This article was downloaded by: [Florida State University] On: 22 December 2014, At: 16:41 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Neurocase: The Neural Basis of Cognition Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/nncs20 The Eye Movements of Japanese Pure Alexic Patients During Single Word and Nonword Reading a b c Masato Kaneko , Takao Fushimi , Akira Uno & Noriko Haruhara a d Department of Rehabilitation, Tokyo Metropolitan Ohtsuka Hospital , Tokyo b Language, Cognition, and Brain Research Group, Tokyo Metropolitan Institute of Gerontology , Tokyo c Institute of Human Science, University of Tsukuba , Ibaraki d Department of Rehabilitation, Saiseikai Central Hospital , Tokyo, Japan Published online: 17 Aug 2010. To cite this article: Masato Kaneko , Takao Fushimi , Akira Uno & Noriko Haruhara (2004) The Eye Movements of Japanese Pure Alexic Patients During Single Word and Nonword Reading, Neurocase: The Neural Basis of Cognition, 10:5, 366-381, DOI: 10.1080/13554790490892266 To link to this article: http://dx.doi.org/10.1080/13554790490892266 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions Neurocase, 10(5): 366–381, 2004 Copyright © Taylor & Francis Inc. 1355-4795/04/1005–366$16.00 Neurocase The Eye Movements of Japanese Pure Alexic Patients During Single Word and Nonword Reading Masato Kaneko1, Takao Fushimi2, Akira Uno3 and Noriko Haruhara4 1 Department of Rehabilitation, Tokyo Metropolitan Ohtsuka Hospital, Tokyo, 2Language, Cognition, and Brain Research Group, Tokyo Metropolitan Institute of Gerontology, Tokyo, 3Institute of Human Science,University of Tsukuba, Ibaraki and 4 Department of Rehabilitation, Saiseikai Central Hospital, Tokyo, Japan Downloaded by [Florida State University] at 16:41 22 December 2014 Abstract Two Japanese patients with pure alexia, SH and YH, who showed right homonymous hemianopia following a left occipital lobe lesion, demonstrated letter-by-letter (LBL) reading in pronouncing Japanese kana words and nonwords. In contrast to alphabetic letters, each Japanese kana character has an invariant and identical pronunciation whether it appears in isolation or as a component of any word and nonword string. It is important to investigate the eye movements as well as reading latency and duration in Japanese-speaking LBL readers. Relative to normal controls, these patients demonstrated a more robust string-length effect, which was characterized by larger increases in reading latency and duration as well as in the number of fixations as the string length increased. We propose that in pure alexia, parallel activation of orthographic representations is abnormally delayed but not completely abolished. Introduction Pure alexia, or letter-by-letter (LBL) reading, is an acquired disorder of reading in which patients require an abnormally long amount of time to read aloud single words as well as text. This reading disorder can be behaviorally defined as a monotonic increase in reading time, which is assessed by reading latency and/or reading duration, as a function of increasing word length. It is assumed that the considerable variation in the slope of this increasing function observed among patients, ranging from a slope of a few hundred milliseconds to several seconds per additional letter, reflects the degree of severity. These patients in tachistoscopic viewing condition and other more severe patients demonstrate this word-length effect in the accuracy of reading aloud words, which is characterized by frequent misreading of the latter part of a word, such as SURF → SURE. Reading accuracy sometimes improves when the patient is allowed to trace the letters with his/her finger, as reported in a phenomenon called Schreibendes Lesen, and it greatly improves when the patient hears the oral spelling of the written word. Pure alexia occurs in the absence of other language impairment involving writing or spelling, although visual anomia sometimes accompanies LBL reading when the lesion in the left occipito-temporal area, which is the typical site associated with pure alexia, extends to a more temporal region. Consistent with this typical site of the lesion, the vast majority of pure alexic patients demonstrate right homonymous hemianopia or quadrantanopia (see Behrmann et al., 1998a; Montant and Behrmann, 2000, Leff et al., 2001, for review of patients). Cognitive neuropsychological studies have proposed at least three possible loci of impairment in pure alexia: a visual processing deficit (Farah and Wallace, 1991; Behrmann et al., 1998a, 1998b), damage to the visual word-form system (Warrington and Shallice, 1980; Warrington and Langdon, 1994, 2002), or disconnection between orthographic and phonological representations (Bowers et al., 1996a, 1996b). One version of the visual deficit theory assumes impairment of a more general-purpose visual processing system and not solely impairment of the visual processing system for reading (Farah and Wallace, 1991; Sekuler and Behrmann, 1996; Behrmann et al., 1995, 1998a, 1998b; Chialant and Caramazza, 1998; Behrmann et al., 2001; Montant and Behrmann, 2001). This hypothesis arose from the facts that pure alexic patients demonstrate a prolonged response time in naming visually complex pictures compared with normal controls (Behrmann et al., 1998b) and impaired performance in recognizing a simple visual stimulus located in a string of similar shapes as well as in recognizing a letter in a letter string (Chialant and Caramazza, 1998). Another version of the visual deficit hypothesis assumes impairment of processing devoted to reading, such as identification of abstract letters, because some patients failed to demonstrate a normal between-case priming effect—an example of a between-case priming effect is when pre-exposure to a lower-case letter such as g facilitates Correspondence to: Masato Kaneko, 1-8-2 Minami-Ohtsuka, Toshima-ku, Tokyo 170-0005, Japan. e-mail: kaneko1234@hotmail.com DOI: 10.1080/13554790490892266 Downloaded by [Florida State University] at 16:41 22 December 2014 Eye movement of Japanese pure alexia subsequent identification of the upper-case letter, G—despite the success of within-case priming (Arguin and Bub, 1993, 1994; Miozzo and Caramazza, 1998). In both versions, which are classified as a peripheral view of impairment in pure alexia, LBL reading is assumed to arise from slow activation of intact orthographic representations (or visual word form), resulting in an attentional shift from letter to letter to restore the identification of individual letters in a string. The second hypothesis proposes that disturbance of the visual word-form system, and not disturbance of letter identification, gives rise to pure alexia (Warrington and Shallice, 1980; Warrington and Langdon, 1994, 2002). Referred to as LBL reading, patients overtly (or covertly) name each letter of a written string (for example, MAKE → /em, ei, kei, i/) to activate a phonological representation of the entire word (in this case, /meik/), although the exact mechanism of this compensative conversion of /em, ei, kei, i → meik/ has yet to be determined. The reading performance of patients who employ this strategy decreases when they are concurrently performing an articulatory suppression task—for example, when patients are asked to count numbers while reading (Warrington and Langdon, 2002). The third hypothesis assumes disconnection between orthographic and phonological representations, and is based on the fact that some patients fail to demonstrate normal homophone priming (for instance, feet → FEAT) despite the success of between-case orthographic priming (for instance, FeEd → FEED) (Bowers et al., 1996a, 1996b; Arguin et al., 1998). In this hypothesis, LBL reading arises from restoration of disrupted activation of phonological representation, which requires a longer amount of time for longer words. The second and third hypotheses are classified as the central view of impairment in pure alexia. All of these hypotheses share the proposition that in normal reading, letters in a string are identified in a parallel manner because normal readers demonstrate a word length effect that is an order of magnitude smaller (a slope of a few tens of milliseconds per additional letter) than that of pure alexic patients (e.g., Weekes, 1997). Parallel activation of orthographic representation from a visual input is also a shared assumption among the current theories of word reading in normal subjects (e.g., Plaut et al., 1996; Coltheart et al., 2001)1. The peripheral view proposed by the visual deficit theory provides a stark contrast with the two central views because it assumes that this parallel identification of letters is impaired in pure alexia. It should be noted that some pure alexic patients demonstrated more rapid or moreaccurate performance for words than nonwords, for high-frequency than low-frequency words (i.e., for frequently-used than infrequently-used words), or for high-imageability than lowimageability words (e.g., for concrete rather than abstract words). Although these effects of lexicality, frequency, and imageability may argue against the visual deficit theory, Behrmann et al. (1998) suggested that these effects are attributable to top-down processing by the central system, which recovers impaired letter identification to some degree. 