COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3), 312 – 323 Semantic errors in deep dyslexia: Does orthographic depth matter? Alan A. Beaton and Nia Wyn Davies Department of Psychology, University of Wales, Swansea, UK Semantic errors of oral reading by aphasic patients are said to be comparatively rare in languages with a shallow orthography. The present report concerns three bilingual brain-damaged patients who prior to their stroke were fluent in both English, an orthographically deep language, and Welsh, an orthographically shallow language. On a picture-naming task, each patient made a similar proportion of semantic errors in the two languages. Similarly, in oral reading of the corresponding words, no patient produced proportionally more semantic paralexias in English than in Welsh. The findings are discussed in relation to the summation hypothesis as invoked by Miceli, Capasso, and Caramazza (1994) to explain apparent differences in frequency of semantic errors of reading in languages differing in orthographic depth. Deep dyslexia is an acquired disorder of oral reading observed in brain-damaged patients, especially Broca’s aphasics with large lesions of the left hemisphere (for anatomical review see Lambon Ralph & Graham, 2000). Semantic errors in reading aloud (semantic paralexias) constitute the cardinal symptom of deep dyslexia (Coltheart, 1980, 2000). In some languages, such as English, the relationship between orthography and phonology is not one-to-one. A given letter string can be pronounced in different ways (for example, the different pronunciations of wind or tear). Conversely, a given sound can be represented in different ways (such as the “ee” sound in bead or freed). The orthography of English and of other languages with similar inconsistency in letter –sound relationships is said to be deep or opaque. In shallow or transparent orthographies such as Turkish, German, Spanish, and Italian, the letter – sound relationships are more regular and consistent. The Welsh orthography is highly transparent (Fear, 1997; Spencer & Hanley, 2003). Welsh is spoken as a first language in Wales by about 19% of the population over the age of 20 (Welsh Language Board, personal communication) but all Welsh speakers also learn English. This means that they grow up reading one orthographically shallow language (Welsh) and another orthographically deep language (English). Reports of deep dyslexia/dysgraphia in a shallow language are relatively rare (see below). This may be because semantic errors of reading Correspondence should be addressed to Alan A. Beaton, Department of Psychology, University of Wales, Swansea, Singleton Park, Swansea SA2 8PP, UK (E-mail: a.a.beaton@swansea.ac.uk). We are grateful to our participants for their willing co-operation and (as always) to Zoe Fisher and Neil Carter for technical assistance. We also express our thanks to Prof. Andrew Ellis and two anonymous referees for their helpful comments on a previous version of this paper. 312 # 2007 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business http://www.psypress.com/cogneuropsychology DOI:10.1080/02643290701251957 DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH occur infrequently in orthographically shallow languages (see Ferreres & Miravalles, 1995) and/ or because they are, for whatever reason, reported relatively infrequently (for review of studies of deep dyslexia in Spanish see Davies & Cuetos, 2005). Discussions concerning the relative frequency of occurrence of semantic errors of reading in deep and shallow orthographies have been very largely based upon comparisons between monolingual brain-damaged readers of these orthographies. Two papers often cited in relation to the apparent rarity of semantic errors in Spanish and Italian are those by Ardila (1991) and Basso and Corno (1994), respectively. Ardila (1991) reported that no semantic errors were made by any of 41 monolingual Spanish-speaking aphasics who were given tests of reading that included a list of 13 single words. Basso and Corno (1994) retrospectively examined the records of 502 brain-damaged Italian patients who were each given a reading test consisting of 10 words. The number of semantic errors was very low. However, deep dyslexic patients do not make errors on every experimental trial. Consequently, 10– 13 is too low a number of trials with which to elicit a reasonable number of semantic paralexias. Indeed, Basso and Corno (1994) begin their discussion by acknowledging that “The small number of stimuli in our test precludes any possibility of making a diagnosis of deep dyslexia” (p. 154). An alternative approach to evaluating the role of orthographic depth in the production of semantic paralexias can be made by comparing oral reading performance of deep dyslexic patients who are fluent readers of both an orthographically deep and an orthographically shallow language. There are, however, relatively few studies of acquired reading