UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Processes and strategies used by normal and disabled readers in analogical reasoning Potter, Margaret 1991

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1991_A2 P67.pdf [ 10.33MB ]
Metadata
JSON: 831-1.0100946.json
JSON-LD: 831-1.0100946-ld.json
RDF/XML (Pretty): 831-1.0100946-rdf.xml
RDF/JSON: 831-1.0100946-rdf.json
Turtle: 831-1.0100946-turtle.txt
N-Triples: 831-1.0100946-rdf-ntriples.txt
Original Record: 831-1.0100946-source.json
Full Text
831-1.0100946-fulltext.txt
Citation
831-1.0100946.ris

Full Text

PROCESSES AND STRATEGIES USED BY NORMAL AND DISABLED READERS IN ANALOGICAL REASONING BY MARGARET POTTER M.A., Lakehead U n i v e r s i t y , 1979 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF EDUCATION i n THE FACULTY OF GRADUATE STUDIES Department o f E d u c a t i o n a l P s y c h o l o g y and S p e c i a l E d u c a t i o n We a c c e p t t h i s t h e s i s as c o n f o r m i n g t o t h e r e q u i r e d s t a n d a r d THE UNIVERSITY OF BRITISH COLUMBIA F e b r u a r y 1991 C) M a r g a r e t P o t t e r , 1991 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of ff/iu^-f fs^ckolo^ j SpetXoJ fcdiA.CjxflOn The University of British Columbia Vancouver, Canada Date <?/ - 0/f. ~ /5~" DE-6 (2/88) Abstract The purpose of t h i s study was: 1) to i d e n t i f y reading d i s a b i l i t y subtypes among a sample of reading-disabled students using two c l a s s i f i c a t i o n methods, 2) to discover the processes and strategies used i n analogical reasoning by i n d i v i d u a l reading disabled and nonreading-disabled students through the method of componential analysis, and 3) to explore the r e l a t i o n s h i p between the processes and strategies used by disabled readers i n analogical reasoning and t h e i r membership i n a reading d i s a b i l i t y subtype. In Phase 1 of the study, groups of normal and disabled readers were established using Grade 5 students attending elementary schools i n a large urban area of Northwestern Ontario. The disabled sample of 77 students comprised 41 males and 3 6 females and the normal reader sample of 2 0 students comprised 7 males and 13 females. In Phase 2, the disabled and normal readers were i n d i v i d u a l l y administered the Boder Test of Reading-Spelling Patterns (Boder & J a r r i c o , 1982), the Peabody Picture Vocabulary Test - Revised (Dunn & Dunn, 1981), and subtests taken from the D u r r e l l Analysis of Reading D i f f i c u l t y (Durrell & Catterson, 1980) . The Schematic Picture Analogies Test (Sternberg & R i f k i n , 1979) was administered to students i n small groups. The f i r s t method of subtyping, the Boder te s t , f a i l e d to i d e n t i f y subtypes among the reading-disabled sample because the students were not as severely disabled as the c l i n i c - r e f e r r e d sample for which the t e s t was designed. The second method, which employed a h i e r a r c h i c a l agglomerative technique of c l u s t e r analysis using students' scores obtained on 23 reading and related variables, d i f f e r e n t i a t e d the normal readers from the disabled readers. Three c l u s t e r s emerged when the reading-disabled data were analyzed alone that were characterized by strengths and weaknesses i n t h e i r reading s k i l l s . Componential analysis of students' analogical reasoning data used mean solution latency as the c r i t e r i o n or dependent variables. Independent or predictor variables were associated with the systematically varied l e v e l of d i f f i c u l t y of each of 24 analogy booklets. Seven models theorized by Sternberg (1977) were f i t t e d to each in d i v i d u a l ' s booklet scores through multiple regression analysis and the preferred model chosen according to f i v e pre-determined c r i t e r i a (Sternberg & R i f k i n , 1979). Disabled readers were grouped according to the processes and strategies they used i n solving analogies. The normal reader group solved analogies as predicted but there was no re l a t i o n s h i p between membership i n a reading d i s a b i l i t y c l u s t e r and membership i n an analogy subgroup. None of the analogy subgroups could be characterized by t h e i r reading performance although the subgroup that used the most e f f i c i e n t model tended to have higher a b i l i t y than the other subgroups. Correlations between solution latency and i v reading and related variables for the normal readers showed that the more p r o f i c i e n t analogical reasoners were faster, more accurate readers and better comprehenders. Few s i g n i f i c a n t c o r relations were detected between solution latency and reading variables for the disabled readers. The lack of r e l a t i o n s h i p between the two systems i s perhaps the most su r p r i s i n g and paradoxical finding of the study. I t i s suggested that t h i s occurred because reading-disabled children, i r r e s p e c t i v e of the c l u s t e r to which they belong, may solve analogies i n a unique way, or because the bottom-up, content-driven nature of the reading task i s so fundamentally d i f f e r e n t from the top-down, content-free nature of the analogical reasoning task. Other explanations suggest that the use of measures at a macro l e v e l to form reading-disabled c l u s t e r s masks any r e l a t i o n s h i p with the analogical reasoning subgroups formed by measures at a micro l e v e l , or that component processing i s so s p e c i f i c to the i n d i v i d u a l that differences are buried within the subtypes implying the existence of subtypes within subtypes. Some of the implications for education are discussed. V TABLE OF CONTENTS Abstract i i LIST OF TABLES v i i i LIST OF FIGURES X LIST OF APPENDICES , . . . . x i i ACKNOWLEDGEMENT x i i i CHAPTER 1: INTRODUCTION 1 Reading D i s a b i l i t i e s 1 D i f f e r e n t i a l Approach 2 Information-Processing Approach 3 The Subtyping Approach 5 Componential Analysis 7 Outline of the Study 8 CHAPTER I I : REVIEW OF LITERATURE 11 Reading D i s a b i l i t i e s 11 Models of Reading 13 S p e c i f i c i t y of Reading D e f i c i t s 16 Decoding 16 P h o n o l o g i c a l A w a r e n e s s D e f i c i t s 16 Short-Term Memory D e f i c i t s 18 Speech Perception D e f i c i t s 19 Name Retrieval D e f i c i t s 2 0 Comprehension 21 Use of Context 21 Short-Term Memory 22 Metacognition 23 Limitations of the Information-Processing Approach 24 Summary of Information-Processing Theory and Research 2 5 Subtyping Research 2 5 The Contribution of Boder 2 6 S t a t i s t i c a l Methods 28 Q-Type Factor Analysis 2 8 H i e r a r c h i c a l Agglomerative Techniques 35 Limitations of the Subtyping Approach 50 v i Componential Analysis 52 Application of Componential Analysis 57 Sternberg's T r i a r c h Theory of Intelligence 73 Summary of Componential Analysis: Theory and Research 77 CHAPTER I I I : PURPOSE OF THE STUDY 79 Research Questions and Hypotheses 8 5 CHAPTER IV: METHODOLOGY 90 Phase 1 90 Target Population 90 Instruments Used 91 Stanford Diagnostic Reading Test 91 Canadian Cognitive A b i l i t y Test 93 Procedure 95 Data Preparation 97 Phase 2 99 Selection of Samples 99 Instruments Used 102 Peabody Picture Vocabulary Test - Revised 102 The Boder Test of Reading-Spe l l i n g Patterns 103 Durr e l l Analysis of Reading D i f f i c u l t y 107 Schematic Picture Analogies 110 Design 116 Procedure 118 Administration of Reading and Related Tests 118 Administration of the Schematic Picture Analogies Test 119 Scoring 121 Data Preparation 123 CHAPTER V: RESULTS 12 6 Reading Typology 12 6 The Boder Test of Reading-Spelling Patterns . . . . 127 Cluster Analysis 127 Box-and-Whisker Plots 12 9 A Comparison of Reading-Disability Clusters . 132 Summary of Clusters 145 v i i Componential Analysis 14 5 Inspection of the Data 14 5 Individual Regression Analysis 147 Individual Analysis of Group D 150 Individual Analysis of Group N 151 Analysis at the Group Level 155 Use of Components 155 Use of the Linear Combination Rule 157 Use of Exhaustive versus Self-terminating Mode 158 Solution Scores 159 Calculation of Component Scores 159 Summary of Componential Analysis 164 The Relationship Between Analogical Reasoning and Reading and Reading-Related S k i l l s 166 Cluster Membership Versus Model Subgroup Membership 166 A Comparison of Reading-Disability Model Subgroups 166 Summary of Model Subgroups 179 Correlations Between Analogy Data and Reading Data 182 Correlations Within group N 183 Correlations within the Model Subgroups . . . 185 Summary 186 CHAPTER VI: DISCUSSION 187 Reading Typology 187 Componential Analysis 188 Regression Models: Components and Strategies . 189 Component Scores 193 The Relationship Between Analogical Reasoning and Reading 194 Factors Underlying Strategy Choice 196 Limitations of the Study 199 Conclusion 2 01 Suggestions for Future Research 2 04 REFERENCES 206 v i i i LIST OF TABLES Table 1 Subtypes i d e n t i f i e d by Q-factor analysis 3 0 2 Subtypes i d e n t i f i e d using neuropsychological variables and h i e r a r c h i c a l agglomerative techniques 41 3 Subtypes i d e n t i f i e d using behavioural variables and h i e r a r c h i c a l agglomerative techniques . . . 46 4 Subtypes i d e n t i f i e d using neuropsychological and/or language variables and h i e r a r c h i c a l agglomerative techniques 4 9 5 Sternberg's t h e o r e t i c a l models 63 6 Sex d i s t r i b u t i o n , age, nonverbal a b i l i t y , and reading comprehension for grade 5 students . . 98 7 Sex d i s t r i b u t i o n , age, nonverbal a b i l i t y , and reading comprehension for Group D and Group N . 101 8 The subtypes of Boder 105 9 Predictor variables for regression 114 10 Models for Regression 115 11 Regression equations 117 12 Analysis of variance among reading d i s a b i l i t y c l u s t e r s on reading and related v a r i a b l e s . . .13 0 13 Descriptive s t a t i s t i c s for c r i t e r i o n variables 14 6 14 Subjects i n Group D with preferred Model 4M 152 15 Subjects i n Group D with preferred Model 1M 153 16 Subjects i n Group N who preferred Models 4M and 1M 154 17 Model f i t s f or Group N and model subgroups within Group D 156 18 Descriptive s t a t i s t i c s on c r i t e r i o n variables for model subgroups within group D 160 ix Table 19 Component latencies calculated from preferred regression models for subgroups within Group D 162 2 0 Frequency table between cl u s t e r s and analogy subgroups 167 21 Correlation of observed solution times (CV1) with a b i l i t y , reading, and related variables 184 X LIST OF FIGURES Figure 1 Picture analogies 59 2 Schematic Picture Analogy 61 3 Overlap of box and whisker plots 131 4 Box and whisker plots for Group D clust e r s on SDRT Decoding variables 13 3 5 Box and whisker pl o t s for Group D cl u s t e r s on SDRT Vocabulary component variable 134 6 Box and whisker plots for Group D cl u s t e r s on SDRT Reading Comprehension variables 135 7 Box and whisker plots for Group D cl u s t e r s on SDRT Reading Rate component variables . . . .13 6 8 Box and whisker plots for Group D cl u s t e r s on Durr e l l reading time variables 138 9 Box and whisker plots for Group D cl u s t e r s on Durr e l l comprehension variables 139 10 Box and whisker plots for Group D cl u s t e r s on Durr e l l oral reading error variables 14 0 11 Box and whisker plots for Group D clust e r s on Durr e l l oral reading error v a r i a b l e s : t o t a l errors, t o t a l s e l f - c o r r e c t e d errors, and proportion self - c o r r e c t e d 14 3 12 Box and whisker plots on verbal and nonverbal a b i l i t y variables 144 13 Box and whisker plots for Group D model subgroups on SDRT Decoding component variables 169 14 Box and whisker plots for Group D model subgroups on SDRT Vocabulary component variable . . . .17 0 15 Box and whisker plots for Group D model subgroups on SDRT Reading Comprehension component variable 172 x i Figure 16 Box and whisker plots for Group D model subgroups on SDRT Reading Rate component va r i a b l e . . . 173 17 Box and whisker plots for Group D model subgroups on Du r r e l l reading time variables 174 18 Box and whisker plots for Group D model subgroups on Du r r e l l comprehension variables 17 6 19 Box and whisker plots for Group D model subgroups on Du r r e l l o r a l reading errors variables . . . 177 2 0 Box and whisker plots for Group D model subgroups on Du r r e l l o r a l reading error v a r i a b l e s : t o t a l errors, t o t a l s e l f - c o r r e c t e d errors, and percentage se l f - c o r r e c t e d 18 0 21 Box and whisker plots for Group D model subgroups on verbal and nonverbal a b i l i t y variables . .181 x i i LIST OF APPENDICES Appendix A Letter of consent 219 B Screening c h e c k l i s t for teachers 222 C Schematic Picture Analogy Booklet 224 D Cards used to i n s t r u c t students i n solving Schematic Picture Analogies 229 E Table l i s t i n g c l u s t e r membership, Boder s p e l l i n g patterns, and CV1 and CV2 model preference for Group D and Group N 231 x i i i ACKNOWLE DGEMENT I am indebted to the members of my committee for t h e i r advice and support that enabled me to carry out t h i s study. Dr. Ron Jarman who encouraged my int e r e s t i n the work of R.J. Sternberg, Dr. Marian Porath who helped me prepare for the or a l defence, and my chairman Dr. David Kendall, who supported me across the miles and was never too busy to give me advice or answer my questions. The i n i t i a l e f f o r t s of Dr. Todd Rogers, who contributed many hours of help and expertise before moving to the University of Alberta, are also acknowledged. I am gratef u l to Dr. R.J. Sternberg who gave me permission to use the Schematic Picture Analogies and supplied me with the t e s t . I am also grat e f u l to the Ontario Educational Research Council who gave me a research grant, and to the administration, teachers, and students of the Board of Education who made the study possible. Lastly, I wish to thank my husband, B i l l , f or h i s love and encouragement that enabled me to bring t h i s d i s s e r t a t i o n to i t s completion. 1 CHAPTER 1: INTRODUCTION Reading D i s a b i l i t i e s For almost a hundred years, parents, educators, psychologists, physicians, and neurologists have wrestled with the problem of children who have adequate i n t e l l i g e n c e but, for a va r i e t y of reasons, experience d i f f i c u l t y learning to read. Downing and Leong (1982) described these children as "reading-disabled." I t i s estimated that i n English speaking countries, t h i s group forms ten to twelve percent of school-age children and that within t h e i r ranks there are as many as ten percent who are severely disabled readers (Downing & Leong, 1982). Reading disabled children experience a major d i f f i c u l t y decoding words. This i s "the process of extracting enough information from word units so that a loc a t i o n i n the mental lexicon i s activated, thus r e s u l t i n g i n semantic information becoming available to the consciousness" (Stanovich, 1982a, p.486). Decoding d i f f i c u l t i e s become apparent early i n the reading-disabled c h i l d ' s school performance and i n l a t e r school years the c h i l d experiences d i f f i c u l t i e s with reading comprehension. A large amount of research has attempted to i d e n t i f y the underlying sources of in d i v i d u a l differences i n the processes that mediate word decoding and text comprehension. T r a d i t i o n a l l y , reading d i s a b i l i t y research has followed either 2 the d i f f e r e n t i a l or the information-processing approach. The d i f f e r e n t i a l approach i s so named because i t i s based on the study of i n d i v i d u a l differences i n task performance. Using aptitude or achievement variables, the focus i s directed at i n t e r i n d i v i d u a l v a r i a t i o n i n behaviour. In d i r e c t contrast, the cognitive or information-processing approach focuses on intragroup v a r i a t i o n that i s created by manipulation of treatment or task variables. D i f f e r e n t i a l Approach The d i f f e r e n t i a l approach produced factor analysis, a procedure which attempts to i d e n t i f y common variances among many vari a b l e s . The a p p l i c a tion of factor analysis to the study of reading d i s a b i l i t i e s has t r a d i t i o n a l l y focused on i n d i v i d u a l differences i n i n t e l l e c t u a l a b i l i t y and reading s k i l l s . One widely used i n t e l l i g e n c e t e s t , the Wechsler Intelligence Scale for Children (Wechsler, 1949, 1974) has been constructed to measure two t h e o r e t i c a l constructs, verbal and visual-perceptual a b i l i t i e s . Factor analysis has been used to c l u s t e r the subtests of t h i s test further, to form s u b a b i l i t i e s such as: verbal conceptualization, acquired knowledge, s p a t i a l a b i l i t y , and sequencing a b i l i t y (Bannatyne, 1974) and verbal comprehension, perceptual organization, and freedom from d i s t r a c t i b i l i t y (Kaufman, 1975). There have been attempts to l i n k learning and reading d i s a b i l i t i e s with a discrepancy i n one or more of these areas. However, Kavale 3 and Forness (1987) claim that there i s no empirical support for t h i s . As f a r as reading i s concerned, tests have been constructed which measure reading s u b s k i l l s such as: phonetic analysis, word decoding, reading speed, and comprehension (e.g. Stanford Diagnostic Reading Test; D u r r e l l Analysis of Reading D i f f i c u l t y ) . These t e s t s were standardized using the scores of representative samples of children and are now used d i a g n o s t i c a l l y to determine strengths and weaknesses of poor readers. Information-Processing Approach The information-processing paradigm has produced a methodology i n which variables are manipulated within tasks to es t a b l i s h i n t e r n a l v a l i d i t y . "Greatly s i m p l i f i e d , the information-processing approach i s conceptualized as a study of how sensory input i s transformed, reduced, elaborated, stored, retrieved and used..." (Swanson, 1987a, p.3). One important idea associated with t h i s approach i s that cognitive processes take time and that the amount of time can indicate how much information i s processed (Sternberg, 1969). Information-processing studies have examined tasks associated with constructs such as attention (LaBerge & Samuels, 1974; Samuels & M i l l e r , 1985), perception, (Massaro, 1975; Venezky & Massaro, 1979), decoding (Frederiksen, 1978; Samuels, LaBerge, & 4 Bremer, 1978), memory (Stanovich, 1988; V e l l u t i n o , 1979), language (Shankweiler & Liberman, 1976), and comprehension (Calfee & Spector, 1981; Samuels & Eisenberg, 1981). Hypotheses have been posited to l i n k a cognitive d e f i c i t i n one of these areas to reading d i s a b i l i t i e s . Word decoding and reading comprehension have received a great deal of attention. In a review of the research into decoding d i f f i c u l t i e s , Stanovich (1982a) concluded that in d i v i d u a l differences i n decoding a b i l i t y were accounted for by phonological processes. In a s i m i l a r review of comprehension d i f f i c u l t i e s , Stanovich (1982b) found that comprehension d i f f i c u l t i e s were related not only to inadequate decoding s k i l l s , but to d e f i c i e n t syntactic a b i l i t i e s and general metacognitive strategies. The d i f f e r e n t i a l and information-processing approaches have t h e i r l i m i t a t i o n s . The f a i l u r e to discover or explicate mental processes and the i n a b i l i t y to discover components of i n t e l l i g e n c e that e x i s t within individuals are the major l i m i t a t i o n s of the d i f f e r e n t i a l , factor a n a l y t i c approach (Sternberg, 1977) . The r e l a t i o n s h i p of these l i m i t a t i o n s to education include: the i n a b i l i t y of the approach to define mental processes necessary for learning, the f a i l u r e to i d e n t i f y subprocesses that underlie academic tasks, and the i n a b i l i t y to provide a l i n k between the theory of cognitive processes i n learning and classroom i n s t r u c t i o n (Wagner & Sternberg, 1984). Limitations associated with the information-processing 5 approach include: the i n a b i l i t y of the method to provide the means for systematic study of the correlates of i n d i v i d u a l differences i n task performance, lack of common language across tasks and across investigators, and an i n a b i l i t y to prevent the overvaluation of t a s k - s p e c i f i c components that have no g e n e r a l i z a b i l i t y to other tasks (Sternberg, 1977). The Subtyping Approach For many years researchers attempted to discover the underlying cause of reading d i s a b i l i t i e s . I n i t i a l l y , they examined reading s k i l l s and neuropsychological processes i n an e f f o r t to i d e n t i f y a single set of causal factors that would explain the disorder. So many d i f f e r e n t theories were proposed that Harris (1982) suggested, "Each of the single-cause proponents has found one part of what i s r e a l l y a very complex s i t u a t i o n " (p.47). The single cause theories have also led to indiscriminate use, i n some c i r c l e s , of the term "dyslexia". O r i g i n a l l y coined to describe unexpected reading f a i l u r e i n individuals of normal i n t e l l i g e n c e , i t has been used to describe many d i f f e r e n t groups of symptoms associated with reading d i s a b i l i t i e s . This has led to the observation that the term i s "not susceptible to precise operational d e f i n i t i o n and serves l i t t l e purpose" (Rutter, 1978, p.5) . The majority of researchers now accept that reading-disabled 6 children form a heterogeneous group (Doehring et a l . , 19 81; Malatesha & Dougan, 1982; Rourke, 1985; Satz & Morris, 1981). In the l a s t f i f t e e n to twenty years, the search for the cause of reading d i s a b i l i t i e s has centred on subtyping methodology. Subtyping i s an attempt to f i n d homogeneous groups within a heterogeneous disorder and thereby i d e n t i f y underlying causes. I t i s c a r r i e d out by grouping subjects according to some commonality among them. Early subtyping techniques included c l i n i c a l observation (Denckla, 1977; Johnson & Myklebust, 1967), analysis of reading/spelling errors (Boder, 1973), and p r o f i l e analysis using neuropsychological variables (Mattis, 1978; Mattis, French, & Rapin, 1975). In the l a s t decade, the majority of subtyping research has involved the use of s t a t i s t i c a l techniques termed c l u s t e r analyses (Morris, B l a s h f i e l d , & Satz, 1981). Cluster analysis i s not a single method but a vari e t y of techniques. One type of clus t e r analysis known as Q-factor analysis has been used by Doehring and his colleagues (Doehring & Hoshko, 1977; Doehring, Hoshko, & Bryans, 1979; Doehring et a l . , 1981) and Rourke and h i s colleagues (Fisk & Rourke, 1979; Petruskas & Rourke, 1979). This method analyzes correlations among subjects and produces factors that describe groups of subjects rather than groups of te s t s (Kavale & Forness, 1987). Although popular for a while, Q-factor analysis i s no longer used as much as other methods of c l u s t e r i n g . In p a r t i c u l a r , h i e r a r c h i c a l agglomerative techniques seem to be 7 preferred (Feagans & Appelbaum, 1986; Lyon & Watson, 1981; Speece, 1987; Speece, McKinney, & Appelbaum, 1985; Spreen & Haaf, 1986; Swanson, 1988). What subtyping research has shown i s that reading d i s a b i l i t i e s do not form a unitary disorder. However, none of the c l a s s i f i c a t i o n techniques used so f a r has been able to discover the inte r n a l cognitive processes and strategies used by subgroups of disabled readers. Subtyping appears to have reached an impasse. The techniques have been unable to generate a model of reading f a i l u r e and no new methodology has emerged to redress the l i m i t a t i o n s . Componential Analysis In 1977, Sternberg proposed a component theory of i n t e l l i g e n c e based upon the idea that separable mental processes underlie i n t e l l i g e n t behaviour. The execution of mental processes takes time and Sternberg theorized that i t was possible to i d e n t i f y the processes used i n a problem solving task by breaking the task into stages, measuring the i n t e r v a l of time taken to complete each stage, performing regression analysis on the i n t e r v a l scores, and comparing the regression equations obtained with t h e o r e t i c a l models. This method of componential analysis was f i r s t applied to analogical reasoning. The task was broken into stages through a 8 method of precuing and the time taken to complete each stage was recorded. Several models were devised that were consistent with the theory and indicated the component processes used i n solving analogies and the order and mode i n which they were executed (Sternberg, 1977). Multivariate analysis was used to see which model best f i t t e d the data. Studies showed a f a i r l y uniform model preference and consistent use of mental processes and strategies at the adult l e v e l (Sternberg, 1977; Sternberg & Nigro, 1980). One analogical reasoning task, i n i t i a l l y used with univ e r s i t y students, was adapted for use with school children (Sternberg & R i f k i n , 1979). Analogies i n picture form were presented to children i n systematically varied booklets, enabling them to be tested i n groups. No sp e c i a l equipment was required and i t was p r a c t i c a l even with very young children because no reading was involved. Results indicated that children use the same components as adults; however, consistency i n t h e i r use of mental components and strategies develops with age (Sternberg & R i f k i n , 1979). Outline of the Study The purpose of t h i s study was to go beyond the l i m i t a t i o n s of subtype analysis and t r y to i d e n t i f y the mental processes and strategies used by subgroups of reading-disabled students i n an analogical reasoning task. I t was anticipated that a comparison of t h i s information with the behaviours that characterized each 9 subgroup would enable disabled readers to be further c l a s s i f i e d and would provide more s p e c i f i c information as to the underlying source of the reading d i s a b i l i t y - information that would prove invaluable i n planning a remedial program. This study attempted to combine the d i f f e r e n t i a l with the cognitive approach through the use of Sternberg's componential analysis. The f i r s t step was to i d e n t i f y subgroups of disabled readers using two c l a s s i f i c a t i o n methods. One method involved subjective techniques using Boder's Test of Reading/Spelling Patterns (Boder & J a r r i c o , 1982), the other involved a s t a t i s t i c a l method of c l u s t e r i n g students using reading and reading-related v a r i a b l e s . Calfee (1977) had theorized that when they read, students make use of a set of independent cognitive and language s k i l l s , so that reading i s c a r r i e d out i n a series of independent stages. He proposed that Sternberg's componential analysis could be used to discover the components and strategies used i n a reading task (Calfee & Drum, 1979). However, while acknowledging that reading may be made up of r e l a t i v e l y independent s k i l l s , Rispens (1982) doubted that reading could be divided into independent subprocesses. He suggested, "The interplay between top-down and bottom-up processing cannot be handled adequately i n componential analysis" (p.187). Swanson (1987b) suggested that reading research should be directed towards i d e n t i f y i n g the mental processes required i n 10 completing selected classroom tasks. He summarized previously l i s t e d c r i t e r i a for s e l e c t i n g such tasks (Gadow & Swanson, 1986) as follows: 1) a task be selected that has a hi s t o r y i n information-processing l i t e r a t u r e ; 2) the task must have a t h e o r e t i c a l r ationale and must assess elementary processing mechanisms; 3) i t must be adaptable to moderately handicapped children; and 4) the task must be related and i n t e r r e l a t e d to a number of academic tasks. Componential analysis of the picture analogies task used by Sternberg and R i f k i n (1979) appears to f u l f i l a l l of Swanson's requirements. Analogical reasoning has been investigated and written about at length i n the information-processing l i t e r a t u r e ; componential analysis can specify the mental components used i n the analogical reasoning task; the picture analogies are suitable for reading-disabled children; and the task i s considered to be a measure of i n t e l l e c t u a l functioning (see Sternberg, 1977). 11 CHAPTER II : REVIEW OF LITERATURE Reading D i s a b i l i t i e s With reference to a reading-disabled population as defined by Downing and Leong (1982), the term "garden-variety" poor readers was used by Stanovich (1988) to d i s t i n g u i s h the group of generally disabled readers from a smaller, more severely disabled group. The terms " s p e c i f i c reading d i s a b i l i t y " , " s p e c i f i c reading retardation", and "developmental dyslexia", have been used interchangeably to denote the smaller group of severely disabled readers (Benton, 1978; Downing & Leong, 1982). Over the years there has been disagreement as to what constitutes a s p e c i f i c reading d i s a b i l i t y . The c r i t e r i a used to se l e c t reading disabled students for research purposes (based upon the d e f i n i t i o n of dyslexia of the World Federation of Neurology (Waites, 1968)) have tended to be exclusionary. Generally, to be considered to have a " s p e c i f i c reading d i s a b i l i t y " a student's reading f a i l u r e could not be attributed to inadequate i n t e l l i g e n c e , poor sight or hearing, neurological or physical impairment, emotional or s o c i a l problems, s o c i a l disadvantage, or lack of adequate i n s t r u c t i o n ( E l l i s , 1985). However, the c r i t e r i a used to judge these exclusionary c h a r a c t e r i s t i c s have varied greatly and there has been a tendency to automatically consider adequate i n s t r u c t i o n a given. 