Regular articleEvaluation of the dual route theory of reading: a metanalysis of 35 neuroimaging studies
Introduction
Despite the large number of neuroimaging studies interested in unraveling brain organization sustaining reading, acquiring a clear picture of cerebral areas involved in visual word access and the specific cognitive processes thought to take place in these regions remains quite a tricky enterprise. The difficulty of such a task arises from many combined elements that contribute to mask the picture one can get by simply looking at the results that are already published.
A first element is the lack of explicit definition of the theoretical framework in which studies are undertaken. This is of prime importance in the interpretation of activation results, since theoretical concepts and the suspected necessary cognitive processes may vary from one study to another. Models of dual route theory provide a framework commonly used in studies of reading. This theory develops the view that word reading can be achieved through two distinct routes, relying on discrete processes (Fig. 1). The graphophonological route, also called indirect route, requires visual words to be transformed into their auditory counterparts, thanks to the application of grapheme-to-phoneme correspondences. The pronunciation of words being available, subjects can then access their meanings. These grapheme-to-phoneme correspondences are however more or less univocal according to the language considered and its degree of transparency: while the same group of letters will invariably lead to the same pronunciation in Spanish, some letter combinations in English will be read alouddifferently, depending on the word in which they are used (e.g., “int” like in “pint” or “mint”). To read correctly those words that do not follow the most usual spelling-to-sound rules, it is necessary to engage another route, called direct or lexicosemantic route. This route relies on the direct association between the visual forms of words (in much the same way as images) and their meanings. This association is thought to build progressively when subjects repetitively meet words, and the direct route is therefore used when subjects read frequent or orthographically irregular words (in which case the association has to be learned, since grapheme-to-phoneme correspondences (GPC) application cannot lead to the proper pronunciation). Learned word forms are then supposed to be stored in a visual word form system, also called orthographical lexicon, whose size and composition would depend on the subject's reading experience. The dual route model originates from the observation of patients that, following brain injury, relied exclusively on one of the two routes for reading. Surface dyslexia, characterized by the impairment of the lexicosemantic route and reading by the graphophonological route, typically leads to regularization errors (e.g., reading “pint” to rhyme with “mint”), homophones confusions (e.g., understand “pear” after reading “pair”), or to accept pseudowords homophones as real words (they accept “rite” as a correct word for “right”). The opposite syndrome, phonological dyslexia, leaves the direct route as the only possible access to written words meanings and leads to difficulties in reading words for which no form–meaning association had previously been established (infrequent words, pseudowords, new names, and also grammatical words, which are devoid of invariant meanings). Despite a general agreement on the processes involved in the two routes, some theoretical variations nevertheless subsist in the conception of the first stages of word access. We developed a conception of dual route models stating the existence of a visual word form system dedicated to the direct route only, which is concordant with the view of some authors Morton, et al., 1980, Coltheart et al., 1993. A concurrent view posits that the visual word form system would be common to both routes and that information would then be relayed to the most appropriate route for the word read (Warrington and Shallice, 1980). According to this latest view, the visual word form system would therefore process not only known words as a whole, but also other infralexical stimuli such as graphemes, syllables, or morphemes. Warrington and Shallice indeed considered that this system would play a role in the “parsing of letter strings into ordered familiar units and the visual categorization of these units” (Warrington and Shallice, 1980, p. 109). According to this view, when a word is presented, its visual grasping could be effected in different ways, depending on the experience of the reader. If this word has been previously seen a sufficient number of times, it may be globally recognized as a prelearned word form (as in the case of an access through the direct, lexicosemantic route). If this is not the case, the reader may then begin to segment the word into sublexical parts, corresponding either to letter combinations he has visually learned because they occur often in his language (such as the unit “tion” that is to be found in many English words) or to simple graphemes (that sometimes need several letters to be considered as a single unit, as with the grapheme “ch”). The main difference thus arises from the fact that this later view would not predict the existence of a specialization for the visual processing of known words, as is done with the concept of “written word lexicon,” but rather a specialization for alphabetical stimuli.
Despite these last theoretical variations on the first processing stages, we decided to insert this metanalysis in the framework of the dual route model of word reading. This model first emerged from cognitive neuropsychological observations of brain-damaged patients (taking therefore cerebral constraints into account), and it has found some support in experimental psychology in normal readers (Coltheart and Rastle, 1994). Additionally, we assisted very recently to a burst of neuroimaging experiments reporting the involvement of different brain areas depending on the specific route thought to underlie word access Herbster et al., 1997, Rumsey et al., 1997, Fiez, 1997, Simos et al., 2002, even though some controversies remain as to the nature of the processes undertaken in some brain regions. It is of prime importance to state that this article does not aim at assessing the appropriateness of this model in comparison to others such as connectionist models, but rather to see how it can accommodate to neuroimaging results.
Besides the possible discrepancies stated above and introduced by theoretical frameworks, differences in sensitivity may also arise out of methodological or technical variations from one laboratory to another, leading to a reduced set of activated regions which can drive authors to neglect the roles of other brain areas. More importantly, the use of varying anatomical labels among studies may also contribute to complicate a general overview of the literature, since it can induce readers to conclude that different brain areas are implicated despite activations that are congruent in terms of spatial coordinates.
The goal of this metanalysis was to clarify results published in the reading literature and to determine (a) whether there exists a system dedicated to the processing of visual word form as postulated by the dual route models, and (b) whether neuroimaging results support the possibility of two distinct routes for accessing words. To answer these questions, we performed an analysis that was as observer independent as possible, based on the core exploitation of coordinates related to cognitive contrasts obtained in 35 different neuroimaging studies. This analysis relied on the use of an automated spatial segregation of activation peaks coordinates, coupled with an automated anatomical labeling of the spatially congruent activations in the Montreal Neurological Institute stereotactic space (Tzourio-Mazoyer et al., 2002).
