lesion in the superior parietal (BA 7) and adjacent occipital (BA 19, precuneus) cortex (a) and underlying white matter. A chronic right frontal lesion is also evident (b). Importantly for the present study, JS was also able to identify incomplete letters and overlapping letters. Average scores were found in a Stroop task (word/color interference [47]) and in the trail making test assessing cognitive flexibility and set shifting [52]. JS performed well in all tasks of number processing (number comparison, reading and writing Arabic numerals, reading number words) and calculation (addition, subtraction, multiplication, division) with the exception of writing number words to dictation. In this last task, again, writing difficulties were apparent. Overall, neuropsychological background tests evidenced severe agraphia for alphabetical stimuli, impaired finger localization and reduced verbal and visual memory. On the other hand, oral language, visual perception, including the recognition of degraded and overlapping letter shapes, imitation of movements, constructive abilities and number processing were found to be intact. M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 2171 Table 1 Neuropsychological background tests Test Score Classification MMSE [20] 26/30 Above the cut-off Estimated intelligence Vocabulary test MWT [35] 25 IQ 95 Average WAIS, short form [11], T scores Information Similarities Picture completion Block design 55 52 58 52 IQ 106 Average Memory Digit span forward (WMS-R; [29]) Digit span backward 7 (53%) 5 (27%) Average Average Verbal memory cerad [45] Verbal learning (items recalled out of 10) Free recall Recognition 4/5/5 2 9 Impaired Impaired Intact Visual memory cerad [45] Constructive abilities Recall 7/11 3 Borderline Impaired Figural memory NAI [47] 4/12 Reduced Language Boston naming test (short form) AAT token test [33] number of errors Writing to dictation Reading words Verbal fluency [44] (animals, fruits, vegetables/min), sum of three categories 14/15 0 0 15/15 39 Intact Intact Impaired Intact Intact Apraxia Test of apraxia, right and left hand Finger localization, number of errors Right/left orientation 65/72 15 20 Intact Reduced Intact Perception and object recognition Visual object and space perception [63] Incomplete letters Silhouettes 19/20 22/30 Above the cut-off Above the cut-off 18/18 17/18 17/18 25/25 23/25 31/32 27/32 30/30 32/32 30/32 Intact Intact Intact Intact Intact Intact Intact Intact Intact Intact Selective attention, cognitive flexibility Trail making test [52] part A Trail making test part B Stroop test (NAI-FWT [47]), difference between naming and interference 69 s 248 s 42 s Average Average Average Number processing Number comparsion (Arabic numerals, spoken number words, written number words) Reading Arabic numerals (one to six digit numerals) Reading number words (corresponding to one to five digit numerals) Writing Arabic numerals to dictation (one to five digit numerals) Writing number words Addition Subtraction Multiplication Division 30/30 18/18 10/10 10/10 0 24/24 24/24 39/42 24/26 Intact Intact Intact Intact Impaired Intact Intact Intact Intact Birmingham object recogntion battery [53] Single letters Overlapping letters Triplets overlapping Minimal feature match Foreshortened match Object decision easy Object decision Association match Item match Picture matching 2172 M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 4. Experimental tasks and results Most of the tasks reported were performed in February and March 2001. In October 2001, the assessment of writing was repeated and a more detailed analysis of writing velocity was performed. Between February and October 2001, JS attended a weekly rehabilitation program and improved his writing abilities. 4.1. Investigation in February 2001 Letter decision: Real letters and non-letter shapes were presented individually in random order. Stimuli were printed in black, bold format (Arial, 120 points) on white cards. Non-letter shapes were randomly composed of parts of real letters cut in two to three pieces. JS had no difficulty to differentiate between printed real and non-letter shapes (52/52 correct; the material was prepared according to the suggestions of Chanoine et al. [7]). Recognition of incomplete letters: Single letters were partially masked and presented on a computer screen (40% covered). JS identified 10 out of 12 stimuli. The two errors resulted from the exchange of letters which may be differentiated from each other only by non-visible, masked details (e.g. Q was classified as O). Verbal–visual letter matching: JS was asked to point to a verbally