Part II Identifying the Deficits Underlying Dyslexia The three papers in this section represent valuable research on three consistent predictors of reading disability: rapid serial naming, phonological memory, and phonological awareness. In the first paper, Lynn Snyder and Doris Downey apply experimentally analytic techniques to ask whether rapid serial naming deficits stem from speed of access to a given label (reaction time) or from post-access speed of articulation. Both components of this task were implicated: schoolchildren with reading disability were significantly slower to respond to the items and took significantly longer to pronounce each name than their normal-reading peers. This study is one of the first to isolate some of the many components of the rapid serial naming task, to ask just what it is that makes this complex task so difficult for the poor reader. Snyder and Downey's success in isolating relevant components invites further analysis of this remarkably consistent correlate of reading success. The second paper is based on an award winning dissertation (The Orton Dyslexia Society 1993) conducted by Brenda Stone under the direction of Susan Brady. This paper features phonological memory as a powerful correlate of reading disability: poor readers displayed m e m o r y deficits in c o m p a r i s o n to I Q - m a t c h e d controls a n d to younger normal readers. The methodological rigor applied in this study should serve to mitigate a number of concerns raised about less well-controlled studies; the results provide compelling evidence that basic phonological processing deficits underlie reading disability. In this study, measures tapping the accuracy of phonological processing were strongly associated with reading performance. However, contrary to the Snyder and Downey results, speed of phonological processing was unrelated either to reading, memory accuracy, or rapid serial naming. We are left with some intriguing questions about the role of speed in explaining phonological problems in less-skilled readers. In Snyder and Downey, with accuracy at ceiling, speed was the telling factor; in Stone and Brady, where performance on tasks was less accurate, speed failed to account for variability in children's scores. While both articles concur in their emphasis on phonological processing, it will be of considerable interest to see how the speed versus accuracy question gets resolved in future research.

29

30

IDENTIFYING THE DEFICITS UNDERLYING DYSLEXIA

The third paper, by Nathlie Badian, explores the predictive value of a small number of kindergarten and first grade measures for literacy in the later elementary years. Although only some of the skills now known to be associated with reading performance were assessed prior to school entry, preschool ability to name 13 letters and to do a visual matching task were found to be strong predictors of reading and spelling in the sixth grade. These early measures become increasingly significant over the years, whereas phonological awareness (first assessed in grade 1) explained more of the variance in reading in the early elementary years; it remains to be determined whether it is the timing or the content of these early measures that determine their later predictive value. This study raises important questions about the changing relevance of underlying cognitive skills for different aspects of reading skill and contributes to the study of the prognostic value of various pre-reading tasks. These three papers make valuable contributions to our understanding of the association between phonological processing and reading skill. They not only strengthen the evidence for a close association between phonology and reading, they also substantially advance our understanding through careful scrutiny of the measures under study. Together, they suggest that different subskills may have distinct roles to play in influencing success on the various components of reading and that these associations may vary over time. It is only with this level of care that we will slowly be able to unravel the necessary prerequisites to success on the various stages and components of reading.

Serial Rapid Naming Skills in Children with Reading Disabilities Lynn S. Snyder California State University Long Beach, California

Doris M. Downey University of Colorado Boulder, Colorado

This study compared the performance of children with reading disability (RD) and normal reading achievement (NRA) on tasks of serial rapid naming, verbal fluency, letter-based word retrieval, and articulatory speed. The groups, composed of children at two discrete age levels, one younger and one older, were matched for age, gender, and neighborhood school. Analyses of the on-line measurement of the children's serial rapid naming indicated that the children with RD had This work was supported in part by Basil O'Connor Research Grant 5-340 from the March of Dimes Birth Defects Foundation, a Scholarly and Creative Activity Award and sabbatical leave from California State University, Long Beach and grant #DCO 1904-01 from the National Institute on Deafness and Other Communication Disorders of the NIH to the first author. The invaluable contributions of Elizabeth Bates, the faculty and staff of the Center for Research in Language at the University of California, San Diego, David Balota, Dawn Godley, Sheila Goetz, Anne Fowler, Hollis Scarborough, two anonymous reviewers and the participation of the Colorado families involved in the Denver Reading Study are ackaiowledged with g_T_atitude. Correspondence may be sent to: L. Snyder, Dept. of Communicative Disorders, California State University, Long Beach, 1250 Bellflower Blvd., Long Beach, CA 90840.