Section snippets
Raw data
The raw data of this metanalysis was constituted by activation peak coordinates reported in 35 neuroimaging studies (using PET and fMRI exclusively, see Table 1) and obtained in contrasts, implying the reading of words or pseudowords. Only studies ranging from years of publication 1990 to 2002 were used, in which stereotactic coordinates were made available exclusively for tasks submitted to nonpathological subjects. These activation tasks were either directly compared to each other or
Results
Thirty-five clusters were located in the left hemisphere; 15 were in the right hemisphere and 5 clusters were constituted by isolated peaks dispersed in the brain with a high spatial standard deviation (many of these peaks were located in white matter and some even fell out of the brain). The aim of this work was to provide a global view of the results published in the last few years concerning two main controversies related to the dual route of reading framework. For this reason, we chose to
Discussion
This method of spatial segregation resulted in a set of clusters of reliable activations that corresponded in large parts to structures previously described as playing a key role in reading. A first question that we wished to investigate in the light of this method of review concerned the possible specialization of a brain area for the processing of visual words.
Conclusions
This metanalysis based on the gathering and spatial segregation of activation peaks obtained in reading studies revealed an overall reliability of published results. Despite methodological and experimental differences inherent to the comparison of studies issued in worldwide laboratories, a consensus can be drawn within the technique's spatial limitation as to which sites are critical for identified processes.
The main result consists of the demonstrated suitability of the dual route model
Acknowledgements
We thank Laure Zago and Emmanuel Mellet for thoughtful comments on the manuscript. We are grateful to the two anonymous referees, the constructive comments of which allowed us to substantially improve the manuscript. This work was presented in part at the 8th International Conference on Functional Mapping of the Human Brain, June 2–6 2002, Sendai, Japan.
References (72)
- et al.
A comparison of neural circuits underlying auditory and visual object categorization
NeuroImage
(2002) - et al.
Cortical mechanisms specific to explicit visual object recognition
Neuron
(2001) - et al.
Functional anatomy of intra- and cross-modal lexical tasks
NeuroImage
(2002) - et al.
The effect of semantic category and knowledge type on lexial-semantic accessa PET study
NeuroImage
(1998) - et al.
Effects of lexicality, frequency, and spelling to sound consistency on the functionnal anatomy of reading
Neuron
(1999) - et al.
Equivalent responses to lexical and nonlexical visual stimuli in occipital cortexa functional magnetic resonance imaging study
NeuroImage
(1997) - et al.
Brain activation during word identification and word recognition
NeuroImage
(1998) - et al.
Reopening the mental imagery debatelessons from functional anatomy
NeuroImage
(1998) - et al.
Encoding words and picturesa positron emission tomography study
Neuropsychologia
(1996) - et al.
Three distinct ventral occipitotemporal regions for reading and object naming
NeuroImage
(1999)
Functional specialization for semantic and phonological processing in the left inferior prefrontal cortex
NeuroImage
The effect of varying stimulus rate and duration on brain activity during reading
NeuroImage
Different cortical activity in reading of kanji words, kana words and kana nonwords
Cognitive Brain Res.
Left posterior BA37 is involved in object recognitiona TMS study
Neuropsychologia
Localization of syntactic comprehension by positron emission tomography
Brain Lang.
A neural basis for category and modality specific of semantic knowledge
Neuropsychologia
Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain
NeuroImage
The neural substrates for concrete, abstract, and emotional word lexicaa positron emission tomography
J. Cog. Neurosci.
A comment on the functional localization of the phonological storage subsystem of working memory
Brain Congn.
The topography of callosal reading pathwaysA case-control analysis
Brain
Regional cerebral blood flow during object naming and word reading
HBM
Regional cerebral blood flow during auditory responsive namingevidence for cross-modality neural activation
NeuroReport
The problem of functional localization in the human brain
Nat. Rev. Neurosci.
A multimodal language region in the ventral visual pathway
Nature
Auditory and visual word processing studied with fMRI
Human Brain Mapping
The visual word form areaspatial and temporal characterization of an initial stage of reading in normal subjects and posterior split brain patients
Brain
Language-specific tuning of visual cortex? Functional properties of the Visual Word Form Area
Brain
Models of reading alouddual-route and parallel-distributed-processing approaches
Psych. Rev.
The visual word form areaa prelexical representation of visual words in the fusiform gyrus
NeuroReport
cerebral mechanisms of word masking and unconcious repetition priming
Nature
fMRI evidence for dual routes to the mental lexicon in visual word recognition
J. Cogn. Neurosci.
Phonology, semantics and the role of the left interior prefrontal cortex
Hum. Brain. Mapp.
A positron emission tomography study of the short-term maintenance of verbal information
J. Neurosci.
The fusiform ‘face area’ is part of a network that processes face at the individual level
J. Cogn. Neurosci.
Cited by (705)
Unraveling individual differences in learning potential: A dynamic framework for the case of reading development
2024, Developmental Cognitive NeuroscienceBrain bases of morphological awareness and longitudinal word reading outcomes
2024, Journal of Experimental Child PsychologyThe neural correlates of sub-lexical semantics and its integration with the lexical meaning in reading Chinese characters
2024, Journal of NeurolinguisticsAn eye movement study on the mechanisms of reading fluency development
2024, Cognitive Development