disorders in bilingual (biscriptal) readers (Paradis, 1994; Ratnavalli et al., 2000). The first published investigation of a bilingual deep dyslexic person appears to be that of Byng, Coltheart, Masterson, Prior, and Riddoch (1984); the patient was deep dyslexic in English while reading was said to be almost impossible in the (syllabic) Devanagari script. More recently, Béland and Mimouni (2001) tested an Arabic/ French bilingual patient, Z.T. This patient was reported to be deep dyslexic in both French and (deep) Arabic but to show more semantic errors in French. Karanth (2002) reported on a bilingual English –Hindi speaking aphasic patient who showed the pattern of deep dyslexia in English but made no semantic errors in Hindi, described as having “high gpc [grapheme– phoneme correspondence]” and “nearly always regular” (p. 144). Reading was said to be poorer overall in Hindi but no quantitative data are provided—perhaps there was insufficient opportunity for semantic errors to arise in Hindi. Raman and Weekes (2005) report the case of a bilingual Turkish – English reader who was deep dysphasic (that is, made semantic errors of repetition) but surface dyslexic in both languages (though when presented with Turkish words to read some of his responses were semantic errors given in English). Miceli, Capasso, and Caramazza (1994) attempted to account for the apparent difference in frequency of semantic errors in oral reading between readers of opaque and transparent orthographies in terms of the summation hypothesis (Caramazza & Hillis, 1990; Hillis & Caramazza, 1991; Miceli, Capasso, & Caramazza, 1999). According to this hypothesis, semantic errors in deep dyslexia are considered to arise if the response from the semantic-lexical route is insufficiently constrained by information from the sublexical route. Miceli et al. (1994) suggest that in orthographically transparent languages these two sources of information are more or less identical. By contrast, in an orthographically deep or opaque language, such as English, the two sources of information will on occasion conflict. For example, the word yacht may Ðbe pronounced in one of two ways (yot versus yæt t) depending on whether it is read as a whole word or decomposed using grapheme– phoneme conversion rules. Damage to a sublexical conversion route would be expected to limit the ability of this route to contribute to correct reading aloud. Miceli et al. (1994) argue that “the same cognitive damage to sublexical conversion mechanisms in a language with transparent and in one with opaque orthography might result in vastly [italics added] different consequences in COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 313 BEATON AND DAVIES reading and writing performance: in opaque languages, it could lead to a severe functional deficit of sublexical conversion; whereas in transparent languages it would interfere only minimally with sublexical conversion. Thus in a language with opaque orthography, damage to a sublexical conversion procedure would result in a situation where the output of the damaged component would consist of relatively unusable information in constraining the activation of lexical entries in the relevant output lexicon. By contrast, in a language with transparent orthography the comparable amount of damage to a sublexical conversion procedure would have relatively minor effects on the information value of the output of the damaged process, at least with respect to its role in constraining the activation of lexical forms in the relevant output lexicon. The obvious consequences of these contrasting situations is that transparent orthography languages are relatively “protected” from the production of semantic paralexias and paragraphias because the sublexical conversion mechanisms even when damaged continue to provide information that is useful in constraining the activation of phonological and orthographic lexical forms” (p. 331). A clear prediction from the argument put forward by Miceli et al. (1994) is that a bilingual English –Welsh deep dyslexic patient will make proportionally more semantic errors in oral reading of English (orthographically deep) than of Welsh (orthographically shallow) words. There should not, however, be any difference between the two languages in the frequency of semantic errors made in picture naming since this task does not engage the sublexical reading route, which therefore cannot contribute to the patient’s performance. This paper reports a test of these predictions. speakers and readers of English and Welsh prior to their stroke, two of the patients (P.D. and J.W.T.) considered English to be their dominant language while the third (M.J.) considered Welsh to be her dominant language. Poststroke all three patients were dysphasic. Speech was effortful and dysfluent with some difficulty in articulation. Aphasia quotients as determined by the Western Aphasia Battery (WAB; Kertesz, 1982) are given below for each