indicated (real) target letter, which was presented with two distractors. Stimuli were printed on a sheet of paper in black, bold format (Arial, 120 points). Two versions of this task were presented. In the across-category task, distractors were non-alphabetical symbols, i.e. digits and geometrical shapes, one being visually similar to the target, the other not (e.g.: target: “S”, distractor visually similar: “5”; distractor visually dissimilar: “ ”). In the within-category task, targets were presented with two letters, one visually similar to the target, the other not (e.g. target: “p”, distractor visually similar: “q”, distractor visually dissimilar: “s”; see [7]). JS perfectly matched the letters named by the examiner with the targets in the across-category task (20/20), but made two spatial errors in the within-category task (18/20). Allographic letter matching: JS was asked to point to the two allographic forms (upper and lower case letters) of verbally indicated target letters which were presented with two distractors. Stimuli were printed on a sheet of paper in black, bold format (Arial, 120 points). Again, two versions of the task were administered. In the across-category task, target letters were presented with non-alphabetical distractors. One distractor was visually similar to a target-allograph, the other visually dissimilar to both of them (e.g. targets: “A, a”; distractor visually similar: “4”; distractor visually dissimilar: “ ”). In the within-category task, targets (e.g. “M, m”) were presented with a lower and an upper case letter, one visually similar to one of the allographs (e.g. “n”), the other dissimilar to the targets (e.g. “S”; see [7]). In both tasks JS identified the corresponding letter pairs without errors (20/20 correct, respectively). Fig. 2. Writing single upper case letters. JS’ repeated attempts to write Z, A and F to dictation. The last letters in every line are copied from a model. Allograph conversion task: JS succeeded to transform 21 out of 26 lower case letters into upper case letters and 18 out of 26 upper case letters into lower case letters. Writing letters to dictation: JS correctly produced 51/120 (42.5%) single upper case letters and 30/78 (38.5%) lower case letters (no significant difference). JS frequently hesitated, stated that he did not recall how to produce the respective letter and produced nothing for several seconds. When encouraged to do so, he made several attempts, usually not resulting in the correct letter (Fig. 2). Failures consisted in single strokes, letter-like shapes and also in letter substitutions which he immediately recognized as errors. Writing with the left hand did not improve the performance. Writing words to dictation: Fifty-five words, consisting of 4–10 letters, were dictated to be written in upper case letters. JS succeeded to write five words (all of them consisting of four letters) and 185/267 (69.3%) letters of the dictated words correctly. In lower case letters, he wrote 10/83 words (consisting of 3–10 letters) and 283/406 (69.7%) letters correctly. As in writing single letters to dictation, JS was not fluent in his written production, frequently hesitated and stopped his attempts (Fig. 3). Error rates in writing single letters to dictation and in writing letters within in words are not directly comparable. In writing words JS was frequently facilitated by copying a repeated letter or by comparing similar letter shapes (e.g. M and N). Oral spelling: JS was asked to orally spell a part of the words given in the writing tasks (words included four to Fig. 3. Writing words to dictation JS’ attempts to write the words LEIM, NULL (upper case letters) and Sohn, Null (lower case letters) to dictation. After some attempts he was provided with a model for copying for the letters M (in LEIM), N (in NULL), h (in Sohn) and N (in Null). M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 seven letters). Though he mostly failed to write them correctly, he was almost perfect in spelling them orally (59/62 words correct). In contrast to his preserved spelling forward, he was unable to spell words backwards (0/8 correct). Letter imagery: JS was asked to describe the visual shape and characteristics of all upper case letters. His descriptions were given to five raters blind to the targets, who recognized between 21/26 and 25/26 letters (mean 23). JS never gave erroneous descriptions, but some descriptions remained too vague to be recognized by all raters. However, also the less precisely described letters (N and Y) were recognized by 2/5 raters. In a second letter imagery task JS had no difficulties