12 Many researchers have been displeased with diagnosis by exclusion. Rutter (1978) referred to the World Federation of Neurology d e f i n i t i o n of dyslexia (Waites, 1968) as "a counsel of despair" (p.12). As part of the F l o r i d a Longitudinal Project, Taylor, Satz, and F r i e l (1979) were able to show that no difference existed between two groups of disabled readers on a wide range of academic tasks and f a m i l i a l variables. One group was selected according to the World Federation of Neurology d e f i n i t i o n and the other was a nondyslexic group. Sporn (1981) questioned the exclusion of children from low SES backgrounds and, using Boder's (1973) method of subtyping, was able to show that subtypes of dyslexia existed among disabled readers who were environmentally disadvantaged. Dorman (1987) modified the subtyping methods of Mattis et a l . (1975), Boder (1973), and Doehring and Hoshko (1977) to accommodate the neurological disorders of 50 children aged 11 to 19 years selected from the i n -and out-patient populations of a hospital and school for the orthopaedically handicapped. These children were handicapped by cerebral palsy, spina b i f i d a , or muscular dystrophy and a l l but four were confined to wheelchairs. Twenty-five of these children were disabled readers, the remainder were normal readers. Dorman was able to show that subtypes, s i m i l a r to those of Mattis et a l . (1975), Boder (1973), and Doehring and Hoshko (1977), existed among t h i s neurologically handicapped sample of disabled readers. A report of Rutter and Yule (1975) seemed to indicate that the 13 incidence of s p e c i f i c reading-disabled students i n school-aged populations formed a "hump" at the lower end of the d i s t r i b u t i o n of reading scores. This implied that the disorder formed a discrete e n t i t y . However, more recent studies (Olson, K l e i g l , Davidson, & Fol t z , 1985; Scarborough, 1984; Seidenburg, Bruck, Fornarolo, & Backman, 1985; Share, McGee, McKenzie, Williams, & S i l v a , 1987) suggested that severely disabled readers do not form a d i s t i n c t subpopulation but are part of the normal continuum. I t was thought that the "hump" of Rutter and Yule (1975) could be caused by f l o o r and c e i l i n g e f f e c t s i n reading scores (Share et a l . , 1987). E l l i s (1985) rejected the model of s p e c i f i c reading d i s a b i l i t y which views i t as a disease rather l i k e measles that one either has or does not have. He compared i t instead with obesity which i s a matter of degree rather than kind. The prevalence of obesity i s dependent upon a r b i t r a r y c r i t e r i a ; s i m i l a r l y , the prevalence of s p e c i f i c reading d i s a b i l i t y , " w i l l depend e n t i r e l y upon where the l i n e i s drawn" ( E l l i s , 1985, p.172). Models of Reading LaBerge and Samuels (1974) theorized a data-driven or bottom-up model of reading which consists of three memory systems: v i s u a l , phonological, and semantic. The graphic input, consisting of l e t t e r s or words, i s recoded into v i s u a l , phonological, or semantic representations and compared with information already stored i n 14 the appropriate memory. The model i s f l e x i b l e and allows optional routes between graphemic input and verbal comprehension. According to t h i s model, each reader has a f i x e d capacity for processing information. Good readers achieve a high degree of automaticity i n decoding v i s u a l information and these automatic decoding s k i l l s enable them to devote most of t h e i r available attention to comprehending what they have read. On the other hand, poor readers who have not acquired t h i s automaticity need to devote most of t h e i r attention to decoding and have l i t t l e capacity l e f t for comprehension. In contrast to t h i s model are the top-down or concept-driven models (Goodman, 1968; Smith, 1971). Goodman (1968) described his model of reading as a "problem-solving" analogue i n which reading i s depicted as guess work. Smith (1971) theorized that reading comprehension i s f a c i l i t a t e d by redundancy - the use of nonvisual information to reduce the amount of v i s u a l processing. This means that the fluent reader uses knowledge of language structure as well as syntactic and semantic cues to make guesses about the text. The reader then checks a few v i s u a l features to see i f the guesses are r i g h t . In other words, fluent readers are depicted as reading only as much of the text as they need i n order to comprehend. These bottom-up and top-down models are s e r i a l processing models i n that they assume reading takes place i n a series of sequentially ordered processes. The i n t e r a c t i v e model of Rumelhart (1977) theorized that decoding and comprehension take place 15 simultaneously and are reci p r o c a l events. He posited four sources of knowledge that the reader uses to comprehend the text: knowledge of s p e l l i n g patterns, knowledge of words, knowledge of sentence patterns, and knowledge of semantic meaning. A l l four sources can be activated simultaneously and i n t e r a c t i v e l y . The model of Calfee (1977) consists of independent stages that intervene between the v i s u a l input and the verbal output. The processing of each stage i s theorized to take a c e r t a i n amount of time so that t o t a l reading time i s the sum of the independent stages. Each stage i s affected by factors s p e c i f i c to that stage. Calfee gave as an example, the task of reading a l i s t of words for l a t e r r e c a l l . This task was theorized to consist of two stages: reading and organizing. Total reaction time would consist of the sum of these independent processes. To confirm independence, i t would be necessary to show lack of i n t e r a c t i o n between the factors and the time parameters should show up as main e f f e c t s . Any in t e r a c t i o n between the factors would imply that the processes were not separable and should be treated as a single complex. In reviewing the strengths and weaknesses of t h e i r e a r l i e r model, Samuels and LaBerge (1983) proposed a feedback loop be added to the model. The feedback loop was designed to show how information i n semantic memory, the l a s t component i n the system, could aid the processing of information i n the other components. This changes the model from a s e r i a l processing model to an i n t e r a c t i v e one. 16 S p e c i f i c i t y of Reading D e f i c i t s Beginning readers concentrate on decoding words. As t h e i r s k i l l s become more automatic and fluency develops, the attention of good readers i s focused on comprehension. Disabled readers struggle to decode words. They do so slowly and even when successful they f a i l to comprehend at the l e v e l of good readers. Researchers have examined decoding and comprehension s k i l l s i n order to discover a brain/cognitive d e f i c i t i n one s p e c i f i c area. The concept of such a d e f i c i t has been named the "assumption of s p e c i f i c i t y " (Stanovich, 1986). Decoding Phonological Awareness D e f i c i t s Phonological awareness i s exhibited i n the a b i l i t y to recognize i n i t i a l sounds i n words, to i d e n t i f y the number of s y l l a b l e s and phonemes i n spoken words, and to match and create rhyming words. Research indicates that students who display poor phonological awareness i n preschool and the early grades are more l i k e l y to have l a t e r reading d i f f i c u l t i e s than t h e i r peers who exhibi t good phonological awareness. For example,: Liberman, Shankweiler, Liberman, Fowler, and Fischer (1977) discovered that c h i l d r e n who could not tap out the number of sounds i n spoken 17 words at s i x years of age were performing at the lower end of t h e i r class i n reading by the time they were i n grade 2. S i m i l a r l y Bradley and Bryant (1985) found that the a b i l i t y of prereaders to i d e n t i f y the i n i t i a l sound i n spoken words and to match rhyming words correlated s i g n i f i c a n t l y with l a t e r reading achievement. Fox and Routh (1980), who tested a group of f i r s t graders on a word segmentation t e s t , found that average readers performed p r o f i c i e n t l y whereas poor readers did badly. Two years l a t e r , when the students were i n grade 3, the word segmentation t e s t was repeated (Fox & Routh, 1983) . A l l the students were able to segment words into phonemes but by t h i s time many of those who were o r i g i n a l l y poor readers had become severely disabled readers. The question has been raised as to the r e l a t i o n s h i p between reading and phonological awareness. Bradley and Bryant (1983, 1985) showed that over a two-year period, t r a i n i n g s i x - and seven-year-old c h i l d r e n to group one s y l l a b l e words according to t h e i r sounds rather than t h e i r meaning placed them approximately s i x months ahead of a control group i n reading and s p e l l i n g s k i l l s . However, Fox and Routh (1984) found that t r a i n i n g kindergarten children to segment and blend words had l i t t l e e f f e c t on l a t e r word recognition. Some researchers believe the r e l a t i o n s h i p i s r e c i p r o c a l and that children do not develop awareness of sounds i n words u n t i l they are faced with words i n p r i n t (Ehri, 1979; Jorm & Share, 1983. This p o s i t i o n i s supported by the study of Morais, Cary, A l g e r i a , and Bertelson (1979) i n which adults who had just 18 learned to read were shown to have better phonological awareness than adults who had never learned to read. Short-Term Memory D e f i c i t s Although not a l l poor readers have short-term memory d e f i c i t s , poor readers generally have d i f f i c u l t y r e t a i n i n g verbal information i n short-term or working memory. Blachman (1983) used kindergarten and grade 1 students to investigate the a b i l i t y of the McCarthy Scales of Children's A b i l i t i e s (McCarthy, 1972) to predict reading achievement. The McCarthy Scales were administered to the students at the beginning of t h e i r kindergarten year and reading readiness measures were administered at the end of the year. The reading readiness measures consisted of the reading section of the Wide Range Achievement Test (Jastak & Jastak, 1978) and an informal test i n which students were asked to name a l l the l e t t e r s of the alphabet i n upper and lower case form and i d e n t i f y the sound associated with each l e t t e r . A s i g n i f i c a n t p o s i t i v e r e l a t i o n s h i p was found between the Memory component of the McCarthy Scales and reading readiness. Blachman suggested that because the Memory component includes subtests that appear to be sensi t i v e to differences i n verbal coding a b i l i t y of good and poor readers, verbal coding a b i l i t y rather than t o t a l memory may have accounted for a s i g n i f i c a n t amount of variance i n reading scores. 19 Liberman et a l . (1977) reported several studies that showed disabled readers were less able to remember str i n g s of unrelated numbers and l e t t e r s presented v i a a tachistoscope. Liberman and Shankweiler (1985) expressed the opinion that incoming written or spoken information i s held i n short-term memory i n phonological form while i t i s being processed. Poor readers therefore, appear to have a d e f i c i t i n the a b i l i t y to form and maintain phonological codes i n short-term memory. Nonverbal information, such as memory for faces does not pose such a problem (Liberman, Mann, Shankweiler, & Werfelman, 1980; Nelson & Warrington, 1980). Speech Perception D e f i c i t s Godfrey, Syrdal-Laskey, Millay, and Knox (1981) reported differences between disabled and nondisabled readers i n the way they categorized contrasting speech sounds. Reading-disabled childr e n were found to have auditory discrimination problems when st i m u l i were presented against background noise (Brady, Shankweiler, & Mann, 1983). T a l l a l (1980) found a group of nonlanguage impaired reading-disabled children who were unable to d i s t i n g u i s h between rapid l y presented tones. T a l l a l and Stark (1982) suggested that among the general population of disabled readers a group e x i s t that s u f f e r from some subtle auditory perceptual d e f i c i t that causes d i f f i c u l t i e s i n recognizing and discriminating between phonemes i n speech. 20 Name Retri e v a l D e f i c i t s The d i f f i c u l t y reading-disabled children have i n naming f a m i l i a r picture items was reported by Jansky and DeHirsch (1972). Denkla and Rudel (1976) also showed disabled readers were less p r o f i c i e n t at naming l i s t s of l e t t e r s , numbers, and pictures than normal readers. Blachman (1983) found that the Rapid Automatized naming t e s t (TRAN) of Denkla and Rudel (1976) combined with the segmentation Task of Liberman, Shankweiler, Fischer, and Carter (1974) was a major component i n predicting l a t e r reading achievement. To di s t i n g u i s h between name r e t r i e v a l speed and s e r i a l naming, Jackson and McClelland (1979) used a l e t t e r matching task with adults i n which items were matched p h y s i c a l l y (AA) or matched by name (Aa). They found s i g n i f i c a n t c o r r e l a t i o n between the naming task and reading a b i l i t y . A study with reading-disabled children yielded s i m i l a r r e s u l t s ( E l l i s , 1981). I t has been suggested that name r e t r i e v a l d e f i c i t s can be linked to d i f f i c u l t y accessing the phonological representation of words i n the lexicon (Brady & Fowler, 1988; Wagner & Torgesen, 1987). This process has been termed phonological recoding with l e x i c a l access. 21 Comprehension Use of Context The d i s t i n c t i o n between good and disabled readers 1 use of contextual cues does not concern whether or not they use them, but how they use them. According to the theories of Goodman (1968) and Smith (1971), good readers make e f f i c i e n t use of contextual cues i n decoding unfamiliar words but disabled readers are i n e f f i c i e n t users of context. However, LaBerge and Samuels (1974) and P e r f e t t i and Lesgold (1979) theorized that an i n d i v i d u a l has a fixed information-processing capacity. I f most of the capacity i s expended on decoding then there i s l i t t l e l e f t over for comprehending. P e r f e t t i , Goldman, and Hogaboam (1979) found that good and poor f i f t h grade readers recognized words i n a contextual se t t i n g f a s t e r than words i n i s o l a t i o n . The recognition times of the poor readers were more affected by the contextual s i t u a t i o n than those of the good readers suggesting that the poor readers made more use of contextual cues than the fluent readers. In the same study, P e r f e t t i et a l . (1979) found that good readers whose recognition time was le a s t affected by context did better on a c l o z e - l i k e task, an i n d i c a t i o n that t h e i r decoding s k i l l s had reached a l e v e l of automaticity that enabled them to complete sentences with ease. Another consideration of the use of context concerns the l e v e l 22 of d i f f i c u l t y of what i s being read. A disabled reader cannot use contextual cues when unable to decode the majority of the words. In a longitudinal study, Stanovich, Cunningham, and Freeman (1984) showed that poor readers demonstrated as much use of contextual cues as good readers when they could read a passage with the same speed and accuracy. Short-Term Memory Poor readers have been shown to perform poorly on a vari e t y of short-term memory tasks (Blachman, 1983; Liberman et a l . 1977; Liberman & Shankweiler, 1985). One explanation, presented e a r l i e r , i s that poor readers are d e f i c i e n t i n phonological coding which r e s u l t s i n slow word-by-word reading which i n turn a f f e c t s t h e i r a b i l i t y to comprehend. P e r f e t t i and Lesgold (1977, 1979) theorized that t h i s slower decoding i n short-term memory reduces the a b i l i t y of the disabled reader to also hold larger units, such as clauses and sentences, i n short-term memory. Another explanation concerns the i n a b i l i t y of poor readers to use deliberate, pla n f u l memorization strategies, such as verbal rehearsal, imagery, and elaboration (see Stanovich, 1982b). These explanations are not mutually exclusive (Torgesen, 1978-79) . There i s evidence to show that poor readers may have d i f f i c u l t y comprehending because of problems with short-term memory as well as the use of i n e f f i c i e n t memorization strategies (see 23 Stanovich, 1982b). Strategy t r a i n i n g has enabled disabled readers to improve t h e i r reading comprehension (Palinscar, 1976; Palinscar & Brown, 1984; Stevens, 1988; Wong, 1979; Wong & Jones, 1982) in d i c a t i n g that the strategies are available to them but they f a i l to use them. Metacognition A large body of research has focused on the role of metacognition and the use of s e l f - d i r e c t i o n i n e f f i c i e n t learning (see Borkowski, Johnson, & Reid, 1987; Wong, 1987). There i s some disagreement as to what constitutes metacognition. Lawson (1984) argues that metacognitive knowledge and the executive processes that d i r e c t that knowledge are separate and d i s t i n c t . F l a v e l (1978) refer s to metacognition as self-awareness and s e l f - r e g u l a t i o n of one's cognitive processes that are under conscious control. For Reeve and Brown (1985) metacognition i s "...a c o l l e c t i o n of problems solving a c t i v i t i e s used by individ u a l s to understand what i s required, to understand t h e i r own c a p a b i l i t i e s , to plan strategies that w i l l allow them to reach the goal, and to monitor and coordinate these a c t i v i t i e s " (pp. 344-345). Many studies have shown that reading-disabled children do not use strategies e f f i c i e n t l y i n both word decoding, reading comprehension, and study s k i l l s (Anderson, 1980; Brown, 1980; Wong & Wong, 1987). Brown (1980) distinguished between deliberate 24 conscious s t r a t e g i c intervention and other i n t e l l i g e n t processing that goes on at a lower, more automatic l e v e l . She used the analogy of "automatic p i l o t " to describe the processing of the s k i l l e d and fluent readers who i s "triggered" into a more conscious s t r a t e g i c state by encountering unexpected and unfamiliar information i n the text. Then deliberate, p l a n f u l , s t r a t e g i c a c t i v i t i e s termed "debugging" take over u n t i l the problem i s solved. According to Brown (1980), debugging a c t i v i t i e s are s k i l l s of metacognition. Less s k i l l e d readers do not use debugging strategies as e f f e c t i v e l y . They have d i f f i c u l t y extracting the main idea from text under both reading and l i s t e n i n g conditions. They f a i l to adopt helping strategies such as underlining; they do not give forgotten material more study time; and they don't know when they are ready to be tested (Brown, 1980). Limitations of the Information-Processing Approach Sternberg (1978) i d e n t i f i e d two major l i m i t a t i o n s of the information-processing approach. The f i r s t concerns the i n a b i l i t y of the method to provide a means for systematically studying correlates of i n d i v i d u a l differences i n task performance. In other words, there i s no procedure for discovering the r e l a t i o n s h i p between components of one task and components of another. Without res o r t i n g to d i f f e r e n t i a l methods. S i m i l a r l y , there i s no procedure 25 for discovering the rel a t i o n s h i p between components of reading tasks and components of i n t e l l i g e n c e without resorting to d i f f e r e n t i a l methods. The second l i m i t a t i o n concerns the overvaluation of task-s p e c i f i c components that have no g e n e r a l i z a b i l i t y to other tasks. A component of a s p e c i f i c task may be i d e n t i f i e d through task manipulation but i t s importance to a domain of tasks can only be i d e n t i f i e d through d i f f e r e n t i a l methods. Summary of Information-Processing Theory and Research In summary, several theories of reading have been devised that view the reading process as data-driven, concept driven, i n t e r a c t i v e , or made up of independent stages. Research indicates that disabled readers have poor decoding s k i l l s which have been linked to d e f i c i t s i n phonological awareness and phonological coding. Disabled readers may also exhibit d e f i c i t s i n short-term memory, speech perception and name r e t r i e v a l . Comprehension d i f f i c u l t i e s are linked to lack of automaticity i n decoding and to poor short-term memory. Poor comprehension i s also linked to i n e f f i c i e n t use of metacognitive s k i l l s . Subtyping Research In 1971, Applebee reviewed s i x reading d i s a b i l i t y models and 26 presented evidence to show why the simplest models did not f i t what was known about the disorder. He suggested that research should be based on the newer, more complex models that corresponded c l o s e l y to the heterogeneity of the disorder. Applebee cautioned "Such a s h i f t w i l l require more sophisticated methods of analysis than have been employed i n the past, and w i l l bring with them a whole new set of problems of int e r p r e t a t i o n and design" (p.119). The f i r s t attempts to c l a s s i f y reading-disabled subjects were based upon c h a r a c t e r i s t i c patterns of d e f i c i t i n neurological and/or psychoeducational processes. The groupings depended upon the instruments used and the c r i t e r i a set by the researcher. They included: Verbal/Performance IQ discrepancy as measured by the Wechsler Adult Intelligence Scale (Wechsler, 1955) and the Wechsler Intelligence Scale for Children (Wechsler, 1949) (Kinsbourne & Warrington, 1963), c l i n i c a l observation of reading and learning c h a r a c t e r i s t i c s (Johnson & Myklebust, 1967), observation of reading errors (Ingram, Mason, & Blackburn, 1970), analysis of reading and/or s p e l l i n g patterns (Boder, 1973; Boder & J a r r i c o , 1982; Sweeney & Rourke, 1978), neuropsychological and educational p r o f i l e analysis (Mattis, French & Rapin, 1975), and c l i n i c a l observation of neuropsychological d e f i c i t s (Denkla, 1977). The Contribution of Boder Boder's (1971, 1973) are considered important contributions 27 to the subtyping l i t e r a t u r e (Gordon, 1984; Hynd & Cohen, 1983; Satz & Morris, 1981). Boder examined the error patterns i n the reading and s p e l l i n g performance of 107 disabled readers, ranging i n age from eight to sixteen years, and i d e n t i f i e d three subtypes. The disabled readers i n Group I, l a b e l l e d "dysphonetic", were characterized by primary d e f i c i t s i n the auditory channel and were unable to analyze words phonetically. The disabled readers i n Group II, l a b e l l e d "dyseidetic", were characterized by a primary d e f i c i t i n the v i s u a l channel and were unable to perceive words as whole words or "v i s u a l g e s t a l t s " . The disabled readers i n Group III were characterized by primary d e f i c i t s i n both channels being both dysphonetic and dyseidetic. Members of t h i s group were the most severely disabled of the three. In addition to these s p e c i f i c reading d i s a b i l i t y subtypes, Boder and J a r r i c o (1982) i d e n t i f i e d two other groups. The members of one were considered to have nonspecific reading retardation and the members of the other to have an undetermined reading d i s a b i l i t y . They suggested that the undetermined group consisted of remediated Group I and Group III subjects. A l l these attempts to c l a s s i f y reading-disabled individuals into subtypes have revealed a great deal of c l i n i c a l information. Satz and Morris (1981) cautioned however, that s e l e c t i o n based on v i s u a l inspection of complex data does not necessarily " r e f l e c t the hidden structure of the data" and the " v a l i d i t y , r e l i a b i l i t y and u t i l i t y of these subtypes have seldom been tested" (p.116). 28 S t a t i s t i c a l Methods In an attempt to reduce the s u b j e c t i v i t y of previous subtyping methods, c l u s t e r analysis was introduced. In the s o c i a l sciences, c l u s t e r analysis i s a r e l a t i v e l y new approach to c l a s s i f i c a t i o n of ind i v i d u a l s , although the methods have been used for several years i n the b i o l o g i c a l sciences (Everitt, 1980). Cluster analysis has been described as "...a q u a s i - s t a t i s t i c a l technique which can be used on multivariate data i n order to create such c l a s s i f i c a t i o n s " (Morris, B l a s h f i e l d , & Satz, 1981, p.79) There are seven major categories of c l u s t e r analysis: 1) h i e r a r c h i c a l agglomerative methods; 2) h i e r a r c h i c a l d i v i s i v e methods; 3) i t e r a t i v e p a r t i t i o n i n g techniques; 4) density searching techniques; 5) factor analysis variants (one of which i s Q-factor a n a l y s i s ) ; 6) clumping techniques; and 7) graphic techniques (Morris, Blashfied, & Satz, 1981). In reading d i s a b i l i t y research, two types of c l u s t e r analysis have been favoured over the others, namely, Q-type factor analysis and h i e r a r c h i c a l agglomerative methods. O-Type Factor Analysis The factor analysis variant, known as Q-factor analysis, analyzes correlations among subjects and produces factors that describe groups of subjects rather than groups of te s t s (Kavale & 29 Forness, 1987) . In the f i e l d of health sciences, Q-factor analysis has been applied to the study of diseases through c l a s s i f i c a t i o n of patients according to t h e i r symptoms and signs of disease (Sneath & Sokal, 1973), i n an attempt to discover the etiology, therapy and prognosis for i n d i v i d u a l patients. For s i m i l a r reasons, several researchers applied the method to the study of reading and learning disorders (Doehring & Hoshko, 1977; Doehring, Hoshko, & Bryans, 1979; Doehring et a l , 1981; Fisk & Rourke, 1979; Petruskas & Rourke, 1979). Results are summarized i n Table 1. I n i t i a l l y , Doehring and Hoshko (1977) used a series of 31 reading-related measures with two groups of children whose ages ranged from eight years up to 16 years 2 months, and who were enroled i n a summer program for children with learning problems. Group R consisted of 3 4 reading-disabled childr e n and Group M consisted of 31 children who exhibited mixed problems. Data i n the form of response latency or error scores were converted to standard scores based upon norms established i n a previous study (Doehring, 1976). The standard scores were Q-factor analyzed to form subtypes. Three subtypes that emerged from the reading-disabled sample continued to emerge when they were combined and analyzed with the mixed group. Subtype 1, which consisted of 3 5 percent of Group R, was characterized by d e f i c i t s i n o r a l word and s y l l a b l e reading. Subtype 2, which consisted of 32 percent of Group R, exhibited d e f i c i t s i n auditory-visual l e t t e r matching. Subtype 3, which consisted of 24 percent of Group R, had d e f i c i t s i n v i s u a l matching 30 Table 1. Subtypes identified by Q-factor analysis Authors/ Sample Number &/or name of subtype & characteristics Reading & related variables Doehring & H o s h k o ( 1 9 7 7 ) R e a d i n g -disabled students Doehring & H o s h k o ( 1 9 7 7 ) R e a d i n g -disabled students Doehring & H o s h k o ( 1 9 7 7 ) R e a d i n g -disabled students T y p e 1 : d i f f i c u l t y with oral word & syllable reading T y p e 1 : d i f f i c u l t y with oral word & syllable reading T y p e 1 : d i f f i c u l t y with oral word & syllable reading T y p e 2 : d i f f i c u l t y with auditory/visual matching T y p e 2 : d i f f i c u l t y with auditory/visual matching T y p e 2 : d i f f i c u l t y with auditory/visual matching Language & neuropsychological variables Petruskas & Rourke ( 1 9 7 9 ) R e a d i n g -d i s a b l e d students Type 1: language d i s t u r b a n c e d i f f i c u l t y with auditory & verbal m e m o r y & auditory/percep-tual s k i l l s T y p e 2 : d i f f i c u l t y with sequencing & f i n g e r localization T y p e 3 : d i f f i c u l t y with visual matching & auditory/visual matching words & syllables T y p e 3 : d i f f i c u l t y with visual matching & auditory/visual matching words & syllables T y p e 3 : d i f f i c u l t y with visual matching & auditory/visual matching words & syllables T y p e 3 : d i f f i c u l t y with c o n c e p t u a l f l e x i b i l i t y , m o t o r i c & retentive s k i l l s Fiske & R o u r k e ( 1 9 7 9 ) Lea r n i n g -disabled students T y p e A : d i f f i c u l t y with f i n g e r l o c a l i z a t i o n T y p e B : d i f f i c u l t y with phonemic hearing & verbal coding T y p e C : d i f f i c u l t y with fingertip writing 31 and auditory-visual matching of words and s y l l a b l e s . Nine percent of the disabled reader group were u n c l a s s i f i e d . Doehring et a l . (1981) used the same 31 reading-related measures and eight additional tests consisting of a colour and a picture naming t e s t , two tests of geometric figure scanning, a test of sentence comprehension and three spelling-type t e s t s . The sample consisted of 85 c l i n i c - r e f e r r e d reading-disabled children aged 8 to 17 years and 3 adults aged 20 to 27 years. Three reading d i s a b i l i t y subtypes also emerged when the standard scores obtained from 39 te s t s were Q-factor analyzed. The subtypes resembled those obtained i n the previous studies (Doehring & Hoshko, 1977; Doehring, Hoshko, & Bryans, 1979). Subtypes 1 (38%), 2 (25%), and 3 (19%) were renamed by Doehring et a l . (1981), Type 0 (Oral Reading D e f i c i t ) , Type A (Association D e f i c i t ) , and Type S (Sequence D e f i c i t ) , respectively. Eighteen percent of the reading-disabled subjects were u n c l a s s i f i e d . Although percentages varied from study to study, i t can be seen that Subtype 1 or Type 0 consistently contained the most subjects and Subtype 3 or Type S, the l e a s t . The l a b e l l i n g of Type S as a sequence d e f i c i t was c r i t i c i s e d by E l l i s (1985) on the grounds that the term was something of a misnomer because i t did not adequately describe the d i f f i c u l t i e s encountered by t h i s group. Doehring et a l . (1981) also administered a battery of language te s t s to t h e i r reading-disabled sample but the data obtained from these t e s t s were analyzed a f t e r the subtypes had been i d e n t i f i e d . 