A Developmental Perspective on Dyslexic Subtypes Franklin R. Manis Patricia A. Szeszulski Liana K. Holt Kathryn Graves University of Southern California Los Angeles
In this paper we apply a developmental model of reading to the question of dyslexic subtypes. Groups of normal readers (n = 40) and dyslexic children (n = 50), matched on reading level and IQ, were given a comprehensive test battery measuring level of development of visual, phonological, and orthographic skills. As a group, dystexics deviated from normal readers of equivalent reading achievement primarily in phonological skills (spelling-to-sound translation and phonemic analysis), although limited differences in knowledge of word-specific spellings were also observed. Dyslexics were superior to the younger normal readers in visual processing of print. Analysis of individual data by reference to the reading level control group revealed three major subgroups: a group with a specific deficit in phonological processing of print (52 percent), a group with deficits in processing both the phonological and orthographic features of printed words (24 percent), and a group with phonological deficits in language (8 percent). The remainder of the sample (I6 percent) had specific deficits in visual or orthographic processing of print, in spelling, or did not differ from the control group. The data support the view that most developmental dyslexics have a specific language disorder involving some aspect of phonological processing. However, small subgroups with very different configurations of reading and nonreading difficulties may exist as well. O n e of the continuing questions of interest in the field of dyslexia concerns the nature of possible s u b t y p e s of reading a n d l a n g u a g e disability a m o n g the dyslexic population. It is n o w clear that dyslexic children This research was supported by an NICHD grant to the first author (USPHSgrant 1 R23 HD20231). We thank the Los Angeles and Orange County chapters of The Orton Dyslexia Society, the Torrance Unified School District, and Landmark West School for their cooperation in obtaining subjects for the study. The test measures were developed jointly by the first two authors. Annals of Dyslexia, Vol. 38, 1988. Copyright ©1988 by The Orton Dyslexia Society ISSN 0474-7534
139
140
THEORETICAL SEEDS
(as they have traditionally been defined) are a heterogeneous group (Boder 1973; Doehring et al. 1981; Lovett 1984; Satz and Morris 1981; Seymour and MacGregor 1984). Many researchers and educators recognize that we must understand this heterogeneity in order to cope more effectively with questions of diagnosis, treatment, and etiology. A number of subtyping schemes have been proposed. Some focus on analysis of reading and spelling performance (e.g., Boder 1973; Doehring et al. 1981;Seymour and MacGregor, 1984), and some on patterns of neuropsychological test performance (Mattis, French, and Rapin 1975; Petrauskas and Rourke 1979; Satz and Morris 1981). At present there is lack of agreement on how best to categorize dyslexic children. A recent and promising approach can be found in the work of Frith and others (Frith 1985; Frith 1986; Seymour and MacGregor 1984; Snowling 1987), who have used a cognitive-developmental framework to analyze variations in reading, spelling, and language skills among dyslexics. In this paper, we will present a developmental model of reading and related skills, use it to derive predictions about the nature of subgroups, describe a comprehensive battery of tests derived from the model, and present data we have collected on dyslexic children and a control group of normal readers.
Model of Reading Development Frith (1985, 1986) and Seymour and MacGregor (1984) assume that individuals learn to read and spell by means of three strategies, arranged in a developmental sequence. The earliest strategy, the logographic strategy, is visually based. The child recognizes words "by sight," that is by means of salient visual features such as the first and last letters, the outline of the word, etc. The second strategy, the phonological (or alphabetic) strategy, involves sequential decoding of spelling units (letters and letter clusters) into the phonemes of the language. Once children have sufficient knowledge of spelling-sound correspondences and decoding rules, they can identify unfamiliar printed words that are in their speaking vocabulary. The third strategy, the orthographic strategy, allows the reader to recognize words based on each word's unique pattern of letter identities and letter order, without the need to translate the letters into a phonological code. It is assumed that the reader stores information about each morpheme (a word or grammatical unit, such as -ed, contained in a word) in a mental lexicon and learns specific associations between the spelling of each morpheme and its meaning. It has been argued that this is the dominant strategy for skilled readers (Jorm and Share 1983; McCusker, Hillenger, and Bias, 1981). According to Frith (1985, 1986), immature readers employ only the logographic strategy. The phonological strategy actually develops first in the context of learning to spell and soon becomes the dominant method of
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
141