to judge whether upper case letters consist mainly of round (e.g. O, S, C) or straight lines (e.g. T, M, A; 17/17 correct). Visual object knowledge: JS performed almost at ceiling (60/61) in a series of tasks assessing visual object knowledge. In these tasks the size of animals has to be judged, the visual characteristics of animals and the spatial orientation of objects have to be assessed. Finally, crucial differences between objects and crucial differences between animals have to be described. All these tasks are presented as series of questions without providing visual models [39]. Drawing: JS showed marked difficulties in drawing objects or animals (Fig. 4). Only 3 out of 10 drawings were correctly recognized by five out of six independent raters (30% correct). He was perfectly aware that his drawings were distorted or incomplete, but was unable to improve them. Similarly, JS had severe difficulties to complete 16 outline drawings [60] where an important detail was lacking (the head of animals or essential details of objects). While drawing JS frequently commented the typical visual characteristics of the animals (e.g. the hare has long ears), but did not succeed to draw the shapes according to his verbal descriptions. Drawings were rated in two categories, first, whether they contained all essential features (scale from 0 to 1; 0 = incomplete, 1 = complete), and second, whether they were well-formed (scale from 0 to 3; 0 = not well-formed and disproportionate, 3 = perfectly well-formed). JS achieved an average score of only 0.5 in the first category and of 1.03 in the second category (six independent raters). No drawing was judged as complete and well-formed by all raters. 2173 Fig. 5. Writing Arabic numerals: production of Arabic numerals in written subtraction problems. Copying: JS fluently wrote well-formed letters in a copy task with unlimited stimulus presentation (10/10 words correct). In three additional copy tasks all letters of the alphabet were flashed on a computer screen for 200, 100 and 50 ms, respectively. JS was able to copy letters presented on the screen for 200 ms (26/26 upper case letters correct), 100 ms (22/26 upper case letters, 24/26 lower case letters correct) and even 50 ms (26/26 upper case letters correct). The very few errors were omissions most likely due to lapses of attention or due to recognition failures of the letters flashed on the screen. Writing Arabic numerals: In writing two to five digit Arabic numerals to dictation, JS wrote 159/170 numerals correctly (Fig. 5). The 170 numerals contained 452 digits, from which 448 were well-formed and fluently produced (99.1%). The difference between writing single letters (81/198 correct; 40.9%) and writing digits was highly significant (χ 2 = 307; P < 0.000). Errors in writing numerals resulted from syntactic difficulties in writing four and five digit numerals (e.g. 4039), in particular in processing internal zeros. Writing Roman numerals: JS was asked to write verbally presented numbers as Roman numerals. Since JS was not very familiar with Roman numerals, he always transcoded the dictated number into the respective letter name (“hundred. . . should be C”) and then tried to write the letter down. In this way he produced only 2/6 Roman numerals correctly, though he was able to name the necessary letters. 4.2. Investigation in October 2001 Writing letters: JS correctly produced 160/212 (75.4%) single upper case letters and 163/212 (76.9%) lower case letters to dictation (no significant difference). As compared to February 2001, JS’ writing accuracy was now significantly higher (χ 2 = 73.75; P < 0.000). Writing digits: JS performed errorless in writing digits to dictation 50/50 (100%). The difference between writing letters and writing digits resulted highly significant (χ 2 = 307; P < 0.000). 4.3. Analysis of writing velocity in February and October 2001 Fig. 4. Drawing JS’ drawing of a giraffe. February 2001: Writing velocity was analyzed using a digitizing tablet (WACOM-UD). In writing to dictation, JS 2174 M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 Table 2 JS: writing velocity in cm/s Writing to dictation (February) Writing to dictation (October) Copying (October) Control Arabic digits Upper case letters Lower case letters Non-letters (errors) 1.90 (7) [0.52] 3.11 (53) [0.98] 3.46 (19) [1.20] 3.64 (41) [0.74] 1.28 (9) [0.57] 2.09 (27) [0.93] 3.57 (24) [0.87] 4.66 (21) [1.25] 1.33 (8) [0.423] 1.53 (23) [0.83] 3.52 (25) [0.93] 4.39 (25) [0.86] 0.92 (30) [0.89] 0.83 (18) [0.79] – – The values are indicated as: mean (n) [S.D.]