Elsevier

Brain and Language

Volume 134, July 2014, Pages 44-67
Brain and Language

The anatomical foundations of acquired reading disorders: A neuropsychological verification of the dual-route model of reading

https://doi.org/10.1016/j.bandl.2014.04.001Get rights and content

Highlights

  • We used an anatomo-correlative procedure to localise the reading impairments of 59 focal brain damaged patients.

  • Two reading tasks, one of words and nonwords, and one of words with unpredictable stress positions, were used.

  • The results of this study are only partially consistent with the current state of the art.

  • Acquired dyslexia may ensue after different cortical damage.

  • White matter disconnection may play a crucial role in some cases.

Abstract

In this study we investigated the neural correlates of acquired reading disorders through an anatomo-correlative procedure of the lesions of 59 focal brain damaged patients suffering from acquired surface, phonological, deep, undifferentiated dyslexia and pure alexia. Two reading tasks, one of words and nonwords and one of words with unpredictable stress position, were used for this study. We found that surface dyslexia was predominantly associated with left temporal lesions, while in phonological dyslexia the lesions overlapped in the left insula and the left inferior frontal gyrus (pars opercularis) and that pure alexia was associated with lesions in the left fusiform gyrus. A number of areas and white matter tracts, which seemed to involve processing along both the lexical and the sublexical routes, were identified for undifferentiated dyslexia. Two cases of deep dyslexia with relatively dissimilar anatomical correlates were studied, one compatible with Coltheart, 1980a, Coltheart, 1980b whereas the other could be interpreted in the context of Morton and Patterson’s (1980), multiply-damaged left-hemisphere hypothesis. In brief, the results of this study are only partially consistent with the current state of the art, and propose new and stimulating challenges; indeed, based on these results we suggest that different types of acquired dyslexia may ensue after different cortical damage, but white matter disconnection may play a crucial role in some cases.

Introduction

The main objective of this paper is to examine the neural correlates of the principal acquired reading impairments in brain damaged patients. Although numerous studies have been conducted on these impairments and documented in detail in the literature, the neuroanatomical foundation of the cognitive behavior of the patients involved is not totally clear as studies were done principally on single-case studies, with only a few group studies being published.

In this section we will review the main cognitive aspects of the acquired reading impairments, while the following section will be dedicated to an assessment of the state of the art in the literature, with the objective of localizing neuropsychological symptoms and describing the functional anatomy of the normal cognitive functions, also in connection with some of the key findings deriving from functional neuroimaging studies. Finally, we will report the main methods employed for drawing inferences from the lesion data obtained from post-mortem analyses of the brain-damaged tissue to voxel-based lesion-symptom mapping.

Prior to the mid-1950s, the classical framework established by the French neurologist Joseph-Jules Dejerine was used to interpret acquired reading disorders. In 1891, he reported the case of a 63-year-old man who was afflicted by reading and spelling impairments (cécité verbale avec agraphie, i.e., verbal blindness with agraphia), but had no other language impairment and, in particular, no object-naming deficit. The post-mortem examination on this patient revealed cerebral damage to the left parietal lobe (including the angular gyrus). In 1892, he described the case of another brain-damaged patient with a reading impairment, but no associated spelling or oral language deficit (cécité verbale pure, i.e., pure alexia, also known as alexia without agraphia). The patient presented a right homonymous hemianopia but did not have any difficulty in naming either objects or colors. In this case the autopsy revealed an occipital and inferior temporal lesion extending to the retroventricular white matter, which had caused a functional disconnection of the trans-callosal pathways. Dejerine suggested that pure alexia was caused by left-hemisphere blindness (due to right hemianopia), right-hemisphere processing of the written stimuli and disconnection of this information from the intact left-hemisphere store of “optical images of letters and words” in the angular gyrus. As Coslett (2000) observed, Dejerine’s ground-breaking accounts of acquired dyslexia – although considered somewhat limited nowadays – are in some aspects the forerunners of contemporary cognitive psychological theories. In his pioneering model of reading, Dejerine assumed the existence of a written word-form area in the left angular gyrus while the right hemisphere is conceived to be word-blind, so that the visual images of words would have to reach the left angular gyrus to match stored orthographic representations. His anatomo-functional account of written language remained undisputed until the second half of the twentieth century (when it was contested by Marshall and Newcombe, 1966, Marshall and Newcombe, 1973), and continues to be considered as the major point of reference for the clinical description of isolated reading (or reading and writing) disorders after brain damage. However, Dejerine’s taxonomy does not account for several aspects of reading disorders that may occur in patients with acquired dyslexia following left-hemisphere lesions, left hemispherectomy or complete cerebral commissurotomy (split-brain patients). These aspects include the emergence of semantic, visual and morphological errors; grammatical class (nouns are read better than verbs and than function words), imageability and word frequency effects. Furthermore, his model cannot account for the dissociated ability to read irregular words or nonwords commonly found in dyslexic patients (surface dyslexia, phonological dyslexia and deep dyslexia).

The study of brain-damaged patients started to play a central role in cognitive neuropsychology with the series of seminal papers such as those by Marshall and Newcombe mentioned above (1966, 1973). These authors made a fundamental contribution to the study of acquired reading disorders with their dual-route processing model, proposed both from a psycholinguistic and a neurolinguistic perspective – see also Morton, 1969, Morton, 1980, Forster and Chambers, 1973, and Morton and Patterson (1980) – suggesting that reading is underpinned by two distinct cognitive procedures, i.e., the lexical and sublexical routes (see Coltheart, Curtis, Atkins, & Haller, 1993 for a review and a comparison with other models). In recent years several single-case and multiple single-case studies have been published, reporting clear-cut dissociations in support of dual-route reading models (e.g., Toraldo, Cattani, Zonca, Saletta, & Luzzatti, 2006). In addition, a computational realization of the dual-route theory of reading, known as the dual-route cascaded (DRC) model, was proposed by Coltheart, Rastle, Perry, Langdon, and Ziegler (2001). This model simulates a number of effects that other computational models of reading were unable to reproduce.