32 The language battery contained tests designed to measure the a b i l i t y to d i f f e r e n t i a t e phonemes within s y l l a b l e s and words, tests of s e r i a l naming, a te s t of short-term memory, a t e s t of verbal comprehension and tests of syntactic-semantic s k i l l s . However, p l o t t i n g the average grade equivalent scores of the language tests for each of the three reading d i s a b i l i t y subtypes did not show d i f f e r e n t patterns of language d e f i c i t . In addition, Q-factor analysis of the language data yielded only two stable factors. Language d e f i c i t Type 1 was characterized by d i f f i c u l t y i n repeating word strings whereas Type 2 was characterized by a d i f f i c u l t y i n naming. Very weak relationships between language d e f i c i t Type 1 and reading d i s a b i l i t y Type 0 and language d e f i c i t Type 2 and reading d i s a b i l i t y subtype A and S existed. Forty-nine percent of the disabled readers were u n c l a s s i f i e d by the language variab l e s . The Q-factor a n a l y t i c technique was used by Petruskas and Rourke(1979) to analyze the neuropsychological p r o f i l e s of samples of disabled and normal readers who obtained a F u l l Scale IQ score between 80 and 120 on the WISC and were aged from 84 to 107 months. One hundred and t h i r t y three reading-disabled children (mean age 92.78 months; mean IQ 96.92) who scored at the 25th percentile or less on the WRAT reading subtest were selected from a c l i n i c -r e ferred population. Twenty-seven normal readers (mean age 96.07 months; mean IQ 107.26) scored at the 45th p e r c e n t i l e or above on the WRAT Reading subtest. A battery of neuropsychological measures 33 designed to assess s k i l l s i n the area of auditory-verbal, sequencing, v i s u a l s p a t i a l , t a c t i l e , motoric, and abstract-conceptual processing were used. The WISC formed part of the battery with each subtest counting as a single measure. Of the 44 o r i g i n a l measures, 20 considered to be the most representative were used i n the Q-factor analysis. Three subtypes emerged. Type 1 (25 %) had large WISC Verbal-Performance discrepancies and showed evidence of language disturbance including d i f f i c u l t i e s i n auditory/verbal memory and auditory/perceptual s k i l l s . For Type 2 (16%), finger l o c a l i z a t i o n and sequencing were the major d i f f i c u l t i e s . The predominant deficiency of Type 3 (8%) occurred i n conceptual f l e x i b i l i t y , p a r t i c u l a r l y l i n g u i s t i c coding. Deficiencies i n motoric and verbal expressive and retentive s k i l l s were also apparent. Fifty-one percent of the reading-disabled sample were u n c l a s s i f i e d . A s i m i l a r study used three groups of learning-disabled students selected from the same c l i n i c a l population (Fisk & Rourke, 1979) . A l l students had F u l l Scale WISC IQs i n the range 86 to 114. The groups were aged 9-10 years (mean 9.95), 11-12 years (mean 11.86), and 13-14 years (mean 13.74); contained 100, 100, and 64 subjects; and had mean IQs of 9 6.18, 9 6.37, and 94.98, respectively. The subjects were i d e n t i f i e d as learning-disabled using a c r i t e r i o n set at the 30th percentile or les s on a l l three WRAT subtests (Arithmetic, Reading, and S p e l l i n g ) . Twenty-one neuropsychological variables s i m i l a r to those of Petruskas and 34 Rourke (1979) were Q-factor analyzed. Fisk and Rourke i d e n t i f i e d three learning d i s a b i l i t y subtypes. The f i r s t , Subtype A (20%) was characterised by d i f f i c u l t y with finger l o c a l i z a t i o n and was s i m i l a r to the Type 2 of Petruskas and Rourke. The l i m i t i n g features of Subtype B (19%) were d e f i c i e n c i e s i n phonemic hearing, verbal coding, and short-term audio-visual memory. Members of t h i s subtype also had a large Verbal-Performance IQ discrepancy s i m i l a r to Type 1 of Petruskas and Rourke. A t h i r d , Subtype C (15%) was only apparent at the two older l e v e l . I t was distinguished by outstandingly poor performance i n f i n g e r t i p writing and was interpreted by the authors as a possible v a r i a t i o n of Subtype A. Forty-six percent of subjects were u n c l a s s i f i e d . Q-factor analysis has some l i m i t a t i o n s . Many of the l i m i t a t i o n s of factor analysis l i s t e d by Sternberg (1977) apply to Q-factor analysis. The f i r s t concerns the i n a b i l i t y of the method to discover the cognitive processes that underlie the behaviour patterns of disabled readers. The second concerns the i n t e r i n d i v i d u a l nature of factor analysis and i t s i n a b i l i t y to indicate processes and strategies used at the i n d i v i d u a l l e v e l . The t h i r d l i m i t a t i o n concerns the fact that t h e o r e t i c a l models are proposed to f i t the data a f t e r the fact, making i t impossible to t e s t a theory or compare one theory with another, A fourth l i m i t a t i o n concerns the a r b i t r a r y r o t a t i o n a l decision that i s made a f t e r factors are extracted, implying that there i s not one 35 s o l u t i o n but as many solutions as there are r o t a t i o n a l p o s s i b i l i t i e s . Q-factor analysis was also c r i t i c i z e d on the grounds that i t f a i l e d to take into account the elevation of the subjects' p r o f i l e s ( F l e i s s , Lawlor, Platman, & Fieve, 1971). However, the McQuitty method of c l u s t e r i n g based on both squared e u c l i d i a n distance c o e f f i c i e n t s and shape distance c o e f f i c i e n t s was used by Doehring, Hoshko, and Bryans (1979) and Doehring et a l . (1981) i n addition to Q-analysis. Very s i m i l a r subtypes were produced s a t i s f y i n g E v e r i t t ' s (1980) suggestion that a good solution should appear under d i f f e r e n t c l u s t e r i n g ,methods. Another l i m i t a t i o n concerns the number of students who were eit h e r u n c l a s s i f i e d or had factor loadings that met the c r i t e r i a for more than one factor. Raising the c r i t e r i a for subtype membership increases the number of u n c l a s s i f i e d subjects but lowering the c r i t e r i a reduces the homogeneity of each subtype. Hi e r a r c h i c a l Agglomerative Techniques The i n i t i a l popularity of Q-factor analysis i n reading and learning d i s a b i l i t y research may be attributed to widespread use of, and f a m i l i a r i t y with, the related R-factor a n a l y t i c technique. However, researchers now seem to prefer h i e r a r c h i c a l agglomerative techniques as few Q-factor an a l y t i c studies have been reported since the work of Doehring et a l . (1981) . H i e r a r c h i c a l 36 agglomerative techniques c l u s t e r i n d i v i d u a l s according to mathematical algorithms. Solutions obtained using Ward's method of average linkage have been shown to be p a r t i c u l a r l y powerful i n comparison to those obtained by other c l u s t e r i n g techniques (Morey, B l a s h f i e l d , & Skinner, 1983). Three groups of researchers involved i n l ongitudinal studies have used h i e r a r c h i c a l agglomerative techniques. Lyon and h i s associates (Lyon & Watson, 1981; Lyon, Rietta, Watson, Porch, & Rhodes, 1981; Lyon, Stewart, & Freedman, 1982) c a r r i e d out a series of subtype i d e n t i f i c a t i o n studies with reading-disabled children i n North Carolina, Alabama, and Georgia. One of the f i r s t investigations (Lyon & Watson, 1981) involved a battery of t e s t s used by Mattis et a l . (1975) designed to measure auditory receptive and expressive language, v i s u a l perception, memory, and visual-motor integration. The battery was administered to 100 s c h o o l - i d e n t i f i e d reading-disabled childr e n (SLD/R) and 50 normal readers (NR) matched for age and IQ. Ages ranged from 11 to 12.5 years with means of 12.3 years (SLD/R) and 12.4 years (NR). F u l l Scale WISC-R scores were within the normal range with means of 105.7 (SLD/R) and 106.1 (NR). A l l members of the SLD/R group exhibited s i g n i f i c a n t d e f i c i t s i n reading, as measured by the Peabody Individual Achievement Test (PIAT), whereas a l l members of group NR were reading at, or above, grade l e v e l . Scores for each variable i n the t e s t battery were converted to standard scores based upon the mean scores of the normal reader 37 group. Cluster analysis using h i e r a r c h i c a l agglomerative techniques that employed a minimum variance c r i t e r i o n produced s i x homogeneous subtypes. Six students (2%) could not be assigned to any of the subgroups. Subgroup 1 (10%) was characterized by d e f i c i t s i n a l l areas tested. Members of t h i s subtype also had poor sight vocabulary, d e f i c i e n t word attack s k i l l s , and were the poorest readers i n the sample. They were likened to Boder's (1971) mixed dysphonetic and dyseidetic s p e c i f i c reading d i s a b i l i t y subtype. Members of Subgroup 2 (12%) exhibited mixed d e f i c i t s i n language comprehension, auditory memory, and visual-motor integration. They were compared with the mixed type of disabled reader described by Johnson and Myklebust (1967) and with a mild form of Boder's mixed dysphonetic and dyseidetic subtype. The language comprehension and sound blending problems of Subgroup 3 (12%) were thought to represent a language disorder with both receptive and expressive components. This subgroup was compared to the language disordered group of Mattis et a l . (1975), the dysphonetic subgroup of Boder (1971) and i n d i r e c t l y to the auditory dyslexic described by Johnson and Myklebust (1967). Subgroup 4 (32%) was characterized by normal language s k i l l s and a d e f i c i t i n visuoperceptive function. The subgroup was compared to Boder's (1971) dyseidetic subtype. I t also had the largest membership, which the authors regarded as unexpected because Mattis et a l . (1975) and other researchers had observed that 38 visuoperceptive disorders occurred more frequently i n younger children. Subgroup 5 (12%) was compared to the aphasic group of Mattis et a l . (1975). They were characterized by d e f i c i t s i n naming a b i l i t y , auditory memory, and sound blending. Members of t h i s subgroup also had the second lowest reading scores i n the sample. Members of Subgroup 6 (16%) had a normal diagnostic p r o f i l e . This pattern was unexpected although these childr e n had the highest reading scores i n the sample. The authors suggested that the poor reading of these children (compared to normal readers) might be associated with s o c i a l , motivational, or pedagogical factors rather than some inherent d e f i c i t . In an attempt to further val i d a t e these subgroups, Lyon et a l . (1981) used external measures of achievement, school history, SES, and parent reports of developmental milestones. The subgroups d i f f e r e d s i g n i f i c a n t l y on the achievement te s t with Subgroups 1 and 5 being more impaired on reading recognition and reading comprehension. Subtype 6, which had exhibited a normal diagnostic p r o f i l e , scores above the others on a l l measures. The socioeconomic and developmental measures did not di s t i n g u i s h among the subgroups. The study was followed by a s i m i l a r one (Lyon et a l . , 1982) using childr e n i n primary grades. The same battery of te s t s was administered to 75 sc h o o l - i d e n t i f i e d reading-disabled children (SLD/R) and 42 normal readers (NR) matched for age and IQ. Ages ranged from 6-5 to 9-9 years with means of 8-2 years (SLD/R) and 39 8-1 years (NR). F u l l Scale WISC-R scores were i n the normal range with means of 102.9 (SLD/R) and 105.3 (NR). The SLD/R students exhibited a s i g n i f i c a n t d e f i c i t i n the PIAT Reading Recognition subtest whereas the NR students read at, or above, grade l e v e l . The Woodcock Reading Mastery t e s t was also administered to both groups of students. H i e r a r c h i c a l agglomerative techniques applied to the t e s t battery standardized scores was c a r r i e d out as i n the previous study. This analysis yielded f i v e c l u s t e r s ; eleven students (15%) could not be assigned to any of them. Subtypes s i m i l a r to f i v e of the s i x i d e n t i f i e d by Lyon and Watson (1981) were noted. Subtype 1 (24%) was characterized by r e l a t i v e strength i n l i n g u i s t i c s k i l l s and d e f i c i t s i n v i s u a l perception, visual-motor integration, and v i s u a l - s p a t i a l s k i l l s (Lyon and Watson's Subgroup 4). This subtype had the lowest Woodcock Word Recognition score. Members of Subtype 2 (13%) exhibited r e l a t i v e strength i n v i s u a l perceptual s k i l l s and d e f i c i t s i n receptive and expressive language (Lyon and Watson's Subgroup 3) . They also scored poorly on the Woodcock Word and Passage Comprehension subtests. Members of Subtype 3 (16%) exhibited a normal p r o f i l e s i m i l a r to that of Lyon and Watson's Subgroup 6 and had the best o v e r a l l Woodcock reading scores. The d e f i c i t s of Subtype 4 (20%) indicated that these students had d i f f i c u l t y remembering, analyzing, and c o r r e c t l y sequencing verbal and v i s u a l information. This subtype was compared with Lyon and Watson's Subgroup 5. Although t h i s subtype had the best Woodcock 40 Letter I d e n t i f i c a t i o n scores they had the lowest Word Attack, Word Comprehension and Passage Comprehension scores. Overall they had the second worst reading scores. The mixed d e f i c i t s of t h i s subtype were compared to Lyon and Watson's Subgroup 1. This group had the second worst reading scores. A subtype s i m i l a r to Lyon & Watson's (1981) Subtype 2 was not present i n the younger group. Lyon et a l . (1982) suggested that Lyon and Watson's Subtype 2 may have consisted of students who were more disabled at an e a r l i e r age i n s p e c i f i c areas, but improved over time u n t i l they exhibited a mixture of milder d e f i c i t s . The subtypes are summarized i n Table 2. The second group of researchers were involved i n the Carolina project. Although the main purpose of t h i s project was to subtype learning disabled students using behavioural variables (Speece, McKinney, & Appelbaum, 1985) and study the s t a b i l i t y of behavioural subtypes over a three-year period (McKinney & Speece, 1986), students were also subtyped using language variables (Feagans & Appelbaum, 1986). The behavioural subtyping studies w i l l be discussed f i r s t . I n i t i a l l y the Classroom Behavior Inventory (CBI) developed by Schaefer (see McKinney, 1984; Speece, McKinney, & Appelbaum, 1985) was used to d i f f e r e n t i a t e learning-disabled children from other groups (Feagans & McKinney, 1981; McKinney, McClure, & Feagans, 1982; McKinney & Forman, 1982). The CBI was used by teachers to rate students on such behaviour as academic competence, 41 Table 2. Subtypes i d e n t i f i e d u s i n g n e u r o p s y c h o l o g i c a l v a r i a b l e s and  h i e r a r c h i c a l agglomerative t e c h n i g u e s A u t h o r s / sample Number &/or name of subtype & c h a r a c t e r i s t i c s L y o n & W a t s o n (1981) 11-12 year-old reading-d i s a b l e d s t u d e n t s S u b t y p e 1 : d i f f i c u l t y w i t h a u d i t o r y r e c e p t i v e & e x p r e s s i v e language, v i s u a l p e r c e p t i o n , memory & v i s u a l -motor s k i l l s S u b t y p e 2: d i f f i c u l t y w i t h l a n g u a g e c o m p r e h e n s i o n , a u d i t o r y memory, & v i s u a l - m o t o r i n t e g r a t i o n S u b t y p e 3: d i f f i c u l t y w i t h l a n g u a g e comprehension & sound b l e n d i n g Lyon e t a l . (1982) 6-9 y e a r -old reading-d i s a b l e d s t u d e n t S u b t y p e 4: d i f f i c u l t y w i t h v i s u a l perception S u b t y p e 1: d i f f i c u l t y w i t h v i s u a l perception S u b t y p e 5 : d i f f i c u l t y w i t h naming, a u d i t o r y memory & sound b l e n d i n g S u b t y p e 2: d i f f i c u l t y w i t h r e c e p t i v e & e x p r e s s i v e language Subtype 6: normal p r o f i l e Subtype 3: normal p r o f i l e S u b t y p e 4 : d i f f i c u l t y w i t h naming, a u d i t o r y m e m o r y , sequencing verbal & v i s u a l i n f o r m a t i o n S u b t y p e 5 : d i f f i c u l t y w i t h r e c e p t i v e & e x p r e s s i v e language 42 d i s t r a c t i b i l i t y , introversion/extroversion, and s o c i a l behaviour. McKinney (1984) described an i n i t i a l study i n which 59 f i r s t -and second-grade, s c h o o l - i d e n t i f i e d learning-disabled childr e n were subtyped using h i e r a r c h i c a l c l u s t e r analysis with measures of i n t e l l e c t u a l a b i l i t y (WISC-R), achievement (PIAT), and classroom behaviour (CBI). Of the four subtypes that emerged a l l showed d e f i c i t s i n some areas of i n t e l l e c t u a l functioning and l e v e l s of achievement. Members of Subtypes I (33%), (II (10%), and III (47%) were perceived by classroom teachers to be low i n academic competence (independence, verbal a b i l i t y , and c u r i o s i t y ) whereas students i n Subtype IV (10%) were perceived to have the same l e v e l of academic competence as average achievers. Members of Subtype IV also had behavioural p r o f i l e s which resembled those of normal children. The three subtypes with behavioural d e f i c i e n c i e s could be d i f f e r e n t i a t e d on task-oriented behaviour, introversion versus extroversion, and h o s t i l i t y versus considerateness with Subtypes II and II e x h i b i t i n g the l e a s t desirable behaviour. The behavioural p r o f i l e s were validated using observations of LD resource teachers and t h e i r responses on the Pupil Rating scale (PRS). None of the teachers i d e n t i f i e d Subtype IV as having behaviour problems. The low achievement of t h i s group could not be explained by either i n t e l l e c t u a l or behavioural problems. Speece, McKinney, and Appelbaum (1985) used only the Classroom Behavior Inventory to subtype learning-disabled students i n the f i r s t stage of a longitudinal study. Their learning-disabled (LD) 43 sample consisted of f i r s t - and second-grade children who had been newly i d e n t i f i e d as learning-disabled by a m u l t i - d i s c i p l i n a r y team. Learning-disabled children were paired with average achieving nonlearning-disabled children (NLD) for comparison purposes. A l l were pa r t i c i p a n t s i n the three-year longitudinal study. The LD sample consisted of 63 children (mean age 86.3 months; mean IQ 96.1) and the NLD sample was made up of 66 children (mean age 85.9 months; mean IQ 107.9). The r e s u l t s of the CBI were subjected to c l u s t e r analysis. The researchers were interested only i n p r o f i l e shape as opposed to elevation or scatter, therefore c o r r e l a t i o n was chosen as the s i m i l a r i t y measure. Clusters were chosen based upon Ward's minimum variance method and the method of complete linkage. Seven c l u s t e r s emerged. Cluster 1 (28.6%) appeared to be a well-adjusted group with mild attention d e f i c i t s . Cluster 2 (25%) and Cluster 5 (9.5%) seemed to be variants of normal classroom behaviour. Members of Cluster 2 were s l i g h t l y more considerate and introverted than members of Cluster 5 who comprised a higher r a t i o of boys than the o v e r a l l sample. Members of Cluster 3 (14.3%) who were exclusively male, had mild attention d e f i c i t s combined with d i s t r a c t i b i l i t y , h o s t i l i t y , and inconsiderateness. In contrast, members of Cluster 4 (11.1%) the majority of whom were female, were characterized as withdrawn, dependent, and introverted. The small number i n Cluster 6 (6.3%) had mixed d e f i c i t s and were regarded as mild versions of Cluster 7 (4.88). Cluster 7 comprised three black males who were 44 r a t e d as s e r i o u s l y i m p a i r e d on a l l s c a l e s . Speece e t a l . , (1985) c a r r i e d o u t i n t e r n a l v a l i d a t i o n o f t h e s e c l u s t e r s u s i n g s p l i t sample r e p l i c a t i o n w h i c h i n v o l v e d r e a n a l y s i n g t h e d a t a o f two t h i r d s o f t h e sample. A second method used d i s c r i m i n a n t a n a l y s i s w i t h c l u s t e r membership as a d i s c r i m i n a n t f u n c t i o n . Both methods c o n f i r m e d t h e seven c l u s t e r s o l u t i o n . E x t e r n a l v a l i d a t i o n was a l s o p r o v i d e d by LD r e s o u r c e t e a c h e r s and a n o t h e r s e t o f b e h a v i o u r a l measures w h i c h p r o v i d e d a d d i t i o n a l i n f o r m a t i o n on how t h e c l u s t e r s d i f f e r e d . A t h r e e - y e a r f o l l o w - u p s t u d y conducted by McKinney and Speece (1986) examined t h e l o n g i t u d i n a l s t a b i l i t y o f t h e s e b e h a v i o u r a l s u b t y p e s and t h e academic consequences o f t h e i r b e h a v i o u r s . There was some l o s s o f s t u d e n t s t h r o u g h a t t r i t i o n s o . t h a t by y e a r 3 , t h e number o f LD s t u d e n t s i n t h e s t u d y had f a l l e n t o 47. Teachers r a t e d t h e b e h a v i o u r o f t h e s u b j e c t s on a y e a r l y b a s i s u s i n g t h e CBI and measures o f r e a d i n g and mathematics. Each y e a r a s e p a r a t e MANOVA was c o n d u c t e d on t h e s u b t y p e s u s i n g t h e t e a c h e r s ' r a t i n g s . The seven o r i g i n a l c l u s t e r s were c o l l a p s e d t o form f o u r c o m p o s i t e subgroups r e p r e s e n t i n g normal b e h a v i o u r p a t t e r n s , a t t e n t i o n d e f i c i t s , w ithdrawn b e h a v i o u r and c l a s s r o o m management prob l e m s . Those c h i l d r e n who had e x h i b i t e d c l a s s r o o m b e h a v i o u r problems and a t t e n t i o n d e f i c i t s d u r i n g t h e i r f i r s t and second g r a d e s had l o w e r achievement l e v e l s t h a n t h o s e who had normal o r wit h d r a w n b e h a v i o u r p r o f i l e s , a l t h o u g h t h e l a t t e r group c o n s i s t e d o f o n l y t h r e e s t u d e n t s . On t h e whole, membership i n t h e f o u r 45 subgroups remained only moderately stable from year to year. The normal behaviour subgroup remained the most stable throughout. The most l i k e l y trend over time was for learning-disabled students to be c l a s s i f i e d i n an at y p i c a l subtype rather than to move to a more adaptive one. Behavioural subtypes are summarized i n Table 3. As part of the Carolina Project, a series of language measures were administered to the o r i g i n a l sample of 63 LD students during the f i r s t year of the study. F i f t y - f i v e members of the LD sample were used to f i n d language subtypes (Feagans & Appelbaum, 198 6) and the o r i g i n a l sample of 66 NLD students was also included i n the study. Six variables taken from the set of language measures were used to subtype the students using h i e r a r c h i c a l agglomerative techniques. Six clust e r s emerged and t h e i r p r o f i l e s plotted, using the scores of the NLD group as standard reference points for each va r i a b l e . Members of Cluster 1 (16%) had acquired basic language structures but were unable to use these s k i l l s i n understanding and paraphrasing narratives. The members of Cluster 2 were characterized by superior vocabularies but l i k e Cluster 1 were unable to use t h i s knowledge. Cluster 3 had members who talked a l o t , but the meaning and substance of t h e i r language was poor. Members of Cluster 4 were the opposite of Cluster 1 i n that they had poor vocabulary and syntax yet were able to paraphrase narratives quite adequately. Cluster 5 was s i m i l a r i n p r o f i l e to Cluster 4 but members had superior language s k i l l s . Cluster 6 was 46 Table 3. Subtypes i d e n t i f i e d u s i n g b e h a v i o u r a l v a r i a b l e s and h i e r a r c h i c a l agglomerative technigues A u t h o r s / sample Number &/or name of subtypes & c h a r a c t e r i s t i c s M c K i n n e y (1984) grade 1 - 2 l e a r n i n g -d i s a b l e d s t u d e n t s Subtype I : poor t a s k - o r i e n t e d b e h a v i o u r , i n t r o v e r t e d & h o s t i l e S u b t y p e I I : behaviour s i m i l a r t o but more undesirable than Subtype 1 Subtype I I I : b e h a v i o u r s i m i l a r t o but more undesirable than Subtype 2 Speece e t a l . (1985) grade 1-2 l e a r n i n g -d i s a b l e d s t u d e n t s S u b t y p e I V : N o r m a l b e h a v i o u r a l p r o f i l e Subtype 1: w e l l -a d j u s t e d w i t h m i l d a t t e n t i o n d e f i c i t s Subtype 2: normal b e h a v i o u r a l p r o f i l e Subtype 3: some d i s t r a c t i b l e , h o s t i l e & i n c o n s i d e r a t e behaviour w i t h m i l d a t t e n t i o n d e f i c i t s and h y p e r a c t i v i t y S u b t y p e 4: w i t h d r a w n , d e p e n d e n t & i n t r o v e r t e d Subtype 5: normal b e h a v i o u r a l p r o f i l e S u b t y p e 6: d e p e n d e n t , withdrawn & i n c o n s i d e r a t e , a m i l d v e r s i o n of Subtype 7 S u b t y p e 7 : w i t h d r a w n , d e p e n d e n t , i n t r o v e r t e d & i n c o n s i d e r a t e , e x t r e m e l y d i s t r a c t i b l e & h o s t i l e 47 s i m i l a r to Cluster 1 i n that members had strong vocabulary and syntax s k i l l s r e l a t i v e to t h e i r narrative s k i l l s . Subtypes 5 and 6 were considered to exhibit normal language patterns. Subtypes were validated using a series of MANOVAs of Nonverbal IQ scores and achievement variables that comprised PIAT Reading Recognition, Reading Comprehension and Math subtests. However, no mention was made of the behavioural subtypes obtained using the same LD group (Speece et a l . , 1986), no comparison was made between behavioural and language subtypes, and no inferences were drawn. The t h i r d project to be discussed was the F l o r i d a project (Satz & Morris, 1981). In an attempt to avoid using exclusionary c r i t e r i a to sel e c t a learning-disabled sample, c l u s t e r analysis was used to define a target subgroup and comparison subgroups i n a population that consisted of 236 white male, grade 6 students. WRAT achievement scores were analyzed using h i e r a r c h i c a l agglomerative techniques. Of the nine subtypes that emerged, two had members with s u f f i c i e n t l y depressed achievement scores that they could be regarded as learning-disabled. There were 89 students i n these two c l u s t e r s . Four neuropsychological tests were administered to these students and the r e s u l t s submitted to h i e r a r c h i c a l agglomerative techniques producing f i v e d i s t i n c t and stable subtypes. The neuropsychological t e s t s were selected because they loaded highly on a language factor and a perceptual factor. Verbal fluency and the WISC S i m i l a r i t i e s 48 subtest comprised the language factor and the t e s t of Visual-Motor Integration and a t e s t of v i s u a l recognition/discrimination made up the perceptual factors. The Peabody Picture Vocabulary Test (PPVT) was used as an IQ marker variable and neurological status, SES, parental reading l e v e l and the Children's Personality Questionnaire (CPQ) were included as c r i t e r i o n v a r i a b l e s . Subtype 1, whose members were severely impaired on both language t e s t s and the PPVT, was considered to represent a global language impairment. Members of Subtype 2 were impaired only on the verbal fluency t e s t and were therefore considered to be a s p e c i f i c language d i s a b i l i t y subtype. Subtype 3 members were impaired on a l l the neuropsychological tests, including the PPVT, and were thought to have a mixed-type global language and perceptual impairment. A perceptual-motor impairment characterized the members of Subtype 4 who performed poorly on the perceptual t e s t s . Subtype 5 was regarded as an unexpected group because i t s members showed no impairment on the neuropsychological variables. The language and neuropsychological subtypes of the Carolina (Feagans & Appelbaum, 1986) and F l o r i d a (Satz & Morris, 1981) studies are summarized i n Table 4. Satz and Morris (1981) recognized that t h e i r study was li m i t e d by : 1) the r e s t r i c t e d range of reading s k i l l s sampled, 2) the use of only four neuropsychological variables for c l u s t e r analysis, 3) the exclusion of a subgroup whose members were impaired only on the Arithmetic subtest, 4) the use of a highly homogeneous white male 49 Table 4. Subtypes i d e n t i f i e d u s i n g n e u r o p s y c h o l o g i c a l and/or language  v a r i a b l e s and h i e r a r c h i c a l agglomerative t e c h n i g u e s A u t h o r s / sample Number &/or name of subtypes & c h a r a c t e r i s t i c s Language v a r i a b l e s Feagans & A p p e l b a u m (1986) Grade 1-2 learning d i s a b l e d s t u d e n t s Subtype 1: poor un d e r s t a n d i n g & p a r a p h r a s i n g of n a r r a t i v e , adequate language s k i l l s S u b t y p e 4 : a d e q u a t e und e r s t a n d i n g & p a r a p h r a s i n g of n a r r a t i v e , poor v o c a b u l a r y Subtype 2: poor understanding & p a r a p h r a s i n g of n a r r a t i v e , s u p e r i o r v o c a b u l a r y S u b t y p e 5: superior language a b i l i t y S u b t y p e 3 : talked a l o t but m e a n i n g & substance of language poor S u b t y p e 6 : a d e q u a t e u n d e r s t a n d i n g & p a r a p h r a s i n g of n a r r a t i v e , good vocabulary & syntax s k i l l s Language & n e u r o p s y c h o l o g i c a l v a r i a b l e s S a t z & M o r r i s (1981) Grade 6 white male l e a r n i n g d i s a b l e d s t u d e n t s Subtype 1: global l a n g u a g e impairment S u b t y p e 2: s p e c i f i c language d i s a b i l i t y (verbal fluency) S u b t y p e 3 : m i x e d - t y p e g l o b a l language and p e r c e p t u a l impairment S u b t y p e 4: perceptual-motor impairment Subtype 5: no n e u r o l o g i c a l impairment 50 sample, 5) the small number of c r i t e r i o n measures used to v a l i d a t e the subtypes, and 6) the use of the PPVT as a measure of i n t e l l e c t u a l a b i l i t y . Limitations of the Subtyping Approach Although i t was hoped that c l u s t e r analysis would redress the l i m i t a t i o n s of the more subjective methods of subtyping t h i s too had i t s l i m i t a t i o n s . The expectations were that the c l u s t e r i n g of reading-disabled students into homogeneous subtypes would lead to the discovery of underlying causes of the disorder and enable remedial programs to be planned. However, problems i n terminology, l a b e l l i n g , a p p l i c a t i o n and v a l i d a t i o n were i d e n t i f i e d (Morris et a l . , 1981). These problems appear to be related to the large number of ava i l a b l e techniques, t h e i r complexity, and researchers' lack of f a m i l i a r i t y with them. One l i m i t a t i o n of c l a s s i f i c a t i o n studies i s that the subtypes can only be described i n terms of the variables used to i d e n t i f y them. Subtypes i d e n t i f i e d using neuropsychological variables are described i n terms of the e f f i c i e n t or i n e f f i c i e n t ways the neuropsychological processes operate, (e.g., Petruskas & Rourke, 1979)). Those i d e n t i f i e d using reading and reading related v a r i a b l e s are described i n terms of strengths and weaknesses in reading s u b s k i l l s (e.g., Doehring et a l , 1981). Those i d e n t i f i e d using behavioural variables are described i n terms of appropriate 51 and inappropriate behaviour patterns (e.g., Speece et a l , 1985). Similar subtypes are produced, even when used with samples of a d i f f e r e n t nature (Doehring & Hoshko, 1977; Doehring et a l . , 1981; Fisk & Rourke, 1979; Petruskas & Rourke, 1979). In contrast, d i f f e r e n t subtypes are produced within the same samples when d i f f e r e n t variables are used (Doehring et a l , . 