Developmental Dyslexia Review PDF

CP11CH11-Peterson ARI 3 March 2015 13:26 ANNUAL...

CP11CH11-Peterson ARI 3 March 2015 13:26 ANNUAL REVIEWS Further Developmental Dyslexia Click here for quick links to Annual Reviews content online, including: Robin L. Peterson1 and Bruce F. Pennington2 Other articles in this volume 1 Department of Rehabilitation Medicine, Children’s Hospital Colorado, Aurora, Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. Top cited articles Colorado 80045; email: [email protected] Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org Top downloaded articles 2 Our comprehensive search Department of Psychology, University of Denver, Denver, Colorado 80208; email: [email protected] Annu. Rev. Clin. Psychol. 2015. 11:283–307 Keywords First published online as a Review in Advance on reading disability, etiology, neuropsychology, language, genetics January 14, 2015 The Annual Review of Clinical Psychology is online at Abstract clinpsy.annualreviews.org This review uses a levels-of-analysis framework to summarize the current This article’s doi: understanding of developmental dyslexia’s etiology, brain bases, neuropsy- 10.1146/annurev-clinpsy-032814-112842 chology, and social context. Dyslexia is caused by multiple genetic and en- Copyright  c 2015 by Annual Reviews. vironmental risk factors as well as their interplay. Several candidate genes All rights reserved have been identified in the past decade. At the brain level, dyslexia is asso- ciated with aberrant structure and function, particularly in left hemisphere reading/language networks. The neurocognitive influences on dyslexia are also multifactorial and involve phonological processing deficits as well as weaknesses in other oral language skills and processing speed. We address contextual issues such as how dyslexia manifests across languages and social classes as well as what treatments are best supported. Throughout the re- view, we highlight exciting new research that cuts across levels of analysis. Such work promises eventually to provide a comprehensive explanation of the disorder as well as its prevention and remediation. 283  CP11CH11-Peterson ARI 3 March 2015 13:26 Contents INTRODUCTION............................................................... 284 DEFINITION OF DYSLEXIA.................................................... 284 EPIDEMIOLOGY................................................................ 286 Socioeconomic Status........................................................... 286 Cross-Cultural Findings........................................................ 287 Comorbidities.................................................................. 288 NEUROPSYCHOLOGY OF DYSLEXIA......................................... 288 History........................................................................ 288 Issues in the Phonological Theory of Dyslexia.................................... 289 Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. Single Versus Multiple Deficit Accounts......................................... 292 Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org BRAIN BASES OF DYSLEXIA.................................................... 292 ETIOLOGY...................................................................... 294 Behavioral Genetics............................................................ 294 Molecular Genetics............................................................. 295 Environmental Influences....................................................... 297 TREATMENT................................................................... 298 CONCLUSIONS................................................................. 299 INTRODUCTION This review uses a levels-of-analysis framework (Pennington 2002) to summarize current knowl- edge regarding developmental dyslexia (henceforth, dyslexia). The guiding principal behind this framework is that a complete explanation of any disorder—or indeed, of any typical developmental phenomenon—requires understanding the phenomenon across multiple levels: its defining symp- toms or behaviors; its neuropsychology (underlying cognitive, emotional, or other psychological processes that are not directly observable and not part of the disorder’s definition); its pathophys- iology (for cognitive disorders, changes in brain structure and function); its etiology, or distal causes, including genetic and environmental risk and protective factors; and its social context. No level is fully reducible to a “lower” level because of the emergence of new phenomena as systems become increasingly complex, and no level has priority over the others in terms of scientific value. For example, obtaining comprehensive knowledge regarding all the genetic and environmental risk factors that contribute to poor reading skills would be extremely valuable but could not replace a thorough understanding of the cognitive processes involved in reading words aloud. The sections below cover dyslexia’s definition and epidemiology (with discussion focusing on both behavioral and social context levels), neuropsychology, brain bases, etiology, and treatment. We particularly highlight exciting recent research that cuts across levels of analysis—for example, from etiology to brain bases, brain bases to