The main difference between Dejerine’s original model and the dual-route models of reading is that the latter provide for the existence of independent input and output orthographic representations, and the process of reading aloud is based on two independent pathways. After an initial visual analysis, letter strings would be processed by an orthographic recognition system, which specifies the abstract letter identity (i.e., non-dependent on letter case or font information) and the position of each letter within the target word. This orthographic information can be converted into the phonological word form by means of three routes of processing running in parallel. Firstly, letter strings are processed along the sub-word-level routine by means of grapheme-to-phoneme conversion rules. This is a serial procedure that can be successfully applied when reading regular words or nonwords but not with irregular words, since it would yield regularization errors. Secondly, words are read along the lexical-semantic route through a three-step procedure, from the orthographic input lexicon to the cognitive system and to the phonological output lexicon. The lexical route provides for successful reading of regular and irregular words and allows access to stored conceptual knowledge (Coltheart et al., 1993, Coltheart et al., 2001). This procedure is only suitable with words whose orthography is stored at lexical level, i.e. cannot process nonwords, and is the only procedure available when reading irregular words. Thirdly, Schwartz, Saffran, and Marin (1980), basing their assumption on the study of a patient who could name irregular words correctly but with no comprehension of their meaning, suggested a direct lexical pathway connecting the orthographic input lexicon and the phonological output lexicon, but bypassing the conceptual system.

It is worth noting that, in normal readers, the nonlexical route is activated in parallel with the lexical routes also when reading irregular words, and this interaction effect leads to longer RTs for low-frequency words (Paap and Noel, 1991, Seidenberg et al., 1984, Taraban and McClelland, 1987). Such interference manifests in the phonological buffer, which is fed by parallel activation of two conflicting phonological strings derived along either the lexical or the sub-word-level reading procedure, and this leads to an RT increase (see Coltheart et al., 2001, p. 221). The same line of reasoning is also valid for the lexical route, which is also activated when reading nonwords, thus leading to a facilitation effect in naming nonwords with large number of orthographic neighbors (Laxon et al., 1992, McCann and Besner, 1987, Peereman and Content, 1995; see Coltheart et al., 2001).

From a cognitive neuropsychological perspective, several cases characterized by dissociation (either classical, strong or trend dissociations, see Shallice, 1988) in their reading performance between irregular words and nonwords have been reported (see Hillis, 2008, Hillis, 2010, Lambon Ralph and Patterson, 2005 for reviews).

Surface dyslexia is the term used to denote selective damage in reading irregular words (e.g., “yacht”, “island”, “colonel”), despite preserved ability in reading both regular words and nonwords along the grapheme-to-phoneme conversion routine (e.g., Behrmann and Bub, 1992, Marshall and Newcombe, 1973, Shallice and Warrington, 1980, Shallice et al., 1983, Temple, 1985, Weekes and Coltheart, 1996 – see Coltheart, Masterson, Byng, Prior, & Riddoch, 1983 for a review). Furthermore, it has been observed that the performance of surface dyslexics might be highly variable both with regard to accuracy and to reading latencies. In opaque orthography languages such as English, the rate of successful reading of words with irregular orthography-to-phonology mapping is a direct measure of the integrity of the lexical route. In Italian and other shallow orthography languages, irregular words are virtually absent in reading, but the position of the major stress in three or more syllable words is ambiguous. This information is not predictable and not diacritically marked, so that it can only be accessed along the lexical and not the sublexical route. Surface dyslexia may be associated with fluent aphasia (as Wernicke’s aphasia, transcortical sensory or anomic aphasia), and it has often been found in conjunction with cases of semantic dementia (Funnell, 1996, Hodges et al., 1992, Jeffries et al., 2004, Patterson and Hodges, 1992, Shallice et al., 1983, Warrington, 1975, but see Blazely, Coltheart, & Casey, 2005 for a comparison of two patients with semantic dementia, one with and the other without surface dyslexia; see also Woollams, Lambon Ralph, Plaut, & Patterson, 2007 for a detailed review of the association between semantic dementia and surface dyslexia).

On the contrary, patients with phonological dyslexia (e.g., Beauvois and Dérouesné, 1979, Coltheart, 1996, Dérouesné and Beauvois, 1979, Saffran and Marin, 1977) cannot read nonwords but are still able to read both regular and irregular words. Phonological dyslexia might be also defined in terms of a strong dissociation between lexical and sublexical reading, performing better on the lexical route. This dissociation arises as damage to the grapheme-to-phoneme conversion procedure (classical phonological dyslexia) but a similar pattern of damage can also be the consequence of a phonological output buffer deficit (Bisiacchi, Cipolotti, & Denes, 1989). Phonological dyslexic patients may also perform poorly when reading aloud morphologically complex words (Funnell, 1983, Hamilton and Coslett, 2008, Patterson, 1982). Moreover, imageability and grammatical class effects, i.e., better performance on concrete nouns than on abstract nouns or function words, have been observed from the early days of research on this topic (see Coltheart, 1996, Marchand and Friedman, 2005). However not all phonological dyslexic patients show lexical effects in word reading (Friedman and Kohn, 1990, Funnell, 1983), and some patients are relatively impaired in reading function words (Glosser and Friedman, 1990, Patterson, 1982). Phonological dyslexia has been observed in mild aphasia (e.g., Dérouesné & Beauvois, 1979) but in general it is associated with severe nonfluent aphasia (e.g., Funnell, 1983) and has been also described in cases of Alzheimer disease (e.g., Caccappolo-van Vliet et al., 2004a, Caccappolo-van Vliet et al., 2004b) and of Primary Progressive Aphasia of the nonfluent type (Brambati, Ogar, Neuhaus, Miller, & Gorno-Tempini, 2009).