1981; Feagans & Appelbaum, 1896; Speece et a l , 1985). In the Carolina project, no mention has been made of a relati o n s h i p , i f any, between the subtypes obtained using behavioural variables and those obtained using language variables. Another l i m i t a t i o n concerns differences between the reading-and learning-disabled samples and i n the c l a s s i f i c a t i o n techniques used, a l l of which make i t d i f f i c u l t to compare subtypes from study to study or generalize findings to other reading-disabled populations. E l l i s (1985) pointed out the anomalies created when subtypes produced by d i f f e r e n t studies are compared with each other and gave the following example. The language disordered subtype of Mattis et a l . (1975) was compared by Doehring and Hoshko (1977) with t h e i r Subtype 1 which was characterized by an or a l reading d e f i c i t (Doehring & Hoshko, 1977). The same subtype was also compared by Boder (1973) with her dysphonetic subtype. Boder's dysphonetic subtype was i n turn compared by Satz and Morris (1981) with t h e i r Subtype 2 which was impaired only on a t e s t of verbal fluency. In E l l i s ' s words, "Something somewhere has obviously gone awry" ( E l l i s , 1985, p. 180). The subtyping l i t e r a t u r e i s studded 52 with s i m i l a r anomalies. E l l i s also noted that some researchers seem to believe that "the dyslexic substrate i s naturally fissured, and that v i r t u a l l y any form of a g i t a t i o n w i l l cause i t to cleave along s i m i l a r l i n e s " (p. 180) . The f u t i l i t y of t h i s point of view i s evident i n the overlap of subtype d e f i c i t s , the subjects that remain u n c l a s s i f i e d , and the subjects that have mixed c l a s s i f i c a t i o n . A major c r i t i c i s m l e v e l l e d by Lovett (1984) i s that i d e n t i f i c a t i o n of subtypes provides l i t t l e information as to how the i d e n t i f i e d d e f i c i t a f f e c t s the reading process. Taylor, Fletcher and Satz (1982) also perceive that there i s a lack of integration between reading subtypes and t h e i r neurological evaluation, and no detailed evaluation of the reading dysfunction i s supplied. Componential Analysis In developing the procedure known as "componential analysis", Sternberg (1977) attempted to c a p i t a l i z e on the strengths of both the d i f f e r e n t i a l and information-processing approaches. The purpose was to i s o l a t e the components of i n t e l l i g e n t performance and discover t h e i r organization and re l a t i o n s h i p . Sternberg based his d e f i n i t i o n of a component on that of Newel and Simon (1972) . "A component i s an elementary information process that operates upon in t e r n a l representation of objects or symbols" (Sternberg, 1977, 53 p. 65). He theorized that cognitive behaviour could be explained by a r e l a t i v e l y small number of component processes. Componential analysis i s divided into f i v e steps: i d e n t i f i c a t i o n of component processes, s p e c i f i c a t i o n of a combination rule for d i f f e r e n t components, s p e c i f i c a t i o n of a combination rule for the same components, discovery of component latencie s , and discovery of r e l a t i o n s of components to each other and to higher mental a b i l i t i e s (Sternberg, 1977). Analogical reasoning i s viewed as an important aspect of i n t e l l i g e n c e and i s often used i n tests of reasoning a b i l i t y (Sternberg, 1977). In reviewing theory and research i n analogical reasoning, Sternberg (1977) found that i n general most theories were incomplete, unsupported by empirical evidence, too s p e c i f i c and narrow, and unable to account for i n d i v i d u a l differences i n information processing. He developed a theory of analogical reasoning to redress most of these weaknesses and used componential analysis of analogical reasoning tasks to provide empirical evidence to support i t (Sternberg, 1977; Sternberg & Nigro, 1 9 8 0 ; Sternberg & R i f k i n , 1979). The analogies take the form A i s to B as C i s to D (A:B::C:D) where D i s eithe r D i o r D2, the answer options. Sternberg proposed s i x components for analogical reasoning: encoding, inference, mapping, application, j u s t i f i c a t i o n , and preparation/response. The encoding component requires the subject to look at each of the terms of the analogy (A, B, C, Di, & D2) and attach meaning to them. 54 In making an inference the subject establishes a re l a t i o n s h i p between the f i r s t and second terms (A & B) and i n mapping, between the f i r s t and t h i r d terms (A & C) . Application requires the subject to e s t a b l i s h a re l a t i o n s h i p between the t h i r d term (C) and one of the answer options (Di& D2) that i s analogous to the re l a t i o n s h i p i n f e r r e d between the f i r s t two terms. J u s t i f i c a t i o n i s the process whereby the subject decides one answer option i s better than another when neither of them exactly f i t s the perceived r e l a t i o n s h i p . In making a response the subject chooses one of the response options (Dior D2) and indicates that choice i n some way. The combination rule may specify processing to be s e r i a l as opposed to p a r a l l e l , exhaustive as opposed to self-terminating, or h o l i s t i c as opposed to an a l y t i c . I f components are executed i n s e r i a l order, the t o t a l time i s the sum of the time taken to process the i n d i v i d u a l components. S i m i l a r l y , the time for multiple executions of one component i s the sum of a l l s i n g l e executions. I f the components are processed i n p a r a l l e l , then the t o t a l time i s equal to that of the most time-consuming component. The time for multiple executions i n p a r a l l e l of one component i s equal to the most time-consuming single execution. Exhaustive and self-terminating processing r e f e r to the mode or strategy used i n executing the components. When exhaustive processing i s s p e c i f i e d , a l l theorized components are executed the maximum number of times. When a self-terminating strategy i s used, components are executed only the necessary number of times. I f the 55 a n a l y t i c method i s theorized, then separable processes are assumed to be used. The h o l i s t i c method of processing makes i t impossible to separate the time taken for the execution of one component process from the others. Analogical reasoning i s theorized to be processed s e r i a l l y which means that solution time for one analogy i s the sum of the time taken to execute each component used i n the solution. I t i s also assumed that components are processed a n a l y t i c a l l y , making i t possible to separate t h e i r execution times. However, there are occasions when multiple executions of the same component are not always assumed to be ana l y t i c . For degenerate forms of analogies ( i . e . , f i r s t and t h i r d terms are i d e n t i c a l making second and fourth terms i d e n t i c a l ) , h o l i s t i c processing i s assumed. The i d e n t i f i c a t i o n of the components and s p e c i f i c a t i o n of a combination r u l e are the t h e o r e t i c a l aspects of componential analysis but theory alone cannot f u l l y explain the behaviour being studied. Therefore, i t i s necessary to devise models that specify the order of execution of the components and the mode of execution. Di f f e r e n t models are theorized for d i f f e r e n t modes of processing; the models of analogical reasoning specify both exhaustive and self-terminating modes. The models are displayed p i c t o r i a l l y as flow charts and mathematically as equations so that they can be tested empirically (see Sternberg, 1977). A task i s broken down into a series of subtasks that generate i n t e r v a l scores or latencies. These scores represent the time taken 56 to process successively less information and the time i n t e r v a l s become successively shorter. Estimation of component time i s achieved through multiple regression analysis using task latencies as the dependent variables. The independent or predictor variables associated with each component represent the amount of processing that must be done to solve the analogy. In analogical reasoning they are dependent upon the semantic distance between terms for verbal analogies, or, i n the case of picture analogies, the number of a t t r i b u t e values that can change from term to term. The regression c o e f f i c i e n t s are interpreted as estimates of duration and d i f f i c u l t y of a single component process. They are parameter estimates f o r i n d i v i d u a l subjects, calculated from the component model which best f i t s the data. Componential analysis enables i n d i v i d u a l differences to be examined at the theory, model, and componential l e v e l s . At the theory l e v e l , i n d i v i d u a l s may d i f f e r i n the components they use and i n the r u l e with which they combine them. At the model l e v e l , i n d i v i d u a l s may d i f f e r i n the order i n which they process the components and i n the mode they adopt. At the component l e v e l , i n d i v i d u a l s may d i f f e r i n t h e i r speed and power of processing. External v a l i d a t i o n of the component scores i s established by showing that they correlate highly with t e s t s of mental a b i l i t y . One v a r i a b l e i s the score on such a t e s t and the other i s component latency or d i f f i c u l t y estimated at the i n d i v i d u a l l e v e l . I f component a b i l i t i e s can account for almost a l l the variance i n the 57 t e s t score, then the nature of the a b i l i t y that i s being measured i s indicated. Components can also be correlated with each other to give an i n d i c a t i o n of the rel a t i o n s h i p between them i n terms of the time taken for each component process and degree of d i f f i c u l t y . A pplication of Componential Analysis In Sternberg's f i r s t studies, reasoning tasks were broken into a series of subtasks through a method of precuing (Sternberg, 1978). In t h i s method, part of a task i s presented i n an untimed s i t u a t i o n (precued), then the amount of time taken to complete the task i s measured. Interval scores that represent stages of processing are formed by varying the amount of precuing. Sternberg made the assumption of a d d i t i v i t y with respect to i n t e r v a l scores and was able to show empirically that such an assumption was tenable. This method was used with analogical reasoning (Sternberg, 1977) , s y l l o g i s t i c reasoning (Sternberg, 1980a; 1980b), c l a s s i f i c a t i o n tasks, series completion tasks, and topology (visual s p a t i a l problems (Sternberg, 1978). Advantages included: 1) disentanglement of components that would otherwise be confounded, 2) comparison of models that would otherwise be indistinguishable, 3) an increase i n the number of data points to be modelled, 4) the necessity for e x p l i c a t i o n of the model i n d e t a i l , and 5) provision of a series of nested i n t e r v a l scores rather than a single solution score. Later methods included 58 presentation of p a r t i a l tasks which was used with s y l l o g i s t i c reasoning (Sternberg, 1980b), a method of stem-splitting used with verbal analogies (Sternberg & Nigro, 1980), and a method of systematically varied booklets used with picture analogies (Sternberg & R i f k i n , 1979). O r i g i n a l l y , the theory of analogical reasoning processes was tested using u n i v e r s i t y students (Sternberg, 1977). Studies using a v a r i e t y of analogical reasoning tasks showed that, generally, those subjects who spent more time on encoding solved the analogies i n l e s s time and were more successful. There was also an i n d i c a t i o n that subjects higher i n reasoning a b i l i t y tended to be more systematic i n t h e i r solution strategy (Sternberg, 1977). Later the theory was used with c h i l d subjects to t e s t i t s g e n e r a l i z a b i l i t y and measure the development of analogical reasoning processes i n children (Sternberg & R i f k i n , 1979). This was done using two kinds of picture analogies (see Figure 1) which were presented v i a systematically varied booklets. These picture analogies enabled subjects to be tested i n groups, required no s p e c i a l equipment, and were p r a c t i c a l even with young children because no reading was required. The Schematic Picture Analogies, shown i n Figure 1, consists of figures that are separable from t h e i r a t t r i b u t e s . This means that the a t t r i b u t e s of the figures (hat colour, s u i t pattern, footwear, and hand baggage) can be removed and the figures w i l l remain i n t a c t . People Piece Analogies (see Figure 1) consist of figures with i n t e g r a l a t t r i b u t e s . The (i) Schematic Picture Analogy Figure 1. Picture Analogies 60 a t t r i b u t e s of these figures (weight, height, gender or colour) are integrated with the figure. The removal of an a t t r i b u t e , therefore, removes or i n t e r f e r e s with the intactness of the figure. The current study i s concerned with Schematic Picture Analogies and discussion w i l l focus on the research concerning these. Schematic Picture Analogies, shown i n Figure 2, vary i n four binary a t t r i b u t e s ; hat colour (white, black), s u i t pattern (striped, dotted), hand baggage (brief case, umbrella), and footwear (shoes, boots). These are the at t r i b u t e s to be encoded. The inference, mapping, and application components are explained i n Figure 2 i n terms of a t t r i b u t e changes from term to term. The combination r u l e for the t o t a l time taken to process the components i s additive. A basic difference between analogies with i n t e g r a l and separable a t t r i b u t e s concerns the psychological mechanisms to which they are subjected (Sternberg & R i f k i n , 1979). To d i s t i n g u i s h the processing of analogies with i n t e g r a l a t t r i b u t e s from those with separable a t t r i b u t e s , Sternberg and R i f k i n adapted the d i f f e r e n t i a t i o n made by P e r f e t t i and Lesgold (1977) between "procedures" and "strategies". "A procedure i s a nonoptional, nonconscious model of information processing: Subjects carry out a procedure with l i t t l e or no awareness of what i s taking place. A strategy i s an optional conscious model of information processing: Subjects carry out a strategy f e e l i n g f u l l y aware of what i s taking place" (Sternberg & R i f k i n , 1979, pp. 199-200). B Dl D2 Components 1. Encoding: 2. Inference: Mapping: Application: 5. Response: Attri b u t e s hat colour s u i t pattern footwear handgear Values black, white striped, dotted shoes, boots umbrella, suitcase the relationship between A and B hat colour (black to white), s u i t pattern (no change), footwear (boots to shoes), handgear (no change) the relationship between A and C hat colour (no change), s u i t pattern dotted to striped, footwear (no change) handgear (no change) the relationship between C and Dl hat colour (white to black), s u i t pattern (no change), footwear (boots to shoes), handgear (no change the relationship between C and D2 hat colour (no change), s u i t pattern (striped to dotted), footwear (no change), handgear (suitcase to umbrella c i r c l e around Di or D2 Combination Rule; Total Time i s the sum of encoding time, inference time, application time, and response time. Figure 2. Schematic Picture Analogy 62 Of the models suggested by Sternberg (1977) , four are procedural models applicable to analogies with i n t e g r a l a t t r i b u t e s and referred to as Models I, I I , I I I , and IV (see Table 5). These procedural models combine exhaustive encoding with other components in exhaustive and self-terminating mode. They also contain a mapping component but there i s no j u s t i f i c a t i o n component as only one of the answer options i s correct. I t i s therefore necessary for subjects to store a l l attributes and corresponding values i n short-term memory to f a c i l i t a t e l a t e r analogy solution. The three other models (Table 5) are modifications of the procedural models referred to as Models IM, II-IIIM, and IVM. These s t r a t e g i c models d i f f e r from the procedural models i n two ways that are linked to the s e p a r a b i l i t y of the att r i b u t e s i n the Schematic Picture Analogies. The f i r s t difference i s the absence of the Mapping component and the second i s the change from exhaustive to self-terminating encoding immediately before the f i r s t s e l f -terminating attribute-comparison process. Thus Model IM d i f f e r s from Model I only through the absence of the mapping component and a l l components are processed exhaustively. Model II-IIIM i s a combination of Models II and III because they are i d e n t i c a l when the Mapping component i s removed. Encoding i s exhaustive for the f i r s t two terms (A & B) but becomes self-terminating when the C term i s encoded. This means that inference i s exhaustive but app l i c a t i o n i s self-terminating. In the IVM model, encoding becomes self-terminating with the f i r s t term of the analogy (A) and the 63 Table 5. Sternberg's t h e o r e t i c a l models Model Components Procedural models exh exh exh exh I encoding + inference + mapping + app l i c a t i o n + response exh exh exh st II encoding + inference + mapping + app l i c a t i o n + response exh exh st st III encoding + inference + mapping + app l i c a t i o n + response exh st st st IV encoding + inference + mapping + app l i c a t i o n + response Strategic models exh exh exh IM encoding + inference + application + response exh/st exh st IIM-IIM encoding + inference + app l i c a t i o n + response st st st IVM encoding + inference + app l i c a t i o n + response exh = st = exhaustive mode self-terminating mode 64 other components are processed i n self-terminating mode. The systematically varied booklets each contain 16 schematic pic t u r e analogies. The booklets are "homogeneous" i n that the number of a t t r i b u t e values that change from A to B, A to C, and Di to D2(the answer options) remain constant within booklets. Students are given 64 seconds per booklet to solve as many analogies as they can. Three dependent variable scores are obtained for each booklet. Mean solution latency correct i s obtained by d i v i d i n g 64 (the number of seconds allowed to solve each booklet) by the number of analogies solved c o r r e c t l y . This r e f l e c t s quantity and q u a l i t y of performance. S i m i l a r l y , a mean solution latency t o t a l i s obtained by d i v i d i n g 64 by the t o t a l number of analogies solved. This r e f l e c t s only quantity of performance. The t h i r d variable, error rate, which i s obtained by d i v i d i n g the number of errors by the number of analogies solved, r e f l e c t s only q u a l i t y of performance, Independent variables are associated with each t h e o r e t i c a l component and consist of "objective" distances between analogy terms and depend upon the mode of processing that i s used. Objective distances i n these analogies are dependent upon the number of a t t r i b u t e values that change from term to term. The degree of d i f f i c u l t y i s related to the number of a t t r i b u t e changes. Therefore, the components with the greater number of att r i b u t e changes require more processing and take more time. 65 When components are processed exhaustively, the values of the variables are d i r e c t l y equal to the number of a t t r i b u t e changes between the appropriate terms of the analogy. When processed i n self-terminating mode, the values of the variables are calculated by applying a m u l t i p l i e r to the number of a t t r i b u t e changes between the appropriate terms. The m u l t i p l i e r i s dependent upon the a t t r i b u t e s that can be processed exhaustively (always 4) and the number of values (1, 2, or 3) of Di that are the same as D2 (the answer options). The basic design of an experiment using Schematic Picture Analogies consists of the crossing of subjects, whose scores represent mean latencies, with the t e s t materials. Estimates of the time taken to process each of the components are obtained from multiple regression analysis using the independent and dependent variables describe previously. Sternberg and R i f k i n (1979) used the method with subjects enroled i n grades 2, 4, and 6 at a r e l i g i o u s day school, and with adult graduate and undergraduate students attending Yale University. Groups were approximately equally divided among the sexes and consisted of 21 second graders (mean age 8 years) , 22 fourth graders (mean age 10 years), 18 s i x t h graders (mean age 12 years), and 18 adult students (mean age 19 years). There was no evidence that the adult students were drawn from the same population as the elementary school students but the authors considered i t reasonable to suppose that the adult students 66 represented a group that was further along i n the developmental sequence. Students were tested i n large groups. They received 2 4 test booklets each containing 16 schematic picture analogies. They also received a geometric analogies t e s t and a t e s t of perceptual speed but Sternberg and R i f k i n considered the r e s u l t s uninteresting and therefore did not discuss them. The school childr e n completed the procedures i n f i v e sessions of about 3 0 minutes and the adults i n one long session with a rest break i n the middle. However, administering the analogies i n so many short sessions may have been unnecessary, as i n Experiment 2 the 24 People Piece booklets were administered i n a single session to a l l except grade 2 students. Administration of the Schematic Picture Analogies and a b i l i t y t e s t s comprised Experiment 1 i n the study of Sternberg and R i f k i n (1979). The People Piece Analogies, which were administered a year l a t e r with d i f f e r e n t student samples from the same population, comprised Experiment 2. As the Schematic Picture Analogies were used i n the current study, discussion w i l l focus on Experiment 1. As expected, the mapping component was found to be unnecessary fo r s o l u t i o n of analogies with separable a t t r i b u t e s , although i t was used by a l l Grade 2 students i n solving analogies with i n t e g r a l a t t r i b u t e s . The modified self-terminating model was the preferred model of analogical reasoning at each age l e v e l . There was evidence that there were better model f i t s at higher age l e v e l than lower ones, suggesting greater o v e r a l l consistency i n the data with 67 increasing age. Sternberg (1977) had showed that more successful adult reasoners took longer to encode s t i m u l i than unsuccessful ones and suggested that there was a trade-off between encoding speed and the speed of performing l a t e r operations. In the Sternberg and R i f k i n study, the same patterns emerged, leading them to state, "The more sophisticated strategy then i s to lengthen one's encoding latency i n order to shorten one's comparison latency (Sternberg & R i f k i n , 1979, p. 230). A study by Wilson (1980) used Schematic Picture Analogies with fourth grade students attending four elementary schools i n a metropolitan area i n southwestern B r i t i s h Columbia, Canada. The students represented a range of socioeconomic l e v e l s and came from a v a r i e t y of ethnic backgrounds. They had been routinely administered the Canadian Test of Basic S k i l l s (CBTS) during t h e i r grade 3 year. Their percentile scores were transformed into Normal Curve Equivalent scores (NCEs) which are equal i n t e r v a l scores (Mean 100, SD 21.06). Three samples of twenty students each were selected to represent low, average, and high a b i l i t y l e v e l s . The low a b i l i t y group had scores from one to three standard deviations below the sample mean (group mean 36.51), the average a b i l i t y group scores ranged between one standard deviation above and below the sample mean (group mean 60.92), and the high a b i l i t y group scores ranged from one to three standard deviations above the sample mean (group 68 mean 89.66). The mean ages for the low, average, and high a b i l i t y groups were 9 years 9 months, 9 years 8 months, and 9 years 8 months, respectively. The Schematic Picture Analogies were administered over two sessions of approximately an hour each to groups of students that ranged i n number from a minimum of 11 to a maximum of 22. Students were given 64 seconds to solve each analogy booklet and t h e i r answers were recorded on a separate sheet. Three c r i t e r i o n variables were calculated but the variance i n C r i t e r i o n Variable 3 (error rate) was judged i n s u f f i c i e n t for analysis to take place. Following the method of Sternberg and R i f k i n (1979) analysis of the remaining data was ca r r i e d out but only the r e s u l t s for C r i t e r i o n Variable 1 (mean latency correct) were reported and discussed. Results for the average and high a b i l i t y groups were s i m i l a r to those obtained by Sternberg and R i f k i n , showing a preference for the modified self-terminating model. However, the low a b i l i t y group appeared to marginally prefer an exhaustive model. In general, the high a b i l i t y group was more accurate, devoted more time to encoding, and used the self-terminating strategy more consistently. None of the groups used the mapping component. There was evidence that one or more additional systematic factors contributed to the variance i n the latency data for the low a b i l i t y group. These additional factors were not accounted for by any of the hypothesized models. I t was suggested 69 that such factors might be att r i b u t a b l e to differences i n response s t y l e . In a study that e s s e n t i a l l y investigated metacognitive processes (Sternberg & Ketron, 1982), the Schematic Picture Analogies and the People Piece Analogies were administered to Yale University students who were trained to use one of three d i f f e r e n t strategies or not trained at a l l . One hundred and s i x t y students were divided into eight groups of twenty. Four of the groups received the Schematic Picture Analogies and four received the People Piece Analogies. In addition they were administered tests that measured reasoning and memory as well as a questionnaire pertaining to the strategy they used to solve the analogies. Discussion w i l l centre on those r e s u l t s that concern Schematic Picture Analogies although the re s u l t s from the People Piece Analogies w i l l be referred to as appropriate. For each set of analogies, students i n three of the groups were trained to use a s p e c i f i c strategy to solve the analogies; those i n the fourth group were t o l d to solve them " i n whatever way you think i s best". The trained groups were each t o l d to proceed through three stages. The f u l l y exhaustive group were t o l d to consider a l l the options at each stage of processing before s e l e c t i n g an answer; those i n the f u l l y self-terminating group were t o l d to process, at each stage, only as much information as needed to a r r i v e at a solution; and those i n the mixed condition group were t o l d to f i n d the rel a t i o n s h i p between a l l the attr i b u t e s of 70 the A and B terms (exhaustive inference) and then process only as much information as needed (self-terminating application) to select an answer. B a s i c a l l y , for analogies with separable a t t r i b u t e s , the f u l l y exhaustive group was trained to use Model IM, the f u l l y s e l f -terminating group was trained to use Model IVM, and the mixed group was trained to use Model II-IIIM. This study indicated that i t was possible to t r a i n students to use a desired strategy or reasonable approximation to that strategy to solve People Piece Analogies. In contrast, the Schematic Picture Analogies r e s u l t s showed that although a few students trained to use Model IM may have used Model IV, the majority of students used Model IVM whether they had been trained to use that strategy or not. Results from the questionnaire indicated that there was an i n t e r a c t i o n between content type and strategy use. Students i n the i n t e g r a l - a t t r i b u t e groups appeared to follow the instructions they were given, and t h e i r descriptions of what they did were consistent with t h e i r strategy use. Those i n the separable a t t r i b u t e groups were unable to follow the instructions given, but t h e i r questionnaire responses indicated that they thought they had indeed followed i n s t r u c t i o n s . Sternberg and Ketron