identifying unfamiliar words and spelling familiar words. Gradually the child amasses enough knowledge of specific morphemes and their spelling patterns to begin using the orthographic strategy to read and eventually to spell. The advantage of using the orthographic strategy is twofold: words can be read and spelled using larger units (such as tion, ight), and words with irregular spelling-sound correspondences (e.g., colonel, island) can be read and spelled successfully. In the present study we assume that the two more advanced strategies continue developing after they are initially acquired. The orthographic strategy increases its range of application by constantly adding new morphemes to the lexicon and, possibly, by increasing the size of morpheme that can be immediately recognized. The phonological strategy increases its range by adding larger and more complex spelling-tosound correspondences. In addition, both strategies continue to increase in efficiency or automaticity with practice. Rapid, efficient word recognition is thought to be one of the principal determinants of progress in reading comprehension in the early grades, and is an important correlate of reading ability at a variety of ages (Carr 1981; Manis and Morrison 1985; Stanovich 1980). The phonological and orthographic reading strategies do not develop in isolation from other language and cognitive skills. To function propefly, both strategies require fundamental visual skills (processing print sequentially from left to right, visually recognizing key spelling patterns, and visually segmenting words into appropriate spelling patterns or morphemes), as well as a variety of language skills (auditory discrimination, phonemic analysis, adequate oral vocabulary, syntactic development, etc.) (Carr 1981; Liberman 1982; Manis and Morrison 1985; Mann 1984; Stanovich 1986). Relevance of the Model to Dyslexic Subtypes
Theoretically, subtypes of reading and spelling failure among dyslexics may occur when individuals fail to acquire mastery over one or more of the three strategies, or experience greater difficulty acquiring one strategy than the others (Frith 1985, 1986; Snowling 1987). The reading deficit may in turn be a function of more basic cognitive or linguistic difficulties. Snowling (1987) and Frith (1985, 1986) hypothesized that most dyslexic children have a specific language disorder that interferes with development of the phonological strategy. They may compensate for the deficit partially by means of the logographic and orthographic strategies or by dint of instruction in phonics, but sensitive enough tests will reveal a problem in processing phonological features of printed and spoken words. A second subgroup may exist with adequate phonological skills, but deficits in processing orthographic features. The source of such a deficit is not clear, but may involve difficulties in visual segmentation and analysis of
142
THEORETICAL SEEDS
printed words (Seymour and MacGregor 1984) or general difficulties in accessing verbal information in long-term memory (Perfetti 1985). Boder (1973) identified subgroups deficient in phonological processing (dysphonetic) and in remembering the visual gestalts of printed words (dyseidetic), as well as a group with both deficiencies, primarily on the basis of spelling errors. Seymour and MacGregor (1984) used reading measures to describe subgroups deficient in phonological processing, morphemic (orthographic) processing, and visual processing of letters. Detailed case studies have been reported that involve a severe phonological deficit (Temple and Marshall 1983) and a deficit in processing of wordspecific visual orthographic information (Coltheart et al. 1983). Each of the earlier studies attempted to subtype dyslexic children by reference to a normal reading group of the same chronological age, or by reference to the dyslexic sample itself. However, this approach leaves open the possibility that some of the deficit patterns observed in the dyslexic sample are a result of the child's reading problems rather than a fundamental cause of them (Bryant and Impey 1986; Snowling 1987). Comparisons to younger normal readers at the same reading level are necessary to establish that the deficits seen in dyslexics are not simply a function of lack of reading experience or the stage of reading development (Backman, Mamen, and Ferguson 1984; Bryant and Impey 1986; Snowling 1987). While there have been no systematic attempts to subtype dyslexics by reference to normal readers with the same level of reading achievement, several previous studies have found that dyslexic children exhibit, as a group, difficulties in phonological processing of printed words (Baddeley et al. 1982; Kochnower, Richardson and DiBenedetto 1983; Olson et al. 1985; Snowling 1980; Szeszulski and Manis 1987) and phonemic analysis of spoken words (Bradley and Bryant 1978; Doehring et al. 1981). In contrast, dyslexics appear to have the same level of skill as reading level controls in the visual-orthographic aspects of reading (Olson et al. 1985; Snowling 1980). However, a possibility not considered by previous studies is that subgroups of dyslexic children may exhibit difficulties in visual or orthographic processing relative to a reading level comparison group. In contrast to the notion of subgroups with specific deficits, several recent authors (e.g., Beech and Harding 1984; Treiman and Hirsh-Pasek 1985) have proposed that dyslexics differ from normal readers only in overall rate of development. These authors appear to assume that dyslexia is a rather broad-based delay in language development that retards reading and spelling development as a whole, rather than affecting specific components of reading and spelling. An important prediction of this developmental lag position is that dyslexics will not differ in relative mastery of reading/spelling strategies from a group of younger normal readers at the same level of reading achievement. In addition, the variability in reading strategies among the two groups should be similar. In the remainder of the paper we will describe a test battery designed
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
143
to measure degree of development of the phonological and orthographic strategies in reading and spelling, as well as associated visual and linguistic skills, and report results of a study applying the test battery to 50 dyslexic children and groups of normal readers matched on reading level.