In 1980 Schwartz, Saffran and Marin observed that their patient WLP was able to read irregular words aloud even though she had no understanding of their meaning. This type of deficit, known as direct dyslexia, led to a revision of the dual-route model suggesting the need of a third reading procedure (direct non-semantic lexical route) connecting the orthographic input lexicon with the phonological output lexicon. Coslett (1991) provided additional support for the existence of a lexical but non-semantic reading procedure, but several authors have proposed other explanations (e.g., Hillis and Caramazza, 1991, McCarthy and Warrington, 1986, Shallice, 1988, Shallice et al., 1983). It is important to note that direct dyslexia has been described in cases of semantic dementia, much less frequently in focal brain-damaged patients.

Deep dyslexia (Coltheart et al., 1980, Marshall and Newcombe, 1973) is a peculiar type of phonological dyslexia (see Sato, Patterson, Fushimi, Maxim, & Bryan, 2008) in which semantic reading errors also occur (semantic paralexia, e.g., HOUND  “dog”; MERRY  “happy”). In addition to semantic errors, deep dyslexic patients usually produce morphological errors (e.g., omission/substitution of prefixes or suffixes, e.g., LOVELY  “love”). This pattern suggests damage to both the orthographic-to-phonological conversion mechanisms (impaired reading of nonwords) and the non-semantic lexical route. This reading pattern is also sensitive to imageability and grammatical class effects; indeed better performances are registered with concrete as opposed to abstract nouns, and nouns are found to be read better than verbs or function words (e.g., Baynes et al., 1998, Saffran and Marin, 1977). It has been suggested that semantic errors in deep dyslexia (see Coltheart, 1980b for an early review) may occur due to failure to inhibit the lexical phonological output representation of other candidates that are semantically related to the target word (Buchanan, Hildebrandt, & MacKinnon, 1994; Colangelo, Buchanan, & Westbury, 2004). When neurologically healthy readers name a single word (e.g., TIGER), this activates a number of words in the semantic system, at different levels (i.e., TIGER, but also LION and CAT). Nevertheless, in normal reading the target word determines a stronger activation towards the phonological output lexicon than the other candidates (which are only partially activated in the conceptual system and would be inhibited from activating the corresponding word form in the phonological output lexicon). This mechanism would be damaged in deep dyslexic patients, who would be unable to select the target word correctly, with respect to the competitors activated in the semantic system and so they produce semantic substitutions as a result. Age of acquisition of words has also been reported as playing a role in the occurrence of semantic errors (Gerhand & Barry, 2000). An interesting link between cognitive neuropsychological data and anatomical foundation of deep dyslexia was proposed by Coltheart, 1980a, Coltheart, 1983, Coltheart, 2000, who suggested that this impairment may result from extensive left-hemisphere perisylvian brain damage, and the consequent emergence of right-hemisphere residual reading abilities, which would be limited to high-frequency concrete nouns. With regard to reading processes, several neuropsychological and neuroimaging studies (e.g., Jones and Martin, 1985, Landis et al., 1983, Schweiger et al., 1989, Weekes et al., 1997) have challenged the classical claim that the right hemisphere is word-blind and completely devoid of lexical and semantic abilities (see Dejerine’s, 1892 interpretation of pure alexia).

Peripheral dyslexias are caused by functional damage upstream from the lexical and sublexical reading routes i.e., they are related to deficits in the visual analysis of letter strings. Such reading impairments include pure alexia (letter-by-letter dyslexia), attentional dyslexia, letter position dyslexia and neglect dyslexia. Only the first of these peripheral impairments will be considered in this study.

Pure alexia (letter-by-letter, LBL, dyslexia) is a peripheral reading impairment originally reported by Dejerine. The label LBL dyslexia has been used in a cognitive neuropsychological framework to describe a peripheral acquired reading deficit with dramatic length effect in which patients are unable to name target words either along the lexical or the sub-lexical route (e.g., Behrmann, Plaut, & Nelson, 1998; see also Binder and Mohr, 1992, Cohen et al., 2003). Some pure alexic patients may be able to name letters within a word, and in these patients reading time increases as the number of letters increases. Patients may be able to process the single letters of a word sequentially by means of a laborious letter naming process and they pronounce the entire word through a slow backward-spelling procedure (i.e., when reading DOG, patients may name each single letter and then reconstruct the entire word form reassembling the single letters starting from the first until they arrive at the final letter, and they recognize the target word from their own spoken output, e.g., DOG → “D .. O .. G … dog”).

At the present state of the art, the neuroanatomical bases of the central and peripheral dyslexias are far from being defined, since most of the studies are single-case descriptions, which do not give a clear and comprehensive picture of the anatomical correlates that are specific for each single reading impairment. Although the number of published neuroimaging studies investigating the anatomical correlates of letter recognition and of the two reading routes, there is no clear-cut evidence regarding which areas or area networks might be critically damaged in association with various reading impairments.