concluded that the accuracy of s e l f - r e p o r t i s dependent, at l e a s t i n part, upon the content of a given task. S i g n i f i c a n t c o rrelations between separable-attribute solution latency and tests of abstract reasoning were observed only for the 71 untrained students and those trained to use a self-terminating mode. Thus t r a i n i n g students to use a strategy they would not have used spontaneously appears to reduce the c o r r e l a t i o n of task latency and psychometric task performance. Correlations between questionnaire responses and task scores for analogies with separable a t t r i b u t e s suggest that those who solved the analogies fas t e s t believed they worked quickly, found the strategy they used easy to learn and maintain, were most conscious of t h e i r own strategy use and were of the opinion that i t placed less demand upon memory. Sternberg and Ketron (1982) suggest that there i s a strong compulsion to solve analogies with separable a t t r i b u t e s using a self-terminating mode that cannot be e a s i l y overcome with t r a i n i n g . They note, from previous research (Sternberg & R i f k i n , 1979), that t h i s i s the model of choice from childhood to adulthood and conclude that spontaneous use of t h i s , the most e f f i c a c i o u s model, i s consistent with the properties of the model. Throughout Sternberg's research using componential analysis, one finding continued to perplex (Sternberg, 1981). This was the consistently high and r e p l i c a b l e c o r r e l a t i o n between the regression constant, which represented the response component, and i n t e l l i g e n c e t e s t scores. Sternberg (1981) theorized that the componential procedures of task decomposition were unable to extract additional c r i t i c a l components of i n t e l l i g e n t performance because they were masked by the response component. In an attempt 72 to study strategy planning and strategy execution he used a nonentrenched or novel task. The task used was a verbal analogical reasoning task but the analogies d i f f e r e d from previous ones i n that they had from one to three analogy terms missing and the p o s i t i o n of the missing terms varied from case to case. The missing terms were selected from supplied answer options. The analogies were analyzed using componential analysis to determine how much time was spent on global planning and how much on l o c a l planning. Global planning referred to a macrostrategy applied to a set of problems whereas l o c a l planning referred to a microstrategy applied to a p a r t i c u l a r problem within the set. Global planning latencies were associated with the sameness of the analogies i n a p a r t i c u l a r set and l o c a l planning latencies with the d i f f i c u l t y of the p a r t i c u l a r item. Latency scores were correlated with the composite of two p e n c i l -and-paper l e t t e r series completion tasks. Results showed that a l l latency scores were moderately correlated with scores on the pencil-and-paper reasoning tasks i n a d i r e c t i o n that indicated shorter latencies were related to higher reasoning scores. In contrast, global planning was moderately correlated with the same scores i n a d i r e c t i o n that indicated longer latencies were related to higher reasoning scores. Local planning was only weakly correlated with reasoning scores. The pattern of latencies showed that better reasoners spent r e l a t i v e l y more time on global planning and r e l a t i v e l y l e s s time on l o c a l 73 planning. F i n a l l y the regression constant continued to be s i g n i f i c a n t l y related to reasoning scores, hence a l l the i n d i v i d u a l difference variance was not extracted by t h i s method. Sternberg (1981) commented that the complexity of the task probably introduced greater complexity into the constant so that the solu t i o n of one problem introduced others. Sternberg's Triarch Theory of Intelligence Sternberg (1985) expanded h i s theory of i n t e l l i g e n c e into a t r i a r c h theory of componential, e x p e r i e n t i a l , and contextual subtheories. The componential subtheory which rel a t e s i n t e l l i g e n c e to the i n t e r n a l world of the i n d i v i d u a l , s p e c i f i e s three basic kinds of components: executive, performance, and knowledge-a c q u i s i t i o n components. K o l l i g i a n and Sternberg (1987) defined metacomponents as "higher order executive processes that are used to plan, monitor, and evaluate one's task performance" (p.9). Strategy s e l e c t i o n was one of the main metacomponents (Sternberg, 1985, 1986). Strategies were referred to as organized c o l l e c t i o n s of component processes .... that can be manipulated to enhance performance on a s p e c i f i c task or set of related tasks" ( K o l l i g i a n & Sternberg, 1987, p. 11). Performance components carry out the orders and d i r e c t i o n s of the executive components i n solving a problem or completing a task. Knowledge-acquisition components are used i n learning new 74 information. They s e l e c t i v e l y encode relevant information, s e l e c t i v e l y assimilate and organize i t to form a new cognitive structure, and s e l e c t i v e l y compare the new with old cognitive structures. The ex p e r i e n t i a l subtheory concerns the application of components within the framework of the ind i v i d u a l ' s experience of the world, the extent to which information processing has become automatic, and the way i n which new s t i m u l i and information are coped with. The contextual subtheory, which rel a t e s i n t e l l i g e n c e to the external world of the i n d i v i d u a l , concerns the way i n which components are applied to experiences i n adapting environments, shaping environments, and s e l e c t i n g new environments. The t r i a r c h theory of i n t e l l i g e n c e encompasses giftedness (Sternberg & Davidson, 1983), mental retardation (Sternberg & Spear, 1985), and learning d i s a b i l i t i e s ( K o l l i g i a n & Sternberg, 1987). With emphasis on the componential subtheory, K o l l i g i a n and Sternberg (1987) outlined each subtheory and explained the l i n k between the subtheory and learning d i s a b i l i t i e s . The main points of K o l l i g i a n and Sternberg's theorizing are presented as follows. The idea of a componential-deficit was used to explain why learning-disabled children experience learning d i f f i c u l t i e s i n areas such as reading, s p e l l i n g , and mathematics. The knowledge a c q u i s i t i o n components i n a s p e c i f i c domain were theorized to be the main suspects of a componential d e f i c i t . These were linked to a poor knowledge base. Sternberg and Suben (1986) had suggested the 75 p o s s i b i l i t y of a feedback loop i n which a r i c h knowledge base could lead to e f f i c i e n t use of knowledge-acquisition components which i n turn could lead to a more enriched knowledge base. K o l l i g i a n and Sternberg suggested that an inadequate knowledge base and i n e f f i c i e n t component processing leads to a more inadequate knowledge base. I n e f f i c i e n t cognitive strategies that stemmed from i n f l e x i b i l i t y i n cognitive s t y l e were also thought to contribute to poor knowledge a c q u i s i t i o n . Learning disabled children are, by d e f i n i t i o n , of normal i n t e l l i g e n c e . For t h i s reason, K o l l i g i a n and Sternberg were of the opinion that the d e f i c i t s could not occur i n the executive or metacomponents as these are global i n nature and would a f f e c t o v e r a l l i n t e l l e c t u a l functioning. However, they suggested an i n d i r e c t l i n k between metacomponential d e f i c i t s and learning d i s a b i l i t i e s through working memory. The implication of short-term memory i n reading d i s a b i l i t i e s has been discussed i n the f i r s t part of t h i s chapter. Working memory has been i d e n t i f i e d as an intermediate storage location for metacomponential information (Swanson, 1982) Therefore, i n e f f i c i e n t storage may lead to d e f i c i e n t componential processing, so that components are poorly executed or some information i s neglected. "In t h i s sense, metacomponents allow the lower order components to continue t h e i r s p e c i f i c d e f i c i e n t operations" ( K o l l i g i a n & Sternberg, 1987, p.10). The r e l a t i o n s h i p between the expe r i e n t i a l subtheory and learning d i s a b i l i t i e s was explained i n terms of automaticity and 76 motivation. I t was thought that lack of automaticity i n processing information through reduced range of experience would not allow for e f f i c i e n c y i n dealing with a novel s i t u a t i o n . This corresponds to the LaBerge and Samuel (1974) theory of automatic information processing i n reading. Motivation d i f f i c u l t i e s and aversion to novel s t i m u l i i n a s p e c i f i c domain were linked to 1) i n e f f i c i e n t functioning of lower order components, 2) d e f i c i e n t cognitive strategies, 3) an inadequate knowledge base, and 4) s p e c i f i c automatization f a i l u r e . The contextual subtheory was not regarded as a primary determinant of learning d i s a b i l i t i e s . However, because i t i s within the school environment that learning d i s a b i l i t i e s are defined, i d e n t i f i e d , and evaluated, the severity of the disorder i s determined by the individual's a b i l i t y to adapt to, shape, and se l e c t aspects of t h e i r environment. K o l l i g i a n and Sternberg suggested that reading-disabled children who are less able to shape t h e i r environment may be more susceptible to having t h e i r bottom-up d e f i c i t s develop into l a t e r top-down d e f i c i t s . I t should be noted that many of K o l l i g i a n and Sternberg's t h e o r e t i c a l explanations have not been tested empirically. However, the authors suggested several areas i n which empirical i n v e s t i g a t i o n i s required. 77 Summary of Componential Analysis: Theory and Research Componential analysis combines the strengths of the d i f f e r e n t i a l and cognitive approaches to the study of i n t e l l i g e n t behaviour. I t has been used extensively with various forms of reasoning tasks; i n p a r t i c u l a r , analogical reasoning. Sternberg (1977) theorized a group of component processes that could be used i n analogical reasoning, an additive rule for t h e i r combination, and a set of models to represent t h e i r mode of execution. He was able to support h i s theory empirically (Sternberg, 1977; 1978). To investigate the development of analogical reasoning i n children, Sternberg and R i f k i n (1979) used a series of picture analogies with separable a t t r i b u t e s , presented v i a systematically varied booklets. They were able to show that, for t h i s type of analogy, childr e n used the same self-terminating model as adults and t h i s use became more consistent with age. Wilson (1980) used the same separable-attribute analogies with childr e n of d i f f e r e n t a b i l i t y l e v e l s . Her study indicated that children with low a b i l i t y marginally preferred an exhaustive mode of processing. Children of average and above average a b i l i t y used the same self-terminating model as the children i n the Sternberg and R i f k i n (197 9) study and consistency of use increased with l e v e l of a b i l i t y . The term "schematic" was applied to these analogies because the separable nature of t h e i r a t t r i b u t e s made soluti o n possible using several strategies. However, indivi d u a l s usually indicated 78 a spontaneous preference for a self-terminating strategy when solving these analogies. Sternberg and Ketron (1982) were able to show that t h i s preference was so strong that i t was not possible to t r a i n u n i v e r s i t y students to solve them using any other strategy. Moreover, indivi d u a l s who had been trained to use another strategy, thought they were using that strategy when i n fact they were using the self-terminating one. Sternberg (1985) expanded his componential theory into a three l e v e l "Triarch Theory of Intelligence". K o l l i g i a n and Sternberg (1987) used t h i s theory to explain the discrepancy between a b i l i t y and academic performance exhibited by learning-disabled children. Such childre n were theorized to have a componential-deficit mainly i n s p e c i f i c domains of knowledge a c q u i s i t i o n that are accompanied by problems of automaticity i n information processing and motivational d i f f i c u l t i e s . The severity of a learning d i s a b i l i t y was interpreted i n terms of the contextual s e t t i n g (school environment) i n which the student i s placed. L i t t l e empirical evidence was given to support t h i s theory although areas needing research were i d e n t i f i e d by the authors. 79 CHAPTER I I I : PURPOSE OF THE STUDY In the previous chapter several unresolved problems were revealed. The f i r s t problem concerns the i d e n t i f i c a t i o n of reading-disabled children i n terms of the severity of the disorder, l e v e l of i n t e l l i g e n c e , and the other exclusionary c r i t e r i a contained i n the World Federation of Neurology d e f i n i t i o n of dyslexia (Waites, 1968). Although several researchers consider that the population of poor readers i s made up of a large group of garden-variety poor readers (Stanovich, 1988) and a smaller group of more severely disabled readers, there i s evidence to suggest that the severity of reading d i s a b i l i t i e s forms a continuum and i s not made up of d i s t i n c t subpopulations;and that where children have a s i m i l a r l e v e l of reading d i s a b i l i t y , those who f i t the World Council d e f i n i t i o n of dyslexia cannot be distinguished by the symptoms they exhibit from those who do not. The majority of studies have examined samples of disabled readers chosen from c l i n i c a l or school-diagnosed sample pools. These sample pools are s i m i l a r i n that both have been diagnosed by psychoeducational assessment. The sample used i n t h i s study did not come from a c l i n i c a l population; rather i t was selected from a school based population of students. This was an attempt to avoid l i m i t i n g the sample to what Benton (1978) theorized might be an a t y p i c a l one with peculiar problems of i t s own. In h i s words, "... i t may be desirable, given our present state of ignorance, to cast 80 a wide net and study children with a wide v a r i e t y of d e f i c i t s including those whose c l i n i c a l pictures do not meet s t r i c t c r i t e r i a of s p e c i f i c reading d i s a b i l i t y " (Benton, 1978, p. 460). Therefore, no attempt was made to exclude children who might be described as c u l t u r a l l y deprived and the i n t e l l e c t u a l c r i t e r i a were set somewhat lower than that advocated by some researchers (e.g., V e l l u t i n o , Spear & Sternberg, 1987). The second problem concerns subtyping and the inconsistencies and confusion created by attempts to compare reading-disabled subtypes i d e n t i f i e d by d i f f e r e n t methods ( E l l i s , 1985). The production of subtypes appears to be more dependent upon the variables that produce them than they are on the nature of the sample or the c l u s t e r i n g method. Clearly subtyping research needs a new methodology to bring some order to the chaos that e x i s t s now. What i s needed i s a methodology that groups reading-disabled students according to a higher l e v e l of mental processing which i s less dependent upon the variables used and more dependent upon the mental c h a r a c t e r i s t i c s of each i n d i v i d u a l i n the sample. Up to the present, t h i s has not been achieved because studies have focused on i n t e r i n d i v i d u a l rather than i n t r a i n d i v i d u a l differences. The purpose of t h i s study was to address these problems by combining an information-processing approach with a d i f f e r e n t i a l methodology. This was to be achieved through two methods of subtyping using reading and related measures that are already f a m i l i a r to school personnel. But the study was designed to go 81 further. I t was designed to examine, through componential analysis, the way i n which reading-disabled and normal readers solved an analogical reasoning task. The advantages associated with componential analysis are that i t allows the t h e o r i s t ; 1) to i d e n t i f y the mental processes and strategies used i n the task, 2) to c a l c u l a t e component scores and indicate the amount of time spent executing each component, and 3) to explore the r e l a t i o n s h i p between the solution scores and external variables through c o r r e l a t i o n a l analysis. Componential analysis can be c a r r i e d out at the i n d i v i d u a l l e v e l as r e s u l t s do not depend upon i n d i v i d u a l differences, the data from many subjects, or the model chosen by other i n d i v i d u a l s . Sternberg (1977) theorized a r e l a t i v e l y small number of component processes for analogical reasoning and combined them into four procedural and three s t r a t e g i c models. The preferred model, which indicates component and strategy use, i s i d e n t i f i e d by inspecting each i n d i v i d u a l ' s regression equations for goodness of f i t . Component scores, which indicate i n d i v i d u a l strengths and weaknesses, are calculated from the regression equation and can be correlated with external variables to e s t a b l i s h relationships between components and other tasks. I t i s important i n s e l e c t i n g an instrument to be used i n componential analysis with a s p e c i f i c group of children that i t has been used empirically for that purpose, consists of tasks that reading-disabled children can perform, and has theorized cognitive 82 processes and strategies that have some re l a t i o n s h i p to reading. Two kinds of picture analogies were used by Sternberg & R i f k i n (1979) to determine the development of analogical reasoning i n children. E i ther would have been useful with reading-disabled childr e n because reading was not required. However, the Schematic Picture Analogies were chosen because the t h e o r e t i c a l models associated with them were s t r a t e g i c and they had been used with children at various developmental (Sternberg & R i f k i n , 1979) and a b i l i t y l e v e l s (Wilson, 1980), allowing conclusions to be made when applying the task to other populations. In a discussion of human reasoning, Sternberg (1986) defined a task as a reasoning task i f s e l e c t i v e encoding, s e l e c t i v e comparison, or se l e c t i v e combination were involved. Inductive reasoning such as analogical reasoning which i d e n t i f i e s a general solu t i o n from a s p e c i f i c example, r e l i e s heavily on se l e c t i v e encoding and s e l e c t i v e inference. Deductive reasoning such as categorical syllogism which i d e n t i f i e s a s p e c i f i c solution from a general example r e l i e s on se l e c t i v e combination. Reading r e l i e s on v i s u a l perceptions of form and d e t a i l which are translated into i n t e r n a l representations and held i n working memory. The top-down or concept driven theories of Goodman (19 68) and Smith (1971) view reading as a problem solving task i n which the reader processes only as much v i s u a l information as necessary to obtain meaning. This implies that s e l e c t i v e encoding can be 83 equated with the term redundancy which was coined by Smith (1971). The information obtained through the process of redundancy i s s e l e c t i v e l y compared with information already i n the system and then s e l e c t i v e l y combined or assimilated to produce new information. The degree of d i f f i c u l t y of a reasoning task i s determined by a serie s of mediators described by Sternberg (198 6) as "any intervening v a r i a b l e that increases or decreases the a v a i l a b i l i t y or a c c e s s i b i l i t y of the i n f e r e n t i a l rules for use i n a p a r t i c u l a r problem" (p. 290). Mediators were theorized to be such intervening variables as p r i o r knowledge of the type of problem and how i t i s solved, working memory capacity, and the ind i v i d u a l ' s a b i l i t y to represent material i n a s p e c i f i c domain (e.g., l i n g u i s t i c , or s p a t i a l ) . Mediators used i n the reading process include f a m i l i a r i t y with the content and format of the text and knowledge of grammatical and syntactic rules. These enable the readers to extract information accurately. The influence of working memory capacity and l i n g u i s t i c a b i l i t y have been implicated i n reading e f f i c i e n c y as have sel e c t i o n and use of appropriate metacognitive strategies. Although schematic picture analogies have been used to i d e n t i f y the cognitive processes and strategies used by high SES elementary school children at d i f f e r e n t age l e v e l s (Sternberg & R i f k i n , 1979) and elementary school children at d i f f e r e n t a b i l i t y l e v e l s Wilson (1980), no research to date has involved disabled 84 readers i n such a task. The contribution of t h i s study was to explore the existence of reading d i s a b i l i t y subtypes i n a n o n c l i n i c a l sample and compare them with subtypes i d e n t i f i e d by previous research. The use of componential analysis added a further dimension to the knowledge of reading d i s a b i l i t i e s through the i d e n t i f i c a t i o n of cognitive processes and strategies used i n a problem solving task and the formation of groups based on process and strategy use. In addition, the re l a t i o n s h i p between reading and analogical reasoning was established by comparing membership i n a reading d i s a b i l i t y subtype with membership i n an analogical reasoning subgroup and by c o r r e l a t i n g solution scores with the reading variables for each analogical reasoning subgroup. This study was unique i n that i t allowed i n t r a i n d i v i d u a l as well as i n t e r i n d i v i d u a l analysis of the s k i l l s of reading-disabled children. This study i s e s s e n t i a l l y an exploratory one as f a r as the disabled readers are concerned because the sample and variables used are d i f f e r e n t from those most often used i n subtyping research. In addition, componential analysis of analogical reasoning data has not been used with a reading disabled sample. However, componential analysis has been used with students of average achievement l e v e l s making i t possible to hypothesize the components and strategies used by the normal reader group i n solving the schematic picture analogies. 85 Research Questions and Hypotheses Question 1 W i l l the Boder Test of Reading-Spelling Patterns reveal the same subtypes as those reported by Boder? The school-based sample used i n t h i s study may not resemble the c l i n i c a l samples used by Boder (1973) and reported i n Boder and J a r r i c o (1982). The broader range of t h i s reading-disabled sample suggests that s p e c i f i c and nonspecific reading d i s a b i l i t y subtypes w i l l be found. I t i s not expected that the undifferentiated subtype w i l l emerge as the grade 5 students used i n t h i s study are approximately ten years of age and the undifferentiated subtype i s associated with older students. Question 2 W i l l i t be possible to d i s t i n g u i s h between the normal reader sample and the reading-disabled sample using a s t a t i s t i c a l c l u s t e r i n g technique with reading and related variables and w i l l the disabled readers form subtypes? Clustering techniques have been used successfully to d i s t i n g u i s h between normal and disabled readers (Doehring et a l . , 1981; Petruskas & Rourke, 1979). 86 Question 3 W i l l there be any correspondence between subtypes obtained using the s t a t i s t i c a l c l u s t e r i n g method and those obtained using the Boder Test of Reading-Spelling Patterns? Although Boder i d e n t i f i e d her subtypes using one measure of reading and two s p e l l i n g measures, v a l i d a t i o n studies suggested further subtype c h a r a c t e r i s t i c s i n terms of reading and related v a r i a b l e s . Question 4 W i l l a l l subjects i n the study use the f i v e components (encoding, inference, mapping, application and response) to solve the schematic picture analogies or w i l l the reading-disabled subjects d i f f e r from the normal readers i n t h e i r choice of components? Previous studies (Sternberg & R i f k i n , 1979; Wilson, 1980) have shown that only encoding, inference, a p p l i c a t i o n and response components were used i n solving these analogies. Mapping, as an optional component, was not used. Hypothesis 4.1 Normal readers w i l l use encoding, inference, application, and response components i n the solution of schematic picture analogies. 87 Question 5 W i l l the reading-disabled students and the normal readers d i f f e r i n the extent to which the l i n e a r combination rule accounts for t h e i r performance i n schematic picture analogy solutions? Evidence from a previous study with u n i v e r s i t y students and high SES elementary school children (Sternberg & R i f k i n , 1979) showed that a l l subjects used the additive combination r u l e and processed the components i n an an a l y t i c fashion. The evidence from Wilson's study (1980) supported use of the l i n e a r combination rule only by the high and average a b i l i t y groups. Hypothesis 5.1 Normal readers w i l l use the l i n e a r combination rule i n the solut i o n of the picture analogies and process i n an an a l y t i c fashion. Question 6 W i l l the reading-disabled students and the normal readers d i f f e r i n the s t r a t e g i c model that they use? Sternberg (1977) and Sternberg and R i f k i n (1980) found that for picture analogies with separable a t t r i b u t e s , students process i n self-terminating mode. 88 Wilson(1980) found that the average and high a b i l i t y groups preferred the self-terminating mode of execution but the low a b i l i t y group marginally preferred am exhaustive mode. Hypothesis 6.1 Normal readers w i l l show preference for a self-terminating mode in solving schematic picture analogies. Question 7 W i l l i t be possible to i d e n t i f y several subgroups of disabled readers using componential analysis? Evidence has established that reading-disabled students do not form a homogeneous group. Therefore i t seems reasonable to suppose that not a l l members of t h i s group w i l l process i n the same fashion or use the l i n e a r combination rule consistently. In addition, at the component l e v e l disabled readers may process more slowly and make more errors. Question 8 W i l l a r e l a t i o n s h i p e x i s t between membership i n one of the reading d i s a b i l i t y subtypes and the model preference indicated for solving the schematic picture analogies? 89 Question 9 W i l l the solution scores of a group of normal readers and subgroups of disabled readers correlate s i g n i f i c a n t l y with the reading and related variables? A solution score indicates the average time taken by a student to solve a single analogy. Sternberg's (1977) research indicated s i g n i f i c a n t correlations between solution scores and a b i l i t y v ariables. Therefore i t i s possible to hypothesize s i g n i f i c a n t c o rrelations for the group of normal readers. The research with disabled readers i s exploratory and no hypothesis can be made for t h i s group. Hypothesis 9.1 The solution scores of the group of normal readers w i l l have s i g n i f i c a n t (p<0.01) p o s i t i v e correlations with reading and error scores and s i g n i f i c a n t negative correlations with a b i l i t y and accuracy scores. 