. Fig. 6. Writing velocity. produced digits significantly faster than letters (U-test; P = 0.02), thus confirming the dissociation between scripts observed in accuracy data. Interestingly, JS wrote the digit 0 at the double velocity of the letter O (1.9 and 0.85 cm/s, respectively). There was no difference in writing velocity for upper and lower case letters. Non-letters (not well-formed letters and writing attempts) were produced significantly slower than correct letters (U-test; P = 0.003). October 2001: In writing to dictation, again, digits were written faster than letters (t-test; P < 0.000). This was true for both, upper case (t-test; P < 0.000) and lower case letters (t-test; P < 0.000). Upper case letters were written faster than lower case letters (t-test; P = 0.031) and letters were produced faster than non-letters (t-test; P < 0.000). As compared to February, JS was faster in producing upper case letters (U-test; P = 0.02), digits (P = 0.002), but not lower case letters (P = 0.44). In copying, the dissociation between scripts disappeared and no significant difference between letters and digits was found. A comparison between writing velocities in copying and dictation shows that JS enhanced his writing velocity in copying upper case letters (t-test; P < 0.000) and lower case letters (t-test; P < 0.000), but not in copying digits. JS’ writing performance was compared to a matched control. His writing was slower for letters (t-test; P < 0.000), but also for digits (t-test; P < 0.005). It is important to note that the control, in contrast to JS, produced upper case letters (t-test; P < 0.000), as well as lower case letters (t-test; P < 0.000) significantly faster than digits (Fig. 6). 5. Discussion JS is a patient with a chronic brain lesion who showed peripheral agraphia in a disproportional fashion; he was severely impaired when trying to write letters, but performed well when writing numbers. He was severely agraphic in the first sessions of the investigation and showed still marked difficulties in the follow-up sessions 8 months later. The overall characteristics of his writing difficulties, for example, the dissociation between scripts, remained stable over time. The improvement of his writing abilities is not attributable to spontaneous recovery, since both testing sessions were performed in the chronic stage (48 and 56 months post-onset, respectively), but more likely to the writing training performed. This training included visual imagery of letters and the mental association of letter shapes with real objects, various copying tasks (copying single letters and words after an increasing delay) and writing to dictation. At home JS daily copied words and sentences from the newspaper (in the beginning immediate copying, later delayed copying). In February 2001, neuropsychological background tests evidenced severe agraphia as JS’ outstanding impairment. In order to localize the writing impairment within the cognitive architecture of handwriting, a series of experimental tasks was performed. Almost perfect oral spelling corroborated the clinical impression of peripheral agraphia and excluded a central deficit in generating the appropriate graphemes. In writing letters and words to dictation severe deficits were apparent, writing attempts often resulting in meaningless strokes or occasionally in letter substitutions. However, JS M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 was always aware of his writing errors and was perfectly able to distinguish between successful writing attempts and ill formed strokes or letter substitutions. His preserved letter identification was also apparent in various tasks assessing the recognition of incomplete letters, masked letters and overlapping letters. Moreover, also his matching between letter names and letter shapes, as well as his matching between corresponding allographs was preserved. JS was also able to imagine upper case letters and to describe their visual characteristics. This evidence, together with the lack of case errors in written production, argues against a deficit at the allographic level where letter case and shape are defined [19,40]. Writing alphabetical script improved dramatically in copying. In the literature on peripheral agraphia several patients have been described who improved their handwriting when copying [3,12,24,34,36,54,64] while other patients did not [1,32,48]. JS fluently wrote well-formed letters, not only when the stimulus material was visible, but also in a tachistoscopic copy task with extremely short presentation times. As pointed out by Margolin [40], copying letters may be accomplished along two pathways, the