Surface dyslexia may be caused by either vascular lesions or neurodegenerative damage. In the first case left temporal lesions (e.g., Patterson and Behrmann, 1997, Temple, 1985), temporo-parietal lesions (e.g., Ferreres et al., 2005, Marshall and Newcombe, 1973, Temple, 1985), parieto-occipital lesions (e.g., Raman and Weekes, 2005, Weekes and Coltheart, 1996) have been reported in the literature, while with regard to neurodegenerative damage, it has been found mainly in semantic dementia, which is usually characterized by atrophy of the anterior temporal region (e.g., Breedin et al., 1994, Hodges et al., 1992, Patterson and Hodges, 1992, Wilson et al., 2009, Wilson et al., 2012).

No consistent information on lesion location has been provided yet for phonological dyslexia (see Lambon Ralph & Graham, 2000 for a review), though it is usually caused by large left fronto-parietal (Denes et al., 1987, Lesch and Martin, 1998) or fronto-temporo-parietal lesions (Goodglass and Budin, 1988, Hamilton and Coslett, 2008, Marchand and Friedman, 2005, Orpwood and Warrington, 1995). There have however been cases of more circumscribed lesions, of the frontal operculum in particular (Fiez and Petersen, 1998, Fiez et al., 2006) and other frontal lesions (Bradley and Thomson, 1984, Dérouesné and Beauvois, 1979; Ferreres, Lopez, & China, 2003). Other authors have reported temporal lesions (Coslett, 1991, Dérouesné and Beauvois, 1979, Sasanuma, 1996), fronto-temporal (Adair et al., 1999, Farah et al., 1996, Friedman, 1996, Kendall et al., 1998, Small et al., 1998), and temporo-parietal lesions (e.g., Friedman, 1996, Friedman and Kohn, 1990, Patterson et al., 1996, Sato et al., 2008). Damage to the superior temporal lobe, angular and supramarginal gyri lesions have been found in most (but not all) patients (Coslett, 2000, Lambon Ralph and Graham, 2000). Furthermore, phonological dyslexia has been observed after left hemispherectomy (Ogden, 1996), after right-sided putaminal hemorrhage (Harley & O’Mara, 2006), and in Alzheimer disease (Caccappolo-van Vliet et al., 2004a, Caccappolo-van Vliet et al., 2004b). Lastly, Rapcsak et al. (2009), working with a large cohort of focal brain-damaged patients suffering from phonological dyslexia, reported damage to a variety of perisylvian cortical regions, which were considered to be consistent with distributed network models of phonological processing.

Deep dyslexia (see Lambon, Ralph & Graham, 2000 for a review) is usually associated with large left fronto-temporo-parietal perisylvian lesions. For instance, Luzzatti, Mondini, and Semenza (2001) described the reading performance of an agrammatic deep dyslexic patient whose CT scan revealed a large left perisylvian frontoinsular lesion extending to the inferior parietal and the anterosuperior temporal areas. Left hemisphere perisylvian lesions causing extensive damage to the language areas and to the underlying white matter have also been reported in other studies (Beaton et al., 1997, Friedman and Perlman, 1982, Glosser and Friedman, 1990, Katz and Lanzoni, 1992, Laine et al., 1990, Nickels, 1992, Nolan and Caramazza, 1982, Price et al., 1998, Roeltgen, 1987, Ruiz et al., 1994, Saffran, 1980, Schweiger et al., 1989, Shallice and Coughlan, 1980, Silverberg et al., 1998). This neuroanatomical evidence is consistent with Coltheart’s account of deep dyslexia based on extensive destruction of the left-hemisphere language areas and the consequent emergence of residual right-hemisphere linguistic abilities (Coltheart, 1980a; see also Saffran, Bogyo, Schwartz, & Marin, 1980). However, deep dyslexia has also been reported in association with less extensive brain damage: fronto-temporal lesions (Balasubramanian, 1996), fronto-parietal lesions (Buchanan et al., 1994, De Bleser et al., 1995, Klein et al., 1994), predominantly parietal lesions (Warrington & Shallice, 1979), temporo-parietal lesions (Byng et al., 1984, Colangelo and Buchanan, 2005, Colangelo et al., 2004, Marshall and Newcombe, 1973, Price et al., 2003, Southwood and Chatterjee, 2001, Weekes et al., 1997), or predominantly subcortical lesions (Coslett et al., 1985, Davies and Cuetos, 2005).

As already mentioned, pure alexia is primarily associated with left occipital and inferior temporal damage, which is consistent with Dejerine’s account of this disorder as a disconnection of right-hemisphere visual analysis from the left angular gyrus store of orthographic lexical information. However, recent neuroimaging studies have suggested that the left inferior temporo-occipital cortex is the neuroanatomical substrate of orthographic processing (visual word form area, VWFA1; Cohen et al., 2000, Cohen et al., 2002, Cohen and Dehaene, 2004, Dehaene and Cohen, 2011, McCandliss et al., 2003). The VWFA would be a region of the visual cortex, which would code “the abstract sequence of letters that composes a written string” (Dehaene & Cohen, 2011, p. 254). This area would constitute “a special case of perceptual expertise, in which extensive visual experience with a class of stimuli drives enhancement of perceptual mechanisms and functional changes in the supporting left fusiform gyrus architecture” (Dehaene & Cohen, 2011, p. 296). Thus, lesions of the left inferior temporal and occipital lobes would cause a primary damage to orthographic processing and not a disconnection from the angular gyrus representations, as Dejerine proposed with his account of pure alexia. The authors also claim that, functionally, pure alexia – LBL dyslexia may not arise from damage to “the visual word-form area itself, but from the deployment of additional top-down processes of serial orientation for spatial attention, associated with activation of the posterior parietal cortex” (Dehaene & Cohen, 2011, p. 257). Clinical observations of certain dysgraphic patients after left inferior temporal and occipital lesions also assigned this area a role as the store of lexical-orthographic representations, both for reading and spelling processing (Rapcsak & Beeson, 2004). It is worth noting that pure alexia may also arise from a disconnection of the VWFA from the left occipital cortex (these two regions are mainly connected by the left inferior longitudinal fasciculus, see Catani & Thiebaut de Schotten, 2008) as in the clinical case reported by Epelbaum et al. (2008).