367 Behrmann et al. (2001) noted that if impairment in parallel identification of letters is compensated by shifting attention from letter to letter, and if this serial attentional shift results in reliance on foveal vision, which has enhanced visual resolution, then the eye movement patterns of pure alexic patients during reading should differ from those of normal readers. Because disturbed eye movements can occur due to hemianopia (Leff et al., 2000) and oculomotor dysfunction (Biscardi et al., 1998), Behrmann et al. prepared two control conditions: first, as control subjects, they employed not only normal readers, but also patients with left homonymous hemianopia due to right hemispheric posterior lesions but who did not suffer from LBL reading. Second, they prepared a visual search task as a control task. Behrmann et al. reported that in an experiment of reading text displayed on a computer screen, the pure alexic patients had a significantly larger number of eye fixations per word, longer duration per fixation, and more frequent occurrence of regressive saccades than both the normal controls and hemianopic controls, whereas the eye movement patterns during the visual search task did not differ among the three groups. These data indicate that LBL reading in pure alexic patients results from serial attentional shift from letter to letter for the purpose of enhancing letter identification by employing foveal vision successively for each letter in the string. To expand the study of Behrmann et al. (2001), we investigated the eye movements in Japanese-speaking pure alexic patients during single word and nonword reading rather than text reading. The data would provide more detailed information about pure alexic reading because we would be able to combine the data on reading latencies and durations and the data on the numbers and durations of fixations. In Japanese kana characters, which are the materials of the present study, the pronunciation of single characters like the following four characters, , is exactly the same as its pron, , , unciation in a word like /hi-ma-wa-ri/meaning sunflower, in contrast to alphabetic letters2. Assessment of reading latencies and durations as well as eye movements is quite useful because the successive sequence of overt naming of each character in a string is indistinguishable from the normal pronunciation of a single word in Japanese kana script. Empirical studies in which Japanese pure alexic patients read aloud have offered scarce data on reading time (Kawamura and Mizobuchi, 1996; Kokubo et al., 2000), whereas data on the accuracy of reading in more severe patients have been frequently reported (e.g., Kashiwagi and Kashiwagi, 1989; Sugishita et al., 1992; Seki et al., 1995; Sakurai et al., 2001). It is of interest to investigate the time course of eye movements and pronunciation of individual characters when Japanese-speaking pure alexic patients read aloud a single kana string. Method Subjects Two pure alexic patients, SH and YH, participated in the experiment of eye movement measurement after informed Downloaded by [Florida State University] at 16:41 22 December 2014 368 M. Kaneko et al. consent was obtained from each patient. SH, a 70-year-old, right-handed Japanese woman, presented in August 1999 with cerebral infarction. A magnetic resonance imaging (MRI) scan in December 2000 revealed a high-intensity area in the left occipital lobe with probable hemorrhagic cerebral infarction as well as a few high-intensity spots in the basal ganglia and thalamus bilaterally (Fig. 1). She showed right homonymous hemianopia without other neurological deficits. An ophthalmologist did not find any basic oculomotor problem such as nystagmus. Neuropsychological assessment was performed three months after the onset of the disease. In the Wechsler Adult Intelligence Scale Revised (WAIS-R), her verbal and performance IQ were 93 and 85, respectively. On the Raven’s Coloured Progressive Matrices (RCPM) test, she scored 34/36. The Japanese Standard Language Test of Aphasia (SLTA) for assessing speech and language pathology revealed that all of her scores were within the normal range, except slightly deteriorated scores on reading aloud single kanji words and reading aloud short sentences comprised of kanji and kana words. However, she showed impairment in reading aloud low-frequency kanji words, and/or kanji words comprised of kanji characters of low character frequency and/or those with high visual complexity3. When she read aloud kana words, she slowly pronounced each character as observed in typical LBL reading, with each character often being traced with her finger before pronunciation. She attended a SH R Fig. 1. MRI of Case SH obtained during the 12th month after onset. YH R Fig. 2. MRI of Case YH obtained during the 14th month after onset. remedial training program for two years at our hospital, and her ability to reading aloud and comprehend advanced to the level of sentence reading, although she still continued to use a kinesthetic strategy when it was difficult to read aloud kana characters. Measurement of eye movements was performed two years after onset. YH, a 73-year-old, right-handed Japanese man, presented in September 1994 with cerebral infarction. An MRI scan in December 1995 revealed a high-intensity area in the left occipital lobe with probable hemorrhagic cerebral infarction (Fig. 2). He showed right homonymous hemianopia without other neurological deficits. An ophthalmologist did not find any basic oculomotor problem. Two years after onset, his verbal and performance IQ were 136 and 108, respectively, on the WAIS-R, and his score on the RCPM test was 34/36, indicating excellent intellectual ability. On the SLTA, all of his scores were within the normal range. YH demonstrated LBL reading with frequent tracing of kana characters with his finger. He attended a remedial training program for five years, and his ability to read aloud kana characters substantially recovered although the speed of reading still seemed slow and within the range of LBL reading. His accuracy in reading aloud kanji characters varied depending on word frequency, character frequency, and/or visual complexity. Measurement of eye movements was performed at five years after onset. Eye movement of Japanese pure alexia Seven Japanese native speakers, aged between 67 and 74 years (mean = 69 years, SD = 3.4), who had no history of neuropsychological disease, served as the normal controls. Their RCPM scores were within the normal range. 20 points 369 80 points × × fixation point Downloaded by [Florida State University] at 16:41 22 December 2014 Stimuli Japanese kana is sub-divided into hiragana (cursive form) and katakana (square form). There is always a pair of hiragana and katakana characters that have the same pronunciation. A kana character usually maps to a single mora, with each character-sound correspondence perfectly transparent. A mora is a phonological unit of spoken Japanese, which comprises a single vowel or a consonant-vowel compound. In terms of typical usage, hiragana is used for most function words, the inflections of verbs and adjectives, and some content words, whereas katakana is used for words that have been adopted from western languages. The experimental stimuli were composed of four types of kana strings: hiragana words, katakana words, katakana pseudohomophones, and pronounceable hiragana nonwords. Each type of kana string consisted of 10 four-character strings and 10 six-character strings. According to the familiarity rating scores on a 7-point scale obtained from the NTT psycholinguistic database (Amano and Kondo, 1999), the mean familiarity with the hiragana and katakana words was 6.4 and 6.5, respectively, indicating that they are highly familiar to Japanese readers. The katakana pseudohomophones were katakana transcriptions of words normally written in kanji. The mean familiarity with the original kanji words was 6.4. Hiragana nonwords were created by reversing the sequential order of constituent characters of hiragana pseudohomophones, which were hiragana transcriptions of words normally written as kanji-kana compounds. Among the words and pseudohomophones, the mean familiarity with the 4-character strings and with the 6character strings was roughly equivalent. From the total number of 80 stimuli, some examples are /himawari/meaning sunflower and /toumorokosi/corn among the hiragana words, /masukomi/mass media and /daiamondo/diamond among the katakana words, /sensou/war and /undoukai/athletic meet among the katakana pseudohomophones, and /zuranaka/and /masausotsigo/ among the hiragana nonwords. Apparatus Eye movements were measured by the “Free View” system (Takei Engineering, Tokyo, Japan), which is an infrared device with a visual resolution of less than 0.2° and a sampling rate of 30 Hz. The system controlled the presentation of stimuli on the visual display and stored the horizontal and vertical coordinates corresponding to the position of the right eye every 33.3 ms. The display was located 100 cm from the subject's eyes. Each kana string was presented on the display at 20-point pixel size in one viewing condition (20P) and at target (Sunflower) Fig. 3. The schema of the stimulus presentations on the display. 