neuropsychology, or social context to etiology. DEFINITION OF DYSLEXIA Individuals with dyslexia have difficulties with accurate or fluent word recognition and spelling despite adequate instruction and intelligence and intact sensory abilities (Lyon et al. 2003). The ultimate goal of reading is comprehension, which is a function of both decoding ability and oral language comprehension (Hoover & Gough 1990). Dyslexia is defined by difficulties with 284 Peterson · Pennington  CP11CH11-Peterson ARI 3 March 2015 13:26 decoding, whereas by comparison, listening comprehension is typically more intact. Thus, although individuals with very limited decoding abilities (i.e., young children or individuals with severe dyslexia) have poor reading comprehension, individuals with milder decoding SES: socioeconomic problems can still support adequate reading comprehension with intact oral language skills (Bruck status 1993). So-called poor comprehenders show the opposite profile of adequate decoding but poor Phonological understanding of what is read. Not surprisingly, poor comprehenders tend to have deficits in oral processing: umbrella language comprehension, and this profile is sometimes considered a type of language disorder term for oral language (Nation et al. 2010). tasks emphasizing Although some previous diagnostic systems have grouped dyslexia and poor reading com- processing of sounds prehension together (e.g., Am. Psychiatr. Assoc. 2000), this review is about dyslexia only. Many researchers use the terms “dyslexia” and “reading disability” interchangeably, although as the preceding discussion makes clear, other learning disorders (e.g., language disorder) can affect Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. reading. Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org Some disorders, such as cystic fibrosis, are categorical (you either have the disorder or you do not), and these categorical disorders often have a discrete etiology, such as a mutation in a sin- gle gene, as is true for cystic fibrosis, phenylketonuria, and Huntington’s dementia. Many other disorders (such as hypertension) do not represent categories, but just extremes on a continuous distribution that ranges from optimal outcomes to poor outcomes, with the underlying causal mechanisms being similar across the whole distribution. Essentially all behaviorally defined disor- ders, including dyslexia, are such continuous disorders. Dyslexia is mainly defined as the low end of a normal distribution of word reading ability (Rodgers 1983, Shaywitz et al. 1992). Thus, in order to diagnose the disorder, a somewhat arbitrary cutoff must be set on a continuous variable. Should the diagnostic threshold for dyslexia be relative to age or intelligence quotient (IQ)? The logic behind IQ-discrepancy definitions is that the cause of poor reading might differ between low-IQ and high-IQ individuals. Specifically, it has been assumed that IQ sets a limit on achieve- ment across domains, and thus children with low IQ are likely to be poor readers because of general learning difficulties rather than a specific decoding problem. Genetic differences contribute more to high-IQ dyslexia than to low-IQ dyslexia (Wadsworth et al. 2010). A related finding is that dyslexia is more genetically based in children from higher socioeconomic status (SES) families than in children from lower SES families (Friend et al. 2008). Together, these results suggest that advantaged children with strong cognitive abilities are likely to be good readers unless they have specific genetic risk factors for poor decoding. There are myriad reasons why other children will struggle with reading. These reasons include environmental influences associated with low SES, and those influences will account for more of the variance in poor reading in children from lower SES families than in children from higher SES families. Although the same risk genes are likely important across the range of SES, they contribute less to poor reading in the presence of environ- mental risk factors associated with lower IQ. We do not yet know which proximal environmental factors are most likely to contribute to low reading ability, though some reasonable possibilities are discussed later in the sections titled Socioeconomic Status and Environmental Influences. Despite this evidence for a different weighting of genetic and environmental risk factors in the etiologies underlying dyslexia in children with high versus low IQ, published work does not support the external validity of the distinction between age-referenced and IQ-referenced definitions in terms of underlying neuropsychology or appropriate treatments. Specifically, poor readers of all general ability levels typically have disproportionately poor skills in phonological processing (processing sounds in language), as discussed further below in the Neuropsychology of Dyslexia section. As a group, children with dyslexia respond best to treatment emphasizing phonics-based reading instruction. Although there are individual differences in how well children with dyslexia respond to such intervention, these differences do not appear to be solely or even primarily a www.annualreviews.org Developmental Dyslexia 285  CP11CH11-Peterson ARI 3 March 2015 13:26 function of IQ ( Jimenez et al. 2009, Stuebing et al. 2009). The two definitions overlap, but some people with clinically significant reading problems meet only IQ-discrepancy criteria (high ability, weaker-than-expected word reading), whereas others meet only age-discrepancy criteria ADHD: attention- deficit/hyperactivity (low ability, poor word reading). The previous version of the Diagnostic and Statistical Manual disorder of Mental Disorders (DSM) (version IV-TR; Am. Psychiatr. Assoc. 