Several functional neuroimaging studies have investigated the neural correlates underlying the reading process of reading words and nonwords (see Brunswick, 2010, Fiez and Petersen, 1998, Mechelli et al., 2003, Price, 1997, Price and Mechelli, 2005, Price et al., 2003 for reviews; see also Jobard et al., 2003, Price, 2012, Turkeltaub et al., 2002 for meta-analyses). Generally, these studies have tried to disentangle the neural pathways involved in: (i) analyzing objects or non-orthographic visual stimuli; (ii) reading words by means of the lexical route; (iii) reading nonwords by means of the sub-word-level route.

According to a number of studies, after an initial processing stage common to words and nonwords (which has been functionally identified in the left occipito-temporal cortex) there are two relatively distinct anatomical pathways for the lexical and the sub-lexical route. The lexical route (which, from an anatomical point of view, has been also referred to as the “ventral orthographic route”) would be associated with the activation of a network of areas including the left occipito-temporal areas, the left inferior temporal cortex (Cohen et al., 2000), the left middle temporal gyrus (posterior portion) (Simos et al., 2002), where there would be areas devoted to the conceptual and semantic processing of words, and the pars triangularis of Broca’s area (Fiebach et al., 2002, Jobard et al., 2003). Grapheme-to-phoneme conversion, on the other hand, would be associated with the activation of a network of areas (in anatomical terms also referred to as the “dorsal phonological route”) involving the activity of the left supramarginal and angular gyrus (Jobard et al., 2003, Price, 1997, Simos et al., 2002), the superior temporal gyrus (Rumsey et al., 1997) and Broca’s area (pars opercularis) (Fiebach et al., 2002, Tagamets et al., 2000, Xu et al., 2001; see Ischebeck et al., 2004, Mechelli et al., 2003, Price, 2012 for a review). These results indicate that the cognitive mechanisms postulated by psychologists and psycholinguistic scholars would have a reliable anatomical counterpart in these ventral and dorsal pathways.

According to other authors, however, the data obtained so far do not support a clear-cut anatomical segregation of these functional pathways (Devlin et al., 2006, Fiebach et al., 2002, Ischebeck et al., 2004). Joubert et al. (2004) for instance, found that reading very high-frequency regular words leads to activation of the left angular and supramarginal gyri, thus supporting the hypothesis that these areas are involved in decoding whole-word representations. These authors also reported that both pseudowords and very low-frequency words lead to activation in the left inferior prefrontal gyrus, which correlates with activation in the left superior and middle temporal gyri. Attempts to localize the neural basis of the lexical and sublexical reading routes through neuroimaging studies would indicate that written words do not activate a specific lexical pathway but rather modulate activation in areas that are associated with semantic processing. Moreover pseudowords would not activate any specific sub-word-level pathway, but rather modulate activation in areas that are associated with phonological processing (Binder et al., 2003). It is worth noting that the anatomo-functional descriptions reported by functional neuroimaging studies challenge the neuroanatomical account proposed by certain lesion studies. For instance, Coslett et al. (1985) suggested that the sublexical route is mediated by a network of left perisylvian areas, whereas the activity of the lexical route is underpinned by left parietal or occipital areas. Rapcsak, Gonzalez Rothi, and Heilman (1987), describing a patient suffering from phonological dyslexia, suggested that the lexical route is underpinned by a dorsal pathway connecting the left inferior visual association areas, the left angular gyrus and Wernicke’s area, while the sublexical route would be mediated by a ventral route in the posterior–inferior portion of the left temporal lobe.

Although still relatively inconsistent, functional neuroimaging studies have revealed the existence of two cortical area networks, dedicated respectively to the processing of written information along the lexical and sublexical route. These networks are interconnected by a complex system of white matter tracts, whose role in reading has been clarified – at least partially – in recent years (see Catani & Thiebaut de Schotten, 2008 for a comprehensive anatomo-functional description of white matter tracts and Vandermosten et al., 2012a for a review of the role of white matter tracts in reading and developmental reading impairments). One of the problems encountered in identifying these tracts is how much they vary from individual to individual. Nevertheless, the neuroanatomical basis of the lexical and sublexical route has been recently studied by Diffusor Tensor Imaging (DTI) Tractography (e.g., Vandermosten, Boets, Wouters, & Ghesquière, 2012b). This is a structural MRI technique by which it is possible to reconstruct the integrity of three-dimensional white matter tracts and white matter properties. Reading, like speaking and spelling, implicates the interaction of a number of cortical areas, which must be functionally connected to carry out these tasks in the best way. For instance, it has also been suggested that developmental dyslexia might reflect poor connectivity between cortical areas due to white matter abnormalities (e.g., Pugh et al., 2000).

In their study of a group of 20 adults with typical reading ability and a second group of 20 developmental dyslexic adult readers using the DTI technique, Vandermosten et al. (2012a) found a correlational double dissociation, which might underpin a neuroanatomical basis for the dual route model of reading. In fact, the sublexical route seems to be founded on the activity of the left arcuate fasciculus, whereas the lexical route relies on the connections mediated by the left inferior occipito-frontal fasciculus. It has also been suggested that the left inferior longitudinal fasciculus mediates the connections underlying the lexical route (see Vandermosten et al., 2012b for a review).