first one without reference to graphomotor patterns, just like a picture may be copied, the second one accessing the spelling system. The short presentation times applied in the present tachistoscopic tasks exclude the use of the first of the two pathways and prevents the subject from slavishly copying single details (as, for example, reported by Meyer-Wahl et al. [43]). In contrast, standard copying tasks do not allow to definitely decide which strategy has been applied by the patient and thus add only limited evidence to the analysis of the writing impairment. JS’ performance in the copying task suggests that a very short presentation of letters enabled him to activate his intact graphomotor programs which were executed fluently without hesitations. Thus, his difficulties in writing without a model point to a lack of activation of graphomotor programs, but not to difficulties in executing them. This view is supported by a kinematic analysis as described further. The analysis of writing velocity evidenced slow and laborious writing to dictation, but preserved writing when a model was provided. This finding, again, points to a problem in activating the appropriate graphomotor patterns. Indeed, fast and fluent copying excludes an impairment in the execution of the motor patterns, which should be evident in both tasks, writing to dictation and copying. Though JS improved his writing abilities from the first to the second series of testing, the difference between writing to dictation and copying remained significant in October 2001. Notably, JS was good in generating mental imagery about letters: thus, the internally generated visual representation and the correct access to the visual letter shape do not seem to be as useful as an externally provided model, even if flashed, to activate the appropriate patterns. JS’ drawings of objects or animals were poorly specified and often incomplete. His deficit in drawing was not 2175 associated to impairments in perception, object recognition or knowledge about the visual characteristics of objects or animals. In fact, several tasks evidenced fully preserved stored visual knowledge. Apparently, JS had difficulties to transform his intact visual knowledge into corresponding drawings. As Zesiger et al. [65] propose, drawing would be mediated by a two-dimensional trajectory generator (which is accessed from visual memory or from the allographic level), while normal handwriting would be mediated by the graphic motor pattern level. In JS’ case impairments in drawing point to deficits at the two-dimensional trajectory generator level. The assumption of such an impairment would explain why JS could not use an alternative “drawing strategy” in order to compensate for the difficulties at the graphic motor pattern level in handwriting. A further point of interest of the present study is the clear-cut dissociation between almost perfectly preserved Arabic script and severely impaired alphabetical script. The kinematic analysis confirmed that JS produced digits significantly faster than letters. Though he improved his writing abilities over 8 months, this dissociation was still evident in the second testing series. In copying, the dissociation between scripts disappeared and no significant difference between letters and digits was found. Seemingly, JS did not profit from the visual model in writing digits (which were already produced fluently), but did so in writing letters. This assumption was confirmed in a direct comparison of writing letters to dictation and of copying (letters and digits, respectively). Though single digits were well-formed and fluently produced, few syntactic errors appeared in complex multi-digit numerals which contained internal zeros (e.g. the number word six thousand one was transcoded as 60001). In fact, writing Arabic script is always a complex process, comprising several levels of elaboration. Though there are considerable differences between theoretical models of transcoding [16,58] all of them agree that there are syntactic and lexical mechanisms involved in the writing process, the former defining the position of the elements within a complex numeral, the latter the single elements’ identity. As for JS, complex numerals with internal zeros are particularly difficult to master for children and for brain damaged patients [15]. Models of number processing make no specific claims on the peripheral aspects of number writing, i.e. they make no assumptions beyond the level of digit selection. Only recently Lochy et al. [38] investigated the temporal, spatial and kinematic characteristics of writing digits. They