2000) required that reading achievement be below the level expected for both age and IQ. The most recent revision of the DSM (DSM-5; Am. Psychiatr. Assoc. 2013) now requires that reading be below age expectations in every case. Although the updated definition should facilitate the identification and remediation of reading problems in children with broader cognitive difficulties, it unfortunately continues to exclude those of high ability who nonetheless have clinically impairing difficulties and could benefit from reading intervention. Indeed, as the above discussion makes clear, the new definition ironically means that fewer children with a stronger genetic etiology will be classified as dyslexic. Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. Thus, for both research and clinical purposes, we think it is more appropriate to identify children Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org who meet either age- or IQ-discrepancy criteria as having dyslexia. EPIDEMIOLOGY Prevalence estimates, of course, depend on definition. A common definition sets the cutoff for reading achievement to 1.5 standard deviations (SD) below the mean for age and identifies 7% of the population; a similar IQ-achievement discrepancy definition identifies a comparable propor- tion (Shaywitz et al. 1990). There is a relatively small but significant male predominance (from 1.5:1 to 3.1:1; Rutter et al. 2004). However, the gender difference in referred samples is even higher (from 3:1 to 6:1; Smith et al. 2001). Boys with dyslexia come to clinical attention more often than girls apparently because they have higher rates of comorbid externalizing disorders, including attention-deficit/hyperactivity disorder (ADHD) (Willcutt & Pennington 2000). Socioeconomic Status As mentioned above, SES is associated with reading skill, as with virtually all other measures of achievement. The effects of SES on literacy are part of the so-called achievement gap that has garnered considerable attention in public policy and education circles. For example, in 2011, the overall reading level of a national sample of fourth graders eligible for free lunch was 0.83 SD lower than that of students not eligible for free or reduced lunch, a large effect size (Natl. Cent. Educ. Stat. 2011). Lower SES is associated with both poorer word reading and poorer reading comprehension, though the effect is larger for reading comprehension, particularly at older ages (MacDonald Wer 2014). The SES–word-reading link means that by definition, a disproportionate number of children from lower SES families will meet diagnostic criteria for developmental dyslexia. Lower SES predicts both poorer early reading skills at the onset of formal literacy instruction and a slower trajectory of literacy growth over the early school years (Hecht et al. 2000). Systematic reviews and meta-analyses have consistently shown that SES accounts for approx- imately 10% of the variance in reading outcome (Scarborough & Dobrich 1994, Sirin 2005, White 1982). Although this effect is statistically significant and moderate, it also means that ap- proximately 90% of the variance in reading outcome is independent of SES and thus that many children from disadvantaged backgrounds will be strong readers while many weak readers will come from advantaged families. SES probably serves as a proxy for several environmental variables that adversely affect literacy development. However, SES is not solely an environmental construct, at least in societies that allow for a degree of social mobility, since there are genetic influences on individual differences that impact people’s ability to achieve higher levels of education, income, and occupational status 286 Peterson · Pennington  CP11CH11-Peterson ARI 3 March 2015 13:26 (Baker et al. 1996). Thus, some of the SES–reading association may be due to a third variable: genes shared by parents and children that influence reading or cognitive abilities more broadly. Adoption studies suggest that of the 10% of reading variance that is related to SES, approximately half is Bioecological gene- mediated by environmental factors and half by genetic factors (Petrill et al. 2005, Wadsworth et al. by-environment 2001). So, about 5% of overall reading outcome can be linked to environmental factors that fall interaction: type of under the SES umbrella. Two important caveats are that (a) existing samples may not have included G × E interaction the very low tail of poor environments for reading development and (b) as discussed previously, shown in dyslexia in which genes have there is evidence for a bioecological gene-by-environment interaction in reading development, greater influence in such that the balance of genetic and environmental influences is not constant across SES levels. more favorable What