The Test Battery Phonological Tasks The first three tasks were reading tasks that measured accuracy and speed in translating printed letter strings into a phonological code. Nonword pronunciation. Subjects were required to pronounce 36 nonwords as rapidly and accurately as possible into a microphone. The stimuli were displayed in lowercase letters on the screen of an Apple II + microcomputer. Stimuli included one and two syllable items that varied in spelling-to-sound complexity and length (e.g., hug, flirt, plune, spow, stining, lutpode, docious). The computerized format enabled both accuracy and reaction time (RT) to be recorded. Nonword verification. The experimenter pronounced a nonword aloud twice and then a nonword was displayed on the computer screen. The subject decided whether the spoken and printed nonwords had the same sound ("grobe"-grobe) or not (" crich'-crish ), and pressed one of two buttons labeled YES and NO. The items displayed on the screen either matched the prompt word exactly or differed by one phoneme. There were 24 trials. Nonword matching. Two nonwords were presented together on the screen, and subjects decided whether they had the same sound (sate-saiD or not (noff-noaf). Once again, matching items had the same sound and nonmatching items differed by only one phoneme. There were 20 trials. Nonword spelling. Subjects were required to spell nonwords of increasing length and complexity (e.g., "dit," "fope," "gleed," "chame," "trouded," "prequebly"). Subjects received a score based on the number of phonetically appropriate spellings of the nonwords. This task was designed to measure the range of application of the phonological strategy, as used in spelling. Sound deletion. Subjects were required to say how a nonword would sound after removing a single phoneme (e.g., pronounce "sparf" without the/p/). There were 22 trials. Subjects received a single score for total number correct. This task was designed to measure skill in phonemic analysis and segmentation in a nonreading format. Rhyme generation. Subjects were given a series of eight prompt words and had to say aloud as many words that rhymed with them as possible in 20 seconds. There were eight prompt words: (hat, said, mail, fame, mend, stumble, flower, and dealing). Any English word, except for proper nouns, was accepted as a response. Subjects received a single score for the total number of items generated. This task measured skill in
144
THEORETICAL SEEDS
phonemic analysis and word retrieval based on sound properties, in a speeded format. Category generation. Subjects were given a series of eight categories (e.g., things to build with, things that can fly). After being prompted with a categor~ subjects had to generate exemplars for the category in the same fashion as the rhyme generation task. This task was designed as a speeded measure of word retrieval that was similar to the rhyme generation task, but did not require attention to the phonological properties of the words. Orthographic Tasks Three tasks were devised that measured accuracy and speed of retrieving exact word spellings from memory. Lexical verification. The experimenter pronounced a word, and the subject was shown a letter string on the screen. The subject had to decide whether the letter string was a correct spelling of the word ("woman"woman) or not ("street"-streat) and press the appropriate button (YES or NO). The incorrect spellings were nonwords that sounded like the target word when pronounced. There were 42 trials. Lexical decision. Subjects decided whether a letter string on the screen was a correctly spelled word (south) or not (stoan), and pressed YES or NO. Again, incorrectly spelled words were nonwords that sounded like a real word when pronounced. There were 42 trials. Homonym matching. The experimenter read a homonym aloud and used it in a sentence. The child decided if the word shown subsequently on the computer screen was the correct homonym or not, and pressed the YES or NO button. An example of a YES trial is: "week-Monday is the first day of the week'-week). An example of a NO trial is: "wait-Wait for me"-weight. There were 42 trials. The rationale behind these three tasks was that the subject must use the orthographic strategy to recognize the correct spelling of a word. Reliance on the phonological strategy would produce many errors, since all of the incorrect choices were homophones of the correct word. Word spelling. Subjects spelled irregular words of ascending difficulty (e.g., two, people, enough, beauty, schedule, psychology). Since irregular words do not follow sound-to-spelling rules of English, it was assumed that subjects had to retrieve the spelling directly from memory rather than by relying on phonological processing. Visual Tasks Two tasks were designed to measure the accuracy and efficiency of fundamental visual skills that are relevant to reading, such as left-to-right scanning and rapid letter recognition. Accuracy and RT were recorded for both tasks. Visual match-letters. Subjects were shown two rows of consonants (4, 5, or 6 letters in length) on the computer screen and had to decide as
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
145
quickly as possible if they matched or not and press the appropriate button. Mismatching rows typically differed by only a single feature of one letter to increase the demands of the task on visual discrimination (e.g., mhkcrq-mbkcrq). Unfamiliar and unpronounceable letter strings were used to obviate use of the phonological or orthographic strategies. Visual match-shapes. Subjects were shown two rows of letter-like shapes (3, 4, or 5 items in length) and decided whether they matched ( ,q 6, G r~ _ ,q 6` G r~ ) or not ( ~" ~- ,q -~ - ~" K- ,,q ½ ). The shapes were composed of features of lowercase letters. It was assumed that the use of novel visual symbols would provide a more sensitive index of visual processing difficulties than familiar letters, as dyslexic subjects may have overlearned the letters.