found that 3 is the most complex and most slowly written digit, probably due to its high curvature and repeated pattern. Moreover, 3 contains horizontal strokes, which are more difficult to execute [42,59]. The easiest digits were 0 and 6, which contain only 2 or 3 strokes and are realized through an anti-clockwise downward stroke [38]. Lochy et al. [37] also studied the writing of more complex, multi-digit numbers and replicated the length effect known for words [62] in 2176 M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 the handwriting of numbers.3 However, what is still lacking is a systematic comparative analysis of handwriting Arabic and alphabetical script. The present study strongly suggests that motor programs are activated separately for the set of letters and for the set of numerals and thus adds evidence to the so far reported dissociations in the processing of Arabic and alphabetical script. JS was also asked to write Roman numerals, a task of particular interest. We hypothesized that the writing of Roman numerals should be preserved, if they are processed as numerals, but should be impaired if they are treated as letters. In fact, in the handwriting of a patient with severe agraphia and preserved number writing [2] the two only letters clearly legible where X and O, both representing numerals in Roman notation, and also representing the digit 0 and the multiplication sign. JS, however, did not succeed to write Roman numerals. From his strategy and his verbal comments it could be inferred that, in the moment he attempted to write, he was searching for letters and did not associate any numerical meaning. JS is not the only case showing a clear-cut dissociation between impaired alphabetical script and preserved Arabic script. In fact, some studies reported preserved number writing in peripheral agraphia without paying specific interest to this finding [12,14,28,34,49,64] and without studying this phenomenon in detail. The best studied case was reported by Anderson et al. [2], a right-handed woman who developed severe alexia and agraphia after a surgical lesion in the left premotor cortex. Anderson et al. [2] concluded that the domain of graphemes and the domain of numbers are, at least in part, processed by separate neuroanatomical structures. As a possible explanation for the functional dissociation and different neuroanatomical representation of scripts they proposed that number concepts are typically acquired in strong association with the digits of the hand. Comparing JS with Anderson et al.’s patient, critical differences become evident. Importantly, their patient’s writing was not only defective in writing to dictation, but also in copying where letters were spatially distorted and frequently overlapping within words. Moreover, their patient’s deficit concerned not only writing, but to the same extent reading of alphabetical script. Though some studies reported preserved number writing in patients with severe peripheral agraphia for alphabetical stimuli, no studies, to our knowledge, reported peripheral agraphia for numbers in the context of preserved alphabetical writing. Since no double dissociation between writing the two scripts has been found, one may hypothesize that numerals are simply easier to write at a peripheral level, possibly because they represent a smaller set of graphomotor patterns than letters. Comparing the visual characteristics of 3 When long and short words are written in one session, letters are written faster in long words than in short words and letters size tends to be decreased. The duration and trajectory length of a letter are thus longer in short than in long words. As Lochy et al. [37] showed, the same is true for digits. digits and letters and the graphomotor patterns they require, there is no apparent reason why digits would be simpler, however, detailed comparative analyses of handwriting the two scripts are needed. Neuropsychological case studies reported selectively preserved, as well as selectively disrupted mathematical abilities and thus suggest that numbers represent a separate category within the semantic system. This claim is also supported by the recent report of a patient [6] who could read and write number words despite being severely impaired in reading and writing non-number words. His performance was attributed to intact semantic representations of numbers, but not of other categories. The present case study of patient JS adds evidence to the so far reported dissociations between the number domain and other domains and shows that numbers and letters are processed separately at a peripheral