patient. There is no comparable aphasia battery for use with Welsh-speaking patients. Patient P.D., male, aged 50 at the time of testing, suffered a left CVA in July, 2000. Computed tomography (CT) scan shortly after admission to hospital showed a large area of infarction in the territory of the left middle cerebral artery (Figure 1). His WAB aphasia quotient was 36. Patient J.W.T., male, aged 69 at the time of testing, suffered a left CVA in June 1998. CT scan (Figure 2) on the day of admission revealed a moderately sized area of infarction involving the left parietal region (left middle cerebral artery territory). According to the radiologist’s report: “There was also another smaller area of infarction at the right posterior parietal region—the latter maybe an old one. No other focal endocranial lesions or midline shift.” WAB aphasia quotient was 40. J.W.T. died suddenly on the 20th July 2006. Method Participants Three bilingual patients were each tested from 5 to 6 years after a left-sided cerebrovascular accident (CVA), at which time their medical condition and cognitive status were stable. Although fluent 314 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) Figure 1. CT scan for P.D. (male, 50 years old). DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH Figure 2. CT scan for J.W.T. (male, 69 years old). Patient M.J., female, aged 71 at the time of testing suffered a left CVA in September 1999. CT scan (Figure 3) two days after her stroke showed a small left parietal infarct. WAB aphasia quotient was 64. M.J. died in January, 2006. Procedure Each participant was given a number of background assessment tests over a series of sessions. These included the Western Aphasia Battery (Kertesz, 1982) and a number of subtests from the Figure 3. CT scan for M.J. (female, 71 years old). Psycholinguistic Assessments of Language Processing in Aphasia (PALPA) battery (Kay, Lesser, & Coltheart, 1992). Some of the PALPA tests were adapted for testing in Welsh, usually by translating the English word into Welsh. Occasionally, however, the Welsh word was different in length to the English equivalent, in which case a Welsh word of the same number of syllables and of approximately equivalent familiarity and imageability (see Fear, 1997) was substituted. There are no irregularly spelled words in Welsh so all Welsh words were regular. Nonwords were devised so as to be equivalent to the corresponding English nonwords. Note, incidentally, that the Welsh alphabet is not exactly the same as the English alphabet. For example, there is no English equivalent of the Welsh “ll”, and there is no “k” in the Welsh alphabet; the Welsh letter “f” is pronounced as a “v” in English; the Welsh “ff” is pronounced as English “f”. The stimulus items for picture naming and oral reading were taken from the Snodgrass and Vanderwart (1980) set. Dual-word items (such as “lamp shade” and “light bulb”) were removed from the stimulus set prior to administration of the tasks. Pictures that had a name agreement in English of less than 60% according to Bates et al. (2003) were also removed, leaving 222 items from the Snodgrass and Vanderwart (1980) set. These 222 items were shown to a panel of 18 native Welsh speakers who were asked to provide the Welsh name for each item. Name agreement was greater than 70% for all items. These stimuli were subsequently presented one at a time to each of our three patients who was asked to provide its name in English or in Welsh or to read the corresponding English or Welsh word depending upon condition. The stimuli were presented in pseudorandom order to each patient across a number of sessions, with the constraint that the same item was not presented for reading and naming within the same session. Within each session, items were blocked according to language of response. That is, participants responded to a set of items in English only or in Welsh only throughout a COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 315 BEATON AND DAVIES Table 1. WAB aphasia scores Scores WAB subtest Spontaneous speech (a) Information content (b) Fluency Comprehension (a) Auditory verbal comprehension (b) Auditory word recognition (c) Sequential commands Repetition Object naming (a) Word fluency (b) Sentence completion (c) Responsive speech Reading (a) Reading comprehension (b) Reading commands (c) Written word–object matching (d) Written word–picture matching (e) Picture stimuli–word matching (f) Spoken word–visual word match (g) Spoken word–visual letter name (h) Oral spelling (by examiner: patient says word spelled) Spelling Writing (a) On request (e.g., name & address) (b) Written output from pics stimuli (c) Writing to dictation (d) Writing words from visual stimuli (e) Writing the alphabet (f) Writing numbers 0–20 (g) Dictated letter names (h) Dictated numbers (i) Copying a sentence from a card Apraxia: Carrying out oral commands Constructional and calculation (a) Construction: drawing (b) Calculation P.D. J.W.T M.J. 