80-point pixel size in the other condition (80P). The positions of the fixation point and the target on the display in these two conditions differed as shown in Fig. 3. In the 20P condition, each string was horizontally presented to the left of the fixation point which appeared at the center of the display so that the stimulus subtended the foveal and parafoveal visual areas in both the normal controls and pure alexic patients who showed right homonymous hemianopia. The minimum visual angle of the 4-character strings was 1.7° and the maximum visual angle of the 6-character strings was 2.5°. In the 80P condition, each string was horizontally presented across the center of the full screen so that the eye movements could be traced with ease. The minimum visual angle of the 4-character strings was 7.4° and the maximum visual angle of the 6character strings was 11.8°. During the experiment, the display image and the subject's oral response of reading aloud were recorded by a digital video recorder for subsequent analyses of eye-movement pattern and reading errors, latencies, and durations. Procedure The subjects were administered the task of reading aloud each string of the words and nonwords. The two patients and four normal controls were tested first under the 80P condition and then under the 20P condition. The other three normal controls were tested first under the 20P condition and then under the 80P condition. Before collection of eye movement data, the subject was trained sufficiently to read aloud a single string presented on the display under each viewing condition. This test string was not among the 4-character and 6-character strings used for data collection. Then, after calibration of the system, the study was started. The administrator controlled the appearance and disappearance of the fixation stimulus and the target through the system. Each experimental trial began with the appearance of the fixation point. When the system’s real-time monitor of eye position detected that the subject fixated on the fixation stimulus, the target was 370 M. Kaneko et al. presented and it remained on the display until the end of the subject’s oral response. Downloaded by [Florida State University] at 16:41 22 December 2014 Data analyses We measured the reading latency and reading duration by displaying video frames along with the sound waves obtained from the digital-video recording. Reading latency was defined as the length of time between presentation of the stimulus and the beginning of the oral response. Reading duration was defined as the length of time between the beginning and the end of the oral response. In the analyses of eye movements, we measured the number of fixations and the duration of each fixation. For each sample obtained every 33.3 ms, the moving average of the velocity of eye movements was calculated from the series of the past three samples including the current one, of horizontal and vertical coordinates. The occurrence of a fixation was defined as at least two consecutive samples each with a velocity of eye movement of less than 5.0 degree/second, corresponding to eye movement within 0.3° (less than one character under the 20P condition) during 66.7 ms. In the statistical analyses, for each string, the number of fixations, the mean duration of a fixation, and the averaged range of the horizontal coordinates of the fixation points (presented as the averaged values of the measured minimum, mean, and maximum values) were calculated in each subject, during the period between the time point at which the subject’s eye position left the fixation point and the time point at which the subject started an oral response (hereinafter, before response) and during the period between the time points at which the subject started and finished an oral response (during response). The eye-movement patterns were further analyzed by visual inspection, in which the trace of eye positions was superimposed on the stimulus view as shown in Fig. 4. When the amount of trace was within a space that a character subtended, it was defined that the subject fixated on that character. Through these observations, we could count the number of occurrences a) b) Fig. 4. Trace of the eye movement that is superimposed on the stimulus view. The target is a 4-character katakana word, /sutereo/, in (a) the 20P and (b) the 80P condition. The lines show a portion of the trace of eye movements before the oral response. of each of the following four eye-movement patterns in each trial: the lack of fixation on the first character, intermediate characters, last character, and fixations on all characters. Results Trials in which the subject produced an error including a dysfluent response, were excluded from the analyses of reading time and eye movement. Normal controls We first report the results of normal controls because detailed data on the reading time and eye movements during reading in Japanese older subjects have not yet been reported to our knowledge. Table 1 presents the mean rate of accuracy, Table 1. Results in the seven normal controls Viewing condition 20 Points Length of string 4 characters Type of string* HW KW 80 Points 6 characters KP HN HW KW 4 characters KP HN HW KW 6 characters KP HN Mean accuracy rate 0.97 0.96 0.94 0.89 0.97 1.00 0.93 0.73 1.00 0.99 0.99 0.90 Reading latency (ms) ** 858 874 990 1455 1114 1013 1323 1659 665 700 775 1147 Reading duration (ms) ** 624 595 654 821 837 772 970 1626 617 597 638 734 Number of fixations ** Before response *** 1.27 1.50 1.97 3.13 2.03 1.99 2.79 3.96 1.24 1.26 1.57 2.47 During response **** 1.31 1.09 1.49 2.18 1.84 1.63 2.20 4.32 1.17 1.03 1.62 1.73 Duration per fixation ** Before response *** 110 101 109 117 102 106 109 120 99 103 98 102 During response **** 112 94 101 106 87 95 103 115 88 103 90 98 HW KW KP HN 1.00 0.97 0.97 0.84 883 719 1039 1399 848 759 899 1341 1.61 2.03 106 103 1.33 1.76 104 95 2.14 2.38 103 96 3.16 3.70 117 108 *HW, hiragana words; KW, katakana words; KP, katakana pseudohomophones; HN, hiragana nonwords. **The mean reading latency, mean reading duration, mean number of fixations, and mean duration per fixation of only the correct responses are shown. ***This period extends from the time of presentation of stimulus to the beginning of the oral response. ****This period extends from the time of the beginning of the oral response to the time of the end of the oral response. Downloaded by [Florida State University] at 16:41 22 December 2014 Eye movement of Japanese pure alexia correct reading latency, reading duration, number of fixations, and duration per fixation as a function of viewing condition, string length and type of string in the normal controls. Because the normal controls had a high level of accuracy (mean = 94%, SD = 3%), the data on reading time and eye movements in Table 1 were obtained from a sufficient number of observations. These dependent variables were submitted to three-way analyses of variance (ANOVAs) with viewing condition (20P vs. 80P condition), type of string (hiragana words, katakana words, katakana pseudohomophones and hiragana nonwords), and string length (4 vs. 6 characters) as factors. The effects are described as significant if p< 0.05 and as marginally significant if p< 0.10. The F-values are shown for significant effects. Bonferroni correction was used in multiple pairwise comparisons between each pair among the four string types. ANOVA on reading latency revealed significant main effects of viewing condition (F1, 6 = 20.81), string type (F3, 6 = 35.09), and string length (F1, 6 = 46.56), as well as a significant interaction between string type and length (F3, 18 = 6.55). The 20P condition yielded a significantly longer mean reading latency (1161 ms) than the 80P condition (916 ms), indicating that reading latency was reduced by larger size of characters rather than by smaller visual angle. In multiple comparisons, the mean reading latency was significantly shortest for hiragana words (880ms) and katakana words (827 ms), intermediate for katakana pseudohomophones (1032 ms), and longest for hiragana nonwords (1415 ms). Analyses of simple main effects showed that the effect of length was significant in each of the four string types, although the size of the length effect, which was assessed by the increase in reading latency per additional character, varied among the hiragana words (118 ms/character), katakana words (40 ms), katakana pseudohomophones (149 ms), and hiragana nonwords (114 ms). Compared with the younger normal readers in the study of Buchanan and Besner (1993) who demonstrated a length effect of less than 20 ms/character both for katakana words and hiragana pseudohomophones, the older normal controls in the present study seemed to have a larger length effect. ANOVA on reading duration revealed significant main effects of viewing condition (F1, 6 = 15.83), string type (F3, 18 = 72.50), and string length (F1, 6 = 108.37). There were significant interactions between viewing condition and type of the string (F3, 18 = 13.99), between viewing condition and length of the string (F1, 6 = 18.29), and between the type and length of the string (F3, 18 = 42.98). The three-way interaction among viewing condition, string type, and string length was also significant (F3, 18 = 6.18). The 20P condition (863 ms) yielded a significantly longer reading duration than the 80P condition (804 ms). Multiple comparisons revealed that the reading duration was significantly shortest for katakana words (681 ms), intermediate for hiragana words (731 ms) and katakana pseudohomophones (790 ms), and longest for hiragana nonwords (1131 ms). The degree of the length effect varied among the hiragana words (112 ms/character), katakana words (85 ms), katakana pseudohomophones (145 ms), 371 and hiragana nonwords (353 ms), each of which yielded a significant length effect. Because the