31

32

IDENTIFYING THE DEFICITS UNDERLYINGDYSLEXIA

significantly larger reaction times and production durations than their NRA peers despite similar levels of accuracy. They also performed significantly worse on the categorical verbal fluency task, the letter-based word retrieval task, and the test of articulatory speed. The findings suggest that both access and post-access processes, such as oral-motor inefficiency that extends the duration of word production, may be implicated in the slower serial rapid naming that has typified many samples of children with RD. Denckla's (1972) classic study of serial rapid naming in children with reading disabilities (RD) initiated more than two decades of research documenting the difficulty these youngsters experience when asked to name a series of items rapidly. Denckla's (1972) and Denckla and Rudel's (1976 a,b) investigations of rapid automatized naming (RAN) demonstrated that children with RD were slower and often less accurate at serial rapid naming than their reading peers. Wolf's studies (1982, 1986), which extended and clarified the nature of these findings, indicated that the slow serial naming speed of children with RD differentiated them from their peers with normal reading achievement (NRA). In fact, the difficulty that children with RD have with rapid naming seems to be apparent long before the children are introduced to reading. Prospective studies conducted by Badian et al. (1990), Wagner, Torgesen, and Rashotte (1994), and Wolf and Obregon (1992) found that very young children's serial rapid naming skills were predictive of their success with reading a couple of years later. Similarly, Walsh, Price, and Gillingham (1988) found that poor naming skill at early grade levels was predictive of later reading problems. The slow serial naming of children with RD also seems to present an enduring problem for this group of youngsters. Evidence from Wolf and Obregon's (1992) longitudinal study and Wolff, Michel, and Ovrut's (1990 a,b) studies of adolescent RD children, taken together with the findings from Felton, Naylor, and Wood's (1990) and Pennington et al's (1990) studies of adult dyslexics, indicates that individuals with RD sustain continuing problems with serial rapid naming. Initially, researchers interpreted findings of slow serial naming as symptomatic of dysnomia or word finding problems in children with RD. Early psychoeducational research conducted by de Hirsch, Jansky, and Langford (1966) and Jansky and de Hirsch (1971) indicated that the accuracy of their perfor-

SERIAL RAPID NAM~NC SKILLS OF READING DISABLED CHILDREN

33

mance on single item naming tasks was lower than that of their NRA peers. On the other hand, later findings from Stanovich, Feeman, and Cunningham's (1983) and Katz and Shankweiler's (1985) experimental studies of naming in children with RD indicated that their reaction times (RTs) did not differ significantly from their normally reading controls on single item naming. Wagner, Torgesen, and Rashotte's (1994) recent work found significant differences between second graders with and without RD on discrete item naming; however, group differences on serial rapid naming speed appeared to reach higher levels of statistical significance. The relative lack of agreement among these findings on single item naming may be related to differences in subject selection criteria and the sampling paradigms employed. In contrast to these conflicting findings, there has been more consistent evidence that slow naming of item series or sequences seemed to differentiate children with RD from the control groups studied. In addition, there has been little agreement about the factors responsible for those observations of reduced naming speed and accuracy of children with RD. German (1982, 1993) has argued that the children may have deficient retrieval strategies, making retrieval of words from the lexicon inefficient and slow. Catts (1986), Katz (1986) and others have suggested that the well-documented difficulty that children with RD experience with phonological processing could result in weakly specified phonological representations for words, making access from the lexicon slower. Fowler (1991) has suggested that the segmental differentiation that occurs during normal phonological development may be deficient or delayed in children with RD, affecting their phonological processing, production, memory, and other aspects of oral and written language. Wolf (1991) has suggested that the subprocesses that serve both reading and serial naming may be similar and an underlying deficit in these subprocesses may be responsible for the observed correlation between reading disability and slow serial rapid naming. Specifically, Wolf suggested that the loci of the deficit may lie in the automatization of responses that require precise timing. She argued that such automatization is characteristic of both the language and motoric domains and hence, the paired problems observed in the motor coordination required for rapid continuous naming and for rapid manual tasks such as tapping patterns. Leonard's (1988) studies of children with specific langnage impairment (SLI) suggested that their slow performance on single item naming tasks may be related to poorly elaborated