Subjects The reading level comparison group consisted of 40 normal readers in grades 1-5 (age range 6:10 to 10:7) recruited from three schools serving a predominantly middle class area. The normal readers had IQ scores ranging from 94 to 123 (mean of 107.3), prorated from the Vocabulary and Block Design subtests of the WISC-R (Wechsler 1974) and were reading at the 50th percentile or above on the Woodcock Word Identification Test (WRMT) (Woodcock 1973) (mean percentile = 79.6). There were 29 boys and 11 girls. The dyslexic sample consisted of 50 children in grades 4-8 (age range 9:2 to 14:9). They were recruited from Special Education classes at 14 public schools in the same district as the normal readers, two private schools for learning-disabled children, and by means of a mailer distributed to local Orton Dyslexia Society members. To qualify for the stud~ they had to have a WISC-R IQ score of 90 or above, and a WRMT word identification score of 35th percentile or less (on the same two tests given the normal readers). The mean IQ of the final sample was 108.4 (range 91-135) and the mean reading percentile was 13.6 (range 1-35). There were 36 boys and 14 girls. According to school records and interviews with parents, none of the children had a history of neurological disease or damage, lack of educational opportunity, or serious emotional disorder. The criterion of 35th percentile or less on the reading measure was somewhat higher than is typical in the dyslexia literature. It was chosen to maximize the probability of finding subgroups within the dyslexic population.
Results We divided the normal readers into three groups based on reading grade level on the Woodcock: 2.2-3.5, 3.6-4.4, and 4.5-5.8. Dyslexics were assigned to one of these three groups. The mean reading grade lev-
146
THEORETICAL SEEDS
els and IQ scores of the dyslexics and normals in each group were very similar. We will report the results of group comparisons, followed by a classification of dyslexics into subgroups based on deviation from the scores of the reading level comparison groups. Group Comparisons Table I shows mean percent correct and mean latency on each of the reading tasks in the battery. Entries for the decision tasks (nonword verification, lexical decision, visual match-letters, etc.) are based on scores for YES and N O trials combined, as there were no important interactions between trial t y p e and subject group (in most conditions, performance was better for all groups on YES trials). Analyses of variance and t-tests were used to evaluate the significance of group differences. The significance level of comparisons b e t w e e n m e a n s by t-test is indicated by asterisks in Table I. The dyslexics were significantly less accurate or slower than the normal readers, collapsing across reading level, on all of the n o n w o r d reading measures except n o n w o r d verification and n o n w o r d matching RT. It can be seen in Table I that a n u m b e r of the comparisons between dyslexics and normal readers at each reading level reached the .01 level by t-test. Dyslexics were inferior to the normal readers, collapsing across reading
Table I Mean RT (in ms) and Percent Correct on the Reading Measures for Dyslexics and Normal Readers at Three Reading Levels Reading Level
1 Variable Nonword Pron. Nonword Verif. Nonword Match Lexical Verif. Lexical Dec. Homonym Verif. Visual Match-L Visual Match-S
2
Dyslexic Normal RT % RT % RT % RT % RT % RT % RT % RT %
4309 42.7 2152 66.4 4917 62.1 1986 71.0 2012 67.2 1798 62.2 2160 88.5 3382 91.8
*p
In this paper we apply a developmental model of reading to the question of dyslexic subtypes. Groups of normal readers (n = 40) and dyslexic children (n = 50), matched on reading level and IQ, were given a comprehensive test battery measuring level of development of visual, phonological, and orthographic skills. As a group, dystexics deviated from normal readers of equivalent reading achievement primarily in phonological skills (spelling-to-sound translation and phonemic analysis), although limited differences in knowledge of word-specific spellings were also observed. Dyslexics were superior to the younger normal readers in visual processing of print. Analysis of individual data by reference to the reading level control group revealed three major subgroups: a group with a specific deficit in phonological processing of print (52 percent), a group with deficits in processing both the phonological and orthographic features of printed words (24 percent), and a group with phonological deficits in language (8 percent). The remainder of the sample (I6 percent) had specific deficits in visual or orthographic processing of print, in spelling, or did not differ from the control group. The data support the view that most developmental dyslexics have a specific language disorder involving some aspect of phonological processing. However, small subgroups with very different configurations of reading and nonreading difficulties may exist as well. O n e of the continuing questions of interest in the field of dyslexia concerns the nature of possible s u b t y p e s of reading a n d l a n g u a g e disability a m o n g the dyslexic population. It is n o w clear that dyslexic children This research was supported by an NICHD grant to the first author (USPHSgrant 1 R23 HD20231). We thank the Los Angeles and Orange County chapters of The Orton Dyslexia Society, the Torrance Unified School District, and Landmark West School for their cooperation in obtaining subjects for the study. The test measures were developed jointly by the first two authors. Annals of Dyslexia, Vol. 38, 1988. Copyright ©1988 by The Orton Dyslexia Society ISSN 0474-7534
