A Test Case for Reading Models

Following brain damage, skilled readers may encounter more severe problems in reading nonwords than familiar words, a type of deficit referred to as phonological dyslexia. We report on 2 individuals with Alzheimer’s disease who show phonological dyslexia. Although highly accurate in reading familiar words aloud (even those with irregular spelling, such as sew), they were quite impaired in nonword reading. Both patients performed well in phonological tasks involving the repetition, identification, and manipulation of phonemes of orally presented words and nonwords. These results challenge the idea, proposed in the context of connectionist and evolutionary theories, that phonological dyslexia originates from a phonological deficit. However, the results are consistent with reading models, such as the dual-route model, that attribute phonological dyslexia to a deficit that selectively affects the reading mechanisms responsible for deriving the sounds of nonwords. According to these models, such a deficit is not necessarily accompanied by a more general phonological impairment. The term phonological dyslexia is used by neuropsychologists to describe reading deficits that affect nonwords (nep, cabe) more severely than familiar words. This disturbance appears in skilled adult readers following cortical brain damage; a developmental form has also been reported in children who have no apparent cortical lesions (e.g., Temple & Marshall, 1983). The understanding of phonological dyslexia not only has obvious clinical consequences, but also has implications for the current debate on reading processing. Current models of reading offer different accounts of phonological dyslexia. Because a better knowledge of this deficit can severely constrain such models, it is of primary theoretical significance. The dual-route model proposes that two types of mechanisms, which are in part neuroanatomically distinct, support reading aloud (see Coltheart, Rastle, Perry, Ziegler, & Langdon, 2001, for a recent instantiation of this account and a discussion of its variants). One series of mechanisms, the lexical route, is implicated in the retrieval of stored information about the orthography, semantics, and phonology of familiar words. An alternate route, the nonlexical route, allows readers to derive the sounds of written words by means of mechanisms that convert letters or letter clusters into their corresponding sounds. The nonlexical route is functionally limited in that it does not provide information about word meaning; nor, in a language like English or Italian, does it guarantee the correct pronunciation of a number of words. Nevertheless, the nonlexical route is responsible for deriving the sounds of nonwords; its selective damage would result in phonological dyslexia (Berndt, Haendiges, Mitchum, & Wayland, 1996; Coltheart, 1985; Derousné & Beauvois, 1985). Another class of models, which we refer to as ‘‘triangle models,’’ offers a different account of phonological dyslexia. According to these models, reading aloud depends on the joint processing of mechanisms that translate orthography into phonology and mechanisms that bind word meaning and phonology. This type of architecture has been proposed in several connectionist models (e.g., Plaut, McClelland, Seidenberg, & Patterson, 1996; Seidenberg & McClelland, 1989). Although differing on a number of details, the various triangle models all share the assumption that identical processes support the reading of words and nonwords. That is, in contrast to the dual-route model, they do not propose any mechanisms specifically involved in the processing of nonwords. Within the framework of triangle models, the cause of phonological dyslexia is assumed to be an impairment in the representation of phonological information (Friedman, 1996; Harm & Seidenberg, 2001; Patterson, 2000; Plaut et al., 1996). Various factors conspire to make nonwords more vulnerable to a phonological impairment than familiar words are. Because nonwords depend more on orthography-phonology mapping, have less stable phonological representations, and do not benefit from the collateral support of semantics, conditions that alter the representation of phonological information are expected to have sizable effects on nonword reading. These models further propose that when phonological impairment is mild, only nonword reading should be affected, thereby accounting for pure cases of phonological dyslexia in which reading of familiar words is spared (e.g., Beauvois & Derousné, 1979; Funnell, 1983; Shallice & Warrington, 1980). We refer to this explanation of phonological dyslexia as the phonological-impairment hypothesis. Farah, Stowe, and Levinson (1996) also appealed to the phonological-impairment hypothesis, although from different theoretical premises. Essentially, their idea was that because