speech duration of a mora is approximately 100–200 ms, the length effect observed in phonological words (hiragana words, katakana words, and katakana pseudohomophones) might be attributable to the effect of speech duration that is not specific to reading. In the analyses of eye movements, the number of fixations and duration per fixation before and during the oral response were subjected to ANOVA separately. ANOVA on the number of fixations before the response revealed significant main effects of string type (F3, 18 = 40.73) and length (F1, 6 = 15.03). Multiple comparisons showed that the mean number of fixations was significantly smallest for hiragana words (1.54 fixations) and katakana words (1.52), intermediate for katakana pseudohomophones (2.12), and largest for hiragana nonwords (3.18). Four-character strings produced a significantly smaller mean number of fixations (1.80) than 6character strings (2.38). The mean length effect—that is,the increase in the number of fixations per additional character— was 0.29: it was 0.28 among the hiragana words, 0.14 among the katakana words, 0.35 among the katakana pseudohomophones, and 0.38 among the hiragana nonwords. ANOVA on the number of fixations during the response revealed significant main effects of string type (F3, 18 = 22.40) and length (F1, 6 = 48.85), and a significant interaction between string type and length (F3, 18 = 10.23). Multiple comparisons showed that the number of fixations was significantly fewer in hiragana words (1.59 fixations), katakana words (1.38), and katakana pseudohomophones (1.92) than in hiragana nonwords (2.98). Four-character strings produced a significantly fewer mean number of fixations (1.46) than 6-character strings (2.48), indicating a length effect of 0.51. Although the effect of length was significant in each of the four string types, the size of the length effect seemed to be smaller for hiragana words (0.35), katakana words (0.32), and katakana pseudohomophones (0.37) than for hiragana nonwords (1.03). ANOVA on the mean duration of a fixation before the response revealed a significant main effect of length (F1, 5 = 9.68), with a shorter mean duration per fixation for 4-character strings (108 ms) than for 6-character strings (112 ms). ANOVA on the mean duration of a fixation during the response revealed a marginally significant three-way interaction of size, length and type (F3, 18 = 2.79, p = 0.07). The mean duration of a fixation was approximately 100 ms in all conditions. Pure alexic patients versus normal controls Because the high error rate observed in the pure alexic patients (SH: 48%, YH: 47%) reduced the number of trials available for analyses, results comparing the pure alexic patients and normal controls are reported, collapsed across the four string types. The reading accuracy of the pure alexic patients did not substantially vary as a function of viewing condition, length, or type: it was 45% for the 20P and 50% for the 80P condition; 48% for the 4-character and 47% for M. Kaneko et al. 10 Normal SH 8 YH 6 4 2 0 4 characters 6 characters 4 characters 20 points 4500 Normal 4000 SH 3500 YH 2500 2000 1500 1000 500 0 6 characters 20 points 4 characters 80 points 10 3000 4 characters 6 characters a) 6 characters Number of fixations during response Reading latency (ms) 5000 Normal SH 8 YH 6 4 2 0 80 points a) 4 6 4 6 characters characters characters characters 5000 20 points 4500 Normal 4000 Reading duration (ms) Downloaded by [Florida State University] at 16:41 22 December 2014 the 6-character strings; and 43–51% across the four types of strings. ANOVA on the reading accuracy of the pure alexic patients revealed that the effects of viewing condition, string type and string length were not significant. Fig. 5 presents the mean correct reading latency (Fig. 5a) and duration (Fig. 5b) of the two pure alexic patients and the means of the seven normal controls as a function of viewing condition and string length. Fig. 6 presents the number of fixations before (Fig. 6a) and during (Fig. 6b) the response. Three-way ANOVAs were performed on reading latency and duration, number of fixations, and duration per fixation, with subject group (pure alexic patients vs. normal controls), viewing condition (the 20P vs. 80P condition), and string length (4 vs. 6 characters) as factors. Number of fixations before response 372 SH YH 3500 80 points b) Fig. 6. Mean number of fixations (a) before and (b) during the oral response in the two pure alexic patients (SH and YH) and the normal controls (mean values among the 7 normal controls), as a function of viewing condition and string length. 3000 2500 2000 1500 1000 500 0 4 characters 6 characters 20 points 4 characters 6 characters 80 points b) Fig. 5. Mean correct reading (a) latency and (b) duration in the two pure alexic patients (SH and YH) and the normal controls (mean values among the 7 normal controls), as a function of viewing condition and string length. ANOVA on reading latencies revealed significant main effects of subject group (F1, 7 = 24.36), viewing condition (F1, 7 = 23.88), and string length (F1, 7 = 48.18), and a significant interaction between subject group and string length (F1, 7 = 6.23). The reading latency of the pure alexic patients (2351 ms) was significantly longer than that of the normal controls (1023 ms). Although analyses of simple main effects showed that the effect of string length was significant both among the pure alexic patients (2141 ms and 2560 ms for the 4- and 6character strings, respectively, with length effect—the increase in reading latency per additional character—of 210 ms/ character) and among the normal controls (924 ms and 1122 ms for the 4- and 6-character strings, respectively, with length effect of 99 ms/character), the effect was significantly larger Downloaded by [Florida State University] at 16:41 22 December 2014 Eye movement of Japanese pure alexia in the pure alexic patients than in the normal controls. ANOVA on reading durations revealed significant main effects of subject group (F1, 7 = 26.84), viewing condition (F1, 7 = 15.74), and length (F1, 7 = 98.83). There were significant interactions between subject group and string length (F1, 7 = 38.70) and between viewing condition and string length (F1, 7 = 6.00, p<0.05). The mean reading duration of the pure alexic patients (2309 ms) was longer than that of the normal controls (820 ms). Although the effect of string length was significant in both the pure alexic patients (1607 ms and 3011 ms for 4- and 6-character strings, respectively, with length effect of 702 ms/character) and normal controls (658 ms and 981 ms for 4- and 6-character strings, respectively, with length effect of 162 ms/character), the effect was significantly larger in the former than in the latter group. Furthermore, the length effect on reading duration, at least in the pure alexic patients, clearly exceeded the range of the length effect on speech duration (100—200 ms increase per additional mora). ANOVA on the number of fixations before the response revealed significant main effects of subject group (F1, 7 = 8.31) and string length (F1, 7 = 20.63). The pure alexic patients produced a significantly larger mean number of fixations before the response(4.34) than the normal controls (2.02). The effect of string length was significant in both the pure alexic patients (3.93 and 4.76 fixations for 4- and 6-character strings, respectively, with length effect—the increase in the number of fixations per additional character—of 0.42/character) and normal controls (1.77 and 2.31 fixations for 4- and 6-character strings, respectively, with length effect of 0.27/character). ANOVA on the number of fixations during the response revealed significant main effects of subject group (F1, 7 = 22.20) and string length (F1, 7 = 134.72), and a significant interaction between subject group and length (F1, 7 = 47.35). The pure alexic patients produced a significantly larger mean number of fixations during the response(5.12) than the normal controls (1.92). Although the length effect was significant in both the pure alexic patients (3.24 and 7.01 fixations for 4- and 6-character strings, respectively, with length effect of 1.89/character) and normal controls (1.44 and 2.41 fixations for 4- and 6-character strings, respectively, with length effect of 0.49/character), the effect was significantly larger in the former group than in the latter group. ANOVA on the duration per fixation either before or during the response did not yield any significant effect or interaction. The mean duration per fixation was approximately 100—106 ms regardless of whether it was measured before or during the response, subject group, viewing condition, and string length. Fig. 7 presents the results of analyses of visual inspection of the eye-movement patterns. To yield a sufficient number of observations, the visual analyses included trials with a selfcorrected response as well as a correct response in each subject. Each panel in Fig. 7 displays the number of occurrences of each eye-movement pattern—the lack of fixation on the first character (LF), the intermediate characters (LI), or the last character (LL), and fixation on all characters (FA)—in the normal controls (mean values) (Fig. 7a), SH (Fig. 7b), and 373 YH (Fig. 7c), as a function of the viewing condition, stimulus length and type. When two kinds of these eye-movement patterns occurred in a single trial (e.g., the lack of fixation on both the first and the