139
140
THEORETICAL SEEDS
(as they have traditionally been defined) are a heterogeneous group (Boder 1973; Doehring et al. 1981; Lovett 1984; Satz and Morris 1981; Seymour and MacGregor 1984). Many researchers and educators recognize that we must understand this heterogeneity in order to cope more effectively with questions of diagnosis, treatment, and etiology. A number of subtyping schemes have been proposed. Some focus on analysis of reading and spelling performance (e.g., Boder 1973; Doehring et al. 1981;Seymour and MacGregor, 1984), and some on patterns of neuropsychological test performance (Mattis, French, and Rapin 1975; Petrauskas and Rourke 1979; Satz and Morris 1981). At present there is lack of agreement on how best to categorize dyslexic children. A recent and promising approach can be found in the work of Frith and others (Frith 1985; Frith 1986; Seymour and MacGregor 1984; Snowling 1987), who have used a cognitive-developmental framework to analyze variations in reading, spelling, and language skills among dyslexics. In this paper, we will present a developmental model of reading and related skills, use it to derive predictions about the nature of subgroups, describe a comprehensive battery of tests derived from the model, and present data we have collected on dyslexic children and a control group of normal readers.
Model of Reading Development Frith (1985, 1986) and Seymour and MacGregor (1984) assume that individuals learn to read and spell by means of three strategies, arranged in a developmental sequence. The earliest strategy, the logographic strategy, is visually based. The child recognizes words "by sight," that is by means of salient visual features such as the first and last letters, the outline of the word, etc. The second strategy, the phonological (or alphabetic) strategy, involves sequential decoding of spelling units (letters and letter clusters) into the phonemes of the language. Once children have sufficient knowledge of spelling-sound correspondences and decoding rules, they can identify unfamiliar printed words that are in their speaking vocabulary. The third strategy, the orthographic strategy, allows the reader to recognize words based on each word's unique pattern of letter identities and letter order, without the need to translate the letters into a phonological code. It is assumed that the reader stores information about each morpheme (a word or grammatical unit, such as -ed, contained in a word) in a mental lexicon and learns specific associations between the spelling of each morpheme and its meaning. It has been argued that this is the dominant strategy for skilled readers (Jorm and Share 1983; McCusker, Hillenger, and Bias, 1981). According to Frith (1985, 1986), immature readers employ only the logographic strategy. The phonological strategy actually develops first in the context of learning to spell and soon becomes the dominant method of
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
141
identifying unfamiliar words and spelling familiar words. Gradually the child amasses enough knowledge of specific morphemes and their spelling patterns to begin using the orthographic strategy to read and eventually to spell. The advantage of using the orthographic strategy is twofold: words can be read and spelled using larger units (such as tion, ight), and words with irregular spelling-sound correspondences (e.g., colonel, island) can be read and spelled successfully. In the present study we assume that the two more advanced strategies continue developing after they are initially acquired. The orthographic strategy increases its range of application by constantly adding new morphemes to the lexicon and, possibly, by increasing the size of morpheme that can be immediately recognized. The phonological strategy increases its range by adding larger and more complex spelling-tosound correspondences. In addition, both strategies continue to increase in efficiency or automaticity with practice. Rapid, efficient word recognition is thought to be one of the principal determinants of progress in reading comprehension in the early grades, and is an important correlate of reading ability at a variety of ages (Carr 1981; Manis and Morrison 1985; Stanovich 1980). The phonological and orthographic reading strategies do not develop in isolation from other language and cognitive skills. To function propefly, both strategies require fundamental visual skills (processing print sequentially from left to right, visually recognizing key spelling patterns, and visually segmenting words into appropriate spelling patterns or morphemes), as well as a variety of language skills (auditory discrimination, phonemic analysis, adequate oral vocabulary, syntactic development, etc.) (Carr 1981; Liberman 1982; Manis and Morrison 1985; Mann 1984; Stanovich 1986). Relevance of the Model to Dyslexic Subtypes