90 CHAPTER IV: METHODOLOGY This study was divided into two phases. In Phase 1, two screening t e s t s were administered to a target population i n order to i d e n t i f y and select samples of disabled and normal readers. In Phase 2 of the study, the students i n each sample were i n d i v i d u a l l y administered reading and reading-related tests and a picture analogy t e s t was given to students i n small groups. This chapter describes the two phases of the study, d e t a i l i n g the instruments and procedures used i n each. The chapter concludes with a description of the preparation of data for analysis. Phase 1 Target Population The target population comprised students enroled i n grade 5. These students attended public elementary schools i n an urban area of Northwestern Ontario. For the purpose of t h i s study, disabled readers were defined as those students of normal i n t e l l i g e n c e who scored at or below the 2 0th percentile on the Comprehension Subtest of the Stanford Diagnostic Reading Test (SDRT), Brown Level, Form A. Percentile ranks were obtained from the d i s t r i b u t i o n of the scores of these grade 5 students. Normal readers were defined as students of normal i n t e l l i g e n c e who scored at or above the 3 5th 91 p e r c e n t i l e on t h i s t e s t . The 35th percentile was chosen to ensure that normal readers were c l e a r l y distinguished from disabled readers. Normal i n t e l l i g e n c e was defined by a Universal Scaled Score i n the range 80 to 12 0 obtained on the Nonverbal Battery of the Canadian Cognitive A b i l i t i e s Test (CCAT), Form 1, Level C. The CCAT Universal Scaled Score, obtained when the t e s t was standardized, has a mean of 100 and a standard deviation of 16. Instruments Used Stanford Diagnostic Reading Test (SDRT) The Stanford Diagnostic Reading Test (Karlsen, Madden, & Gardner, 1976) i s a group-administered multiple-choice t e s t which measures s p e c i f i c reading s k i l l s . The Brown Level, Form A was used i n t h i s study as i t i s intended for grades 5 through 8 and low achievers i n the higher grades. This t e s t was selected because i t i s designed to sample s k i l l s i n four reading components: Vocabulary, Comprehension, Decoding, and Rate. The Vocabulary component consists of one subtest, Auditory Vocabulary. This measures the a b i l i t y to match a sentence to one of three words written i n the t e s t booklet, each of which i s pronounced o r a l l y by the examiner. There are fo r t y items i n t h i s subtest. The Comprehension component consists of one subtest that 92 includes two measures of reading comprehension. One corresponds to factual comprehension and the other to i n f e r e n t i a l comprehension. The t e s t employs the paragraph format i n which the student i s required to read a paragraph and then answer written questions. Two subtests make up the Decoding component. The f i r s t , Phonetic Analysis, measures the a b i l i t y to make sound-symbol associations. One group or group of l e t t e r s i s underlined i n a word and the student must select, from a choice of three, the printed word that contains the same sound. Consonant sounds are distinguished from vowel sounds v i a the scoring key and t h e i r scores recorded separately. The second Decoding subtest, Structural Analysis, measures the a b i l i t y to use s y l l a b l e s , prefixes, root words and blends. I t has two parts. In part one, the student indicates where t h r e e - s y l l a b l e words should be divided into s y l l a b l e s . In part two, the items consist of four s y l l a b l e s , three of which can be blended to make a word. The student i s required to indicate which s y l l a b l e does not f i t the word. The l a s t s k i l l area, Reading Rate, consists of one subtest which measures the a b i l i t y to read material of a lower grade l e v e l , quickly and accurately. This subtest requires the student to read a sentence and choose a missing word or group of words from a choice of three. Students are t o l d before they begin that the purpose of the t e s t i s to see how fast and accurately they can read. The t e s t y i e l d s two sets of scores; one indicates the t o t a l 93 number of responses made and the other indicates the number of correct responses. Raw scores are the number of correct responses, or, i n the Reading Rate subtest, the number of items completed. A frequency d i s t r i b u t i o n was produced for the combined Factual and I n f e r e n t i a l scores of the Reading Comprehension subtest. The raw scores were converted to percentile rank and used to s e l e c t samples of disabled and normal readers. In the manual, in t e r n a l consistency was reported at above .9 for a l l subtests at grade l e v e l s 5 to 9, except for Auditory Vocabulary at grades 5 and 6. Internal consistency for t h i s subtest was reported to be .84 for grade 5 and .88 for grade 6. C r i t e r i o n v a l i d i t y was established by c o r r e l a t i n g the subtests with those of the Stanford Achievement Test and ranged from .39 to .98. Canadian Cognitive A b i l i t y Test (CCAT) The Canadian Cognitive A b i l i t y Test (Thorndike, Hagen, Lorge, & Wright, 1974) i s a group-administered multiple-choice t e s t and one of the more popular tests of i t s kind i n use i n Canada (Randhawa, Hunt, & Rawlyk, 1974). Although i t was standardized i n Canada, the content, except for items that contain imperial units, i s the same as the Cognitive A b i l i t i e s Test (CAT) used i n schools i n the United States. These items have been replaced with "neutral" 94 rather than metric items because Canadian school c h i l d r e n were not a l l f a m i l i a r with metric units when the t e s t was standardized. The m u l t i l e v e l e d i t i o n of the CCAT, Form 1, consists of Verbal, Quantitative, and Nonverbal Batteries designed for use with grades 3 to 12. A series of graded items are divided into eight separate but overlapping l e v e l s contained within a sing l e reusable booklet. Level C was selected as the most suitable f o r use with Grade 5 students. Each item i s multiple-choice with answers recorded on a separate hand-scorable sheet. The Nonverbal Battery of the CCAT does not contain any written verbal material, making i t s u itable for use with disabled readers. I t was chosen for t h i s reason and as well as for the fact that i t was normed on a Canadian population. The Nonverbal Battery has three subtests: 1) Figure C l a s s i f i c a t i o n , 2) Figure Analogies, and 3) Figure Synthesis. In these t e s t s the student i s required to: 1) f i n d a drawing that f i t s a class of drawings, 2) solve figure analogies i n the form A i s to B as C i s to D (A:B::C:D), and 3) s e l e c t figures that can be made from a group of given parts. The raw score i s the number of correct responses. Tables are provided for converting raw scores to standard scores, percentile rank, and stanines based upon chronological age. The score of i n t e r e s t i n t h i s study was the Universal Scaled Score which has a mean of 100, a standard deviation of 16, and i s standardized by age. The CCAT was normed on approximately 27,000 English-speaking 95 Canadian school children i n grades 3 to 9. As the obtained sample did not match the d i s t r i b u t i o n of students throughout the provinces and t e r r i t o r i e s , sample responses were weighted to remove in e q u i t i e s . Construct v a l i d i t y was inferred from factor analysis of data c o l l e c t e d from 300 students who had also taken the Canadian Lorge-Thorndike Intelligence Test (CLTIT), an e a r l i e r version of the CCAT. Three d i s t i n c t factors were reported. Randhawa et a l . (1974) pointed out that the factors were unrotated and therefore d i f f i c u l t to i n t e r p r e t . They undertook to rotate them using Varimax. The re s u l t s indicate a general factor and a nonverbal factor common to both t e s t s . Further investigation of the construct v a l i d i t y using the WISC (Randhawa et a l . , 1974) revealed four factors. Three of the factors loaded on both tests but a fourth factor loaded on the CCAT alone. The factors common to both tests were i d e n t i f i e d as Verbal, Nonverbal, and Reasoning factors. The emergence of a common nonverbal factor gives support to the use of the CCAT Nonverbal Battery rather than the WISC as t h i s would prove impractical with the large numbers of subjects tested i n the present study. Procedure Administration of the SDRT was conducted as part of a standardization program to obtain l o c a l norms for the t e s t . Schools selected to take part i n the study had an enrolment of at lea s t 15 96 grade 5 students or, where French Immersion was an option, at least 15 students receiving i n s t r u c t i o n i n English. Thirty-two schools met t h i s c r i t e r i a . They had a t o t a l of 966 grade 5 students i n a proportion representative of urban and r u r a l sections of the population. The students came from a wide socio-economic range and a v a r i e t y of ethnic backgrounds. In accordance with School Board po l i c y , l e t t e r s requesting permission for students to take part i n the screening phase of the study were sent by in d i v i d u a l p r i n c i p a l s to parent or guardians. Permission was obtained for 801 students to take part. Forty-five of these students did not complete the SDRT reading comprehension subtest, reducing the sample pool to 757. The Stanford Diagnostic Reading Test (SDRT) and the Nonverbal Battery of the Canadian Cognitive A b i l i t i e s Test (CCAT) were administered i n three sessions of approximately an hour and a half each. Testing was arranged to s u i t each class timetable and where possible was conducted on three successive school days or within one school week. The SDRT Subtests 1 and 2 were administered i n the f i r s t session, followed by Subtests 3, 4, and 5 i n the second, and the CCAT Nonverbal Battery i n the t h i r d . Adminstration of each subtest was followed by a f i v e to ten minute break. The t e s t s were administered by the p r i n c i p a l researcher assisted by a team of graduate students who were enroled i n the Master's program i n C l i n i c a l Psychology at a l o c a l u n i v ersity. Testing usually took place i n the student's own grade 5 classroom 97 with one proctor supervising up to f i f t e e n students. Classroom teachers were not involved i n the t e s t administration or required to be present during t e s t i n g although some elected to remain i n the classroom and worked qui e t l y i n the background. Where possible, large groups of students were divided into smaller groups to reduce the l i k e l i h o o d of behaviour problems. A few behaviour problems occurred during breaks between subtests but did not disrupt the administration of the t e s t s . Data Preparation The p r i n c i p a l researcher and the graduate students hand-scored the t e s t s they administered. An additional person who was unconnected with the administration of the t e s t s checked a l l protocols for scoring accuracy. A l l CCAT and SDRT Reading Comprehension protocols were rescored by the p r i n c i p a l researcher. The data were analyzed using S t a t i s t i c a l Package for the So c i a l Sciences - Version X (SPSS ) subprogram Frequencies (SPSS User's Guide, 1986). Separate frequency d i s t r i b u t i o n s and summary were obtained for each SDRT subtest but only the d i s t r i b u t i o n for the Reading Comprehension subtest was used i n t h i s study. S t a t i s t i c s are presented i n Table 6. The SDRT reading comprehension subtest contained equal numbers of factual and i n f e r e n t i a l items. S p l i t - h a l f r e l i a b i l i t y of the subtest, corrected for length using the Spearman-Brown formula, was .96. 98 Table 6. Sex d i s t r i b u t i o n , age, nonverbal a b i l i t y , and reading  comprehension for grade 5 students Variable # of missing cases Sex d i s t r i b u t i o n Males 395 0 Females 406 0 Total 801 0 Age (in months) Mean 128.4 0 SD 7.6 Range 109.0 - 158.0 CCAT (Universal Score) Mean 98.1 50 SD 16.9 Range 50.0 - 13 8.0 Reading Comprehension (Raw Score) Mean 37.8 45 SD 13.5 Range 3.0 - 60.0 99 Phase 2 Selection of Samples One hundred and f i f t y - o n e students had a Reading Comprehension Score at or below the twentieth percentile and met the reading c r i t e r i o n for membership i n Group D but f i f t y - e i g h t had CCAT scores below 80. The remaining ninety-three students who met both these c r i t e r i a for membership i n Group D were a l l asked to take part i n Phase 2. Random number tables were used to s e l e c t twenty-five students from those who met the c r i t e r i a for membership i n Group N. Parental permission was requested by the p r i n c i p a l researcher for student p a r t i c i p a t i o n i n Phase 2 of the study (Appendix A). Teachers, without being informed as to which group the students belonged, were asked to complete a c h e c k l i s t (Appendix B) on behalf of t h e i r students. The c h e c k l i s t was designed to i d e n t i f y students who: 1) had been educated i n a language other than English since Grade 1, 2) had a hearing problem, 3) had an uncorrected v i s u a l problem, 4) were infrequent attenders, or 5) had behavioural problems. The f i r s t four questions were designed to exclude students for whom the answer was "Yes". The f i f t h question was included to provide confirmatory information should a student's behaviour prove so disruptive at some point i n the study that t e s t i n g had to be 100 discontinued. This information was to be withheld from the examiners to avoid prejudicing the student/examiner rapport. However, t h i s was a precaution that proved unnecessary as none of the students were i d e n t i f i e d as po t e n t i a l behaviour problems. Teachers were also asked to give t h e i r opinion of a student's reading l e v e l i n comparison with the peer group. Their opinions were based upon observation of the student and professional j udgement. Parental permission to take further part i n the study was not given for 13 pot e n t i a l members of the reading-disabled group and three students were i d e n t i f i e d by the teacher c h e c k l i s t as i n e l i g i b l e to take part. Parental permission was refused for three p o t e n t i a l members of the normal group and t e s t i n g was not completed for two others before the end of the school year. The t o t a l sample, therefore, consisted of 97 Grade 5 students, of whom 77 were disabled readers and 20 were normal readers. The reading-disabled group (D) contained 41 males and 3 6 females. The normal readers group (N) contained 7 males and 13 females. Table 7 contains d e s c r i p t i v e s t a t i s t i c s of Group D and Group N for sex d i s t r i b u t i o n , age, nonverbal a b i l i t y and SDRT Reading Comprehension score. Teachers i d e n t i f i e d 62 out of 77 students, or 81 percent of group D as below-average readers compared to t h e i r peers; the others were regarded as average. This implied that some teachers had a lower c r i t e r i o n for i d e n t i f y i n g below average readers than the d e f i n i t i o n used i n t h i s study. They also i d e n t i f i e d 19 out of 2 0 i n Group N 101 Table 7. Sex d i s t r i b u t i o n , age, nonverbal a b i l i t y , and  reading comprehension for Group D and Group N Variable Group D Group N Sex d i s t r i b u t i o n Males 41 7 Females 36 13 Total 77 20 Age (in Months) Mean 129.0 126.6 SD 7.6 6.7 Range 116 - 156 111 - 140 CCAT (Universal Score) Mean 92.0 104.3 SD 10.3 11.7 Range 80 - 117 80 - 119 Reading Comprehension (raw score) Mean 18.4 43.4 SD 3.8 8.7 Range 10 - 23 32 - 58 102 as average readers; the remaining student was i d e n t i f i e d as below average. Instruments Used  Peabody Picture Vocabulary Test - Revised The Peabody Picture Vocabulary Test - Revised (Dunn & Dunn, 1981), designed to measure a subject's receptive vocabulary, comes i n two forms, L and M. In t h i s study Form L was used. I t i s an i n d i v i d u a l l y administered t e s t that can be given to persons from two-and-a-half to fo r t y years of age. The t e s t consists of a series of plates bound into an easel k i t for ease of presentation. Each plate i s divided into four quarters with a picture displayed i n each quarter. The examiner says a word and the subject i d e n t i f i e s the matching picture. Different s t a r t i n g points are indicated appropriate to the subject's age. A basal of eight correct consecutive responses i s established and the t e s t i s discontinued when the subject makes s i x errors i n eight consecutive responses. The raw score i s the t o t a l number of correct responses plus a l l the items below the basal l e v e l which are assumed to be correct. Tables are provided for converting raw scores to standard score equivalents, percentile ranks and stanines, based on age i n t e r v a l s of s i x months. Standard score equivalents with a mean of 100 and standard deviation of 15 were used i n t h i s study. 103 The t e s t was standardized i n the U.S. on a representative sample of young persons aged two-and-a h a l f to eighteen years. The Rasch-Wright standardization procedure (Wright, 1977) was adopted, requiring the administration of only one form of the t e s t per subject. Forms L and M were given a l t e r n a t e l y . Three kinds of r e l i a b i l i t y data were reported; s p l i t - h a l f r e l i a b i l i t y , immediate t e s t - r e t e s t , and delayed t e s t - r e t e s t . R e l i a b i l i t y c o e f f i c i e n t s for Form L at ages ten to twelve years ranged from .77 to .86. The authors report construct and c r i t e r i o n v a l i d i t y of the unrevised PPVT to be high and by implication also the PPVT-R. No construct or c r i t e r i o n - r e l a t e d v a l i d i t y data are reported i n the manual for the revised t e s t . The Boder Test of Reading-Spelling Patterns The Boder Test of Reading-Spelling Patterns (Boder & J a r r i c o , 1982) i s an i n d i v i d u a l l y administered t e s t designed to d i s t i n g u i s h disabled readers from normal readers and to c l a s s i f y them into subgroups according to t h e i r reading-spelling patterns. The t e s t consists of graded word l i s t s that have equal numbers of phonetic and nonphonetic words. The word l i s t s are used i n both the reading and s p e l l i n g t e s t s . The examiner presents the student with the word l i s t s and notes whether the words are recognized within one second (flash) or within ten seconds (untimed) or not read at a l l (not read). The 104 l i s t f o r which the student reads 50 percent of the words on f l a s h presentation indicates the student's grade l e v e l . Instructions i n the manual advise that the t e s t should be continued through two more grade l e v e l s to give s u f f i c i e n t unknown phonetic and nonphonetic words for the s p e l l i n g t e s t . While the student works on paper and p e n c i l tasks suggested i n the manual, the examiner selects a l i s t of ten known and ten unknown words for the s p e l l i n g t e s t each containing f i v e phonetic and f i v e nonphonetic words. The examiner presents each known word i n i s o l a t i o n then s p e c i f i e s i t s meaning by using i t i n a sentence. Unknown words are presented i n i s o l a t i o n as the purpose i s to see i f the student can write a good phonetic equivalent of the word. In scoring the s p e l l i n g t e s t , the known words are scored for correctness only; whereas, the unknown words are scored for t h e i r phonetic equivalence to the dictated words, which include words that are c o r r e c t l y spelled. Guidelines are given i n the manual for i d e n t i f y i n g good phonetic equivalents (GFEs). The Boder t e s t uses a Reading Quotient (RQ) to d i s t i n g u i s h normal from disabled readers and to d i f f e r e n t i a t e between disabled-reader subtypes. The RQ i n i t s simplest form i s a r a t i o of reading age to chronological age. Reading age i s calculated by adding s i x years to the student's grade equivalent score. The reading-spelling patterns that d i s t i n g u i s h the various categories are summarized i n Table 8. Boder based her claim of high r e l i a b i l i t y and v a l i d i t y on the use of the t e s t with more than 3,000 children i n a v a r i e t y of 105 Table 8. The subtypes of Boder Reader Type Spel l i n g Pattern Known vs Unknown Words Typical Reading Quotient Normal >50% >50% >100 Dysphonetic* >50% <50% 70 - 85 Dyseidetic* <50% >50% 50 - 70 Mixed Group* <50% <50% <67 Nonspecific >50% >50% 80 - 90 Undetermined <50% >50% 80 - 90 * S p e c i f i c reading d i s a b i l i t y 106 c l i n i c a l settings. Using the t e s t protocols of 54 children, t e s t -r e t e s t r e l i a b i l i t y was established by gender, two age l e v e l s , and long-term ret e s t for four components of the t e s t . For reading l e v e l , a l l c o e f f i c i e n t s were between .96 and .99 except for long-term ret e s t which was .81. R e l i a b i l i t y for c o r r e c t l y spelled known words ranged from .56 to .84 and for GFEs from .72 to .89. The Chi-square s t a t i s t i c which indicated the r e l a t i o n s h i p between the subtype c l a s s i f i c a t i o n at f i r s t and second t e s t i n g was s i g n i f i c a n t at the p<.003 l e v e l . Internal consistency was measured using randomly selected t e s t protocols of 46 children. S p l i t - h a l f r e l i a b i l i t y was calculated using odd/even items where possible, otherwise the f i r s t h a l f of the items were compared with the l a s t h a l f . R e l i a b i l i t y c o e f f i c i e n t s , corrected for length were .99, .97, .82, and .92 for f l a s h presentation, reading l e v e l , unknown word l i s t and GFEs, respectively. However, the r e l i a b i l i t y c o e f f i c i e n t of .99 for f l a s h presentation i s open to question as i n t e r n a l consistency i s inappropriate for speeded tests and produces spuriously high r e s u l t s . To support her claim of v a l i d i t y for the t e s t and subtypes, Boder demonstrated that the subtypes were consistently i d e n t i f i e d i n roughly the same proportions by a number of independent studies. Support was also offered from a v a r i e t y of studies for the v a l i d i t y of the subtypes. The studies employed such tests as the WISC-R, the WRAT, tes t s of auditory and v i s u a l memory and other psychological 107 te s t s . The subtypes were distinguishable from each other and performed on these tests as expected from t h e i r c h a r a c t e r i s t i c s . For example i n a memory for faces t e s t the dyseidetic group i d e n t i f i e d s i g n i f i c a n t l y fewer faces than the dysphonetics and i n a d i g i t span t e s t the dysphonetics r e c a l l e d s i g n i f i c a n t l y fewer d i g i t s i n sequence than, the dyseidetics. Construct v a l i d i t y was established through correlations with the WRAT. Correlation c o e f f i c i e n t s ranged from .74 with a group of reading-disabled boys to .95 for a mixed group of normal and disabled readers. The scoring scheme for the t e s t i s based on the assumption that normal readers s p e l l words i n t h e i r sight vocabulary with an accuracy rate of 7 0 percent or more. Boder presented evidence that normal readers at d i f f e r e n t grade l e v e l s had scores consistent with t h i s assumption thus establishing c r i t e r i o n - r e l a t e d v a l i d i t y of normal s p e l l i n g patterns. D u r r e l l Analysis of Reading D i f f i c u l t y The D u r r e l l Analysis of Reading D i f f i c u l t y (Durrell & Catterson, 1980) i s a diagnostic reading t e s t that consists of a series of reading, s p e l l i n g , memory and phonics subtests. The tes t i s designed to be used from preprimer to grade 6 l e v e l . The tes t was standardized on students i n f i v e states i n the United States. R e l i a b i l i t y of the subtests was reported as ranging from .63 to .97. Construct v a l i d i t y was established by c o r r e l a t i n g the subtests 108 with the Metropolitan Reading Tests. Correlations ranged from .36 to .85. Oral and S i l e n t Reading subtests each consist of four timed paragraphs arranged i n ascending order of d i f f i c u l t y . In Oral Reading, when a student f a i l s to pronounce a word, the examiner allows f i v e seconds to elapse before supplying the word. This allows the reading to proceed at a reasonable pace and enables students to answer the comprehension questions without being penalized because they could not read some of the words. The nature of the S i l e n t Reading task makes timing less precise. Timing ends when students indicate that they have finished reading each paragraph. I t i s possible for students to skim quickly over the passage, to ponder unknown words at great length, or to reread parts of sentences. Thus the length of time taken i s determined by the student's s i l e n t reading strategy. Paragraphs to be read o r a l l y are accompanied by questions that measure l i t e r a l comprehension. The manual provides guidelines for scoring and gives examples of acceptable responses. There are no s p e c i f i c questions accompanying the paragraphs to be read s i l e n t l y but i n s t r u c t i o n s are provided for e l i c i t i n g unaided and aided r e c a l l of what was read. I t was anticipated that s i l e n t and oral reading paragraphs at the grade 5 l e v e l could prove too d i f f i c u l t f or many members of Group D and f a i l to discriminate among them. In contrast, paragraphs that were too easy could produce a low c e i l i n g e f f e c t 109 and f a i l to discriminate between readers at the upper end of Group D and those i n Group N. For t h i s reason, Oral and S i l e n t Reading paragraphs 2A (grade 3), 2B (grade 4), and 3 (grade 5) were used with both groups to provide a range that was neither too easy nor too d i f f i c u l t but would discriminate among the readers. The Listening Comprehension subtest consists of graded paragraphs to be read by the examiner to the student, followed by l i t e r a l comprehension questions. The manual cautions that older children may answer poorly on passages that are too easy; therefore, grade 4 and grade 5 passages were chosen for t h i s study. The intention was to provide paragraphs close to the students 1 enroled grade without making the task too d i f f i c u l t y Oral Reading yielded three d i f f e r e n t types of scores. The f i r s t was the t o t a l time (in seconds) that i t took to read the passages. The second indicated the t o t a l number of correct or p a r t i a l l y correct answers to comprehension questions. Correct answers were given a score of one point; p a r t i a l l y correct were given a score of h a l f a point. The t h i r d score was the t o t a l number of o r a l reading errors. These were scored somewhat d i f f e r e n t l y from ins t r u c t i o n s i n the manual. Four types of errors were i d e n t i f i e d : r e p e t i t i o n , omission, sub s t i t u t i o n , and addition. A f i f t h category was devised to record s e l f - c o r r e c t e d errors. Whole-word and part-word errors were recorded separately, giving a t o t a l of nine scores connected with 110 or a l reading errors. Each error and sel f - c o r r e c t e d error was given a score of one. S i l e n t Reading yielded two scores. Again, the f i r s t was the t o t a l time (in seconds) that i t took to read the passages. The second was the t o t a l number of r e c a l l e d facts i n the s t o r i e s . The examiner was permitted to aid students' r e c a l l through i n d i r e c t questioning but these responses were not scored. Listening Comprehension yielded one score for the two selected passages. Correct responses to comprehension questions scored one point and p a r t i a l l y correct responses scored h a l f a point. Schematic Picture Analogies This t e s t provides students with a task that does not require any reading. The analogies (Sternberg & R i f k i n , 1979) are of the form A i s to B as C i s to D (A:B::C:D). Each term depicts a figure that can vary i n four separable at t r i b u t e s (See Figure 2, p.61): hat colour (black, white), s u i t pattern (striped, dotted), handgear (briefcase, umbrella), and footwear (shoes, boots). To solve the analogy the l a s t term D i s chosen from one of two options: Di or Sixteen separate analogies are contained i n one booklet (Appendix C). The booklets are "homogeneous" which means that the number of a t t r i b u t e values that change from A to B, A to C, and Di to D2 remain constant within booklets although no two analogies are I l l i d e n t i c a l . This ensures that the amount of processing needed to solve an analogy remains the same throughout a booklet. The booklets are arranged so that the number of a t t r i b u t e differences between A and B, A and C, and Di and D2, change from booklet to booklet and no two booklets contain i d e n t i c a l combinations of differences. The time allowed to solve one booklet i s sixty-four seconds and each subject receives a t o t a l of twenty-four booklets. The analogies are randomly ordered within booklets and the booklets are randomly presented to avoid any order e f f e c t . Three scores are obtained for each subject on each booklet: the number of analogies solved c o r r e c t l y , the number of analogies solved i n c o r r e c t l y , and the t o t a l number of analogies completed. Each booklet can have a maximum of sixteen and a minimum of zero for a l l three scores. These scores are used to calc u l a t e mean latency and mean error rate for each booklet. Mean latency represents the average amount of time taken to c o r r e c t l y solve one analogy (mean latency correct) or the average amount of time taken to complete one analogy (mean latency t o t a l ) . Mean latency scores are calculated by taking the appropriate score for a subject on one of the booklets and d i v i d i n g i t into 64 - the number of seconds allowed to complete the booklet. Mean error rate i s obtained by di v i d i n g the number of analogies solved i n c o r r e c t l y by the t o t a l number of analogies completed. In the event that a booklet score i s zero, i t i s adjusted to equal one, making i t possible to compute mean latency score. 112 The mean latency scores and error rate are the three dependent variables and each subject has 24 such scores for each vari a b l e . Mean latency correct i s C r i t e r i o n Variable 1 (CV1), mean latency t o t a l i s C r i t e r i o n Variable 2 (CV2) arid mean error rate i s C r i t e r i o n Variable 3 (CV3). The independent or predictor variables are associated with the t h e o r e t i c a l components (encoding, inference, mapping, ap p l i c a t i o n ) . Their values are determined by the "objective" distance between analogy terms A and B, A and C, and C to Dtrue expressed i n terms of the number of a t t r i b u t e values that change from term to term and the strategy employed by the student. When the exhaustive mode i s used the subject compares a l l the at t r i b u t e values encoded for a given p a i r of s t i m u l i and the predictor variable score i s equal to the number of at t r i b u t e changes between them. I f the self-terminating mode i s used, only a subset of possible comparisons i s made between a p a i r of s t i m u l i . The distance between Dtrue and Dfalse terms provides a m u l t i p l i e r that i s used to calculate the values of the predictor variable for self-terminating operations. Exhaustive encoding has a constant value of 5, the number of terms to be encoded. The values for exhaustive inference, mapping, and application range from 1 to 3. The m u l t i p l i e r used to calculate the value for the s e l f -terminating mode i s obtained from the equation: (N + 1)/(N) (N - h + 1) , 113 where N i s the number of attr i b u t e s that can be processed exhaustively and h i s the number of a t t r i b u t e values of Dtrue that are the same as Dfalseand ranges from 0 to 3 (Sternberg & R i f k i n , 1979). Since N i s always 4, the m u l t i p l i e r s i m p l i f i e s to 5/(20 -4h) . Thus i t can be seen that, for each analogy booklet, the predictor variables can have two values, one for exhaustive mode and one for self-terminating mode. These values change from booklet to booklet by fixed predetermined amounts. The values for the predictor variables (1 to 8 ) are l i s t e d i n Table 9. The odd-numbered variables are exhaustive and the even-numbered are self-terminating. The mean values of these variables are l i s t e d at the bottom of the table. Sternberg theorized seven possible regression models that could be used to solve analogies. These are shown i n Table 10. Models 1 to 4 (procedural models) have a mapping component and encoding i s exhaustive. These models contain a l l possible combinations of exhaustive and self-terminating inference, application, and mapping components. Models 1 to 4 are modified to produce Models IM, 2-3M, and 4M by the removal of the mapping component. The correct analogy solution i s the one that makes the rel a t i o n s h i p between the C and D terms the same as the rel a t i o n s h i p between the A and B terms. For each booklet, the value of the predictor variable associated with the inference (A to B) and appl i c a t i o n (C to Dtrue)components i s the same for the same 114 Table 9. Predictor variables for regression Applica-Analogy Inference Mapping t i o n Encoding Dt=DfMultiplier Booklet A - B A - C C - Dt 1 2 3 4 5 6 7 8 9 10 exh. St. exh. St. exh. St. exh. st . (h) 5/(20 -4h) 1 1 .63 1 .63 1 . 