last characters), both kinds were tallied. Thus, the total number of occurrences of each type of eye movement was occasionally greater than the number of stimuli, that is, 10. The performance of the normal controls varied according to the properties of the string: the pattern of fixation on all characters was dominant when viewing hiragana nonwords; and fixations were often omitted when viewing katakana words, hiragana words, and katakana pseudohomophones. On the other hand, the two pure alexic patients overwhelmingly had an eye-movement pattern with fixations on every character in each string regardless of the properties of the string. Finally, to combine the data of reading time and eye movements, Figs. 8a–h display typical examples of the horizontal coordinates of the eye position when reading the katakana words as a function of time from onset of the stimulus to triggering of the response (separated into three periods by the initiation of saccade to the string stimulus and the initiation of oral response), separately for each subject group, viewing condition, and string length. Furthermore, Fig. 9 displays the range of the horizontal coordinates (i.e., the average of the minimum, mean, and maximum of the horizontal coordinates) of the fixation points before and during the response of the two pure alexic patients and the normal controls (mean values), as a function of viewing condition and string length. In the normal controls, the fixation points that were detected before the oral response were mainly located in the left half of the string, although those detected during the oral response were often found in the right half of the string. These tendencies were clearly depicted in the 80P condition, and the data from the 20P condition demonstrated a similar pattern. These patterns showed that the normal controls had not yet fixated on the characters in the latter part of the string when initiating the oral response, even though string length and lexicality had significant effects on reading latency. In the pure alexic patients, although the range of the fixation points before (and also during) the oral response extended more rightward than that in the normal controls, the range of the fixation points before the oral response did not reach the end of the string (except in YH for 4-character strings in the 20P condition), even though string length had a significant effect on reading latency in the pure alexic patients. Discussion Two pure alexic patients, SH and YH, were impaired in reading aloud Japanese kana words and nonwords as assessed by the accuracy of pronunciation. Even in the correct responses, they demonstrated LBL reading, with a significantly more robust string-length effect than the normal controls, that was characterized by larger increases in reading latency and duration as the string length increased. This LBL reading pattern was accompanied by a larger number of fixations before and during the oral response and also a disproportionate increase 374 M. Kaneko et al. Normal controls 20P condition 10 5 FA 0 HW KW KP HN HW KW KP HN LL LI 80P condition 15 LF 10 5 0 HW KW KP HN HW KW 4 characters KP HN 6 characters a) Case SH Case YH 20Pcondition 10 5 FA 0 HW KW KP HN 15 HN LL 80P condition LI HW KW KP 20P condition 15 LF 10 5 Mean number of occurrences 15 Mean number of occurrences Downloaded by [Florida State University] at 16:41 22 December 2014 Mean number of occurrences 15 10 5 FA 0 HW KW KP HN HW KW KP HN 80P condition 15 LL LI LF 10 5 0 HW KW KP 4 characters HN HW KW KP 0 HN 6 characters b) c) HW KW KP HN 4 characters HW KW KP HN 6 characters Fig. 7. Results of analyses of visual inspection of eye-movement patterns. The number of occurrences of FA, LL, LI and LF in (a) the normal controls (mean values among the 7 normal controls), (b) Case SH, and (c) Case YH. FA, fixation on all characters; LL, lack of fixation on the last character; LI, lack of fixation on the intermediate characters; LF, lack of fixation on the first character. HW, hiragana word; KW, katakana word; KP, katakana pseudohomophone; HN, hiragana nonword. in the number of fixations during the response as the string length increased, relative to respective values in the normal controls. These results are in general agreement with those observed in pure alexic patients who read aloud words in the alphabetic writing system in terms of both reading time (see Behrmann et al., 1998a; Montant and Behrmann, 2000, Leff et al., 2001, for review of patients) and eye-movement pattern in reading text (Behrmann et al., 2001). Visual analyses of the eye-movement patterns revealed that whereas the normal controls often omitted their fixations when reading aloud words and pseudohomophones, the pure alexic patients overwhelmingly fixated on every character in a string irrespective of its property. However, the horizontal coordinates of the fixation points before the oral response did not reach the end of the strings in the pure alexic patients similar to the normal controls, despite the finding that in both groups string length had a significant effect on reading latency. On the basis of the data summarized above, we shall consider the implications of the current results on normal and LBL reading of Japanese kana strings. Effect of string length on reading time To interpret the finding that string length had a significant effect on reading time in both the normal controls and the pure alexic patients in the present study, we begin by referring to the interpretations of the effect of string length in current models of word reading in the alphabetic writing system. As Eye movement of Japanese pure alexia mentioned earlier, it is generally agreed that in normal reading the string input activates its orthographic representations in a parallel manner. This assumption is held in both the dualroute (e.g., Coltheart et al., 2001) and the single route (Plaut et al., 1996) models of reading. In the dual-route framework, the string input activates visual feature units and then letter units across individual letter positions. Subsequently, in the lexical route, the activation of letter units feeds to word units Normal, 20P 4 characters 8 375 Normal, 80P 6 characters 4 characters 6 characters 6 Horizontal coordinate (degree) 2 0 –2 –4 A A B –8 A C –6 B C B C A B 0 1000 0 1000 Time (ms) Time (ms) a) 0 b) 1000 Time (ms) c) 0 C 500 Time (ms) d) SH, 20P 4 characters 8 6 Horizontal coordinate (degree) Downloaded by [Florida State University] at 16:41 22 December 2014 4 4 2 0 –2 –4 A B C –6 –8 0 1000 2000 3000 Time (ms) e) Fig. 8. Examples of the trace of the horizontal coordinate of the position of the right eye, as a function of time from stimulus onset, in each stimulus condition of (a–d) a normal control and (e–g) the pure alexic patient SH. In both the normal control and the pure alexic patient, the 4-character word is (the mass media) and the 6-character word is (diamond). A, The period between the time point of presentation of the stimulus and the time point at which the subject’s eye left the fixation point. The coordinates are depicted in diamonds. B, The period between the time point at which the subject’s eye left the fixation point and the time point at which the subject started the oral response. The coordinates are depicted in squares. C, The period between the time points at which the subject started and finished the oral response. The coordinates are depicted in triangles. The coordinates that fulfilled the criterion of eye fixation are shown as filled markers across the three periods. 376 M. Kaneko et al. SH, 20P 6 characters 8 Downloaded by [Florida State University] at 16:41 22 December 2014 Horizontal coordinate (degree) 6 4 2 0 –2 –4 A –6 B C –8 0 1000 2000 3000 Time (ms) f) Horizontal coordinate (degree) SH, 80P 4 characters 8 6 4 2 0 –3 –4 A B –6 C –8 0 2000 1000 Time (ms) g) Horizontal coordinate (degree) 8 6 4 2 0 –2 –4 –6 –8 h) Fig. 8. Continued SH, 80P 6 characters A 0 B 1000 2000 Time (ms) C 3000 3000 Eye movement of Japanese pure alexia 8.0 Horizontal coordinate (degree) 20 Points 80 Points Before responses normal SH YH 6.0 4.0 2.0 0.0 During responses normal SH YH –2.0 –4.0 –6.0 4 characters 6 characters 4 characters 6 characters Downloaded by [Florida State University] at 16:41 22 December 2014 –8.0 Fig. 9. The average of the minimum (the lower boundary of the error bar), mean (the marker), and maximum (the upper boundary of the error bar) of the horizontal coordinates of the fixation points before and during the oral response of the two pure alexic patients and the normal controls (mean values among the 7 normal controls), as a function of viewing condition and string length. in the orthographic input lexicon, resulting in incomplete activation of word units of orthographically similar words as well as complete activation of the word unit of the target word. The spreading activations of orthographic representations comprising visual feature units, letter units, and word units proceed in a parallel manner across letter positions and bidirectionally among three layers of different kinds of units. In the nonlexical route, each letter unit is translated into its phonological counterpart in a serial manner and with a leftto-right order on the basis of a set of pronunciation rules. The outputs from the lexical and the nonlexical routes accumulate and are combined in the phoneme system (or the response buffer) until a sufficient level of activation across phoneme units comprising a whole