reading is, from an evolutionary perspective, a recently acquired function, it cannot Address correspondence to Michele Miozzo, Department of Psychology, Columbia University, 1190 Amsterdam Ave., New York, NY 10027; e-mail: [email protected]. PSYCHOLOGICAL SCIENCE Volume 15—Number 9 583 Copyright r 2004 American Psychological Society depend on specialized brain regions. Consequently, brain damage cannot cause selective reading deficits. However, damage to cognitive functions with dedicated brain regions (e.g., vision or phonology) can affect reading. In particular, phonological deficits should have selective effects on nonword reading essentially for the reasons already explained: because nonwords require additional phonological processing for word sounds to be assembled. Proponents of the dual-route model do not deny that phonological deficits could affect reading, particularly nonword reading. A point of divergence concerns whether such deficits should invariably accompany this form of dyslexia, as proposed by the phonological-impairment hypothesis, or whether other factors can also yield phonological dyslexia, as proposed by the dual-route model. It is striking that patients with acquired phonological dyslexia almost invariably fail in a range of phonological tests, including, for example, nonword repetition and phoneme-manipulation tasks (e.g., ‘‘repeat the word sharp without the initial phoneme’’). This association of deficits lends support to the phonological-deficit hypothesis. However, there have also been individuals with acquired phonological dyslexia who have performed well on phonological tasks, although these cases have been reported much more rarely. In fact, only 2 or 3 such patients have been reported. One such individual is LB (Derousné & Beauvois, 1985), a Frenchspeaking patient who, for example, could assemble nonwords if provided with the individual phonemes (e.g., /g/, /r/, /a/ ! /gra/) and could pronounce the last phoneme of a word spoken by the experimenter. The second individual, the Italian speaker RR (Bisiacchi, Cipolotti, & Denes, 1989), was able to pronounce the first phoneme of aurally presented words and performed within control subjects’ range in the difficult spoonerism task, which requires swapping word onsets (as in Niccolò Macchiavelli ! Miccolò Nacchiavelli). More recently, we (Caccappolo-van Vliet, Miozzo, & Stern, in press) documented phonological dyslexia in a subject (RG) diagnosed with Alzheimer’s disease. Because of a general cognitive decline, RG could perform phonological tasks only if they required relatively simple instructions: word and nonword repetition, rhyme judgment (‘‘Do the words pair and bear rhyme?’’), and production of words that share onset phonemes (shot ! sugar) or rhyme (table ! cable). RG performed flawlessly on all these tasks. The dissociations between impaired nonword reading and preserved performance on phonological tasks documented in these patients are problematic for the phonological-deficit hypothesis. These dissociations, however, are in line with the dual-route model, which does not assume phonological deficits to be the principal cause of phonological dyslexia and thus does not anticipate a recurrent association between phonological deficit and nonword reading impairment. Upon closer scrutiny, however, these dissociations might appear to be less compelling. Several questions have been raised about these patients’ data. For example, Patterson (2000) questioned whether LB was a convincing case of phonological dyslexia, because his accuracy was not far better for words (range: 74–94%) than for nonwords (range: 48–85%). Patterson raised similar concerns about RR, as there were indications that this patient could read short nonwords but had some problems with low-frequency abstract familiar words. Harm and Seidenberg (2001) suspected that LB’s relatively good performance in the phonological tasks resulted from rehabilitative training that emphasized phonological reading and improved LB’s phonological awareness. Regarding RG, data were limited to phonological tasks with procedures simple enough to be understandable to this patient. In short, if previous neuropsychological data do not allow one to make conclusive claims about whether phonological dyslexia and phonological deficits dissociate, it also appears clear that this issue needs further investigation. An opportunity to address this issue is offered by the 2 patients we report on here, M.O. and I.B. Both have progressive dementia and encountered relatively severe problems in reading nonwords. Their word reading was highly accurate, and they also performed remarkably well on a wide variety of phonological tasks. Their serious cognitive damage prevented us from administering a few tasks; still, our results provide significant constraints on current accounts of acquired phonological dyslexia.