5 (10) 2 (10) 4 (10) 2 (10) 8 (10) 6 (10) 51 (60) 45 (60) 11 (80) 17 (100) 26 (60) 1 (20) 2 (10) 4 (10) 57 (60) 55 (60) 40 (80) 24 (100) 27 (60) 5 (20) 6 (10) 5 (10) 54 (60) 54 (60) 50 (80) 44 (100) 36 (60) 5 (20) 8 (10) 8 (10) 14 (40) 6 (20) 6 (6) 6 (6) 6 (6) 3 (4) 4 (6) 0 (6) 0 (6) 22 (40) 8 (20) 6 (6) 6 (6) 6 (6) 4 (4) 5 (6) 1 (6) 0 (6) 22 (40) 8 (20) 6 (6) 6 (6) 6 (6) 3 (4) 5 (6) 1 (6) 2 (6) 3 (6) 1 (34) 0 (10) 0 (10) 4 (12.5) 10 (10) 1 (2.5) 2 (5) 8 (10) 40 (60) 4 (6) 0 (34) 2 (10) 2 (10) 6 (12.5) 10 (10) 2 (2.5) 4 (5) 7 (10) 60 (60) 3 (6) 6 (34) 2 (10) 4 (10) 9 (12.5) 9.5 (10) 2 (2.5) 4 (5) 8 (10) 60 (60) 10 (30) 16 (24) 14.5(30) 24 (24) — Note: WAB ¼ Western Aphasia Battery. Maximum score in parentheses. session. If the patient gave the plural rather than singular form of a stimulus item it was accepted as correct. Results The patients’ scores on subtests of the Western Aphasia Battery are shown in Table 1, and their 316 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) scores on PALPA and other tests are shown in Table 2. Patients P.D. and J.W.T. showed an effect of imageability in reading words, poorer performance in reading function words than content words (for each patient and each language chi-square significant in each case at p , .02 or beyond), and an DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH Table 2. Scores on tests administered to P.D, J.W.T, and M.J Tests Pyramid & Palm Trees (a) Picture–picture version (b) Word–word version PALPA 47 (spoken word–picture matching) PALPA 48 (written word– picture matching) PALPA 31 (word reading aloud: imageability & freq) WELSH (translated version) PALPA 32(grammatical class reading) Nouns Adjectives Verbs Function WELSH (translated version) Nouns Adjectives Verbs Function PALPA 35 (Regularity of reading English) Regular words Irregular (exception) words PALPA 36 (Nonword reading in English) WELSH (nonword reading translated version) PALPA 8 (Nonword repetition in English) WELSH (translated version) PALPA 9 (imageability & frequency word repetition) WELSH (translated version) Letter–sound conversion (written letter–spoken sound) English Welsh HI/HF HI/LF LI/HF LI/LF HI/HF HI/LF LI/HF LI/LF HI/HF HI/LF LI/HF LI/LF HI/HF HI/LF LI/HF LI/LF P.D J.W.T M.J 49 (52) 47 (52) 37/40 34/40 49 (52) 49 (52) 38/40 37/40 50 (52) 49 (52) — 32/40 16/20 10/20 2/20 0/20 15/20 9/20 5/20 1/20 14/20 13/20 9/20 6/20 17/20 13/20 5/20 2/20 18/20 17/20 13/20 11/20 — 8/20 7/20 4/20 0/20 12/20 12/20 11/20 4/20 13/20 15/20 12/20 4/20 9/20 5/20 2/20 1/20 10/20 8/20 4/20 3/20 — 17/30 16/30 0/24 1/24 21/30 18/30 3/24 3/24 — 9/24 6/24 0/30 0/30 1/30 0/30 2/30 2/30 17/20 14/20 10/20 9/20 16/20 15/20 5/20 3/20 — 18/20 14/20 12/20 5/20 19/20 15/20 14/20 5/20 — 4/26 7/28 5/26 8/28 10/26 8/18 Note: HI ¼ high imageability. LI ¼ low imageability. HF ¼ high frequency. LF ¼ low frequency. PALPA ¼ Psycholinguistic Assessments of Language Processing in Aphasia. Maximum score in parentheses. COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 317 BEATON AND DAVIES Table 3. Picture naming in English and Welsh P.D. J.W.T. M.J. English Welsh English Welsh English Welsh Correct Omission Semantic error Name of visually similar object Circumlocution Cross-linguistic semantic errora Phonemic error Morphological errorb Name correct language incorrectc 99 40 46 3 26 3 2 — 3 71 39 46 2 12 16 — — 36 124 39 26 — 19 6 2 1 5 104 36 36 — 4 6 — — 36 124 22 39 4 19 6 — — 8 132 19 38 3 5 9 — — 16 Total 222 222 222 222 222 222 a Semantic error given in nontarget language; borange !oranger; ccorrect name given in nontarget language. impairment in reading, writing, and repeating nonwords of both languages. Patient M.J. could not be tested so extensively but the pattern of her performance was similar (imageability effect in reading English words; poor nonword reading and nonword repetition in English and Welsh; poor function word reading in English). Letter – sound conversion was impaired in all three patients in both English and Welsh, and they all made semantic errors of reading (and naming) in each language (see Tables 3 and 4 below). These findings establish that all three patients were deep dyslexic in both English and Welsh. Picture naming The data for picture naming in English and Welsh are shown in Table 3. Examples of errors referred to as naming visually similar items are giving the response wheel to the stimulus item button or screw to needle. McNemar tests showed that for patient P.D., picture naming was significantly better in Table 4. Reading in English and Welsh P.D. J.W.T. M.J. English Welsh English Welsh English Welsh Correct Omission Semantic error Cross-linguistic semantic errora Visual then semantic errorb Name of visually similar object Circumlocution Phonemic error Morphological error Name correct language incorrectc 147 30 15 1 1 15 10 1 2 — 83 65 30 6 — 21 7 — 1 9 152 17 12 2 — 26 10 — 1 2 121 37 21 7 — 27 9 — — — 151 21 14 6 — 21 8 — — 1 159 31 16 2 — 7 5 — — 2 Total 222 222 222 222 222 222 a Semantic error given in nontarget language; bhanger ! harbour ! anchor; ccorrect name given in nontarget language. 