Theoretically, subtypes of reading and spelling failure among dyslexics may occur when individuals fail to acquire mastery over one or more of the three strategies, or experience greater difficulty acquiring one strategy than the others (Frith 1985, 1986; Snowling 1987). The reading deficit may in turn be a function of more basic cognitive or linguistic difficulties. Snowling (1987) and Frith (1985, 1986) hypothesized that most dyslexic children have a specific language disorder that interferes with development of the phonological strategy. They may compensate for the deficit partially by means of the logographic and orthographic strategies or by dint of instruction in phonics, but sensitive enough tests will reveal a problem in processing phonological features of printed and spoken words. A second subgroup may exist with adequate phonological skills, but deficits in processing orthographic features. The source of such a deficit is not clear, but may involve difficulties in visual segmentation and analysis of
142
THEORETICAL SEEDS
printed words (Seymour and MacGregor 1984) or general difficulties in accessing verbal information in long-term memory (Perfetti 1985). Boder (1973) identified subgroups deficient in phonological processing (dysphonetic) and in remembering the visual gestalts of printed words (dyseidetic), as well as a group with both deficiencies, primarily on the basis of spelling errors. Seymour and MacGregor (1984) used reading measures to describe subgroups deficient in phonological processing, morphemic (orthographic) processing, and visual processing of letters. Detailed case studies have been reported that involve a severe phonological deficit (Temple and Marshall 1983) and a deficit in processing of wordspecific visual orthographic information (Coltheart et al. 1983). Each of the earlier studies attempted to subtype dyslexic children by reference to a normal reading group of the same chronological age, or by reference to the dyslexic sample itself. However, this approach leaves open the possibility that some of the deficit patterns observed in the dyslexic sample are a result of the child's reading problems rather than a fundamental cause of them (Bryant and Impey 1986; Snowling 1987). Comparisons to younger normal readers at the same reading level are necessary to establish that the deficits seen in dyslexics are not simply a function of lack of reading experience or the stage of reading development (Backman, Mamen, and Ferguson 1984; Bryant and Impey 1986; Snowling 1987). While there have been no systematic attempts to subtype dyslexics by reference to normal readers with the same level of reading achievement, several previous studies have found that dyslexic children exhibit, as a group, difficulties in phonological processing of printed words (Baddeley et al. 1982; Kochnower, Richardson and DiBenedetto 1983; Olson et al. 1985; Snowling 1980; Szeszulski and Manis 1987) and phonemic analysis of spoken words (Bradley and Bryant 1978; Doehring et al. 1981). In contrast, dyslexics appear to have the same level of skill as reading level controls in the visual-orthographic aspects of reading (Olson et al. 1985; Snowling 1980). However, a possibility not considered by previous studies is that subgroups of dyslexic children may exhibit difficulties in visual or orthographic processing relative to a reading level comparison group. In contrast to the notion of subgroups with specific deficits, several recent authors (e.g., Beech and Harding 1984; Treiman and Hirsh-Pasek 1985) have proposed that dyslexics differ from normal readers only in overall rate of development. These authors appear to assume that dyslexia is a rather broad-based delay in language development that retards reading and spelling development as a whole, rather than affecting specific components of reading and spelling. An important prediction of this developmental lag position is that dyslexics will not differ in relative mastery of reading/spelling strategies from a group of younger normal readers at the same level of reading achievement. In addition, the variability in reading strategies among the two groups should be similar. In the remainder of the paper we will describe a test battery designed
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
143
to measure degree of development of the phonological and orthographic strategies in reading and spelling, as well as associated visual and linguistic skills, and report results of a study applying the test battery to 50 dyslexic children and groups of normal readers matched on reading level.