string is achieved to initiate articulation. The effect of string length in normal reading is assumed to arise from the nonlexical route, in which phonemic activation of an entire string derived from this route is delayed as a function of increasing string length. In the theories of pure alexia, however, it is assumed that the stringlength effect observed in LBL reading does not arise from the abnormally slow function of the nonlexical route because (1) overt LBL reading (for instance, MAKE → /em, ei, kei, i/) does not correspond to the output from the nonlexical route (for instance, MAKE → /m, ei, k/); (2) pure alexic patients demonstrate an abnormal string-length effect for words as well as nonwords; (3) the size of the length effect in pure alexic patients is an order of magnitude larger than that in normal readers; and (4) Coltheart et al. (2001) demonstrated that in their computational model of the dual route theory, namely, the Dual-Route Cascaded (DRC) model, an abnormally slow nonlexical route yielded phonological alexia rather than pure alexia. According to the single route model of word reading, in which a string input activates its orthographic representation in a parallel manner and then the orthographic representation 377 is translated into its phonological representation by parallel distributed processing, the string-length effect in normal reading arises from the fact that longer strings have a smaller number of orthographic neighbors (e.g., words differing by one letter, such as SURE for SURF) because the experience of similar inputs helps the learning of computation in a distributed system. Additionally, given that the human visual acuity sharply drops as eccentricity increases even in foveal vision (Anstis, 1974; Jacobs, 1979), the string-length effect may occur because stimuli of longer words exert a stronger constraint on visual processing4. Thus, even when the activation of orthographic and phonological representations proceeds in a parallel manner in the system, a longer string may have a demerit in the computation of phonology from orthography. The presence of stronger visual constraints on longer strings may also apply to the activation of orthographic representations in the framework of the dual-route models. If pure alexic patients are impaired in this parallel activation of orthographic representations, their deficit should emerge more prominently in strings (and letter positions in a string) that exert stronger visual constraints. Consistent with this assumption, it has been reported that when a string is presented too briefly for pure alexic patients to identify each letter in a string in a LBL fashion, the patients are less accurate in reporting or detecting the letters around the final position of the string than those around the initial position, and sometimes this deficit is more pronounced for longer words (Kay and Hanley, 1991; Rapp and Caramazza, 1991; Buxbaum and Coslett, 1996; Chialant and Caramazza, 1998). Furthermore, Montant et al. (1998) reported that when a word was presented briefly with the fixation point at various letter positions within the word, a pure alexic patient demonstrated a decrease in accuracy of reading aloud words as a function of increasing eccentricity of the fixation point5. The standard view of reading Japanese kana strings is based on the dual-route framework of reading (e.g., Sasanuma, 1987), in which high-frequency kana words are processed by the lexical route whereas low-frequency kana words and kana nonwords are pronounced by the nonlexical route comprising the kana-mora correspondence rule. An implicit assumption of this dual-route framework is that orthographic representations of kana strings are activated in a parallel manner. Nevertheless, it is possible (and sometimes claimed) that Japanese readers pronounce each kana character in any kana string without any reference to other kana characters in that string— we refer to this successive and independent pronunciation of each character in a string as LBL pronunciation—because LBL pronunciation of a string is apparently indistinguishable from the normal pronunciation of the string. However, this is clearly not the case at least in normal readers because previous studies on reading aloud Japanese kana strings reported that lexicality (Besner and Hildebrandt, 1988), word frequency (Hino and Lupker, 1998), and string length (Buchanan and Besner, 1993) had significant effects on reading latency in normal young readers, and because the current study demonstrated that both string type (including orthographic and Downloaded by [Florida State University] at 16:41 22 December 2014 378 M. Kaneko et al. phonological lexicality) and string length have significant effects on reading latency in normal older subjects. Given the parallel activation of orthographic representations of kana strings, the string-length effect observed in the reading latency of the normal controls of the present study is considered to arise from the nonlexical route in the framework of the dualroute model. If one adopts the single-route model for Japanese kana characters, the string-length effect arises from the fact that longer strings impose a cost on the parallel processing of orthographic and phonological representations due to the smaller number of neighbors and/or larger degree of visual eccentricity. Our pure alexic patients also demonstrated a significant string-length effect on reading latency, with a significantly larger length effect in the patients than in the normal controls6. This indicates that the patients as well as the normal controls initiated articulation after processing of the entire string had begun, with the speed of this processing slower in the patients than in the normal controls. However, although the effect of string length on reading duration in the normal controls was within the effect of string length on speech duration, the size of the length effect on reading duration in the patients was not only significantly larger than that in the normal controls, but also clearly beyond the effect of string length on speech duration. Thus, in the pure alexic patients, the effect of length on reading latency indicates parallel activation of orthographic representation, whereas the effect of length on reading duration seems to support that these patients employed LBL pronunciation. To interpret these apparently discrepant results, we propose that in pure alexia, the parallel activation of orthographic representations is abnormally delayed, but not completely abolished, resulting in an exaggerated string-length effect. The data on eye movements provide further support for our proposal. Effect of string length on eye-movement pattern First, it should be noted that our normal controls showed a larger mean number of fixations in a single string—2.02 before and 1.92 during an oral response, for a total of 3.94 times—compared with the number of fixations per word in text reading reported in the literature (less than 1.5, for example, in Rayner and Pollatsek, 1987; Behrmann et al., 2001) and also that observed in single-word processing tasks (e.g., 2.05 in Pynte, 1996). Furthermore, the mean duration of a fixation in the normal controls (around 100 ms) seemed to be shorter than those reported in the literature on text reading (e.g., 200–250 ms for English text reported in Rayner and Pollatsek, 1987 and Behrmann et al., 2001 and for Japanese text including kana and kanji characters reported by Kanbe, 1998). These discrepancies may be due to differences in the subjects’ age, apparatus, stimuli, tasks, and/or how a fixation was defined and calculated. For example, the finding of a larger number of fixations and shorter fixation duration in our data suggests that we employed a rather strict criterion of fixation, namely, a rather small movement: a fixation was defined as eye movement within 0.3° during 66.7 ms. Although further eye movement studies should be performed to elucidate these discrepancies, we believe that our data provide reliable information at least in terms of relative differences between normal individuals and pure alexic patients7. Our normal controls demonstrated the significant effects of string type and length on the number of fixations both before and during the oral response. These effects on the number of fixations mirrored the effects on reading latencies and durations. Before an oral response, the number of fixations significantly increased as a function of decreasing lexicality from words to pseudohomophones to nonhomophonic nonwords and as a function of increasing string length. Because these fixation points were mainly in the left half of the string as illustrated in Figs. 8a–d and Fig. 9, the significant effects of string type and length on the number of these fixations suggest that in the normal controls, the entire string (including characters on which the subject had not yet fixated) had already begun to be processed before the oral response. This interpretation is consistent with the proposal based on reading latencies that a string input activates its orthographic representation in a parallel fashion in normal readers. The normal controls fixated on the latter characters during the oral response perhaps because there was insufficient information on the characters on which the subject had not yet fixated before the oral response, namely the latter part of the string, to support actual articulation. Thus, although the normal controls demonstrated