are the specific environmental variables that directly influence reading development? environments Methodologically rigorous research on this question is still at an early stage of development, Rapid serial naming: but the causal factors are probably many and act at the child, family, neighborhood, school, and type of Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. broader community levels. In terms of family factors, research indicates that the quality and nature neuropsychological Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org of language interaction between parents and children vary across SES levels (Hoff 2003), including task requiring speeded naming of a matrix of around specific preliteracy and literacy activities (Phillips & Lonigan 2009, Robins et al. 2014). familiar objects, color We return to the question of genetic and environmental influences on dyslexia in the Etiology swatches, letters, or section. Importantly, although this research tells us about the distal causes of individual differences numbers in reading, it does not tell us about the extent to which particular environmental treatments (such as providing evidence-based reading instruction) can shift the average score for a group with poorer- than-average reading, such as children from lower SES backgrounds. This issue evokes the Flynn effect that has been demonstrated for IQ, in which the mean IQ of the whole distribution increases over time, and this effect appears to be carried disproportionately by improvements in the low tail of the distribution (perhaps because of basic public health improvements, such as better nutrition; Lynn & Hampson 1986). There probably has been a Flynn effect for reading as well over the last century. Despite these group-level changes, which are driven by environmental factors, the etiology of individual differences may well have remained the same, and these differences include substantial genetic influence. Cross-Cultural Findings Although research on dyslexia initially focused primarily on reading difficulties in English, recently a good deal of attention has been paid to the nature of dyslexia across languages. Here, we briefly summarize what is known about how dyslexia manifests across languages showing two different types of variability: first, among alphabetic orthographies that vary in the degree of consistency of letter-sound correspondences; and second, in alphabetic versus logographic orthographies. Children at the low end of reading ability distribution in languages with more consistent mappings between letters and sounds (e.g., Italian or Finnish) have less severe reading problems than those learning to read less consistent languages (i.e., English), at least in terms of accuracy (Landerl et al. 1997). Difficulties with reading fluency (speed of reading connected text) seem similar across languages (Caravolas et al. 2005). Several studies have noted important universal features in normal and disordered reading across cultures despite linguistic differences. Cognitive predictors of early reading were similar for five European orthographies (Finnish, Hungarian, Dutch, Portuguese, and French), in agreement with previous results in English. Particularly, phonological awareness was the main predictor of reading in each language, although it had more of an effect in less consistent than in consistent orthographies. Other predictors, such as rapid serial naming, vocabulary knowledge, and verbal short-term memory, made smaller contributions than did phonological awareness, except in Finnish (the most consistent language), in which vocabulary had at least as large an effect on reading (Ziegler et al. 2010). www.annualreviews.org Developmental Dyslexia 287  CP11CH11-Peterson ARI 3 March 2015 13:26 Cross-cultural similarities appear to extend in large part to logographic languages as well, such as Chinese. By contrast with alphabetic languages, in which letters represent phonemes, the smallest written units in Chinese are characters representing monosyllabic morphemes (units LI: language impairment of language that convey meaning). However, phonology is not irrelevant to reading in Chinese. Chinese characters have phonological elements, and skilled readers of the language show phono- SSD: speech sound disorder logical effects on word recognition (Pollatsek 2014). Phonological awareness is a key correlate and predictor of reading skill in Chinese just as in alphabetic orthographies. However, in contrast to alphabetic languages in which awareness of phonemes is critically important, morphological and syllabic awareness play a large role in learning to read Chinese (McBride-Chang et al. 2005). This finding is not surprising given the differences in how the orthographies represent language. Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. Comorbidities Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org In addition to its comorbidity with ADHD, dyslexia is comorbid with two other disorders of language development: language impairment (LI) and speech sound disorder (SSD) (Nittrouer & Pennington 2010). LI is defined by problems in the development of structural language, in- cluding syntax (grammar) and semantics (vocabulary), whereas for SSD the defining problem is in the ability to accurately and intelligibly produce the sounds of one’s native language. In each case, evidence indicates that the comorbidity with dyslexia is mediated by shared etiologic and neurocognitive risk factors (Pennington & Bishop 2009, Willcutt et al. 2010). The comorbidities are clinically significant because dyslexia is not diagnosed until after a child has been exposed to formal literacy instruction, but ADHD, SSD, and LI are all likely to be apparent earlier and can thus indicate a child’s risk for later reading problems. Children with dyslexia are also at increased risk for learning disabilities affecting other academic skills, including those impacting reading comprehension (Christopher et al. 2012), math (Landerl & Moll 2010), and writing (Berninger et al. 2001). Thus, many students with dyslexia have more than a “specific” reading disability and can be expected to struggle broadly in school. The most recent revision of the DSM (version 5) acknowledges these high levels of comorbidity by grouping all learning disabilities under a single umbrella diagnosis (“Specific Learning Disorder”) and allowing the clinician to select a modifier(s) describing the particular academic skill(s) affected (Am. Psychiatr. Assoc. 2013). This approach has the advantage of explicitly recognizing the overlap among learning disorders that again arises from shared risk factors. However, equally strong evidence does not support the validity of all learning disorders. Although many decades of work have elucidated the causes, neurobiology, neuropsychology, and appropriate treatments for dyslexia, the scientific study of difficulties in written expression or mathematical problem solving is at a much earlier stage of development. NEUROPSYCHOLOGY OF DYSLEXIA History Reading disabilities in children were first described more than a century ago by Pringle-Morgan (1896) and Kerr (1897) and were labeled dyslexia because of the parallels with the loss of reading ability following brain injury in previously skilled adult readers. In the first part of the twentieth century, Samuel Orton (1925, 1937) focused on the letter and word reversal errors commonly seen in children with dyslexia (such as b/d confusions or reading “was” for the word saw) and hypothesized that a visual problem arose because of a hemispheric dominance failure in which mirror images of visual stimuli were not inhibited. Vellutino (1979) demonstrated that such reversal errors were restricted to processing print in one’s own language and were thus really linguistic 288 Peterson · Pennington  CP11CH11-Peterson ARI 3 March 2015 13:26 rather than visual in nature. Unfortunately, the perception that dyslexia primarily reflects a visual problem persists among many in the lay public and continues to form the basis of therapies for the disorder that lack empirical support (Pennington 2011). Although it remains possible that Phonological grain some sorts of visual processing problems correlate with dyslexia, the scientific consensus for the size: size of the last several decades has been that dyslexia is a language-based disorder whose primary underlying linguistic unit most deficit involves problems in phonological processing (processing of sounds in oral language) that informative in the lead to later problems processing written language (Vellutino et al. 2004). Despite this long mapping between spoken and written history, this and other fundamental controversies about the neuropsychology of dyslexia have language (i.e., been recapitulated several times over the years, as discussed further below. phonemes, In the phonological theory of dyslexia, the ability to attend to and manipulate linguistic sounds morphemes, or is crucial for the establishment and automatization of letter–sound correspondences, which in syllables) turn underlie accurate and fluent word recognition through the process of phonological coding. Phoneme awareness: Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. Phonological processes are important not only for learning to read alphabetic orthographies (in meta-linguistic Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org which the script represents phonemes, or individual speech sounds) but also for learning to read awareness of individual speech sounds logographic orthographies (in which the script represents language at the morpheme/syllable (phonemes) level), although the phonological grain size most important for skilled reading varies across scripts (Perfetti et al. 1992). Issues in the Phonological Theory of Dyslexia Access versus representations. One old controversy in the phonological theory of dyslexia that has recently been revisited concerns the nature of the phonological deficit. The initial assumption was based on a Chomskyan view (Chomsky & Halle 1968) and held that the ability to discrimi- nate phonemes was innate and universal, and so implicit phonological representations should be intact in dyslexia. Early investigations into the phonological deficit thus focused on metacognitive awareness of phonemes (e.g., Liberman 1973). Phoneme awareness develops over several years and is strongly correlated with literacy skill, making it an attractive initial core deficit candidate for dyslexia. However, research on early language development eventually demonstrated