The inferior occipito-frontal fasciculus is not easily identifiable as it runs parallel to the inferior longitudinal fasciculus, connecting the ventral portion of the occipital lobe with the orbitofrontal cortex (Catani & Thiebaut de Schotten, 2008). Recently, Martino, Brogna, Robles, Vergani, and Duffau (2010) suggested that the inferior occipito-frontal fasciculus might be sub-divided into two components: the first would connect the frontal lobe with the middle occipital gyrus and the second would connect the frontal lobe with the inferior occipital gyrus. The inferior longitudinal fasciculus on the other hand is a bundle of white matter lying in the central portion of the occipital and temporal lobe, connecting the former with the latter (Catani & Thiebaut de Schotten, 2008). This bundle originates in the extrastriate visual association areas, connecting them with the temporal lobe (both laterally and medially). The white matter tracts of the left inferior occipito-frontal fasciculus and left inferior longitudinal fasciculus play a very important role with regard to the reading processes; in fact, the former tract connects the VWFA anteriorly with the left inferior frontal cortex, whereas the latter connects it posteriorly with the left occipital lobe (Vandermosten et al., 2012b).

The arcuate fasciculus is part of the longitudinal fasciculus, which is the main white matter tract of the brain, connecting posterior with anterior cortical areas. The anatomy of the arcuate fasciculus is not immediately visible; Catani, Jones, & ffytche, 2005 studied it with the DTI technique and suggested that it is constituted of three different pathways: a medial tract, which joins Wernicke’s and Broca’s areas and two lateral tracts, one anterior and one posterior, connecting the inferior parietal lobule with Broca’s and Wernicke’s areas respectively.

Although functional neuroimaging studies – which measure brain activity in healthy subjects – are now in common use, anatomo-functional correlation continues to be a fundamental method for inferring the neural bases of cognitive functions (Rorden & Karnath, 2004). This method allows researchers to correlate the cortical and subcortical structures that have been damaged by a brain lesion with a patient’s symptoms. The lesion approach has its roots in the early decades of the 19th century (Bouillaud, 1825 – see Broca, 1861, Luzzatti and Whitaker, 2001). For more than one century the post-mortem approach was the only method for drawing inferences regarding the anatomical correlates of cognitive functions. However, since its introduction in clinical practice, computerized tomography (CT) and magnetic resonance imaging (MRI) have become the most important instruments for examining the morphological structure of the brain. These techniques have the added value of allowing the researcher to plan and conduct group studies, overlapping the lesions of patients with a similar cognitive deficit. In a group study, patients’ brain volumes are transformed to a standard stereotaxic space and lesions are mapped and overlaid in order to identify common areas of damage underlying a cognitive deficit (see Rorden & Karnath, 2004 for the main limitations of the lesion method and a comparison with the functional neuroimaging procedures).

The overlap method is useful for examining regions of brain damage in a relevant segment of a patients’ sample presenting a certain symptom. Nevertheless, this method has a significant limitation: overlay plots may highlight damaged areas with higher anatomical vulnerability (e.g., due to the vascular structure) and not those that are functionally most related to the disorder being examined. As suggested by Rorden and Karnath (2004), this problem can be solved by introducing a control sample (i.e., a group of neurological patients who do not exhibit the deficit being studied) in the experimental design. The overlay plot of the control group is then subtracted from the overlay plot of experimental group patients. The result will highlight the areas of the brain that are related to the deficit being studied through a functional relation, and not a mere anatomical contingency.

The voxel-based lesion-symptom mapping approach (VLSM, Bates et al., 2003) can also be used to assess whether the presence of a lesion can be considered a consistent predictor of a certain behavioral outcome. This method uses the same voxel-based procedure that is normally employed in the analysis of functional neuroimaging studies. The VLSM approach has three main advantages: (i) it represents the “gold standard” in terms of spatial resolution; (ii) it allows the researcher to draw statistical inferences on a voxel basis; (iii) it allows a comparison with the results of functional neuroimaging. VLSM divides the voxels within a lesion into two groups, damaged and undamaged; a statistical test such as a t-test, a Brunner–Munzel’s test for continuous measures or a Liebermeister test in the case of sample proportions (see Rorden, Karnath, & Bonilha, 2007) are then used to compare the two groups’ outcome behavioral measures.

The principal aim is to produce evidence as to which left-hemisphere brain areas are critically involved in the genesis of acquired central reading disorders and pure alexia (direct dyslexia, neglect dyslexia, letter position dyslexia and attentional dyslexia have not been considered in this study). We also correlated our neuroanatomical findings with normal processing of written words and nonwords. As mentioned in the introduction, the dual route model holds that reading is underpinned by two different and relatively encapsulated routes: the lexical and the sublexical one. However, the anatomical counterparts of the two routes are far from clear and another objective of the present study is to provide further information from the study of brain-damaged patients.

The study has focused on surface dyslexia, with the objective of ascertaining whether it is more probable that this reading impairment will occur after left temporal lesions (e.g., Patterson & Behrmann, 1997), left temporo-parietal lesions (e.g., Ferreres et al., 2005) or left parieto-occipital lesions (e.g., Raman & Weekes, 2005); it will also attempt to establish the role of the left anterior temporal areas in the origin of the reading impairment. These areas are typically damaged in semantic dementia (e.g., Wilson et al., 2012), but it might be not directly correlated with a defect of the lexical route, since conceptual (and not lexical) knowledge is the first to be damaged in this deficit: i.e., left anterior temporal damage in semantic dementia may be a poor indicator of the areas that are involved in processing words along the lexical route, since the deterioration of lexical representations is usually secondary to that of semantic knowledge. Furthermore, evidence from brain-damaged patients may help to ascertain and clarify the role of the areas of activation that have been identified by functional neuroimaging studies as the neuroanatomical correlate of the lexical route, i.e., the left occipito-temporal areas, the left inferior temporal cortex, the posterior portion of the left middle temporal gyrus and Broca’s area (pars triangularis).