34

IDENTIFYING THE DEFICITS UNDERLYINGDYSLEXIA

lexical stores, i.e., a storage-elaboration deficit or delay. In this instance, slower n a m i n g w o u l d be related to the d e l a y e d vocabulary development of children with SLI which results in vocabulary words that have been stored in their lexicons for shorter periods of time, with less of an opportunity to be elaborated by experience. The design of m a n y studies examining the serial rapid naming of children with RD has not permitted the investigation of all of the factors responsible for their slow naming because of the nature of the questions being addressed. To date, those studies using serial rapid naming tasks have employed the subjects' total naming time as the index of naming speed. Because existing studies have used this s u m m a r y variable, factors responsible for the slow speed may have been masked. It is unclear whether the children with RD studied have taken longer to respond in general to the items, to retrieve the names of the items, to produce each item a n d / o r whether their production of non-target behaviors such as fillers, e.g., "I kinda think it's a . . . , " "well, I think it's a . . . " and back channel responses, e.g., "err," "uhmmmm," has inflated the total naming time. Some or all of these variables may be implicated in the slow serial naming speed observed in children with RD. Recent studies of naming in adults have been effective in identifying different loci of disruptions in naming behavior including post-access processes such as the speed of retrieval of motor plans and articulatory efficiency. These factors have been found to slow response times and extend the duration of the production of target words (Balota 1990; Balota and Duchek 1988). This finding has particular relevance for the study of naming in children because it has been known for some time that children's production durations (PDs) decrease as their articulatory productions mature and stabilize (Hutterflocher 1984). Children with RD have been reported to have some difficulty with some speech production tasks. For example, studies by Snowling (1981) and Catts (1986) found that children with RD make significantly more speech production errors than their reading skilled peers on phonologically complex or infrequently occurring stimuli, such as pseudowords. They suggested that the elevated error rates of children with RD may be related to underlying phonological processing deficits such as weakly specified phonological representations. Ackerman, Dykman, and Gardner's (1990) research also focused on the articulation rates of children with RD, but they examined their performance with well learned information. They found that children with

SERIAL RAPID NAMING SKILLS OF READING DISABLED CHILDREN

35

more severe dyslexia were slower in repetitive counting and naming alternating digits and letters than children with milder reading delays. It is of interest, however, that they found that their subjects' counting rate, but not their naming rate, was highly correlated with measures of phonological sensitivity. In light of the persisting difficulty that children with RD experience with rapid fine motor coordination of other kinds, e.g., finger tapping (Wolff, Michel, and Ovrut 1990a, b), it might be reasonable to expect to find their oral-motor production similarly affected secondary to a more general delay or deficit in fine motor coordination. Further, it is possible that post-access motor production inefficiency may affect the serial rapid naming of children with RD, in addition to other naming relevant factors such as general access, retrieval, specification of phonological representations, and the elaboration of lexical stores. Consequently, this study was undertaken to examine the serial rapid naming performance of children, supplementing the classical summary measure with on-line measures so that artifactual variables such as fillers and false starts could be controlled and the potential contribution of any post-access motor production factors to total naming time could be documented. Thus, the serial rapid naming performance of a sample of children with RD was compared to that of a group of NRA peers with respect to their response times, production durations, total naming times, and accuracy. METHODS SUBJECTS The subjects for this study were 30 children with RD and 30 children with NRA selected from the subject pool of the Denver Reading Study (Snyder and Downey 1991) at two different age levels. Subjects were sampled from the Denver Reading Study pool by taking all subjects from the RD group at each of the targeted age levels for whom there were complete files and who h a d NRA m a t c h e s w h o s e files were s i m i l a r l y complete. Further, all subjects' audio recordings of their performance on the serial rapid naming and articulatory rate tasks had to have a sufficiently high quality to permit accurate on-line measurement. This sampling heuristic resulted in 15 younger children with RD between the ages of 8 yr. 6 too. and 10 yr. 3 mo., with a m e a n age of 9 yrs. 4 too., a n d 15 older children w i t h RD between the ages of 12 years and 13 yr. 4 too., with a mean age

36

IDENTIFYING THE DEFICrFS UNDERLYING DYSLEXIA

of 12 yr. 7 mo. The RD group had been matched with children with NRA for age, gender, and neighborhood school. The NRA matches included 15 younger children with NRA with a mean age of 9 yr. 7 mo. and 15 older children with NRA with a mean age of 12 yr. 8 mo. There were 24 boys and 6 girls in each group. All subjects came from middle-income monolingual English homes in the Denver-Boulder, Colorado metropolitan area. All subjects demonstrated normal psychosocial development according to district staffing team reports, had normal auditory acuity, and no uncorrected visual acuity deficits. Further, all subjects demonstrated general intellectual performance in the normal range, with Full Scale IQs ranging from 92 to 135 on the Weschler Intelligence Scale for Children - Revised (WISC-R), (Weschler 1974). All children with NRA demonstrated at least grade level performance on one or more of the school districts' administered tests of silent reading comprehension. All of the children with RD had demonstrated reading performance at least two grade levels below current grade placement on the same school district measures. All RD children had also met the State of C o l o r a d o ' s d i s c r e p a n c y criterion, a 40% discrepancy between their performance on one or more standardized measures of reading achievement and the level of achievement predicted by their Full Scale IQ scores. MEASURES