that phone- mic representations are not innate and instead develop in response to linguistic input. A detailed review of early phonological development is beyond the scope of this article, but briefly, it ap- pears that for all children, phonological representations start out as fairly holistic and become gradually more detailed or segmented over time. Babies likely represent most words as single entities. With language development, phonological representations begin to emphasize syllables, then subsyllabic distinctions, and ultimately individual phonemes (e.g., Metsala & Walley 1998). In a seminal paper, Fowler (1991) argued that this implicit phonological development underlies parallel development in metaphonological awareness (Treiman 1985) and helps account for the fact that preschool children generally cannot perform tasks that require explicit manipulation of individual phonemes, although they can perform phonological awareness tasks at a larger grain size (e.g., rhyming). Through the 1990s and first decade of the 2000s, empirical evidence accrued highlighting faulty implicit phonological representations in dyslexia (Boada & Pennington 2006, Elbro et al. 1998, Manis et al. 1997). More specifically, children with dyslexia show deficits on phonological processing tasks that do not require explicit attention to or manipulation of phonemes, including categorical speech perception, priming, and lexical gating tasks. Various research groups have described the implicit phonological deficit in different ways, including that phonological repre- sentations in dyslexia are imprecise, poorly segmented, or otherwise degraded. Recently, the debate about the nature of the phonological deficit and whether it involves phono- logical representations themselves or only metacognitive access to those representations has been www.annualreviews.org Developmental Dyslexia 289  CP11CH11-Peterson ARI 3 March 2015 13:26 rekindled. Proponents of the updated “access-only” argument point out that adults with dyslexia perform normally on many implicit phonological tasks, including those emphasizing phonotactics and stress perception (Dickie et al. 2013, Ramus & Szenkovits 2008). Furthermore, findings on some other implicit tasks (particularly categorical speech perception) have been inconsistent, with some scientists arguing that speech perception problems are more clearly linked to LI than to dyslexia ( Joanisse et al. 2000). Most of the studies in support of the access-only view have used be- havioral methods, although recently Boets and colleagues (2013) published a neuroimaging study showing that brain response to phonetic contrasts in primary and secondary auditory cortex was similar in adults with and without dyslexia, but structural and functional connectivity between au- ditory cortex and left inferior frontal gyrus were reduced in those with dyslexia. Boets et al. (2013) argued that these findings were consistent with “intact but less accessible phonetic representations in dyslexia” (p. 1251; see also Ramus 2014). Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. This rekindled argument has garnered a good deal of attention, with some of the papers in Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org favor of the access-only view being published in top journals and gaining numerous citations. Certainly, many of the experiments showing intact aspects of phonological processing in dyslexia have been elegantly designed and have yielded important results for the field to integrate into current explanations of reading difficulties. However, we think this argument faces a number of challenges that its proponents have not yet adequately addressed. First, the history of the access- only view and its basis in a nativist, Chomskyan account of phonology have not been explored. Second, there is a striking lack of consensus regarding which tasks are most appropriate to measure implicit phonological representations. The fact that dyslexia appears to be associated with atypical performance on lexical gating tasks (Boada & Pennington 2006) but not stress perception (Dickie et al. 2013, Mundy & Carroll 2012) is probably meaningful, especially if these results could be replicated in a within-subjects design. Third, we disagree with the conclusion that normal brain response to phonemes in auditory cortex implies normal phonological representations. In fact, phonological representations can be thought of as emerging from the mappings between acoustic inputs and motoric outputs (Plaut & Kello 1999). Given this view, deficient connectivity between more posterior auditory regions and left inferior frontal gyrus is exactly what we would expect to find in children with poor phonological representations. Causal direction. A second issue in the phonological theory of dyslexia concerns the direction of effect between phonological development and reading (Castles et al. 2011). Because written language is parasitic on oral language, and formal literacy instruction does not begin until chil- dren have mastered most of the fundamentals of a spoken language, it seems reasonable that the causal direction flows from phonology to reading rather than vice versa. Several lines of evidence support this conclusion. First, preschool children who will