No clear predictions or hypotheses can be put forward with respect to phonological dyslexia, as a wide number of critical left-hemisphere areas have been identified by previous researches, ranging from fronto-parietal lesions (e.g., Lesch & Martin, 1998), fronto-temporo-parietal lesions (e.g., Hamilton & Coslett, 2008), frontal lesions (e.g., Ferreres, López, & China, 2003), temporal lesions (e.g., Sasanuma, 1996), fronto-temporal lesions (e.g., Adair et al., 1999), or temporo-parietal lesions (e.g., Sato et al., 2008). Our hypothesis is that phonological dyslexia may arise after damage to a number of left perisylvian areas, which might constitute a neural network underlying the sub-word-level reading procedure. Therefore we also intend to ascertain the role of the cortical areas and the white matter tracts of this network in the pathogenesis of this reading impairment; the role of the left frontal operculum, which has been identified by Rapcsak et al. (2009) as playing a fundamental role in this network, also deserves particular attention and further investigation. In addition, functional neuroimaging studies suggest that sub-word-level processing might be based on a number of cortical areas (the left supramarginal gyrus, the left superior temporal gyrus and Broca’s area, pars opercularis) and connections between these areas. Finally, we raised the question as to whether the grammatical class and concreteness effects that often appear in phonological dyslexia might have a reliable anatomical counterpart. We hypothesize that patients showing these effects may have lesions involving specific sites of the sublexical network (and not simply larger lesions than those of patients that do not show such effects).

Deep dyslexia may also occur after lesions of a network of left perisylvian areas, as found in previous studies, which have reported either large perisylvian (e.g., Luzzatti et al., 2001), fronto-temporal (e.g., Balasubramanian, 1996), fronto-parietal (e.g., De Bleser et al., 1995) or temporo-parietal lesions (e.g., Colangelo & Buchanan, 2005).

With regard to pure alexia, in our opinion the two anatomo-functional frameworks described above lead to two different predictions with respect to the anatomical basis of the impairment. According to Dejerine’s original model (1892), both the “centre de mémoire visuelle des lettres” and the “centre de mémoire visuelle des mots” are stored in the left angular gyrus, and pure alexia is caused by left occipital lesions which also involve the retroventricular white matter, thus causing left hemisphere blindness and disconnecting the left angular gyrus from visual information projected to the intact right-hemisphere visual areas. On the contrary, the anatomo-functional account of the VWFA proposed by Cohen and Dehaene (e.g., Cohen & Dehaene, 2004) asserts that the left VWFA would constitute a functional specialization within the visual ventral stream, which would code the abstract sequence of letters that compose a written string. This predicts that pure alexia would occur after left occipital and inferior temporal lesions or from disconnection of the left occipital pole from the occipito-temporal sulcus (as in the patient described by Epelbaum et al., 2008). Contrary to Dejerine’s original proposal, the retroventricular white matter connecting the visual information projected to the right hemisphere with the left angular gyrus would not play a crucial role in the genesis of pure alexia.

Section snippets

Participants

The participants were recruited from a number of focal left-hemisphere brain-damaged patients consecutively admitted to three Rehabilitation Units in Northern Italy, some of whom had been discharged at time of testing, and were being examined for their reading impairment at a minimum of one month post-stroke (median = 4 months). All were able to give their informed consent to participate in the research project and to understand task instructions at the time of testing. The inclusion criteria for

Results

The behavioral results obtained on the aphasic sample with the reading task are reported in the first subsection; the following subsections describe the findings emerging from the neuroimaging study, according to the methodological approach used to answer experimental questions. Firstly lesions were overlapped, clustering patients by type of reading impairment (surface dyslexia, phonological dyslexia, deep dyslexia, undifferentiated dyslexia and pure alexia). Although this approach has a major

Discussion

The principal objective of this study was to identify the neuroanatomical bases of acquired reading disorders and the neural substrate of the lexical and sublexical routes of reading. In the past, the functional neuroanatomy of the written language has been mainly investigated through single-case studies, which however have not provided a clear anatomical picture or evidence that clearly converges with that obtained from functional neuroimaging studies. This paper adopted an anatomo-correlative

Conclusion

The results of this study indicate that surface dyslexia has clear neuroanatomical correlates including the left middle occipital gyrus, the left superior, middle and inferior temporal gyrus, the left insula (posterior and inferior portions), with involvement of the inferior occipito-frontal fasciculus. These areas and this fasciculus constitute an anatomo-functional network underlying the activity of the lexical route. The role of the inferior occipito-frontal fasciculus in the lexical route

Acknowledgments

We wish to thank Otto Karnath, Olufunsho Faseyitan and Carolyn Wilshire for their insightful comments and helpful suggestions on a previous draft of this manuscript. Preliminary results of this study were presented at the 50th Annual Meeting of the Academy of Aphasia (San Francisco, 2012). The study was supported by a FAR grant of the University of Milano-Bicocca and by a grant of Finlombarda (ASTIL 2010) to C.L.

References (171)

  • H.B. Coslett et al.

    Reading: Dissociation of the lexical and phonologic mechanisms

    Brain and Language

    (1985)
  • R. De Bleser et al.