The Test Battery Phonological Tasks The first three tasks were reading tasks that measured accuracy and speed in translating printed letter strings into a phonological code. Nonword pronunciation. Subjects were required to pronounce 36 nonwords as rapidly and accurately as possible into a microphone. The stimuli were displayed in lowercase letters on the screen of an Apple II + microcomputer. Stimuli included one and two syllable items that varied in spelling-to-sound complexity and length (e.g., hug, flirt, plune, spow, stining, lutpode, docious). The computerized format enabled both accuracy and reaction time (RT) to be recorded. Nonword verification. The experimenter pronounced a nonword aloud twice and then a nonword was displayed on the computer screen. The subject decided whether the spoken and printed nonwords had the same sound ("grobe"-grobe) or not (" crich'-crish ), and pressed one of two buttons labeled YES and NO. The items displayed on the screen either matched the prompt word exactly or differed by one phoneme. There were 24 trials. Nonword matching. Two nonwords were presented together on the screen, and subjects decided whether they had the same sound (sate-saiD or not (noff-noaf). Once again, matching items had the same sound and nonmatching items differed by only one phoneme. There were 20 trials. Nonword spelling. Subjects were required to spell nonwords of increasing length and complexity (e.g., "dit," "fope," "gleed," "chame," "trouded," "prequebly"). Subjects received a score based on the number of phonetically appropriate spellings of the nonwords. This task was designed to measure the range of application of the phonological strategy, as used in spelling. Sound deletion. Subjects were required to say how a nonword would sound after removing a single phoneme (e.g., pronounce "sparf" without the/p/). There were 22 trials. Subjects received a single score for total number correct. This task was designed to measure skill in phonemic analysis and segmentation in a nonreading format. Rhyme generation. Subjects were given a series of eight prompt words and had to say aloud as many words that rhymed with them as possible in 20 seconds. There were eight prompt words: (hat, said, mail, fame, mend, stumble, flower, and dealing). Any English word, except for proper nouns, was accepted as a response. Subjects received a single score for the total number of items generated. This task measured skill in
144
THEORETICAL SEEDS
phonemic analysis and word retrieval based on sound properties, in a speeded format. Category generation. Subjects were given a series of eight categories (e.g., things to build with, things that can fly). After being prompted with a categor~ subjects had to generate exemplars for the category in the same fashion as the rhyme generation task. This task was designed as a speeded measure of word retrieval that was similar to the rhyme generation task, but did not require attention to the phonological properties of the words. Orthographic Tasks Three tasks were devised that measured accuracy and speed of retrieving exact word spellings from memory. Lexical verification. The experimenter pronounced a word, and the subject was shown a letter string on the screen. The subject had to decide whether the letter string was a correct spelling of the word ("woman"woman) or not ("street"-streat) and press the appropriate button (YES or NO). The incorrect spellings were nonwords that sounded like the target word when pronounced. There were 42 trials. Lexical decision. Subjects decided whether a letter string on the screen was a correctly spelled word (south) or not (stoan), and pressed YES or NO. Again, incorrectly spelled words were nonwords that sounded like a real word when pronounced. There were 42 trials. Homonym matching. The experimenter read a homonym aloud and used it in a sentence. The child decided if the word shown subsequently on the computer screen was the correct homonym or not, and pressed the YES or NO button. An example of a YES trial is: "week-Monday is the first day of the week'-week). An example of a NO trial is: "wait-Wait for me"-weight. There were 42 trials. The rationale behind these three tasks was that the subject must use the orthographic strategy to recognize the correct spelling of a word. Reliance on the phonological strategy would produce many errors, since all of the incorrect choices were homophones of the correct word. Word spelling. Subjects spelled irregular words of ascending difficulty (e.g., two, people, enough, beauty, schedule, psychology). Since irregular words do not follow sound-to-spelling rules of English, it was assumed that subjects had to retrieve the spelling directly from memory rather than by relying on phonological processing. Visual Tasks Two tasks were designed to measure the accuracy and efficiency of fundamental visual skills that are relevant to reading, such as left-to-right scanning and rapid letter recognition. Accuracy and RT were recorded for both tasks. Visual match-letters. Subjects were shown two rows of consonants (4, 5, or 6 letters in length) on the computer screen and had to decide as
A DEVELOPMENTALPERSPECTIVEON SUBTYPES
145
quickly as possible if they matched or not and press the appropriate button. Mismatching rows typically differed by only a single feature of one letter to increase the demands of the task on visual discrimination (e.g., mhkcrq-mbkcrq). Unfamiliar and unpronounceable letter strings were used to obviate use of the phonological or orthographic strategies. Visual match-shapes. Subjects were shown two rows of letter-like shapes (3, 4, or 5 items in length) and decided whether they matched ( ,q 6, G r~ _ ,q 6` G r~ ) or not ( ~" ~- ,q -~ - ~" K- ,,q ½ ). The shapes were composed of features of lowercase letters. It was assumed that the use of novel visual symbols would provide a more sensitive index of visual processing difficulties than familiar letters, as dyslexic subjects may have overlearned the letters.