later develop dyslexia show deficits on various phonological tasks, with phoneme awareness being particularly predictive of later literacy attainment by around kindergarten age (Pennington & Lefly 2001, Scarborough 1991, Snowling et al. 2003). Furthermore, children with dyslexia underperform even younger, typically developing children matched on reading level on phoneme awareness tasks (Wagner & Torgesen 1987), and these deficits tend to persist in adults with dyslexia who have otherwise compensated well for the disorder (Bruck 1992, Hatcher et al. 2002). The conclusion that phoneme awareness deficits have a unidirectional causal link to read- ing problems is oversimplified for several reasons, however. Speech scientists complain about the tyranny of the phoneme (Greenberg 2004) because these idealized representations have be- come reified and likely mislead us about what dimensions in the speech stream are important in development and how those dimensions are flexibly integrated to recover linguistic structures, such as words. There are longstanding controversies about the units of speech perception (Goldinger & 290 Peterson · Pennington  CP11CH11-Peterson ARI 3 March 2015 13:26 Azuma 2003), and recent evidence demonstrates that speech representations preserve much more than phonemes. This work has led to a proposal that phonemes are not the targets of speech per- ception and are mainly important in the context of learning an alphabetic written language (Port Orthography: 2007). Related work with adult natural illiterates (who are cognitively normal but have no formal written system for a schooling) confirms that phoneme-level representations do not arise automatically in language language. Alphabetic development (Castro-Caldas et al. 1998, Morais et al. 1979). In other words, as literate adults orthographies we think that individual phonemes exist in the speech signal like beads on a string, but this is an represent individual sounds with letters; illusion that arises from our extensive experience with an alphabetic script. Thus, difficulties in logographic phonological development in dyslexia are probably not restricted to phonemic or segmental rep- orthographies resentations and must lie in other dimensions of the speech stream, at least initially. Despite this represent words or evidence that learning to read changes phonological development, methodologically rigorous work morphemes with demonstrates that phoneme awareness training, when combined with direct reading instruction, symbols Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. improves literacy in early school-age children (Hulme et al. 2012). The most accurate conclusion Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org therefore appears to be that the relationship between phonology and literacy is bidirectional. Orthographic learning. In recent years there has been burgeoning interest in an orthographic learning account of reading problems, which emphasizes not phonological representations them- selves but rather the ability to establish mappings between phonemes and graphemes, or letters and sounds (e.g., Aravena et al. 2013). This explanation has strong face validity to explain dyslexia, which is essentially defined by problems decoding print. Good neurophysiological evidence in- dicates that skilled readers treat letters as single audiovisual objects (Blau et al. 2010), and the orthographic learning hypothesis states that problems developing such integrated representations interfere with the emergence of fluent reading. Limited behavioral support for this hypothesis comes from research comparing the performance of children with and without dyslexia when they were asked to learn associations between sounds in their native language and an unfamiliar or- thography. Although both groups learned the associations, the children with dyslexia performed more poorly than controls under time pressure (Aravena et al. 2013). In addition to its face validity, the orthographic learning hypothesis has a number of strengths. It represents an admirable attempt to integrate across the brain and neuropsychological levels of analysis to explain reading development and difficulties. Furthermore, it avoids some of the reductionistic errors of other accounts that have been put forward as alternatives to the dominant phonological view, such as auditory and visual explanations. However, this hypothesis also faces some serious problems. Most critically, it does not account for the early language development of predyslexic children, who have subtle difficulties with spoken language long before they encounter a written script. Babies who will become dyslexic show a brain response to speech stimuli that differs from that of babies who will not become dyslexic (Guttorm et al. 2005). As toddlers, these children lag behind their peers in vocabulary and syntax (grammar) development, and in preschool, they have difficulties with phonological awareness (Scarborough 1991, Torppa et al. 2010). Notably, this same criticism can be applied to the access-only phonological view. A related point is that it is difficult to test a pure integration account of phoneme-grapheme binding because we know that children with dyslexia are not equivalent to their typically developing peers