level of the writing process. Acknowledgement The present research has been supported by grant HPRN-CT-2000-00076 of the European Community. References [1] Alexander MP, Fischer RS, Friedman R. Lesion localization in apractic agraphia. Archives of Neurology 1992;49:246–51. [2] Anderson SW, Damasio AR, Damasio H. Troubled letters but not numbers: domain specific cognitive impairments following focal damage in frontal cortex. Brain 1990;113:749–66. [3] Baxter DM, Warrington EK. Ideational agraphia: a single case study. Journal of Neurology, Neurosurgery and Psychiatry 1986;49:369–74. [4] Beauvois MF, Dérouesné J. Lexical or orthographic agraphia. Brain 1981;104:21–49. [5] Butterworth B. The mathematical brain. London: Macmillan, 1999. [6] Butterworth B, Cappelletti M, Kopelman M. Category specificity in reading and writing: the case of number words. Nature Neuroscience 2001;8:784–6. [7] Chanoine V, Teixeira Ferreira C, Demonet JF, Nespoulous JL, Poncet M. Optic aphasia with pure alexia: a mild form of visual associative agnosia: a case study. Cortex 1998;34:437–48. [8] Coslett HB, Gonzalez Rothi LJ, Valenstein E, Heilman KM. Dissociations of writing and praxis: two cases in point. Brain and Language 1986;28:357–69. [9] Crary MA, Heilman KM. Letter imagery in a case of pure apraxic agraphia. Brain and Language 1988;34:147–56. [10] Croisile B, Laurent B, Michel D, Trillet M. Pure agraphia after deep left hemisphere haematoma. Journal of Neurology, Neurosurgery and Psychiatry 1990;53:263–5. [11] Dahl G. Reduzierter wechsler intelligenztest. Meisenheim am Glan: Verlag Anton Hain, 1972. [12] Davies EJ, Coughlan T, Ellis AW. Peripheral dysgraphia with impaired processing of musical and other symbols. Journal of Neurolinguistics 1997;10:11–7. [13] Déjerine J. Contribution à l’étude anatomopathologique et clinique des différentes variétés de cécité verbale. Mémoires de la Société Biologique 1882;4:61–90. [14] Del Grosso Destreri N, Farina E, Alberoni M, Pomati S, Nichelli P, Mariani P. Selective uppercase dysgraphia with loss of visual imagery of letter forms: a window on the organisation of graphomotor patterns. Brain and Language 2000;71:353–72. M. Delazer et al. / Neuropsychologia 40 (2002) 2167–2177 [15] Delazer M, Denes G. Writing Arabic numerals in an agraphic patient. Brain and Language 1998;64:257–66. [16] Delazer M, Girelli L. When alpha Romeo facilitates 164: semantic effects in verbal number production. Neurocase 1997;3:461–75. [17] Ellis AW. Slips of the pen. Visible Language 1979;13:265–82. [18] Ellis AW. Spelling and writing (and reading and speaking). In: Ellis AW, editor. Normality and pathology in cognitive functions. London: Academic Press, 1982. p. 113–46. [19] Ellis AW. Normal writing processes and peripheral acquired dysgraphias. Language and Cognitive Processes 1988;3:99–127. [20] Folstein MF, Folstein SE, McHugh PR. Mini mental state. Journal of Psychiatric Research 1975;12:189–98. [21] Friedman RB, Alexander M. Written spelling agraphia. Brain and Language 1989;36:503–17. [22] Gerstmann J. Fingeragnosie und isolierte agraphie: ein neues syndrom. Zeitschrift für die gesamte Neurologie und Psychiatrie 1927;108:152–77. [23] Goodman RA, Caramazza A. Aspects of the spelling process: evidence from a case of acquired dysgraphia. Language and Cognitive Processes 1986;1:263–96. [24] Graná A, Girelli L, Gattinoni F, Semenza C. Letters and number writing in agraphia: a single case study. Brain and Cognition 2001;46:149–54. [25] Grewel F. Acalculia. Brain 1952;75:397–407. [26] Grewel F. The acalculias. In: Vinken PJ, Bruyn GW, editors. Handbook of clinical neurology. Amsterdam: North Holland, 1969. p. 181–96. [27] Grossman M, Libon DJ, Ding XS, Cloud B, Jaggi J, Morrison D, et al. Progressive peripheral agraphia. Neurocase 2001;7:339–49. [28] Hanley JR, Peters SA. dissociation between the ability to print and write cursively in lower case letters. Cortex 1996;32:737–45. [29] Härting C, Markowitsch HJ, Neufeld H, Calabrese P, Deisinger K, Kessler J. WMS-R: Deutsche adaptation der revidierten fassung der wechsler memory scale. Bern: Verlag, 1999. [30] Heilman KM, Gonyea EF, Geschwind N. Apraxia and agraphia in a right-hander. Cortex 1977;10:284–8. [31] Henschen SE. Über sprach-, musik- und rechenmechanismen und ihre lokalisation im Großhirn. Zeitschrift für die gesamte Neurologie und Psychiatrie 1919;52:273–98. [32] Hodges JR. Pure apraxic agraphia with recovery after drainage of a left frontal cyst. Cortex 1991;27:469–73. [33] Huber W, Poeck K, Weniger D, Willmes K. Aachener aphasie test. Göttingen: Verlag für Psychologie Hogrefe, 1983. [34] Kapur K, Lawton