318 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH English than in Welsh (x2 ¼ 9.12, p , .01). The same was true for J.W.T. (x2 ¼ 4.76, p , .05). For patient M.J., naming was slightly but not significantly better in Welsh (x2 ¼ 1.33, p . .05). (For all tests reported in this paper degrees of freedom ¼ 1, and the direction is two-tailed.) Chi-square comparison of semantic errors relative to all other errors of naming (excluding omissions) in English compared with Welsh showed no significant effect for any of the three patients (p . .05). This was also the case when the analyses were repeated taking account of failures to respond (omissions) as well as errors. Reading aloud The data for reading in English and Welsh are shown in Table 4. McNemar tests showed that for patients P.D. (x2 ¼ 42.67, p , .001) and J.W.T. (x2 ¼ 10.80, p , .01) reading was significantly better in English than in Welsh. For patient M.J. reading was slightly but not significantly better in Welsh than in English (x2 , 1). Chi-square tests comparing semantic errors relative to all other errors (excluding omissions) of reading in Welsh versus English gave the following results: For P.D., x2 , 1, for J.W.T., x2 ¼ 1.02; for M.J., x2 ¼ 3.18 (p . .05 in each case). Equivalent results were obtained when the analyses were repeated taking account of failures to respond (omissions) as well as errors. Thus semantic errors were not proportionally more frequent in one language than in the other. Reading versus picture naming We compared reading with naming performance in each of the two languages of our patients. McNemar tests showed that for P.D., reading in English was significantly more accurate than picture naming (x2 ¼ 28.10; p , .001) but not in Welsh (x2 ¼ 1.90, p . .05). For J.W.T., too, reading in English was significantly better than naming (x2 ¼ 9.33, p , .01); the difference between the two tasks narrowly failed to reach significance in Welsh (x2 ¼ 3.48, p . .05). For patient M.J., reading was significantly better than naming in both English (x2 ¼ 10.57, p , .01) and Welsh (x2 ¼ 9.72, p , .01). For responses given in English, comparison of semantic and other errors in picture naming versus reading (see Tables 3 and 4) gave the following results. For P.D., x2 ¼ 1.36 (p . .05); for J.W.T., x2 ¼ 4.80 (p , .05); for M.J., x2 ¼ 5.80 (p , .05). Thus in English the latter two patients made proportionally more semantic errors, relative to other types of error, in naming than in reading. For responses given in Welsh, the corresponding chi-square values for semantic errors in naming versus reading were not significant for any of the three patients (p . .05 in each case). The chi-square analyses comparing semantic errors in picture naming versus reading in each of the two languages were repeated taking account of failures to respond (omissions) as well as errors. In English, P.D. but not the other two patients showed significantly more semantic errors in picture naming than in reading (x2 ¼ 5.83, p , .05). In Welsh, only M.J. showed a significantly greater proportion of semantic errors in picture naming (x2 ¼ 3.89, p , .05). Discussion It has been argued that the summation hypothesis can account for an apparent difference between transparent and opaque orthographies in the frequency of semantic errors of reading made by brain-damaged subjects. According to Miceli et al. (1994), a transparent orthography is relatively protected from the production of semantic paralexias. We therefore compared reading in English and Welsh by bilingual deep dyslexic participants. If Miceli et al. are correct, we should have observed relatively more semantic paralexias in English than in Welsh. Our data provide a clear refutation of this prediction; there were not more semantic errors made by our patients in reading English than in reading the corresponding Welsh words. The production of semantic errors of reading and/or writing has been held to require the presence of impaired sublexical conversion COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 319 BEATON AND DAVIES mechanisms, since otherwise intact conversion procedures would “block” the output of semantic errors (Patterson & Marcel, 1977; see papers in Caramazza & Hillis, 1990; Coltheart, Patterson, & Marshall, 1980; Friedman, 1996; Hillis & Caramazza, 1991; Miceli et al., 1999; Southwood & Chatterjee, 2000; Nickels, 1992). If this is true, then there would be no reason to expect any difference in semantic paralexias between deep and shallow orthographies when the orthographic-phonological sublexical conversion route is totally abolished since the means of “blocking” semantic errors would be unavailable for both languages. All else being equal, semantic errors in the two languages would be “free” to arise with approximately equal frequency in reading (as in picture