    Rapid recovery of aphasia and deep dyslexia after cerebrovascular left-hemisphere damage in childhood

    Journal of Neurolinguistics

    (1995)
  • S. Dehaene et al.

    The unique role of the visual word form area in reading

    Trends in Cognitive Sciences

    (2011)
  • S. Eickhoff et al.

    A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data

    Neuroimage

    (2005)
  • S. Epelbaum et al.

    Pure alexia as a disconnection syndrome: New diffusion imaging evidence for an old concept

    Cortex

    (2008)
  • A.R. Ferreres et al.

    Phonological alexia with vowel–consonant dissociation in non-word reading

    Brain and Language

    (2003)
  • J.A. Fiez et al.

    Specific reading and phonological processing deficits are associated with damage to the left frontal operculum

    Cortex

    (2006)
  • K.I. Forster et al.

    Lexical access and naming time

    Journal of Verbal Learning and Verbal Behavior

    (1973)
  • R.B. Friedman et al.

    Impaired activation of the phonological lexicon: Effects upon oral reading

    Brain and Language

    (1990)
  • R.B. Friedman et al.

    On the underlying causes of semantic paralexias in a patient with deep dyslexia

    Neuropsychologia

    (1982)
  • N. Friedmann et al.

    Developmental surface dyslexias

    Cortex

    (2008)
  • C.R. Genovese et al.

    Thresholding of statistical maps in functional neuroimaging using the false discovery rate

    Neuroimage

    (2002)
  • S. Gerhand et al.

    When does a deep dyslexic make a semantic error? The roles of age of acquisition, concreteness, and frequency

    Brain and Language

    (2000)
  • G. Glosser et al.

    The continuum of deep/phonological dyslexia

    Cortex

    (1990)
  • H. Goodglass et al.

    Category and modality specific dissociations in word comprehension and concurrent phonological dyslexia

    Neuropsychologia

    (1988)
  • J.R. Hanley et al.

    Reading speed in pure alexia

    Neuropsychologia

    (1996)
  • A.E. Hillis et al.

    Mechanisms for accessing lexical representations for output: Evidence from a specific semantic deficit

    Brain and Language

    (1991)
  • A.E. Hillis

    Cognitive processes underlying reading and writing and their neural substrates

  • G. Jobard et al.

    Evaluation of the dual route theory of reading: A metanalysis of 35 neuroimaging studies

    Neuroimage

    (2003)
  • G.V. Jones et al.

    Deep dyslexia and the right-hemisphere hypothesis for semantic paralexia: A reply to Marshall and Patterson

    Neuropsychologia

    (1985)
  • S. Joubert et al.

    Neural correlates of lexical and sublexical processing in reading

    Brain and Language

    (2004)
  • M. Laine et al.

    Semantic errors in a deep dyslexic

    Brain and Language

    (1990)
  • T. Landis et al.

    Semantic paralexia: A release of right hemisphere function from left hemisphere control

    Neuropsychologia

    (1983)
  • C. Luzzatti et al.

    Lexical representation and processing of morphologically complex words: Evidence from the reading performance of an Italian agrammatic patient

    Brain and Language

    (2001)
  • J.C. Adair et al.

    Articulatory processes and phonologic dyslexia

    Neuropsychiatry Neuropsychology and Behavioural Neurology

    (1999)
  • V. Balasubramanian

    Deep dyslexia and dysgraphia in a Broca’s aphasic

    Brain and Language

    (1996)
  • E. Bates et al.

    Voxel-based lesion-symptom mapping

    Nature Neuroscience

    (2003)
  • K. Baynes et al.

    Modular organization of cognitive systems masked by interhemispheric integration

    Science

    (1998)
  • A. Beaton et al.

    Semantic errors of naming, reading, writing, and drawing following left-hemisphere infarction

    Cognitive Neuropsychology

    (1997)
  • M.F. Beauvois et al.

    Phonological alexia: Three dissociations

    Journal of Neurology, Neurosurgery and Psychiatry

    (1979)
  • M. Behrmann et al.

    Surface dyslexia and dysgraphia: Dual routes, a single lexicon

    Cognitive Neuropsychology

    (1992)
  • M. Behrmann et al.

    A review and new data supporting an interactive account of letter-by-letter reading

    Cognitive Neuropsychology

    (1998)
  • J.R. Binder et al.

    Neural correlates of lexical access during visual word recognition

    Journal of Cognitive Neuroscience

    (2003)
  • J.R. Binder et al.

    The topography of callosal reading pathways: A case control analysis

    Brain

    (1992)
  • P.S. Bisiacchi et al.

    Impairment in processing meaningless verbal material in several modalities: the relationship between short term memory and phonological skills

    The Quarterly Journal of Experimental Psychology

    (1989)
  • A. Blazely et al.

    Semantic impairment with and without surface dyslexia: Implications for models of reading

    Cognitive Neuropsychology

    (2005)
  • J.B. Bouillaud

    Recherches cliniques propres à démontrer que la perte de la parole correspond à la lésion des lobules antérieurs du cerveau et à confirmer l’opinion de M. Gall, sur le siège de l’organe du langage articulé

    Archives Génerales de Médicine

    (1825)
  • S.D. Breedin et al.

    Reversal of the concreteness effect in a patient with semantic dementia

    Cognitive Neuropsychology

    (1994)
  • P. Broca

    Perte de la parole, ramollissement chronique et destruction partielle du lobe antérieur gauche du cerveau

    Bulletin de la Société Anthropologique

    (1861)
  • E. Brunner et al.

    The nonparametric Behrens-Fisher problem: Asymptotic theory and small-sample approximation

    Biometrical Journal

    (2000)
  • Cited by (0)

    View full text