NF. Dysgraphia for letters: a form of motor memory deficit. Journal of Neurology, Neurosurgery and Psychiatry 1983;46:573–5. [35] Lehrl S. MWT-B: Mehrfachwahl–Wortschatz intelligenztest. Erlangen: Verlag, 1977. [36] Levine DN, Mani RB, Calvanio R. Pure agraphia and Gerstmann’s syndrome as a visuospatial language dissociation: an experimental case study. Brain and Language 1988;35:172–96. [37] Lochy A, Pillon A, Zesiger P, Seron X. Verbal structure of numerals and digits handwriting: new evidence from kinematics. The Quarterly Journal of Experimental Psychology A 2002;55A:263–88. [38] Lochy A, Zesiger P, Seron X. Motor complexity of Arabic digits: a descriptive study through a new kinematic approach. Current Psychology of Cognition 2001;20:19–36. [39] Luzzatti C, Vecchi T, Agazzi D, Cesa-Bianchi M, Vergani C. A neurological dissociation between preserved visual and impaired spatial processing in mental imagery. Cortex 1998;34:461–9. [40] Margolin DI. The neuropsychology of writing and spelling: semantic, phonological, motor and perceptual processes. Quarterly Journal of Experimental Psychology 1984;36a:459–89. 2177 [41] Margolin DI, Binder L. Multiple component agraphia in a patient with atypical cerebral dominance: an error analysis. Brain and Language 1984;22:26–40. [42] Meulenbroek RGJ, van Galen GP. Perceptual-motor complexity of printed and cursive letters. Journal of Experimental Education 1990;58:95–110. [43] Meyer-Wahl R, Rijntjes M, Wolfram H, Wegner K, Weiller C, Dettmers C. Die ideationale Schreibstörung—Übersicht und Fallbericht. Neurologische Rehabilitation 2000;6:235–44. [44] Monsch AU, Bondi MW, Butters N, Salmon DP, Katzman R, Thal LJ. Comparisons of verbal fluency tasks in the detection of dementia of the Alzheimer’s type. Archives of Neurology 1992;49:1253–8. [45] Morris JC, Mohs RC, Rogers H, Fillenbaum G, Heyman A. Consortium to establish a registry for Alzheimer’s disease (CERAD) clinical and neuropsychological assessment of Alzheimer’s disease. Psychopharmacological Bulletin 1988;24:641–52. [46] Ohno T, Bando M, Nagura H, Ishii K, Yamanouchi H. Apraxic agraphia due to thalamic infarction. Neurology 2000;54:2336–9. [47] Oswald WD, Fleischmann UM. NAI—Nürnberger-Alters-Inventar. Göttingen: Verlag, 1997. [48] Papagno C. A case of peripheral dysgraphia. Cognitive Neuropsychology 1992;9:259–70. [49] Patterson K, Wing AM. Processes in handwriting: a case for case. Cognitive Neuropsychology 1989;6:1–23. [50] Power RJD, Dal Matello MF. The dictation of Italian numerals. Language and Cognitive Processes 1990;5:237–54. [51] Rapp B, Caramazza A. From graphemes to abstract letter shapes: levels of representation in written spelling. Journal of Experimental Psychology: Human Perception and Performance 1997;23:1130–52. [52] Reitan RM. Validity of the trail making test as an indicator of organic brain damage. Perceptual and Motor Skills 1958;8:271–6. [53] Riddoch MJ, Humphreys G. BORB: the Birmingham object recognition battery. Hove: Lawrence Earlbaum, 1993. [54] Roeltgen DP, Heilman KM. Apractic agraphia in a patient with normal praxis. Brain and Language 1983;18:35–46. [55] Rothi LJ, Heilman KM. Alexia and agraphia with spared spelling and letter recognition abilities. Brain and Language 1981;12:1–13. [56] Schonauer K, Denes G. Graphemic jargon: a case report. Brain and Language 1994;47:279–99. [57] Seron X, Fayol M. Number transcoding in children: a functional analysis. British Journal of Developmental Psychology 1994;12:281– 300. [58] Seron X, Noël M-P. Transcoding numbers from the Arabic code to the verbal one or vice versa: how many routes? Mathematical Cognition 1995;1:215–45. [59] Serratrice G, Habib M. L’écriture et le cerveau: Mécanismes neurophysiologiques. Paris: Masson, 1993. [60] Snodgrass JG, Vanderwart MA. A standardized set of 280 pictures: norms for name agreement, image agreement, familiarity and visual complexity. Journal of Experimental Psychology: Human Learning 1980;6:174–215. [61] Valenstein E, Heilman KM. Apraxic agraphia with neglect-induced paragraphia. Archives of Neurology 1979;36:506–8. [62] van der Plaats RE, van Galen GP. Effects of spatial and motor demands in handwriting. Journal of Motor Behavior 1990;22:361–5. [63] Warrington EK, James M. A new test of object decision: twodimensional silhouettes featuring a minimal view. Cortex 1991;27: 370–83. [64] Zangwill OL. Agraphia due to a left parietal glioma in a left-handed man. Brain 1954;77:510–20. [65] Zesiger P, Martory MD, Mayer E. Writing without graphic motor patterns: a case of dysgraphia for letters and digits sparing shorthand writing. Cognitive Neuropsychology 1997;14:743–63.