a large number of fixations per word and they sometimes fixated on every character in a string (especially for nonwords) as indicated in Fig. 7a, this does not necessarily mean that they employed LBL pronunciation. In our pure alexic patients, the number of fixations before the oral response was greater than that in the normal controls, and itincreased as a function of increasing string length. However, the spatial range of these fixation points did not reach the end of the strings as in the case of the normal controls (Figs. 8e–h, Fig. 9). The significant effect of string length on the number of fixations before the oral response suggests that, even in the pure alexic patients, processing of the entire string has already begun before the initiation of the oral response, indicating residual function of the parallel activation of orthographic representation. However, the remaining data on eye-movement patterns seem to support the view that the patients employed LBL pronunciation. In the pure alexic patients, the number of fixations during the oral response was greater than that in the normal controls, and it increased as a function of increasing string length, with a significantly larger effect of string length in the patients than in the normal controls. Furthermore, the visual analyses demonstrated that the patients overwhelmingly fixated on every character in a string irrespective of the properties of the string. Our findings that string length had a significant effect on the number of fixations before the oral response and that the pure alexic patients had an abnormally increased number of fixations on each character in a string further support that in pure alexia, parallel activation of orthographic representations Downloaded by [Florida State University] at 16:41 22 December 2014 Eye movement of Japanese pure alexia is abnormally delayed, but not completely abolished. Thus, although the pure alexic patients sequentially fixated on each character in a string to complete the overt pronunciation of the entire string, the residual activation by the not-yet-fixated character plays a role in the time course of the patients' pronunciation and eye movements. This proposal is consistent with reports in the literature that properties of the entire string such as word frequency and imageability have an interacting effect with word length on reading latency in some pure alexic patients: for example, the word-length effect is more robust for low-frequency words and the word-frequency effect is more salient for longer words (Behrmann et al, 1998a). Behrmann et al. (1998a) interpreted this kind of interacting effect as arising from both the gradual bottom-up activation of orthographic representation from a string input and the top-down facilitation from lexical information in the framework of an Interactive Activation model (IA model: McClelland and Rumelhart, 1981), which is comparable to the three levels of orthographic representations in the DRC model—the visual feature, letter, and word levels. Because the processing among these levels is cascaded and bidirectional, even the incomplete activation at the letter level, for example, feeds forward to and back from the word level, resulting in restoration of letter activation especially for high-frequency words. In the pure alexic patients in our study, lexicality did not have a significant effect on their accuracy of reading aloud kana strings. As a result of the high error rate in the patients, it was difficult to probe the effect of lexicality on reading time and eye movements. However, we do not conclude that the effect of lexicality (or frequency) is absent in our pure alexic patients because, if present, it would be possible to observe this effect by increasing the number of strings in the experimental material. Furthermore, the clinical observations showed that these two patients demonstrated the frequency effect in reading aloud kanji words as described in the Subjects section. Although the presence of effects of whole string property other than length apparently indicates more central impairment than abnormally delayed parallel activation of orthographic representation, the assumption of bidirectional communication between lower and higher levels of representation may account for these effects as suggested by Behrmann et al. (1998a). Concluding remarks Our interpretation of the results of the current study can be summarized as follows: Although each Japanese kana character in a word and nonword can be pronounced independently of other characters in that string, the normal controls pronounced these kana strings by parallel activation of orthographic representation arising from the string input. While the pure alexic patients fixated on almost all characters in kana strings, the residual function of parallel activation of orthographic representations prevented the patients from pronouncing each kana character in a string without any reference to other characters in that string. 379 Although these interpretations are more consonant with the visual deficit hypothesis of pure alexia than with the hypotheses assuming impairment of visual word-form system or disconnection between orthographic and phonological representations, we have not attempted to rule out these other hypotheses because we have not yet addressed the phenomena on which these other theories are based. We did not prepare an experimental condition that investigates eye-movement patterns in reading tasks other than overt pronunciation, for example, lexical decision, semantic decision, or letter identification, which has been useful in studies on pure alexia, as well as nonreading tasks. Furthermore, because we did not investigate how reading difficulty per se influences eye movements, it is important to compare eye-movement patterns across different types of alexia including pure alexia, phonological alexia, surface alexia, and so on. However, despite the many issues that should be addressed in future studies, we hope that the current study provides detailed information on the reading time and eye-movement patterns in Japanese patients with pure alexia. Notes 1 These models of word reading differ with regard to processing after parallel identification of individual letters: One model assumes that these identified letters are translated into phonology within a single system that operates in a parallel manner (e.g., Plaut et al., 1996), whereas another model assumes that phonological translation is accomplished by two separate routes, in which the lexical route addresses wholeword phonology from whole-word orthography in a parallel manner and the nonlexical route converts each letter into its phonological counterpart in a serial manner based on a set of pronunciation rules (e.g., Coltheart et al., 2001). 2 Japanese orthography is comprised of two scripts—morphographic kanji and phonographic kana (see Patterson et al., 1995, among others, for further description). Whereas the majority of kanji characters have multiple legitimate pronunciations, each kana character has a single transparent and invariant pronunciation. 3 It should be noted that character frequency is not synonymous with word frequency. Kanji comprises approximately 6,000 characters, with about half of these used in daily life. Each kanji character appears as a single-character noun, as a component of multiple-character kanji nouns, kanji-kana compound verbs and adjectives, or one, two, or three of these written word classes. Because word frequency, character frequency and visual complexity are correlated with one another in kanji script, it is not easy to disentangle these effects in clinical observations. 4 During reading, the efficiency of letter identification is assumed to deteriorate asymmetrically, with more deterioration for letters that are located on the left than on the right of the fixation point (Nazir et al., 1991). This assumption has been empirically supported by, for example, the data demonstrating that the reading latency increased as a function of the Downloaded by [Florida State University] at 16:41 22 December 2014 380 M. Kaneko et al. distance of the first fixated letter from the center (or just left of the center) of a string, with a more robust effect when the fixated letter is located on the right than on the left of the center of the string (O'Regan et al., 1984). 5 In pure alexic patients, the difficulty in identifying letters located around the final part of a string cannot be simply attributed to the decrease in visual acuity as a function of increasing retinotopic eccentricity. For example, less accurate identification of letters around the final part is also observed when a string is presented briefly to the left of the fixation point, and thus the final letter appears at the center of foveal vision (Kay and Hanley, 1991; Chialant and Caramazza, 1998). Furthermore, in some cases or conditions, poorer letter identification is observed in the central rather than the final part of a string (Reuter-Lorentz and Brunn, 1990; Chialant and Caramazza, 1998). The exact nature of the visual constraint in pure alexic patients should be further investigated. 