naming). Certainly all of our patients (in common with other deep dyslexics) had a severely impaired sublexical conversion route. However, some residual capacity for converting from letter to sound remained available; all three patients performed poorly but at above-chance level on letter –sound conversion in both languages. Thus there is little or no reason to argue that our patients did not provide a suitable test of the summation hypothesis as applied to reading in a shallow (transparent) and a deep (opaque) orthography. Our failure to confirm the prediction of the summation hypothesis concerning semantic errors therefore argues against any straightforward version of its application to languages differing in orthographic depth as outlined by Miceli et al. (1994). Indeed, the hypothesis predicts that there should be more semantic errors made in response to regular words than to irregular words. A glance at the published literature (e.g. Beaton, Guest, & Ved, 1997) suggests that this is not the case, and it certainly did not apply in the present study. When all is said and done concerning the irregularity of the English language, it remains improbable that phonological recoding of irregular words (which would produce regularization errors in reading) would not provide some useful information. Between the correct pronunciation and a regularized pronunciation there will always be 320 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) considerable phonemic overlap even in the case of such highly irregular or exception words as yacht or colonel. This being so, it is likely that there will be only very few occasions on which a damaged conversion procedure will provide less useful information in a deep orthography than in a shallow one. Even if Miceli et al. (1994) are correct that, following brain damage, more sublexical information will be preserved in a transparent (shallow) than an opaque (deep) orthography, almost always sufficient information from each orthography would survive to block the output of many semantic errors. If this is so, then arguably there is no basis for the prediction that there will be fewer semantic errors made in reading a shallow than a deep orthography. The interpretation of results obtained from bilingual patients is not without a number of potential problems. These include the level of competence achieved in each language, the relative age and order of acquisition of the two languages, and so on (for discussion see Paradis, 2004). Of our three patients, two considered themselves to be dominant in English (and named and read more items correctly in English than Welsh) while the third patient considered Welsh to be her dominant language (and named and read more items correctly in Welsh than English). In all three cases, however, there is no evidence of proportionally more semantic errors of reading in one language than in the other. Neither language dominance nor orthographic depth therefore appears to have influenced the relative frequency of production of semantic errors in our patients. For all three of our patients, reading accuracy was better than picture-naming performance in both languages (significantly so for English but in Welsh significant only for M.J. though close to significance for J.W.T.). This is consistent with the availability of some sublexical support in reading. Furthermore, it is possible that some degree of implicit phonological processing was also available to each of our patients (Buchanan, Hildebrandt, & MacKinnon, 1994, 1996; Buchanan, McEwen, Westbury, & Libben, 2003; Katz & Lanzoni, 1992). An alternative DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH explanation of the better reading than naming accuracy in our patients draws on the notion that the semantic system is not unimodal (or supramodal) but is organized by “modality” of input (Beavois, 1982; Shallice, 1988; for contrary view see Caramazza, Hillis, Rapp, & Romani, 1990; Hillis & Caramazza, 1995; for review and alternative formulation see Plaut, 2002). On this account, information presented via the spoken or written word enters a verbal semantic system whereas (nonverbal) information received through vision enters a visual semantic system. The verbal semantic system has direct access to the speech output mechanisms while the visual nonverbal system has only indirect access via communication with the verbal system. These two systems are assumed to be in contact but in certain cases of brain damage there may be some impediment to transfer of information between the two systems. Direct access to speech output mechanisms from the verbal semantic system would explain better reading than naming performance for corresponding items. Turning now to another feature of our results, all three patients made proportionally more semantic errors, relative to other types of error, in naming than in reading when responding in English (though the statistical significance of the effect varied according