| |
Alexia, or acquired impairment of reading, is extremely common after stroke, dementia, or traumatic brain injury. Reading can be affected in a variety of different ways, leading to a number of different clinical syndromes or types of alexia. To understand these alexic syndromes, it is necessary to appreciate the cognitive processes underlying the task of reading words. Reading aloud a familiar word, such as leopard, normally entails at the very least seeing and perceiving the entire written letter string, recognizing the word as a known word (by accessing a stored representation of the learned spelling of the word, in the orthographic input lexicon), accessing its meaning (or semantic representation), and accessing the pronunciation (the stored sound of the word, in the phonological output lexicon), as well as activating motor speech mechanisms for articulating the word. Even the first of these components, seeing and perceiving the entire written letter string, requires complex visual-perceptual skills, including computation of several levels of spatial representation, before the stored representations can be accessed. Furthermore, reading an unfamiliar word—say, an unfamiliar surname—entails access to print-to-sound conversion, or grapheme-to-phoneme conversion (GPC), mechanisms. Familiar words can also be read via GPC mechanisms, but the accuracy of pronunciation will depend on the “regularity” of the word—the extent to which the word conforms to typical GPC rules. For example, sail but not yacht can be read accurately via GPC mechanisms.
These components underlying the reading process are schematically depicted in Figure 1 (see Hillis and Caramazza, 1992, or Hillis, 2002, for a review of the evidence for various components of this model). Various features of this model are controversial, such as the precise nature and arrangement of the components, the degree to which they interact, and whether the various levels of representation are accessed in parallel or serially (Shallice, 1988; Hillis and Caramazza, 1991; Plaut and Shallice, 1993; Hillis, 2002). Nevertheless, most models of naming include most of the components depicted in Figure 1. Neurological impairment can selectively impair any one or more of these components of the reading process, with different consequences in terms of the types of errors produced and the types of stimuli that are affected. In addition, because several of the components of reading are cognitive mechanisms that are also involved in other tasks, damage to one of these shared components will have predictable consequences for reading and other tasks. For example, impairment at the level of semantics, or word meaning, will affect not only the reading of familiar irregular words but also the naming and comprehension of words. Thus, it is possible to identify what component of the reading system is impaired by considering the types of errors produced by the individual, the types of words that elicit errors, and the accuracy of performance across other language tasks, such as spoken naming and comprehension. The consequences of damage to each level of the reading process are discussed here.
Figure 1..
The cognitive processes underlying reading.
Visual Attention and Perception
To read a familiar or unfamiliar word, the string of letters must be accurately perceived in the correct order and converted to a series of graphemes (abstract letter identities, without a particular case or font). Perception of the printed word can break down when there is (1) poor visual acuity or visual field cut, (2) impairment in distinguishing individual letters or symbols in a string (attentional dyslexia; Shallice, 1988), or (3) impairment in perception of more than one object or feature at a time (called simultagnosia; Parkin, 1996). An individual with simultagnosia might read the word chair as “h” or “i”. Finally, individuals with damage to the nondominant (usually right) hemisphere of the brain often fail to perceive the side of a visual stimulus like a word that is contralateral to the brain damage. Such an impairment, known as neglect dyslexia, results in reading errors such as chair → “fair,” spool → “pool,” and love → “glove,” errors that entail substitution, deletion, or insertion of letters on the left side (or initial letters) of words (Kinsbourne and Warrington, 1962; see papers in Riddoch, 1991, for reviews). Depending on the level of spatial representation affected, reading accuracy is sometimes improved by moving the printed word to the unaffected side of space or by spelling the word aloud to the person (see Hillis and Caramazza, 1995, for a discussion of various types of neglect dyslexia resulting from damage to distinct levels of spatial representation that are computed prior to accessing a stored graphemic representation). All types of reading stimuli are likely to be affected in neglect dyslexia, although words that have final letter strings in common with other words often elicit the most errors. For example, the words light, fight, might, right, tight, sight, blight, bright, slight, and so on are all likely to be read as the same word, since only the final letters are perceived and used to access the stored graphemic representation for recognition. Pseudo-words also elicit comparable errors (e.g., glamp → “lamp” or “damp”). An individual with impairment in computing one or more levels of spatial representation will usually also make errors in perceiving the left side of nonlinguistic visual stimuli (Hillis and Caramazza, 1995), although exceptional cases of pure neglect dyslexia, without other features of hemispatial neglect, have been reported (Costello and Warrington, 1987; Patterson and Wilson, 1990).
Orthographic Input Lexicon
Impairment at the level of accessing learned spellings of familiar words, or stored graphemic representations that constitute the “orthographic input lexicon,” results in impaired recognition of written words despite accurate perception of the letters. The individual with this impairment will often fail to distinguish familiar from unfamiliar words, or pseudo-words (e.g., glamp), in a task known as lexical decision. Sometimes such an individual will read each letter in the string aloud serially, which seems to facilitate access to the orthographic input lexicon (resulting in letter-by-letter reading; see papers in Coltheart, 1998). If GPC mechanisms are intact, these mechanisms may be used to read even familiar words, resulting in “regularization” of irregular words (e.g., one → “own”). Other errors are predominantly visually similar words (e.g., though → “touch”). Oral reading of all types of words may be affected, although very familiar words—those frequently encountered in reading—may be relatively spared. Since the orthographic input lexicon is not involved in other linguistic tasks, damage to this cognitive process does not cause errors in other tasks. Therefore, individuals with impairment of this component are said to have pure alexia.
Semantic System
Disruption of semantic representations is often incomplete, such that the meanings that are accessed are often impoverished, and only certain categories of words are affected. For example, an alexic patient may read dog as “cat” if an incomplete semantic representation of dog is accessed that specifies only 〈animal〉, or 〈mammal〉, 〈domesticated〉, 〈quadraped〉, etc., without information about what differentiates a dog from a cat. Thus, most errors are semantically related words, such as robin → “cardinal” or robin → “bird” (errors called semantic paralexias). However, if GPC mechanisms are available, these may be used to read all types of words, or used to block semantic paralexias. Or GPC mechanisms may be combined with partial semantic information to access the correct phonological representation in the output lexicon, so that the individual can read aloud words better than he or she can understand words (Hillis and Caramazza, 1991). Often there is especially incomplete semantic information to distinguish abstract words, so that abstract words are read less accurately than concrete words. Similarly, functors are read least accurately, often with one functor substituted for another (e.g., therefore → “because”); verbs are read less accurately than adjectives; and adjectives are read less accurately than nouns (Coltheart, Patterson, and Marshall, 1980). If GPC mechanisms are also impaired (a commonly co-occurring deficit), pseudo-words and unfamiliar words cannot be read. Since the semantic system is shared by the tasks of naming and comprehension, semantic errors are also made in oral and written naming and in comprehension of spoken and written words (Hillis et al., 1990).
Phonological Output Lexicon
Impairment in accessing phonological representations for output results in poor oral reading despite accurate comprehension of printed words. For example, gray might be read as “blue” but defined as “the color of hair when you get old” (from Caramazza and Hillis, 1990). Again, if GPC mechanisms are available, these mechanisms may be used to read aloud all types of words, resulting in regularization errors on irregular words. Otherwise, errors may be semantically related (e.g., fork → “spoon”), or phonologically related to the target (e.g., choir → “queer”). Phonological representations of words that are used more frequently may be more accessible than other words, so that high-frequency words are read more accurately than low-frequency words. Since the phonological output lexicon is also essential for oral naming and spontaneous speech, the person will make similar errors on these tasks as in reading.
| |