Subjects The reading level comparison group consisted of 40 normal readers in grades 1-5 (age range 6:10 to 10:7) recruited from three schools serving a predominantly middle class area. The normal readers had IQ scores ranging from 94 to 123 (mean of 107.3), prorated from the Vocabulary and Block Design subtests of the WISC-R (Wechsler 1974) and were reading at the 50th percentile or above on the Woodcock Word Identification Test (WRMT) (Woodcock 1973) (mean percentile = 79.6). There were 29 boys and 11 girls. The dyslexic sample consisted of 50 children in grades 4-8 (age range 9:2 to 14:9). They were recruited from Special Education classes at 14 public schools in the same district as the normal readers, two private schools for learning-disabled children, and by means of a mailer distributed to local Orton Dyslexia Society members. To qualify for the stud~ they had to have a WISC-R IQ score of 90 or above, and a WRMT word identification score of 35th percentile or less (on the same two tests given the normal readers). The mean IQ of the final sample was 108.4 (range 91-135) and the mean reading percentile was 13.6 (range 1-35). There were 36 boys and 14 girls. According to school records and interviews with parents, none of the children had a history of neurological disease or damage, lack of educational opportunity, or serious emotional disorder. The criterion of 35th percentile or less on the reading measure was somewhat higher than is typical in the dyslexia literature. It was chosen to maximize the probability of finding subgroups within the dyslexic population.
Results We divided the normal readers into three groups based on reading grade level on the Woodcock: 2.2-3.5, 3.6-4.4, and 4.5-5.8. Dyslexics were assigned to one of these three groups. The mean reading grade lev-
146
THEORETICAL SEEDS
els and IQ scores of the dyslexics and normals in each group were very similar. We will report the results of group comparisons, followed by a classification of dyslexics into subgroups based on deviation from the scores of the reading level comparison groups. Group Comparisons Table I shows mean percent correct and mean latency on each of the reading tasks in the battery. Entries for the decision tasks (nonword verification, lexical decision, visual match-letters, etc.) are based on scores for YES and N O trials combined, as there were no important interactions between trial t y p e and subject group (in most conditions, performance was better for all groups on YES trials). Analyses of variance and t-tests were used to evaluate the significance of group differences. The significance level of comparisons b e t w e e n m e a n s by t-test is indicated by asterisks in Table I. The dyslexics were significantly less accurate or slower than the normal readers, collapsing across reading level, on all of the n o n w o r d reading measures except n o n w o r d verification and n o n w o r d matching RT. It can be seen in Table I that a n u m b e r of the comparisons between dyslexics and normal readers at each reading level reached the .01 level by t-test. Dyslexics were inferior to the normal readers, collapsing across reading
Table I Mean RT (in ms) and Percent Correct on the Reading Measures for Dyslexics and Normal Readers at Three Reading Levels Reading Level
1 Variable Nonword Pron. Nonword Verif. Nonword Match Lexical Verif. Lexical Dec. Homonym Verif. Visual Match-L Visual Match-S
2
Dyslexic Normal RT % RT % RT % RT % RT % RT % RT % RT %
4309 42.7 2152 66.4 4917 62.1 1986 71.0 2012 67.2 1798 62.2 2160 88.5 3382 91.8
*p