6 The string-length effect on the reading latency in our pure alexic patients (i.e., 210 ms increase per additional character) seems to be small compared with that reported in patients in alphabetic writing systems (see Leff et al., 2001 for a discussion about the absolute size of word-length effect in pure alexia). This is perhaps due to the experimental set-up we employed, in which a voice key was not used to erase the presented stimuli from the display as soon as it detected the subject’s oral response. In the experimental set-up used in our study, the patient may tend to initiate articulation before completion of phonological processing of the entire string, resulting in reduction of the effect of string length on reading latency. In agreement with this assumption, the effect of string length on reading duration was sufficiently large in our patients. 7 The criterion of 0.3° was employed to detect a fixation on a character in a string presented in the 20P condition, in which each character subtended 0.4°. However, in some studies, fixations less than 100 ms were eliminated from the analyses due to the possibility that such short fixation times reflect oculomotor programming not specific to reading (e.g., Morrison, 1984; Kambe et al., 2001). When we did not track fixations on each character in the 20P condition and also eliminated these short fixations from the analyses—that is, when fixation was defined as at least three consecutive samples each with a velocity of eye movement of less than 10.0 degree/second, corresponding to eye movement within 1.0° during 100 ms—our normal controls showed a mean duration of fixation of 206 ms before and 197 ms during the oral response, which were similar to those obtained in previous studies, although the mean number of fixations on a single string (1.96 before and 1.80 during an oral response, for a total of 3.76 times) was still large. Upon comparison of the mean number of fixations between the normal controls and the pure alexic patients using this criterion, we obtained the same statistical results as described in the main text: significant main effects of subject group and string length on the number of fixations both before and during the oral response, and a significant interaction between subject group and string length on the number of fixations during the oral response. Reference Amano S, Kondo T. NTT database series: Lexical properties of Japanese. Tokyo: Sanseido,1999 [in Japanese]. Antis SM. A chart illustrating variations in acuity with retinal position. Vision Res 1974; 14: 589–592. Arguin M, Bub D. Single character processing in a case of pure alexia. Neuropsychologia 1993; 31: 435–458. Arguin M, Bub D. Pure alexia: Attempted rehabilitation and its implications for interpretation of the deficit. Brain Lang 1994; 47: 233–268. Arguin M, Bub D, Bowers J. Extent and limits of covert lexical activation in letter-by-letter reading. Cognitive Neuropsychology 1998; 15: 53–92. Arguin M, Fiset S, Bub D. Sequential and parallel letter processing in letterby-letter dyslexia. Cognitive Neuropsychology 2002; 19: 535–555. Behrmann M, Nelson J, Sekuler EB. Visual complexity in letter-by-letter reading: “Pure” alexia is not pure. Neuropsychologia 1998b; 36: 1115–1132. Behrmann M, Plaut DC, Nelson, J. A literature review and new data supporting an interactive account of letter-by-letter reading. Cognitive Neuropsychology 1998a; 15: 7–51. Behrmann M, Shallice T. Pure alexia: A nonspatial visual disorder affecting letter activation. Cognitive Neuropsychology 1995; 12 : 409–454. Behrmann M, Shomstein SS, Black SE, Barton JJS. The eye movements of pure alexic patients during reading and nonreading tasks. Neuropsychologia 2001; 39: 983–1002. Besner D, Hildebrandt, N. Orthographic and phonological codes in the oral reading of Japanese kana. J Exp Psychol Learn Mem Cogn 1986; 13: 335–343. Biscaldi M, Gezeck S, Stuhr V. Poor saccadic control correlates with dyslexia. Neuropsychologia 1998; 36: 1189–1202. Bowers JS, Arguin M, Bub DN. Fast and specific access to orthographic knowledge in a case of letter-by-letter surface alexia. Cognitive Neuropsychology 1996b; 13: 525–567. Bowers JS, Bub DN, Arguin M. A characterization of the word superiority effect in a case of letter-by-letter surface alexia. Cognitive Neuropsychology 1996a; 13: 415–441. Buchanan L, Besner D. Reading aloud: Evidence for the use of a whole word nonsemantic pathway. Can J Exp Psychol 1993; 47: 133–152. Buxbaum LJ, Coslett HB. Deep dyslexic phenomena in a letter-by-letter reader. Brain Lang 1996; 54: 136–167. Chialant D, Caramazza A. Perceptual and lexical factors in a case of letter-byletter reading. Cognitive Neuropsychology 1998; 15: 167–201. Coltheart M, Rastle K, Perry C, Ziegeler J. DRC: A dual route cascaded model of visual word recognition and reading aloud. Psychol Rev 2001; 108: 204–256. Farah MJ, Wallace M. Pure alexia as a visual impairment: A reconsideration. Cognitive Neuropsychology 1991; 8: 313–334. Hino Y, Lupker, SJ. The effects of word frequency for Japanese kana and kanji words in naming and lexical decision: Can the dual-route model save the lexicalselection account. J Exp Psychol Hum Percept Perform 1998; 24: 1431–1453. Jacobs RJ. Visual resolution and contour interaction in the fovea and periphery. Vision Res 1979; 19: 1187–1196. Kambe G, Rayner K, Duffy S. Global context effects on processing lexically ambiguous words: Evidence from eye fixations. Mem Cognit 2001; 29, 363–372. Kanbe N. Reading Japanese and eye movements. In: Osaka N, editor. Reading: Information processing of brain and mind. Tokyo: Asakura shoten, 1998; 1– 16 [in Japanese]. Kashiwagi T, Kashiwagi A. Recovery process of a Japanese alexic without agraphia. Aphasiology 1989; 3: 75–91. Kawamura M, Mizobuchi J. Brain mechanism for reading. Osaka N, editor. Reading: Information processing of brain and mind. Tokyo: Asakura shoten, 1998; 185–216 [in Japanese]. Kay J, Hanley R. Simultaneous form perception and serial letter recognition in a case of letter-by-letter reading. Cognitive Neuropsychology 1991; 8: 249–273. Kokubo K, Suzuki K, Yamadori A, Imaizumu S, Onuma T. Mechanism of “letter-by-letter” reading in a Japanese case of pure alexia. Japanese Journal of Neuropsychology 2000; 16: 145–153 [in Japanese with an English abstract]. Leff AP, Crewes H, Plaut GT, Scott SK, Kennard C, Wise RJS. The functional anatomy of single-word reading in patients with hemianopic and pure alexia. Brain 2001; 124: 510–521. Leff AP, Scott SK, Crewes H, Hodgson TL, Cowey A. Howard D, Wise RJS. Impaired reading in patients with right hemianopia. Ann Neurol 2000; 47: 171–178. Downloaded by [Florida State University] at 16:41 22 December 2014 Eye movement of Japanese pure alexia McClelland JL, Rumelhart DE. An interactive activation model of context effects in letter perception: Part 1. An account of basic findings. Psychol Rev 1981; 88: 375–407. Miozzo M, Caramazza A. Varieties of pure alexia: The case of failure to access graphemic representations. Cognitive Neuropsychology 1998; 15: 203–238. Montant M, Behrmann M. Pure Alexia; Pure alexia. Neurocase 2000; 6: 265–294. Montant M, Behrmann M. Phonological activation in pure alexia. Cognitive Neuropsychology 2001; 18: 697–727. Montant M, Nazir TA, Poncet M. Pure alexia and the viewing position effect in printed words. Cognitive Neuropsychology 1998; 15: 93–140. Morrison RE. Manipulation of stimulus onset delay in reading: Evidence for parallel programming for saccades. J Exp Psychol Hum Percept Perform 1984; 10: 667–682. Nazir TA, O'Regan JK, Jacobs AM. On words and their letters. Bulletin of the Psychonomic Society 1991; 29: 171–174. O’Regan JK, Levy-Schoen A, Pynte J, Brugaillere B. Convenient fixation location within isolated words of different length and structure. J Exp Psychol Hum Percept Perform 1984; 10: 250–257. Patterson K, Suzuki T, Wydell T, Sasanuma S. Progressive aphasia and surface alexia in Japanese. Neurocase 1995; 1; 155–165. Plaut DC, McClelland JL, Seidenberg MS, Patterson K. Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychol Rev 1996; 103: 56–115. Pynte, J. Lexical control within-word eye movements. J Exp Psychol Hum Percept Perform 1996; 22: 958–969. 381 Rapp BC, Caramazza A. Spatially determined deficits in letter and word processing. Cognitive Neuropsychology 1991; 8: 275–311. Rayner K, Pollatsek A. Eye movements in reading: A tutorial review. In: Coltheart M, editor. Attention and performance XII: The psychology of reading. London, UK: Lawrence Erlbaum, 1987; 327–62. Reuter-Lorentz PA, Brunn JL. A prelexical basis for letter-by-letter reading: A case study. Cognitve Neuropsychology 1990; 7, 1–20. Sakurai Y., Ichikawa Y, Mannen T. Pure alexia from a posterior occipital lesion. Neurology 2001; 56: 778–781. Sasanuma S. Reading disorder resulting from brain damage. In: Goryo, K. Reading. Tokyo: Tokyo Daigaku Shuppan-kai, 1987; 175–226 [In Japanese]. Seki K, Yajima M, Sugishita M. The efficacy of kinesthetic reading treatment for pure alexia. Neuropsychologia 1995; 33; 595–609. Sekuler EB, Behrmann M. Perceptual cues in pure alexia. Cognitive Neuropsychology 1996; 13: 941–974. Sugishita M, Otomo K, Kabe S, Yunoki K. A critical reappraisal of neuropsychological correlates of Japanese ideogram (kanji) and phonogram (kana) reading. Brain 1992; 115: 1563–1585. Warrington EK, Langdon DW. Spelling dyslexia: A deficit of the visual wordform. J Neurol Neurosurg Psychiatry 1994; 57: 211–216. Warrington EK, Langdon DW. Does the spelling dyslexic read by recognizing orally spelled words? An investigation of a letter-by-letter reader. Neurocase 2002; 8: 210–21. Warrington EK, Shallice T. Word form dyslexia. Brain 1980; 103: 99–112. Weekes BS. Differential effects of number of letters on word and nonword naming latency. Q J Exp Psychol 1997; 50A: 439–456.