All subjects were given a test of general cognitive abilities, the Weschler Intelligence Scale for Children - Revised, (WISC-R) (Weschler 1974), a reading comprehension measure, the Reading test of the Metropolitan Achievement Test (1976) and a word recognition test, the Reading Recognition subtest of the Peabody Individual Achievement Tests (PIAT) (Dunn and Markwardt 1970) to confirm group membership. In order to assess the children's serial naming, associated verbal (lexical) fluency, and letter-based word retrieval, the Producing Names on Confrontation, the Producing Word Associations subtests from the Clinical Evaluation of Language Function (CELF) (Wiig and Semel 1980), and the Controlled Word Association test from Benton's neuropsychological battery (Benton 1973) were administered. In addition, articulatory speed was assessed with a clinical test of oral diadochokinetic rate (Fletcher 1972). The accuracy and total time scores from the serial rapid naming task (reported for the larger sample in Snyder and Downey 1991), were also recorded for entry into the analyses for the present study.

SERIAL RAPID NAMING SKILLS OF READING DISABLED CHILDREN

37

The serial rapid n a m i n g subtest involved asking the child to n a m e as quickly as possible the colored shapes that a p p e a r in a series on a test booklet page placed before them. The test was preceded by two trials. O n the first trial the child was asked to name, as quickly as possible, the colors that appeared in rows of circles on a single test booklet page. On the second trial, the child was asked to name, as rapidly as possible, the geometric shapes that appeared in rows of black line d r a w i n g s on another test booklet page. The test plate for the serial rapid n a m i n g task contained six rows of six colored shapes, e.g., red circle, blue square, etc. on a single page. The child was asked to n a m e the color and shape of each of the 36 colored shapes as quickly as possible. The total n u m b e r of correct responses on the serial rapid n a m i n g task was recorded as well as the total time in seconds that it took the child to complete the task. Performance on the serial rapid n a m i n g task was audiotape recorded. The audiotapes w e r e played on a recorder interfaced with an IBM compatible c o m p u t e r interfaced with a voice activated relay. A p r o g r a m written in A s s e m b l y l a n g u a g e for an IBM compatible c o m p u t e r r e c o r d e d in milliseconds the subjects' reaction times, the time it takes the subjects to begin to produce each item, and production durations, the time it takes to p r o n o u n c e each item on the serial rapid n a m i n g task. The voice activated program m e a s u r e d subjects' reaction times from the voice offset of one item to the voice onset of the next item. Production durations w e r e m e a s u r e d from the onset of the subject's voice response to its offset. The p r o g r a m also p e r m i t t e d the deletion of all back channel, filler, a n d circumlocutionary responses that could inflate actual n a m i n g time. Calibration checks w e r e performed every four subjects to insure the precision of measurement. The w o r d fluency test involved asking the child to n a m e as m a n y animals as possible in a one m i n u t e time period. The child's score consisted of the n u m b e r of nonduplicated n a m e s p r o d u c e d d u r i n g the time period. The test task was p r e c e d e d b y one d e m o n s t r a t i o n trial in w h i c h the child w a s a s k e d to n a m e as m a n y foods as possible in one minute. The letter-based w o r d retrieval task e m p l o y e d , Benton's FAS Set test, involved asking the child to n a m e rapidly as m a n y w o r d s as possible that begin with "f" within a designated time period (one minute). The child was given a practice trial using the letter "t'. The child's score consisted of the n u m b e r of correct w o r d s p r o d u c e d within the time period. Lastly, each subject was administered a classical clinical test

38

IDENTIFYING THE DEFICITS UNDERLYINGDYSLEXIA

of articulatory speed. This task was preceded by three practice trials in which the examiner first modeled the target behavior, the repetition of a single syllable, e.g., [pAl, as quickly as possible. Then, the examiner asked the child to repeat that syllable as many times as possible within 3 seconds. The examiner indicated to the child when to start and stop. This practice trial was repeated two more times, once with the syllable [tA] and once with [kA]. Following these trials, the child was then asked to repeat the syllable string [pAtAkA] as many times as possible within 5 seconds. The child's score consisted of the rate recorded using the count-time method (Fletcher 1972), i.e., the number of times the entire three syllable string was produced during that time period. All subjects were tested individually and their responses were recorded on standardized test protocols and on audiotape. INSTRUMENTS A N D PROCEDURES

Audiotape recordings of the subjects' performance on the serial rapid naming task were played on a Marantz PMD 211 audiotape recorder whose output was interfaced with a Gerbrands 1341T voice activated relay interfaced with an IBM compatible 386/33 DX computer. RESULTS READING A N D IQ

Initial analyses of variance (ANOVAs) indicated significant differences between the RD and NRA groups on both reading measures and the IQ measures. The means and standard deviations for each reading and age group by age level appear in table I. The results of the ANOVA indicated significant reading group differences on the r e a d i n g c o m p r e h e n s i o n m e a s u r e (F(1,55) = 101.86, p =