in processing phonemes of their native language. So, the meaning of the fact that they are slower in learning phoneme-grapheme mappings is ambiguous; it could arise directly from the unimodal phonological deficit. To show that there is an additional contribution of cross-modal letter-sound processing over and above the well-established phonological processing problem in dyslexia, we will need behavioral studies that can somehow control for unimodal phonological and orthographic processing across groups. This will be an important issue to be addressed by future research. www.annualreviews.org Developmental Dyslexia 291  CP11CH11-Peterson ARI 3 March 2015 13:26 Single Versus Multiple Deficit Accounts Despite the large body of evidence supporting the phonological theory of dyslexia, it is now Processing speed: clear that a single core phonological deficit is insufficient to explain reading difficulties. As we speeded and others have discussed elsewhere (Bishop & Rutter 2009, Pennington 2006), the etiology of neuropsychological dyslexia (like all behavioral disorders) is complex and multifactorial, so it should not be surprising task emphasizing that the underlying neuropsychology also involves several interacting risk and protective factors. information processing (e.g., visual Consistent with a multiple deficit hypothesis, results of family-risk designs (which follow children perception; learning a who are at genetic risk for dyslexia based on their family history but who are too young to have new symbolic code). been diagnosed with the disorder) and longitudinal studies of children with early speech/language Graphomotor output disorders consistently indicate that many children develop normal-range literacy skills despite often required preschool phonological deficits similar in magnitude to those of children who ultimately develop dyslexia (Bishop et al. 2009, Peterson et al. 2009, Snowling et al. 2003). These children appear to Access provided by University of Zurich - Hauptbibliothek on 11/22/22. For personal use only. be protected from dyslexia because of relative strengths in other cognitive skills associated with Annu. Rev. Clin. Psychol. 2015.11:283-307. Downloaded from www.annualreviews.org reading. Conversely, children with multiple cognitive deficits are at much higher risk for dyslexia. Across countries and languages, many cognitive-linguistic constructs predict later dyslexia. The constructs that are most consistently implicated include phonological awareness, rapid serial nam- ing, verbal short-term memory, vocabulary and other aspects of broader oral language skill, and graphomotor processing speed (McGrath et al. 2011, Pennington et al. 2012, Scarborough 1998, Wolf & Bowers 1999). The most powerful individual predictor varies with developmental stage. In toddlers and young children, broader language development is most strongly linked to later reading; by age 4 or 5 years, phonological awareness is the dominant predictor; and tasks em- phasizing speed (i.e., rapid serial naming and processing speed) become increasingly important as literacy development progresses, probably because they are more linked to reading fluency than to single- word reading accuracy (Pennington & Lefly 2001, Puolakanaho et al. 2008, Scarborough 1990, Snowling et al. 2003, Torppa et al. 2010). Longitudinal research suggests that these deficits make a causal contribution to reading problems and are not fully accounted for by comorbidities or the cumulative effects of reading difficulties. Research has made clear for many years that dyslexia does not result from disturbances in basic visual perception (Ramus 2003, Vellutino 1979). However, recently there has been renewed interest in the possible role of visual attentional deficits in reading difficulties (Facoetti et al. 2010). Visual attention is measured through serial search, orienting/cueing paradigms, or crowd- ing paradigms that require participants to recognize pictures amid varying degrees of visual clutter; some of these skills probably contribute to performance on nonlinguistic processing speed tasks known to be correlated with reading. A recent study demonstrated that performance on visual attention tasks in preschool significantly predicted reading ability two years later, after accounting for the influence of reading-related phonological processing skills (Franceschini et al. 2012). Initial evidence suggests a similar pattern of results across writing systems with varying degrees of consis- tency in letter-sound relationships (i.e., Italian and French) (Zorzi et al. 2012). Although a deficit in visual attention does not easily account for the early speech-language phenotype in predyslexic children, it might represent an additional cognitive difficulty that interacts with language problems to cause reading failure. Further research is needed on this question. BRAIN BASES OF DYSLEXIA Because reading is a linguistic skill, we would expect it to involve activation of brain structures used in oral language processing and some additional structures associated with visual-object processing and establishment of visual-linguistic mappings (see Figure 1). Indeed, functional 292 Peterson · Pennington  CP11CH11-Peterson ARI 3 March 2015 13:26 PARIE PA RIETT AL CORT CO RTEE X