to whether failures to respond—omissions—were included in the analysis). When tested in Welsh, however, only one of the patients showed significantly more semantic errors in naming than in reading, and that only when omissions were included. It is possible that residual phonological support was greater for English than for Welsh reading, although the data shown in Table 2 (and inspection of error responses to nonword stimuli) do not, in fact, support this idea. Nonword reading and letter – sound conversion were approximately equivalent for the two languages. Finally, in contrast to the dual-route account of reading, the so-called triangle model (Patterson & Lambon-Ralph, 1999; Seidenberg & McClelland, 1989) does not include separate lexical and sublexical routes. Instead a single route is used for reading both familiar words and unfamiliar letter strings. In this model, there are connections between orthography and phonology and between each of these and the semantic system. Differences in orthographic depth are captured by differences in degree of consistency in the mapping between spelling (orthography) and pronunciation (phonology), which are picked up by the learning algorithm and encoded by the strength of the connections (Seidenberg, 1992). The computation of phonology is constrained by semantic processing: The weaker the connections between orthography and phonology, the greater the influence of semantics. Damage to phonological-processing mechanisms would produce increased reliance on semantic processing. An impairment of semantic processing, a feature of dual-route accounts of at least some (central) types of deep dyslexia (Dickerson & Johnson, 2004), in addition to impaired phonological-processing mechanisms, implies that semantic errors will be produced in response to some target words, even in the context of some degree of preserved phonological processing. The triangle model might therefore predict that inherently weaker connections between orthography and phonology for an orthographically deep than for a shallow language, combined with a given degree of semantic impairment, would lead to a greater likelihood of semantic errors being made by a deep dyslexic patient reading an opaque (deep) than a transparent (shallow) language. That is, both the triangle model and the “summation version” of the dual-route model at first blush might be thought to make the same prediction—more semantic errors in reading a deep than a transparent language. However, for the reasons outlined above, such a prediction may not, in fact, be appropriate. Certainly, our findings do not support such a prediction though, of course, they do not constitute evidence against the summation (or triangle) hypothesis as such. Manuscript received 9 May 2006 Revised manuscript received 9 January 2007 Revised manuscript accepted 30 January 2007 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 321 BEATON AND DAVIES REFERENCES Ardila, A. (1991). Errors resembling semantic paralexias in Spanish-speaking aphasics. Brain and Language, 41, 437– 445. Basso, A., & Corno, M. (1994). Semantic errors in transcoding tasks in a shallow orthography: A retrospective study on 52 Italian vascular patients. Journal of Neurolinguistics, 8, 149– 156. Bates, E., D’Amico, S., Jacobsen, T., Székely, A., Andonova, E., Devescovi, A., et al. (2003). Timed picture naming in seven languages. Psychonomic Bulletin and Review, 10, 344– 380. Beaton, A. A., Guest, J., & Ved, R. (1997). Semantic errors of naming, reading, writing and drawing following left hemisphere infarction. Cognitive Neuropsychology, 14, 459– 478. Beauvois, M.-F. (1982). Optic aphasia: A process of interaction between vision and language. Philosophical Transactions of the Royal Society of London B, 298, 35 – 47. Béland, R., & Mimouni, Z. (2001). Deep dyslexia in the two languages of an Arabic/French bilingual patient. Cognition, 82, 77 – 126. Buchanan, L., Hildebrandt, N., & MacKinnon, G. E. (1994). Phonological processing of non-words by a deep dyslexic patient: A rowse is implicitly a rose. Journal of Neurolinguistics, 8, 163– 181. Buchanan, L., Hildebrandt, N., & MacKinnon, G. E. (1996). Phonological processing of non-words in deep dyslexia: Typical and independent? Journal of Neurolinguistics, 9, 113– 133. Buchanan, L., McEwen, S., Westbury, C., & Libben, G. (2003). Semantics and semantic errors: Implicit access to semantic information from words and non-words in deep dyslexia. Brain and Language, 84, 65 – 83. Byng, S., Coltheart, M., Masterson, J., Prior, M., & Riddoch, J. (1984). Bilingual biscriptal deep dyslexia. The Quarterly Journal of Experimental Psychology, 36A, 417– 433. Caramazza, A., & Hillis, A. E. (1990). Where do semantic errors come from? Cortex, 26, 95 –122. Caramazza, A., Hillis, A. E., Rapp, B. C., & Romani, C. (1990). The multiple semantics hypothesis: Multiple confusions? Cognitive Neuropsychology, 7, 161– 189. Coltheart, M. (1980). Deep dyslexia: A review of the syndrome. In M. Coltheart, K. E. Patterson, & J. C. Marshall (Eds.), Deep dyslexia (pp. 23 –47). London: Routledge & Kegan Paul. 322 COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) Coltheart, M. (2000). Deep dyslexia is right hemisphere reading. Brain and Language, 71, 299– 309. Coltheart, M., Patterson, K. E., & Marshall, J. C. (Eds.). (1980). Deep dyslexia. London: Routledge & Kegan Paul. Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual route cascaded model of visual word recognition and reading aloud. Psychological Review, 108, 204– 256. Davies, R., & Cuetos, F. (2005). Acquired dyslexia in Spanish: A review and some observations on a new case of deep dyslexia. Behavioural Neurology, 16, 85 – 101. Dickerson, J., & Johnson, H. (2004). Sub-types of deep dyslexia: A case study of central deep dyslexia. Neurocase, 10, 39 – 47. Fear, W. (1997). Ratings for Welsh words and their English equivalents. Behavior Research Methods, Instruments and Computers, 29, 425– 445. Ferreres, A. R., & Miravalles, G. (1995). The production of semantic paralexias in a Spanish-speaking aphasic. Brain and Language, 49, 152– 172. Friedman, R. (1996). Recovery from deep alexia to phonological alexia: Points on a continuum. Brain and Language, 52, 114– 128. Hillis, A. E., & Caramazza, A. (1991). Mechanisms for accessing lexical representations for output: Evidence from a category-specific semantic deficit. Brain & Language, 40, 106– 144. Hillis, A. E., & Caramazza, A. (1995). Cognitive and neural mechanisms underlying visual and semantic processing: Implications from “optic aphasia”. Journal of Cognitive Neuroscience, 7, 457– 478. Hillis, A. E., Rapp, B. C., & Caramazza, A. (1999). When a rose is a rose in speech but a tulip in writing. Cortex, 35, 337– 356. Karanth, P. (2002). The search for deep dyslexia in syllabic writing systems. Journal of Neurolinguistics, 15, 143– 155. Katz, R. B., & Lanzoni, S. M. (1992). Automatic activation of word phonology from print in deep dyslexia. Quarterly Journal of Experimental Psychology, 45A, 575– 608. Kay, J., Lesser, R., & Coltheart, M. (1992). Psycholinguistic Assessment of Language Processing in Aphasia. Hove, UK: Psychology Press. Kertesz, A. (1982). Western Aphasia Battery. Psychological Corporation: Harcourt Brace Jovanovich, Inc. Lambon Ralph, M., & Graham, N. L. (2000). Acquired phonological and deep dyslexia. Neurocase, 6, 141– 178. DEEP DYSLEXIA AND ORTHOGRAPHIC DEPTH Miceli, G., Capasso, R., & Caramazza, A. (1994). The interaction of lexical and sublexical processes in reading, writing and repetition. Neuropsychologia, 32, 317– 333. Miceli, G., Capasso, R., & Caramazza, A. (1999). Sublexical conversion procedures and the interaction of phonological and orthographic lexical forms. Cognitive Neuropsychology, 16, 557– 572. Nickels, L. (1992). The autocue? Self-generated phonemic cues in the treatment of a disorder of reading and naming. Cognitive Neuropsychology, 9, 155– 182. Paradis, M. (1994). Neurolinguistics aspects of implicit and explicit memory: Implications for bilingualism and SLA. In N. C. Ellis (Ed.), Implicit and explicit learning of languages (pp. 393– 418). London: Academic Press. Paradis, M. (2004). A neurolinguistic theory of bilingualism. John Benjamins. Patterson, K. E., & Lambon-Ralph, M. (1999). Selective disorders of reading? Cognitive Neuroscience, 9, 235– 239. Patterson, K. E., & Marcel, A. J. (1977). Aphasia, dyslexia and the phonological coding of written words. Quarterly Journal of Experimental Psychology, 29, 307– 318. Plaut, D. C. (2002). Graded modality-specific specialisation in semantics: A computational account of optic aphasia. Cognitive Neuropsychology, 19, 603– 639. Raman, I., & Weekes, B. S. (2005). Acquired dyslexia in a Turkish – English speaker. Annals of Dyslexia, 55, 79 – 104. Ratnavalli, E., Murthy, G. G., Nagaraja, D., Veerendrakumar, M., Jayaram, M., & Jayakumar, P. N. (2000). Alexia in Indian bilinguals. Journal of Neurolinguistics, 37 – 46. Seidenberg, M. S. (1992). Beyond orthographic depth in reading: Equitable division of labour. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 85-118). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96, 523– 568. Shallice, T. (1988). From neuropsychology to mental structure. Cambridge, UK: Cambridge University Press. Snodgrass, J. G., & Vanderwart, M. (1980). A standardized set of 260 picture norms for name agreement, image agreement, familiarity and visual complexity. Journal of Experimental Psychology. Human Learning and Memory, 6, 174– 215. Southwood, M. H., & Chatterjee, A. (2000). The interaction of multiple routes in oral reading: Evidence from dissociations in naming and oral reading in phonological dyslexia. Brain and Language, 72, 14 – 39. Spencer, Ll., & Hanley, R. (2003). Effects of orthographic transparency on reading and phoneme awareness in children learning to read in Wales. British Journal of Psychology, 94, 1 – 28. COGNITIVE NEUROPSYCHOLOGY, 2007, 24 (3) 323