63 5 3.13 3 5/8 ( . 63) 2 1 .63 2 1.25 1 . 63 5 3 .13 3 5/8 3 1 .63 3 1.88 1 . 63 5 3 .13 3 5/8 4 2 1.25 1 . 63 2 1. 25 5 3 .13 3 5/8 5 2 1.25 2 1.25 2 1.25 5 3 .13 3 5/8 6 3 1.88 1 .63 3 1.88 5 3 .13 3 5/8 7 1 .42 1 .42 1 .42 5 2 . 08 2 5/12 (.42) 8 1 .42 2 .83 1 .42 5 2. 08 2 5/12 9 1 .42 3 1.25 1 .42 5 2 . 08 2 5/12 10 2 .83 1 .42 2 .83 5 2 . 08 2 5/12 11 2 .83 2 .83 2 .83 5 2 . 08 2 5/12 12 3 1.25 1 .42 3 1.25 5 2.08 2 5/12 13 1 .31 1 .31 1 .31 5 1.56 1 5/16 (.31) 14 1 .31 2 .63 1 .31 5 1. 56 1 5/16 15 1 .31 3 .94 1 .31 5 1.56 1 5/16 16 2 . 63 1 . 31 2 . 63 5 1. 56 1 5/16 17 2 . 63 2 . 63 2 .63 5 1. 56 1 5/16 18 3 .94 1 .31 3 .94 5 1. 56 1 5/16 19 1 .25 1 .25 1 .25 5 1.25 0 5/20 (.25) 20 1 .25 2 .50 1 .25 5 1.25 0 5/20 21 1 .25 3 .75 1 .25 5 1.25 0 5/20 22 2 . 50 1 .25 2 . 50 5 1.25 0 5/20 23 2 .50 2 .50 2 .50 5 1.25 0 5/20 24 3 .75 1 .25 3 .75 5 1.25 0 5/20 Mean 1. 67 . 67 1.67 . 67 1.67 .67 5 2 . 01 Note: exh. st . = exhaustive execution = self-terminating execution 115 Table 10. Models for Regression Model Components Y' 2 Y 1 3 Y» 4 encoding-response bo encoding-response bO encoding-response bO encoding-response bO exhaustive inference* + b l V l exhaustive inference + b l V l exhaustive inference + b l V l + exhaustive mapping b3V3 + exhaustive mapping b3V3 + exhaustive ap p l i c a t i o n * b5V5 self-terminating application b6V6 self-terminating self-terminating mapping application + b4V4 + b6V6 self-terminating self-terminating self-terminating inference* mapping app l i c a t i o n * + b2V2 b4V4 + b6V6 1M encoding-response Y' = bO + 2-3M response exhaustive inference* b l V l exhaustive inference bO + b l V l + exhaustive app l i c a t i o n * b5V5 self-terminating self-terminating application encoding b6V6 + b8V8 4M response self-terminating self-terminating self-terminating inference* a p p l i c a t i o n * encoding bO + b2V2 + b6V6 + b8V8 * = confounded components 116 mode of processing. This means exhaustive inference i s confounded with exhaustive application, as i n Models 1 and IM, and s e l f -terminating inference i s confounded with self-terminating a p p l i c a t i o n , as i n Models 4 and 4M. S i m i l a r l y , whenever encoding i s exhaustive i t has a constant value and i s therefore confounded with the response component which i s also a constant, as i n Models 1, 2, 3, 4, and IM. In the regression models, the regression c o e f f i c i e n t s b l to b8 are the mean values of the predictor variables 1 to 8 (see Table 9). Columns 1 and 5 are i d e n t i c a l as are columns 2 and 6. The regression models (see Table 10) collapse into the form i l l u s t r a t e d by Table 11 i n which a single term represents the confounded components. Design In t h i s design a factor representing subjects i s crossed with a task factor consisting of repeated measures using 24 picture analogy booklets. There are 24 independent measures for the three predictor variables and seven possible models which are entered into regression analysis for a t o t a l of 21 times S regressions (where S equals the number of subjects). Each model i s a regression equation i n which the value of the regression c o e f f i c i e n t s associated with the t h e o r e t i c a l components have been predetermined (see Table 11). The time taken to process each of the components 117 Table 11. Regression equations Model Components 1 Y 1 bO + 1. 67V1/V5 + 1.67V3 exh. encoding/ exh. inference/ exh. response* app l i c a t i o n * mapping 2 Y' bO + 1.67V1 + 1.67V3 + 0.67V6 exh. encoding/ exh. St. St. response* inference mapping application 3 Y' bO + 1.67V1 + 0.67V3 + 0.67V6 exh. encoding/ exh. St. St. response* inference mapping application 4 Y 1 bO + 0. 67V2/V6 + 0.67V3 exh. encoding/ St. inference/ St. response* app l i c a t i o n * mapping 1M Y' bO + 1. 67V1/V5 exh. encoding/ exh . inference/ response* app l i c a t i o n * 2-3M Y 1 bO + 1.67V1 + 0.67V6 + 2.01V8 response exh. s t . s t . inference application encoding 4M Y 1 = bO + 0.67V2/V6 + 2.01V8 response s t . inference/ s t . appli c a t i o n * encoding exh. = exhaustive s t . = self-terminating * = confounded components 118 i s unknown; only the mean time taken to solve an analogy i s known. Multiple regression analysis p a r t i t i o n s t h i s mean time among the t h e o r e t i c a l components. The model which best f i t s the data i s the one which accounts for most of the task variance. This model enables componential parameters to be calculated. Procedure Administration of Reading and Related Tests The examiners were a special education teacher and three graduate students experienced i n t e s t i n g . Training was given i n a session of approximately three hours and opportunity to practise administering the tests was provided. The examiners were not t o l d to which of the two groups students belonged. The students were i n d i v i d u a l l y administered the Peabody Picture Vocabulary Test, the D u r r e l l subtests, and the Boder t e s t i n a single session of approximately one hour. Oral reading was tape-recorded, enabling the number of reading errors and other scores to be checked by the p r i n c i p a l researcher. The examiners marked the Boder word l i s t s as they proceeded. When a student could read no more than ten words i n a graded wordlist (50 percent) on f l a s h presentation, two more l i s t s were presented to provide a large sample of unknown words for the 119 s p e l l i n g t e s t . Students were asked to draw a clock face from memory while the examiner selected twenty words for the s p e l l i n g t e s t . The twenty-word s p e l l i n g t e s t was then administered without any further marking on the part of the examiners. The word l i s t s were checked and the s p e l l i n g t e s t s marked by the p r i n c i p a l researcher using the guidelines i n the manual for i d e n t i f i c a t i o n of Good Phonetic Equivalents (GFEs). The clock drawings were examined for evidence of reversals i n writing and placing numbers or evidence of number sequencing problems. Although some of the drawings were crudely executed, there was no evidence of any of these problems. The drawings were therefore not referred to again or used i n interpretation of l a t e r analyses. Administration of the Picture Analogies Test The examiner who administered the analogy booklets was a q u a l i f i e d schoolteacher who had no other involvement i n the tes t i n g and was unaware of the student's group membership. Training was given to the examiner over a period of approximately three hours. The schematic picture analogies were administered to subjects i n groups of not more than f i v e . Where a school had more than f i v e students taking part i n Phase 2, they were divided into two smaller groups as appropriate. I t was possible to administer the twenty-four booklets i n a single session of approximately one hour. The sessions began with 120 a period of i n s t r u c t i o n followed by administration of ten booklets. A f t e r a f i v e minute break, instructions were b r i e f l y reviewed and the remaining fourteen booklets completed. The time taken for i n i t i a l i n s t r u c t i o n was usually no longer than ten minutes and the time needed for giving-out and c o l l e c t i n g - i n the booklets amounted to about ten minutes. During the i n s t r u c t i o n sessions, three 8 by 11 inch cards were shown to the students (Appendix D) . The f i r s t consisted of two figures approximately 5 inches high and was used to demonstrate the way i n which the attr i b u t e s could vary. The second and t h i r d cards each showed an analogy problem i n which the f i r s t two terms d i f f e r e d by one and two attr i b u t e s respectively. The students were made f a m i l i a r with the at t r i b u t e s and t h e i r optional values through the use of Card 1; the format and goal of an analogy problem were explained using Cards 2 and 3. Emphasis was placed on choosing from the response options the figure that was as l i k e and as d i f f e r e n t from the t h i r d term as the second term was as l i k e and as d i f f e r e n t from the f i r s t term. Questions were answered and further explanations given u n t i l a l l students appeared to understand what was expected of them. Following a b r i e f practice session i n which students completed a sheet containing four sample analogies, the f i r s t set of analogy booklets were placed face-down i n front of the students. They were instructed to turn the booklets over when t o l d and solve them i n the same way as they had practised by c i r c l i n g one of the two answer options. Students were advised 121 that i f they wished to change an answer, the f i r s t response should be crossed out and the other one c i r c l e d . Students were t o l d that they might be unable to complete a l l the analogies i n a booklet and should not worry when t h i s occurred. I t was also pointed out that i t was advisable not to take notice of the other students as they were working on d i f f e r e n t booklets at any given time. Scoring: Peabody Picture Vocabulary Test - Revised. This t e s t was marked by the examiner who administered i t . Test protocols were checked by the p r i n c i p a l researcher to make sure the basal and c e i l i n g scores had been i d e n t i f i e d c o r r e c t l y and the raw scores calculated accurately. No marking or c a l c u l a t i n g errors were detected. The raw scores were transformed into a standard score (based on age) with a mean of 100 and standard deviation of 15. Boder Test of Reading - Spelling Patterns. While t h i s t e s t was being administered, the examiners marked the words i n each graded word l i s t as they were read o r a l l y . These were l a t e r checked by the p r i n c i p a l researcher to make sure the 122 s p e l l i n g t e s t s had been constructed c o r r e c t l y . The p r i n c i p a l researcher then marked the s p e l l i n g t e s t noting the number of words i n each l i s t that were spelled accurately or met the c r i t e r i a for GFEs. Du r r e l l Analysis of Reading D i f f i c u l t y . In t h i s t e s t , a student's o r a l reading was tape-recorded. This enabled the p r i n c i p a l researcher to assess the number of oral reading errors i n each category and the number of self- c o r r e c t e d errors. Examiners also tape-recorded the responses to comprehension questions making i t possible for the p r i n c i p a l researcher to score these answers. The t o t a l times for oral and s i l e n t reading, the comprehension scores and number and type of o r a l reading errors were recorded on the i n d i v i d u a l record sheets. Schematic Picture Analogies. This t e s t was marked by the p r i n c i p a l researcher. The students recorded t h e i r responses ion the t e s t booklets. When the booklets were marked, there was no confusion as to a student's intended response. Occasionally a student changed an answer but t h i s was c l e a r l y indicated. A l l answer booklets were checked by a second person and any booklets i n which a marking error was detected were 123 rescored a t h i r d time. Only three marking errors were detected i n 2328 (97 x 24) booklets. Data Preparation As tests were marked and scores v e r i f i e d , the scores were entered on the in d i v i d u a l record sheets. Once the scores for the Phase II te s t s were entered into the computer, Phase I and Phase II f i l e s were merged. Readino: and reading-related variables. Twenty-three scores were obtained for each student on the reading and related t e s t s . The f i v e subtests of the Stanford Diagnostic Reading Test provided nine scores. Four of these scores were obtained from the Decoding component which measured phonetic analysis of consonants, phonetic analysis of vowels, s t r u c t u r a l analysis involving s y l l a b i c a t i o n , and s t r u c t u r a l analysis involving blending. The SDRT Auditory Vocabulary subtest provided one score fo r the Vocabulary component. Factual comprehension and i n f e r e n t i a l comprehension provided two scores for the Comprehension component. Two scores variables obtained from the Reading Rate component r e f l e c t e d the accuracy of comprehension at speed and t o t a l comprehension at speed. 124 Fourteen scores were obtained from three subtests of the D u r r e l l Analysis of Reading D i f f i c u l t y . Oral and s i l e n t reading provided two measures of time and two measures of comprehension and the l i s t e n i n g t e s t provided one measure of comprehension. These measures were; or a l reading time, s i l e n t reading time, o r a l reading comprehension, s i l e n t reading comprehension, and l i s t e n i n g comprehension. In addition the o r a l reading t e s t provided nine measures associated with oral reading errors. They were part- and whole-word r e p e t i t i o n , part- and whole-word omission, part- and whole-word substitution, part- and whole-word addition, and percentage of s e l f - c o r r e c t e d errors. The l a t t e r score equalled the r a t i o of t o t a l s e l f - c o r r e c t e d errors to the t o t a l errors. The i d e n t i f i c a t i o n number of each student, the scores from the CCAT, PPVT, SDRT, and Durrell t e s t s were entered into a computer data f i l e by the p r i n c i p a l researcher. A graduate student helped to compare a printout of the data with the i n d i v i d u a l record sheets to make sure the entries were one hundred percent accurate. The percentage of self - c o r r e c t e d o r a l reading errors for each y student was generated using SPSS subprogram Numeric Transformations (SPSS X User's Guide, 1986). Printouts of these computer-generated scores were checked for accuracy. 125 Schematic Picture Analogies. The number of correct and incorrect responses was recorded on an i n d i v i d u a l record sheet and entered into a computer data f i l e . A computer printout of the data was also checked with the in d i v i d u a l record sheet to make sure the entries were one hundred percent accurate. For each student the t o t a l number of responses for each booklet was computed by adding the number correct to the number of errors. Mean latencies were computed by d i v i d i n g the number solved c o r r e c t l y and the t o t a l number of responses into 64, the number of seconds allowed for completion of each booklet. Scores of zero were recoded to equal one to make d i v i s i o n possible. The mean error rate for each booklet was computed by d i v i d i n g the number of errors by the number of analogies completed. 126 CHAPTER V: RESULTS The r e s u l t s and discussion presented i n t h i s chapter are organized into sections. In the f i r s t section an overview i s presented of the reading and related data and the various methods used i n attempts to subtype disabled readers. The second section explains how the models preferred by i n d i v i d u a l students i n solving schematic picture analogies were determined and how subgroups were formed by grouping students who showed preference for the same model. This i s followed by a description of the analogy subgroups i n terms of components and strategies that were used and solution and component scores. The fourth part examines the r e l a t i o n s h i p between the reading and analogy data of Group D i n terms of c l u s t e r versus analogy subgroup membership, c h a r a c t e r i s t i c s of the analogy subgroups, and correlations between observed solu t i o n times ( c r i t e r i o n v a r i a ble 1) and reading variables. Reading Typology Two methods of subtyping were used with Group D. The f i r s t involved the Boder t e s t which was designed to categorize students according to t h e i r reading and s p e l l i n g patterns. The second, a method of c l u s t e r analysis, involved a h i e r a r c h i c a l agglomerative technique and used the data c o l l e c t e d from the SDRT and Durrell subtests. Both methods are reported but only the s t a t i s t i c a l 127 c l u s t e r i n g was able to i d e n t i f y reading-disabled subgroups. The Boder Test of Reading-Spelling Patterns In t h i s study the reading-disabled sample proved to be c h a r a c t e r i s t i c a l l y very d i f f e r e n t from the c l i n i c a l samples used by Boder. The majority of disabled readers read the graded word l i s t s at a much higher l e v e l than expected. Boder defined disabled readers as those who had Reading Quotients (RQs) below 100 but only 20 members of Group D met t h i s c r i t e r i o n . The RQs of those who did were not s u f f i c i e n t l y low to be c l a s s i f i e d into any of the s p e c i f i c reading d i s a b i l i t y subtypes (Table 8, p. 106). Therefore i t was not possible to i d e n t i f y s p e c i f i c reading d i s a b i l i t y subtypes using t h i s method (Question 1). Examination of s p e l l i n g patterns alone indicated that 25 were si m i l a r to those of the reading d i s a b i l i t y subtypes i d e n t i f i e d by Boder, but the severity of t h e i r d i s a b i l i t y as indicated by the Reading Quotient did not match the s p e c i f i c reading d i s a b i l i t y subtypes. The s p e l l i n g patterns are l i s t e d i n Appendix E. Cluster Analysis The c l u s t e r analysis was ca r r i e d out using the UBC CGroup (Lai, 1982) computer program. The technique used i s a h i e r a r c h i c a l agglomerative technique (see Morris, B l a s h f i e l d , & Satz, 1981) 128 involving Ward's (1963) algorithm for minimum variance. In t h i s method, subjects are combined i n a series of stepwise groupings using a c r i t e r i o n based on p r o f i l e s i m i l a r i t y . At each step, groups are paired, reducing the number by one, u n t i l a l l subjects are contained i n one f i n a l group. A p i c t o r i a l representation of the pairings i s provided by a h i e r a r c h i c a l tree graph. The program also provides a p l o t of the logarithmic error term, the slope of which i s proportional to the rate of increase i n error. A marked change i n the slope of the graph indicates a large increase i n error. The decision as to how many groups should be selected i s aided by the h i e r a r c h i c a l tree graph and the error p l o t . In the present study, two c l u s t e r i n g analyses were completed. I n i t i a l l y Group N and Group D were combined to see i f Group N could be distinguished from Group D using the 2 3 reading and reading-related variables considered i n t h i s study. Of the f i v e groups i d e n t i f i e d i n t h i s c l u s t e r analysis, Cluster 5 c l e a r l y d i f f e r e n t i a t e d Group N from Group D as i t contained a l l but two members of Group N. In the second analysis Group D was analyzed alone. A three-c l u s t e r solution gave the best r e s u l t (Question 2). Clusters 1, 2, and 3 contained 26, 20, and 31 members respectively. As the Boder t e s t f a i l e d to i d e n t i f y subtypes within Group D no comparison could be made between these c l u s t e r s and the r e s u l t s of that test (Question 3). Analysis of variance indicated s i g n i f i c a n t differences (p<.05) 129 among the c l u s t e r s on 17 of the reading and related variables (Table 12) . Sixteen of the differences exceeded a s i g n i f i c a n c e l e v e l of .01. This r e s u l t confirms that the groups created by the h i e r a r c h i c a l agglomerative technique were s u b s t a n t i a l l y d i f f e r e n t on the majority of reading and related v a r i a b l e s . Included i n Table 12 are the CCAT and PPVT variables which represent nonverbal and verbal a b i l i t y , respectively. Box-and-Whisker Plots Box-and-Whisker plots (Tukey, 1977) were used to make comparisons among the c l u s t e r s . This method was used because i t permits a d e t a i l e d i l l u s t r a t i o n of the d i s t r i b u t i o n of scores within each c l u s t e r . The s i m i l a r i t i e s and differences i n batches of data are noted through a comparison of t h e i r ranges, the amount of overlap of t h e i r boxes, the value of t h e i r medians, and the number of t h e i r outside and far-out values. The boxes contain the middle 50 percent of subject scores for each batch of data with the remaining 50 percent d i s t r i b u t e d , 25 percent above the box and 25 percent below. This makes i t possible to estimate the amount of overlap between d i s t r i b u t i o n s . When comparing d i s t r i b u t i o n s : 1) i f the boxes do not overlap then there i s no s i m i l a r i t y between the middle 50 percent of the d i s t r i b u t i o n s (see Figure 3a), (2) i f a l l the scores of one d i s t r i b u t i o n f a l l above or below the median of the other then they do not overlap at l e a s t f i f t y percent of the scores of that d i s t r i b u t i o n (see Figure 3b), and (3) i f neither 130 Table 12. Analysis of variance aittoncf reading d i s a b i l i t y c l u s t e r s  on reading and related variables Variable SS Error MS Error F SS MS (3 ,76) Decoding SDRT PA (consonants) 219. ,28 670. ,54 109. , 64 9. 06 12 . 10** PA (vowels) 436. ,42 819. ,30 218. ,21 11. 07 19 .71** SA (syllables) 1085. ,55 4012 . 97 542. ,78 54. 23 10 . 01** SA (blending) 710. ,70 1466. , 16 355. ,35 19. 81 17 .94** RR SDRT Accuracy 185. ,69 1203 . 75 92. .84 16 . 27 5 .71** Speed 151. , 07 2053 . 30 75. ,53 27 . 75 2 .72 Time (Durrell) OR time 31431. ,61 40342. .21 15715. ,80 545 . 17 28 .83** SR time 20870. , 18 86384. .99 10435. , 09 1167 .37 8 .94** Comp. (SDRT) AV 46. ,73 1670. .5 23. ,37 22 .58 1 . 04 Factual 31. , 58 546. .78 15. ,79 7 . 39 2 . 14 In f e r e n t i a l 7. , 11 498. .61 3 . ,55 6 .74 . 53 Comp. (Durrell) OR 51. ,45 318, .86 25. ,72 4 .31 5 .97** SR 370. ,89 5969. .89 185. ,44 80 . 67 2 .30 List.Comp. 117. ,20 670. .22 58. , 60 9 . 06 6 .47** OR errors P-w rep. 3 . ,20 12. .93 1. , 60 . 17 9 . 16** W-w rep. 1132 . ,47 1567, .53 566. ,23 21 . 18 26 .73** P-w omm. 120. ,43 255, .93 60. .22 3 . 45 17 .41** W-w omm. 33, ,74 308, .20 16. .87 4 . 17 4 . 05* P-w sub. 92 . , 75 294, .06 46. . 38 3 . 97 11 . 67** W-w sub. 259. . 01 481, . 06 129. . 51 6 . 50 19 . 92** P-w add. 24. .54 121, .59 12, .27 1 .64 7 .47** W-w add. 49. .61 256, .91 24. .80 3 .47 7 . 15** S-C (%) , 05 1, .24 . 02 .20 1 .45 A b i l i t y CCAT 457. .83 3787, .10 228. .92 102 . 19 2 .24 PPVT 77. .99 7561, .98 39. . 00 88 .88 .44 * = p<.05 **= p<.01 131 (a) No overlap between boxes of Groups A and B X # c o r r e c t T } Groups IK 8 (b) No overlap between d i s t r i b u t i o n of Group A and the median of Group B # c o r r e c t Groups 8 ) I „ T \\ooL J _ (c) No overlap of the boxes of Groups A and B with the medians of the other d i s t r i b u t i o n # c o r r e c t I 8 XT?. Groups Figure 3. Overlap of box and whisker pl o t s 132 box overlaps the median of the other there i s no s i m i l a r i t y between at l e a s t h a l f of the middle f i f t y percent of the d i s t r i b u t i o n s (see Figure 3c) . In the present study, batches of data for any given v a r i a b l e are considered to be more d i s s i m i l a r than a l i k e i f any of these conditions are s a t i s f i e d . A Comparison of Reading-Disability Clusters Figure 4 shows the plots of the four SDRT variables for the Decoding component. Cluster 1 had more d i f f i c u l t y decoding consonants and vowels than Clusters 2 and 3, and Clusters 1 and 2 had more d i f f i c u l t y separating words into s y l l a b l e s and blending s y l l a b l e s into words than Cluster 3. Cluster 2 had phonetic analysis s k i l l s that resembled those of Cluster 3 but Cluster 3 had better s t r u c t u r a l analysis s k i l l s . The SDRT Vocabulary component, which was represented by the Auditory Vocabulary subtest, did not discriminate among the c l u s t e r s (Figure 5). The median values for Clusters 2 and 3 were the same and considerable overlap of the d i s t r i b u t i o n s was evident. Figure 6 shows the plots of the SDRT Reading Comprehension component: factual and i n f e r e n t i a l comprehension. Members of Cluster 3 tended to have higher factual comprehension scores than members of Cluster 1 but Clusters 1 and 2 had the same median and t h e i r d i s t r i b u t i o n s were s i m i l a r . Neither factual nor i n f e r e n t i a l comprehension discriminated among the c l u s t e r s . 133 P h o n e t i c A n a l y s i s ( C o n s o n a n t s ) P h o n e t i c A n a l y s i s ( V o w e l s ) # c o r r e c t A. 0k <*3 \ 2- 3 C l u s t e r s S t r u c t u r a l A n a l y s i s ( S y l l a b i c a t i o n ) ¥1 # c o r r e c t to 3 I X 3 C l u s t e r s l% # c o r r e c t 8<L I Z C l u s t e r s S t r u c t u r a l A n a l y s i s ( B l e n d i n g ) at # c o r r e c t C l u s t e r s F i g u r e 4. B o x - a n d - w h i s k e r p l o t s f o r Group D c l u s t e r s on SDRT D e c o d i n g component v a r i a b l e s 134 A u d i t o r y V o c a b u l a r y 37 11 3 - I I Z 3 Clusters Figure 5. Box-and-whisker plots for Group D clusters on SDRT Vocabulary component variable 135 F a c t u a l C o m p r e h e n s i o n if c o r r e c t e a x 0 C l u s t e r s I n f e r e n t i a l C o m p r e h e n s i o n c o r r e c t r-3 C l u s t e r s F i g u r e 6. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n SDRT R e a d i n g C o m p r e h e n s i o n c o m p o n e n t v a r i a b l e s 136 C o r r e c t R e s p o n s e s 33 c o r r e c t F A X T " i J — / 2. C l u s t e r s T o t a l R e s p o n s e s 3*r N u m b e r o f R e s p o n s e s 6 ..1. 5 I . S B I 7. 3 C l u s t e r s F i g u r e 7. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n SDRT R e a d i n g R a t e c o m p o n e n t v a r i a b l e s 137 The two variables associated with the SDRT Reading Rate component measured speed and accuracy of reading r e l a t i v e l y easy material. The number of correct responses and the t o t a l number of responses at speed are shown i n Figure 7. Cluster 3 was more accurate than Cluster 2. Speed of reading did not d i f f e r e n t i a t e between the c l u s t e r s . In Figure 8, the box plots of Du r r e l l Oral and S i l e n t reading time are shown. Members of Cluster 3 were faster o r a l readers than eithe r of the other c l u s t e r s . Members of Cluster 1 were faster oral readers than Cluster 3. Members of Cluster 2 were the slowest s i l e n t readers. Clusters 1 and 3 were s i m i l a r i n t h e i r s i l e n t reading speed. The box plots of Du r r e l l Oral Reading, S i l e n t reading and Listening Comprehension are shown i n Figure 9. Members of Cluster 2 had better Oral Reading Comprehension than those i n the other c l u s t e r s and there was less v a r i a b i l i t y i n t h e i r scores. Overlap of the d i s t r i b u t i o n s for S i l e n t Reading comprehension shows that the c l u s t e r s were a l i k e on t h i s variable. Members of Cluster 1 had better l i s t e n i n g comprehension than members of Cluster 3. The box plo t s for oral reading error variables are shown i n Figure 10. Of the eight types of errors, Cluster 2 made more part-and whole-word re p e t i t i o n s , whole-word substitutions, and part- and whole-word additions than eith e r Clusters 1 or 3, and more part-and whole-word omissions and part-word substitutions than Cluster 3. Members of Cluster 2 also displayed the most v a r i a b i l i t y i n 138 O r a l R e a d i n g T i m e y3e T i m e i n S e c o n d s f 7 ax C l u s t e r s S i l e n t R e a d i n g T i m e T i m e i n S e c o n d s C I u s t e r s F i g u r e 8 - B o x - a n d - w h i s k e r p l o t s f o r G r o u p 0 c l u s t e r s o n D u r r e l l R e a d i n g T i m e v a r i a b l e s 139 O r a l R e a d i n g C o m p r e h e n s i o n 2.3 # c o r r e c t ft ; % 3 C l u s t e r s S i l e n t R e a d i n g C o m p r e h e n s i o n # c o r r e c t C I u s t e r s L i s t e n i n g C o m p r e h e n s i o n # c o r r e c t 3* > I a 3 C I u s t e r s F i g u r e 9. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n D u r r e l l C o m p r e h e n s i o n v a r i a b l e s 140 P a r t - w o r d R e p e t i t i o n E r r o r s # o f e r r o r s i3 C l u s t e r s W h o l e - w o r d R e p e t i t i o n E r r o r s # Of e r r o r s a a I X. 3 C l u s t e r s P a r t - w o r d O m i s s i o n E r r o r s # o f e r r o r s S i 8 1 X C I u s t e r s W h o l e - w o r d O m i s s i o n E r r o r s S # o f e r r o r s a 8 E .1 1 * C l u s t e r s F i g u r e 10. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n D u r r e l l O r a l R e a d i n g E r r o r v a r i a b l e s 141 Part-word Substitution Errors II r # of errors 8 CIusters Whole-word Substitution Errors # of errors i _a. 3 Clusters Part-word Addition Errors # of errors S B2 Clusters Whole-word Addition Errors # of errors 8 ax X J . : / a.. 3 Clusters Figure 10 (cont.) Box-and-whisker plots for Group D clusters on Durrell Oral Reading Error variables 142 t h e i r errors scores for a l l error variables except part-word sub s t i t u t i o n . Cluster 1 made more part- and whole-word omissions, part-word substitutions and part-word additions than Cluster 3. The t o t a l number of errors and sel f - c o r r e c t e d errors were used to c a l c u l a t e the percentage of sel f - c o r r e c t e d errors. They are shown i n Figure 11. No overlap exists between any of the boxes for t o t a l errors confirming that the cl u s t e r s were d i s s i m i l a r on t h i s v a r i a b l e . Cluster 2 made more oral reading errors than either Clusters 1 or 3 and Cluster 1 made more errors than Cluster 3. The box-plot for sel f - c o r r e c t e d errors shows that, although Cluster 2 made the most errors, members of t h i s c l u s t e r also s e l f - c o r r e c t e d the most errors. The proportion of sel f - c o r r e c t e d errors ranged from zero to 58 percent. Cluster 3 had the largest range but the amount of overlap shows that the cl u s t e r s were s i m i l a r . Although the amount of overlap shows that the cl u s t e r s were s i m i l a r . Although members of Cluster 2 made the most errors, members s e l f -corrected i n a proportion s i m i l a r to members of the other c l u s t e r s . The c l u s t e r s were also compared on the nonverbal a b i l i t y v a r i a b le (CCAT) and the verbal a b i l i t y v a r i a ble (PPVT) (see Figure 12). Neither of these variables was able to discriminate among the cl u s t e r s although the scores of Cluster 2 show the most v a r i a b i l i t y f o r the CCAT and the least v a r i a b i l i t y for the PPVT. 143 T o t a l E r r o r s 5% e r r o r s \ X 3 C l u s t e r s S e l f -C o r r e c t e d E r r o r s I S # S e l f -C o r r e c t e d P e r c e n t S e l f -C o r r e c t e d E r r o r s *81 3 P e r c e n t C l u s t e r s ' X 3 CI u s t e r s r n g u r e n - B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n D u r r e l l O r a l R e a d i n g E r r o r v a r i a b l e s : T o t a l E r r o r s , T o t a l S e l f -C o r r e c t e d E r r o r s a n d P r o p o r t i o n S e l f - C o r r e c t e d E r r o r s 144 C C A T 1/7 U n i v e r s a l A g e S c o r e r i a 3 C l u s t e r s P P V T S c a l e d S c o r e .A. a .L-I * 3 C l u s t e r s F i g u r e 12. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D c l u s t e r s o n V e r b a l a n d N o n v e r b a l A b i l i t y V a r i a b l e s 145 Summary of Clusters Members of Cluster 1 were characterized by the poorest s k i l l s i n the decoding component. They were second i n the number of oral reading errors made and second i n or a l reading speed. Members of Cluster 2 were characterized by the greatest number of or a l reading errors and the slowest reading speed. Despite t h i s , t h e i r oral reading comprehension was better than that of the other groups. When required to read at speed, they made fewer responses than Cluster 1 and were less accurate than either of the cl u s t e r s even though the material was easier. Members of Cluster 3 were characterized by the best decoding s k i l l s , the fast e s t o r a l reading time and the l e a s t o r a l reading errors, yet they did not have higher o r a l or s i l e n t reading comprehension scores than the other c l u s t e r s . Componential Analysis Inspection of the Data There were three c r i t e r i o n variables for each booklet. C r i t e r i o n Variable 1 (CV1) was the mean latency for analogies solved c o r r e c t l y , C r i t e r i o n Variable 2(CV2) was the mean latency for the t o t a l number of analogies completed, and C r i t e r i o n Variable 3 (CV3) was the mean error rate. The mean score, standard 146 Table 13. Descriptive s t a t i s t i c s for c r i t e r i o n variables Group Variable Mean SD Range Total CV1 9.52 4.74 4.79-30.03 n=97 CV2 6.48 2.12 4.02-16.76 CV3 0.19 0.16 0.00 - 0.58 Group D CV1 10.09 4.35 4.79-30.03 n=77 CV2 6.62 2.20 4.02 - 16.76 CV3 0.25 0.13 0.00 - 0.58 Group N CV1 7.3 0 2.28 4.74-12.28 n=20 CV2 5.93 1.69 4.06 - 11.21 CV3 0.16 0.10 0.00 - 0.43 147 deviation, and range for each of these variables are shown i n Table 13. Before componential analysis was undertaken, inspection of the data was c a r r i e d out to determine the number of variables to be analyzed. C r i t e r i o n Variables 1 and 2 both had s u f f i c i e n t variance to make further analysis worthwhile but CV3 had variance close to zero, making further analysis impractical. Although CV1 and CV2 were both analyzed, only CV1 i s discussed at length because t h i s r e f l e c t s q u a l i t y and quantity of performance whereas CV2 r e f l e c t s only quantity. Results for CV2 are l i s t e d i n Appendix E. Individual Regression Analysis There were 24 scores for each i n d i v i d u a l for each c r i t e r i o n v a r i a b l e . Regression analysis was ca r r i e d out using the SPSS X subprogram Regression (SPSS User's Guide, 1986) with forced entry of the independent variables for each of the seven models theorized by Sternberg. The f i r s t decision to be made i n se l e c t i n g the preferred model was acceptance or r e j e c t i o n of Models 1 to 4. This was based upon the s i g n i f i c a n c e of self-terminating encoding i n Models 2-3M and 4M and the si g n i f i c a n c e of the mapping component i n Models 1 to 4. Models 1 to 4 were automatically rejected i f the mapping component was not s t a t i s t i c a l l y s i g n i f i c a n t . I f the self-terminating encoding i n Models 2-3M and 4M was s i g n i f i c a n t Models 1 to 4 were again 148 rejected as self-terminating mode was preferred to exhaustive mode. Aft e r the decision was made to accept or r e j e c t Models 1 to 4, the models were evaluated according to f i v e c r i t e r i a (Sternberg & R i f k i n , 1979). 2 1. Significance of increase i n R . The model preferred on t h i s c r i t e r i o n i s the one with the 2 highest value of R . Similar models such as 2-3M and 4M tend to 2 have s i m i l a r values of R so i t i s necessary to consider other c r i t e r i a to di s t i n g u i s h between them. When one model with larger 2 . . R d i f f e r s from another by the addition of a component, the model 2 • 2 with the larger R i s only preferred i f the increase i n R , due to the addi t i o n a l component, i s s i g n i f i c a n t . This i s judged by the sig n i f i c a n c e of the value of F for the regression c o e f f i c i e n t of that component. 2. The value of regression F. The value of F for each regression equation i s related to the 2 value of R and the number of components i n the model. Where two models d i f f e r by one component, as i n the case of 2-3M and 4M, the model with the additional component has the smaller value of F even though i t may have the larger value for R2. Therefore, the sig n i f i c a n c e l e v e l not the siz e of regression F i s considered i n 149 se l e c t i n g the preferred model. I f two models have the same l e v e l of s i g n i f i c a n c e they are judged to be equally preferred. 3. The s i g n i f i c a n c e of the regression c o e f f i c i e n t s . The value of F for a regression c o e f f i c i e n t indicates whether i t d i f f e r s s i g n i f i c a n t l y from zero or not. I f a component i s considered to have been used, then i t must d i f f e r s i g n i f i c a n t l y from zero. In in d i v i d u a l analysis, s i g n i f i c a n c e i s set as p<.10 as the models tend to be less stable (Sternberg, July 27, 1987; personal communication). 4. Decrease i n standard error of estimate. The standard error of estimate indicates "badness of f i t " (Sternberg & R i f k i n , 1979). As the standard error of estimate 2 . decreases from one model to another, the value of R increases. I f the standard error of estimate for a model decreases with the addition of a component, then the proportion of decrease i s considered to determine whether the model with the additional component i s preferred. I f the decrease i s small the more parsimonious model i s preferred. Sternberg & R i f k i n (1979) considered proportional decreases i n the standard error of estimate between Models 2-3M and 4M of .12, .11, and .03 to be unimportant, whereas a proportional decrease of .19 was considered large enough 150 to warrant the i n c l u s i o n of the additional parameter i n 2-3M. 5. The nature of component estimates. A component estimate that d i f f e r s s i g n i f i c a n t l y from zero indicates that the component has probably been used. Small i n s i g n i f i c a n t negative parameter estimates are regarded i n the same way as small p o s i t i v e i n s i g n i f i c a n t parameter estimates. However, component estimates represent the mean time taken to process the component. As negative time i s not possible, small negative parameter estimates are probably a t t r i b u t a b l e to sampling error, whereas large negative values indicate that the l i n e a r model i s i n v a l i d and the additive combination rule has probably been v i o l a t e d . Models with s i g n i f i c a n t p o s i t i v e components are preferred to those with s i g n i f i c a n t negative estimates. Individual Analysis of Group D 2 The values of R and regression F were smaller than those reported by Sternberg & R i f k i n (1979) and Wilson (1980) r e f l e c t i n g the reduction i n the amount of variance available for analysis at the i n d i v i d u a l l e v e l . Results also tended to be less stable than at the group l e v e l so the s i g n i f i c a n c e l e v e l was set at .10 (Sternberg, personal communication, July 27, 1987). A preliminary inspection of the regression c o e f f i c i e n t s for mapping and s e l f -151 terminating encoding indicated that none of the subjects had shown preference f o r Models 1 to 4 so s e l e c t i o n c r i t e r i a were applied to Models IM, 2-3M and 4M only. Individual analysis of CV1 indicated that 29 students i n Group D (37.7%) preferred Model 4M (Table 14) and 12 students (15.6%) preferred Model IM (Table 15). These groups are hereafter referred to as Subgroup 4M and Subgroup IM respectively. T h i r t y - s i x students (46.8%) had no preferred model. Within t h i s group, two subgroups could be detected. Eighteen students (23.3%) had s i g n i f i c a n t regression F values but t h e i r models contained large negative parameter estimates that were often s i g n i f i c a n t and large standard error of estimate i n d i c a t i n g a very poor f i t of the data to the models. This group w i l l be referred to as Subgroup NM1 (No Model 1) . The remaining eighteen students (23.3%) had few s i g n i f i c a n t parameter estimates and no s i g n i f i c a n t values for regression F i n any model. This group w i l l be referred to as Subgroup NM2 (No Model 2) . Individual Analysis of Group N Individual analysis of CV1 for Group N indicated that fourteen students (70%) preferred Model 4M and one student (5%) preferred Model IM (see Table 16). Two students (10%) had regression models s i m i l a r to those of the members of Subgroup NM1 and three (15%) were s i m i l a r to those of students i n Subgroup NM2. 152 Table 14. Subjects i n Group D with preferred Model 4M Component estimates Student # St. Inference/ Application St. Encoding Response R 2 F a est D 1 2.62+ 1.17 7.01 .45 8.75** 1.97 D 2 1. 69 1.54 3.39 .30 4 .40* 2 . 39 D 4 -0.27 2.49* 2.30 .31 4.74* 2.77 D 5 -1.25 3.05* 3.44+ .29 4.26* 3.23 D 9 2.50 4.29* -0.56 .43 7.81** 4.64 D14 0.86 1.15** 2.09** .76 33.86** 0. 63 D16 9.16* -1.94 7.41* .29 4.24* 5. 06 D19 0.30 1.99** 2.01+ .48 9.62** 1. 67 D22 0.56 1.76* 2.93* .33 5.05* 2 .16 D2 3 0.32 0.76+ 3.45** .27 3 .79* 1.11 D24 -0.16 1.35** 2.70** .72 26.58** 0. 62 D25 2.25 0.51 6.45** .23 3 .23 + 2 .22 D31 1.58 1.52* 2 .10+ .41 7.31* 1.99 D3 5 -1.14 1. 60** 3.29** .46 8.75** 1.11 D36 0. 03 0.87** 3.22** .49 10.05** 0. 69 D43 3.15+ 2.75* -0.86 .55 12.73** 2.55 D49 -0.43 1.96** 2.00+ .39 6.61** 1.76 D50 1.21 1.53 1.77 .43 8.03** 1.76 D51 2 . 76 1.50 4.35* .31 4 . 64* 3 .11 D57 8.56** 0.99 2.63* .73 27.65** 2 . 55 D58 0. 64 0.96* 3.21** .35 5.58* 1.26 D59 -0.28 1.60* 2.44* . 32 5.04* 1.66 D60 3 .18 4.16** 0.55 .40 7.05** 5. 04 D61 3.00* 2.30* 0.85 . 60 15.71** 2 .22 D62 1.83 1.00 6.45** . 30 4 .39* 2 .13 D63 4.24* 1.00 2.86 .40 6.90** 2 . 88 D64 0.92 2.17+ 4 . 57* . 27 3 . 89* 3 .15 D71 -1.74 2.59** 3 . 92* . 32 4.82* 2 .42 D74 5.97* 0.67 3.56* . 64 18.30** 2 .17 Total 1.80** 1. 63** 3.08** .95 184.88** 0.43 + = p<.01 * = p<.05 **= p<.01 St. = self-terminating execution 153 Table 15. Subjects i n Group D with preferred Model 1M Component estimates Student Exh.Inference/ Exh. Encoding/ R F a est # Application Response D 6 1.87+ 5.50** . 13 3 .22 + 3.80 D12 3.32* 3.29 .20 5.47* 5.17 D15 5.63** -0.48 .48 20.47** 4 . 54 D30 2 . 00** 4 . 32** . 53 25.06** 1.46 D3 3 7.11** 0.11 .46 18.55** 6. 02 D39 4 . 37** 2.20 . 61 33.84** 2.74 D44 2.14** 4.20** .27 8.19** 2.74 D48 5.19** 0.01 .39 14.22** 5. 02 D54 1.68** 5.93** .42 15.71** 1.55 D65 4.69** 5.18+ .31 10.05** 5.40 D67 1. 32+ 6.18** .15 3.80+ 2.47 D70 0.88+ 6.09** . 13 3 .17+ 1.80 Total 3.35** 3.55** . 72 56.23** 4 .11 + = p<. 1 * = p<.05 **= p<.01 Exh. = exhaustive execution 154 Table 16. Subjects i n Group N who preferred Models 4M and IM Component estimates for Model 4M Student # St. Inference/ S.T. Response R 2 F a est Application Encoding N 2 0.83 1. 39** 1.79** .69 23 .84** 0.86 N 3 0.96 1.31** 2.48** .55 12 .93** 1.16 N 6 0.18 2.43** 1.78** .52 11 .26** 1.84 N 7 0.54 2.83** 2.23 + .54 12 .21** 1.89 N 9 0.91 1.50** 2.75** . 50 10 . 30** 0.97 N i l 0.27 0.49 3.58** .22 2 .98 + 0.85 N12 0.06 1.44** 3.08** .40 7 . 02** 1. 37 N13 1. 63 1.73+ 3.22** .33 5 .28* 2.57 N14 0. 01 1.11* 2.91** . 37 6 . 16** 1.11 N15 1.03 0.73+ 3.22** . 38 6 . 30** 1.15 N16 2.29+ 0. 66 3.45** .31 4 .72* 1.99 N17 0.48 2.52** 1.59* .48 9 . 51** 1.90 N18 1.50 0.20 4.33** .12 1 .42 + 2 . 01 N19 0.22 3.19** 3 .86* .41 7 .2** 3.00 Component estimates for Model IM Student —> # Exh.Inference/ Exh. Encoding/ R 2 F a est Application Response N 4 3 . 95+ 2.91+ 16 4 .26+ 7.00 + = p<.10 * = p<.05 **= p<.01 155 Analysis at the Group Level When the data were collapsed across the groups (see Table 17), Group N showed a preference for Model 4M. Within Group D, Subgroup 1M continued to prefer Model 1M and Subgroup 4M continued to prefer Model 4M, with a stronger group model emerging i n each case. Despite the apparent f a i l u r e of members of Subgroup NM2 to use a model, a group preference for Model 1M emerged. However, t h i s regression equation indicated only a weak preference for 1M. Although the value of regression F and the component estimates were s i g n i f i c a n t at p<.01, t h i s model accounted for only twenty percent of the variance. The 1M regression equation for Subgroup NM1 was the one that 2 best f i t the data. However, despite an R of .72, i t had a large standard error of estimate of 4.11. The component estimate for inference/application was p o s i t i v e and s i g n i f i c a n t but the component estimate for encoding/response was negative and i n s i g n i f i c a n t . Use of Components The r e s u l t s of i n d i v i d u a l analysis of CV1 data indicate that 53.2 percent of the students i n Group D showed preference for a model compared to 75 percent i n Group N. The r a t i o of students p r e f e r r i n g Model 4M and 1M were 29 and 12 respectively for disabled 156 Table 17. Model f i t s f or Group N and model subgroups for Group N  and Group D Group (Preferred (Model) Inf. Component estimates Inf./ App. Enc. Enc./ Resp. App. Resp. R 2 F a est Group N (Model 4M) n=2 0 2.88** 0.61 4.15** . 69 22 . 98** 1.05 Group D Subgroup 4M (Model 4M) n=29 1.80** 1.63** 3.08** .95 184. 88** .43 Subgroup IM (Model IM) n=12 3.35** 3.55** .72 56. 23** 1. 63 Subgroup NM1 (Model IM) n=18 8.42** -.69 . 72 56. 09** 4.11 Subgroup NM2 1.67* 8.76** .20 5. 65* 2.57 (Model IM) n=18 * = p<.05 **= p<.01 Inf. = Inference App. = Application Enc. = Encoding Resp.= Response 157 readers and 14 to 1 for normal readers. The students who indicated preference f o r a model used the encoding, inference, application, and response components hypothesised by Sternberg's component theory of analogical reasoning (Question 4) although inference was confounded with application i n Models 4M and IM and encoding was confounded with response i n Model IM. The mapping component was not used. T h i r t y - s i x students i n Group D and f i v e i n Group N did not indicate preference for any of the models so that i t i s impossible to say which components they used, i f any. The majority of Group N used encoding, inference, application, and response components, therefore Hypothesis 4.1 i s not rejected. Use of the Linear Combination Rule Examination of the amount of variance explained by the regression equation of the preferred model for each subgroup within Group D (Table 17) indicates the extent to which members used the 2 l i n e a r combination rule (Question 5). An R equal to .95 for Subgroup 4M supports the view that the l i n e a r combination rule was 2 used by t h i s group. An R equal to .72 for Subgroup IM indicates that the l i n e a r combination rule was used by t h i s group but to a les s e r extent. Indeed, at the ind i v i d u a l l e v e l regression F for some of t h e i r equations was s i g n i f i c a n t only at p<.10 l e v e l . 2 Although Subgroup NM1 also had an R equal to .72 for t h e i r preferred group model IM, the exhaustive encoding/response 158 component was negative and nonsignificant. In addition the large negative and often s i g n i f i c a n t parameter estimates and large standard error of estimate of t h e i r i n d i v i d u a l models indicates that the assumption of l i n e a r processing was not fe a s i b l e for t h i s 2 group. Subgroup NM2, with an R equal to .20 also appears to have v i o l a t e d the l i n e a r combination ru l e . The normal reader group had a regression model that accounted for 69% of the variance i n the data and use of the l i n e a r combination rule i s indicated. Therefore, Hypothesis 5.1 i s not rejected. Use of Exhaustive versus Self-terminating Mode Preference for Model 4M indicates the use of a s e l f -terminating mode of processing whereas exhaustive processing was used by those who showed preference for Model IM. The majority of students i n Group D and Group N who used a model preferred s e l f -terminating mode. The groups d i f f e r e d i n the r a t i o of members who showed preference for t h i s mode. The r a t i o of disabled readers using self-terminating mode to those using exhaustive mode was 29 to 12. For normal readers the r a t i o was 14 to 1. A greater proportion of normal readers appeared to prefer a self-terminating strategy than disabled readers (Question 6) and Hypothesis 6.1 i s not rejected. However, t h i s r e s u l t would be more conclusive had the number of normal readers been s i m i l a r to the number of disabled readers. 159 Solution Scores Descriptive s t a t i s t i c s for each of the model groups are shown i n Table 18. Although CV3 data were not subjected to regression analysis, i t i s included i n the descriptive s t a t i s t i c s because i t enables a comparison to be made among the subgroups. As the c r i t e r i o n v a r i a ble CV2 represents only speed of processing, examination of CV2 data indicates that, on average, Subgroup 1M solved the t o t a l number of analogies i n the shortest time and Subgroup NM1 took the most time. Comparison of CV2 with CV1 which represents speed and accuracy and inspection of CV3 indicates that Subgroup 4M was the most accurate with l i t t l e d ifference between the other three. Calculation of Component Scores In the regression equations the independent or predictor variables associated with each component indicate the mean number of times that component i s used i n solving analogies. The regression c o e f f i c i e n t s associated with each component represent the mean time taken for a single execution of that component. In a sing l e regression analysis, the c r i t e r i o n v a r i a ble which represents mean solution latency (for an i n d i v i d u a l or group) i s p a r t i t i o n e d among the components i n the regression equation. These pa r t i t i o n e d scores are the component scores and they represent the 160 Table 18. Descriptive s t a t i s t i c s on c r i t e r i o n variables  for model subgroups within group D Group Variable Mean SD Range 4M CV1 7.57 1.96 4.79 - 11.12 n=29 CV2 6.58 2.03 4.02 - 11. 12 CV3 0.16 0.08 0.00 — 0.41 IM CV1 9.13 1. 68 7.56 — 13 . 00 n=12 CV2 5.76 1.48 4.10 - 8.35 CV3 0.31 0.13 0. 04 — 0.48 NM1 CV1 13 .35 5. 02 6. 54 — 30. 03 n=18 CV2 7.26 3 . 00 4.19 - 16.76 CV3 0.30 0.10 0. 10 - 0.40 NM2 CV1 11.54 5.11 6.42 — 26.31 n=18 CV2 6. 60 1.94 4.49 - 10. 08 CV3 0.31 0.16 0.11 - 0.58 161 mean time taken to execute a component i n solving the analogies. They are calculated by multiplying each predictor variable i n the equation by i t s regression c o e f f i c i e n t . The regression equation for Model 4M i s : Y1 = bO + .67V2/V6 + 2.01V8 where Y 1 i s the predicted solution time, bO i s the regression constant, V2/V6 represent the time taken for a single execution of self-terminating inference confounded with self-terminating application, and V8 represents the time taken for a single execution of self-terminating encoding (see Table 11). For the 4M subgroup within Group D t h i s regression equation becomes: Y 1 = 3.08 + (.67)(1.80) + (2.01)(1.63) A sing l e execution of the confounded inference/application component (V2/V6) took 1.80 seconds (see Table 17), a single execution of the encoding component (V8) took 1.63 seconds, and the response component was executed i n 3.08 seconds. This regression equation indicates that on average t h i s group spent 1.21 seconds processing inference/application, 3.28 seconds encoding, and 3.08 seconds responding, for a composite time of 7.57 seconds. Component scores and composite times for the other subgroups are shown i n Table 19. As IM was the group model that best f i t the data for Subgroup NM1, t h i s has been used to cal c u l a t e component scores. The composite solution times obtained by summing the component times are the same as the observed solution times (CV1) for a l l subgroups except NM1 (see Table 19). The discrepancy between 162 Table 19. Component latencies calculated from preferred, regression  models for subgroups within Group D Name of Analogy subgroup 4M IM NM1 NM2 Model used to calculate component latencies 4M IM IM IM Self-terminating encoding 3.28 NA NA NA Response 3 . 08 NA NA NA Self-terminating inf./app. 1.21 NA NA NA Exhaustive enc./resp. NA 3.55 0. 00 8.76 Exhaustive inf./app. NA 5. 60 14 . 06 2.79 Composite 7.57 9 .15 14.06 11. 55 NA - not applicable i n f . = inference app. = a p p l i c a t i o n enc. = encoding resp.= response 163 observed and calculated solution time for Subgroup NM1 i s caused by the encoding/response component's negative regression c o e f f i c i e n t which r e s u l t s i n an estimate of negative time. To balance t h i s , the time pa r t i t i o n e d for execution of the exhaustive inference/application component i s longer than the mean observed time. This overestimate for the inference/application component r e s u l t s from a poor model f i t to the data. I t was previously noted that, i n d i v i d u a l l y , t h i s group did not seem to process i n a s e r i a l fashion and probably v i o l a t e d the l i n e a r combination rul e . In estimating component times for Subgroups 4M and IM, the model preferred by each subgroup was used. The amount of time spent by Subgroup 4M executing the self-terminating encoding component (3.28 seconds) was almost equal to the time spent executing the response component (3.08 seconds) and approximately two seconds more than the time spent executing the confounded inference/application component (see Table 19). In contrast, Subgroup IM spent less time executing the encoding/response component (3.55 seconds) than executing the inference application component (5.60 seconds). In estimating component times for Subgroups NM1 and NM2, the regression equation for Model IM was used. As stated previously, the model for Subgroup NM1 overestimated the time taken to execute the inference/application component. Also, the encoding/response component was equated with zero because negative processing time i s not possible. Comparison of component times for these two 164 subgroups indicates NM2 spent longer executing the encoding component than confounded inference/application, whereas the reverse seemed to be true for Subgroup NM1. Although encoding/response components of Subgroup NM1 were equated with zero, encoding i s regarded as a mandatory process on which analogy solution depends, therefore some time must have been spent on the process. However, because the regression c o e f f i c i e n t was negative and not s i g n i f i c a n t l y d i f f e r e n t from zero, i t was not possible to estimate how much time was spent on t h i s process. The NM2 subgroup appears to have spent a r e l a t i v e l y longer time executing the exhaustive encoding/response component and a r e l a t i v e l y s h o r t e r time executing the exhaustive inference/application component. These estimates should be regarded with caution as both No Model groups have poor model f i t s and appear to have v i o l a t e d the l i n e a r additive r u l e . Summary of Componential Analysis Individual analysis of CV1 data for Group D and Group N indicated four subgroups within Group D. At the i n d i v i d u a l l e v e l , members of two subgroups showed preference for Model 4M and IM respectively while members of the two remaining subgroups l a b e l l e d NM1 and NM2, had no model preference. Members of Subgroup NM1 2 tended to have large values of R , s i g n i f i c a n t values of F, and a large standard error of estimate i n t h e i r regression equations. 165 In addition, many of the s t a t i s t i c a l l y s i g n i f i c a n t parameter estimates were negative. Members of Subgroup NM2 tended to have low 2 . . . values of R , no s i g n i f i c a n t values of F, and few s i g n i f i c a n t component estimates i n t h e i r regression equations. When data were collapsed across subjects, the preferred model for Subgroups 4M and IM continued to emerge. A weak Model IM emerged for Subgroups NM1 and NM2. Similar i n d i v i d u a l analysis of Group N indicated that fourteen members showed preference for Model 4M, one had preference for Model IM and the remainder resembled members of ei t h e r Subgroup NM1 or NM2. When the data were collapsed across subjects, preference for Model 4M continued to emerge. Through t h e i r preferred models, students i n the reading d i s a b i l i t y subgroups 4M and IM and the majority of students i n Group N indicated that they had used a l l Sternberg's theorized components except mapping. The self-terminating mode was used by the majority of students who showed preference for a model i n group D and Group N but the normal reader group had a higher proportion of students p r e f e r r i n g t h i s mode of processing. Subgroup 4M used the l i n e a r combination rule and to a lesser extent so did Subgroup IM and Group N. The poor f i t of the data to the models for Subgroups NM1 and NM2 did not support use of the l i n e a r combination r u l e by eithe r group. Examination of solution and component scores indicated that Subgroups 4M and IM spent less time solving analogies than 166 Subgroups NM1 and NM2. Subgroup 4M spent more time executing the encoding and response components than inference/application whereas Subgroup IM had the shortest encoding/response time. Component scores were calculated for Subgroups NM1 and NM2 using Model IM, but the poor f i t of the data to the model suggests that these scores are i n v a l i d . The Relationship Between Analogical Reasoning and Reading and Reading-Related S k i l l s Cluster Membership Versus Model Subgroup Membership Cluster membership (for Group D) was compared with model subgroup membership through the construction of a two way frequency table (see Table 20) . Members of the c l u s t e r s were d i s t r i b u t e d among the analogy groups i n very s i m i l a r proportions which seems to indicate that for members of Group D, there i s no re l a t i o n s h i p between the processes and strategies used to solve analogies and t h e i r reading and reading-related s k i l l s (Question 8). A Comparison of Reading-Disability Model Subgroups The d i s t r i b u t i o n of scores for the reading and a b i l i t y v ariables f o r each of the Group D subgroups formed on the basis of 167 Table 20. Frequency table between cl u s t e r s and analogy subgroups Analogy subgroups Clusters Total 4M IM NM1 NM2 1 9 4 8 5 26 2 9 2 4 5 20 3 11 6 6 8 31 Total 29 12 18 18 77 168 t h e i r analogical reasoning were examined using Tukey's Box-and-Whisker Plots. The variables were placed i n the same groups and i n the same order as they were for the r e a d i n g - d i s a b i l i t y c l u s t e r s . As before, differences between subgroups were evaluated by the amount t h e i r d i s t r i b u t i o n s overlapped. Subgroups were considered to be more d i f f e r e n t than a l i k e i f : (1) there was no overlap between t h e i r boxes, (2) a l l the scores of one d i s t r i b u t i o n f e l l above or below the median of the other, and (3) neither box overlapped the median of the other. Figure 13 shows the box-plots for the variables of the SDRT Decoding component. None of the analogy subgroups could be distinguished from the others based upon the stated c r i t e r i a . I t was noted that, for each of the four variables, Subgroup IM had the narrowest range of scores for a l l but s t r u c t u r a l analysis involving s y l l a b i c a t i o n and Subgroup 4M had the widest range of scores for a l l but s t r u c t u r a l analysis involving blending. Despite t h i s , Subgroup 4M had l e a s t v a r i a b i l i t y of scores for a l l but phonetic analysis of consonants. There was l i t t l e difference among the subgroups for the SDRT Auditory Vocabulary variable (see Figure 14). Their d i s t r i b u t i o n s had the same median value; the upper half of the d i s t r i b u t i o n of Subgroup 4M extended higher than the others. Subgroup 4M also had the widest range of scores and Subgroup NM2 the narrowest. However, Subgroup NM2 had the most v a r i a b i l i t y i n the scores and Subgroup 169 P h o n e t i c A n a l y s i s ( C o n s o n a n t s ) # c o r r e c t ± P h o n e t i c A n a l y s i s ( V o w e l s ) # c o r r e c t M o d e l G r o u p s M o d e l G r o u p s S t r u c t u r a l A n a l y s i s ( S y l 1 a b i c a t i o n ) S t r u c t u r a l A n a l y s i s ( B l e n d i n g ) 47 c o r r e c t i 8 JM JfM iiMi una. M o d e l G r o u p s # c o r r e c t M o d e l G r o u p s F i g u r e 1 3 . B o x - a n d - w h i s k e r - p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n SDRT d e c o d i n g c o m p o n e n t v a r i a b l e s 17 0 A u d i t o r y V o c a b u l a r y 37 c o r r e c t nr. x X M o d e l G r o u p s F i g u r e 1 4 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n S D R T V o c a b u l a r y c o m p o n e n t v a r i a b l e s 171 NM1 had the l e a s t . There was considerable overlap between the boxes of the subgroup f o r the SDRT factual comprehension va r i a b l e (Figure 15) in d i c a t i n g s i m i l a r i t y i n t h e i r d i s t r i b u t i o n s . Here Subgroup 4M had the widest range of scores and the lea s t v a r i a b i l i t y . V a r i a b i l i t y was s i m i l a r for the other three subgroups. Subgroup IM had the narrowest range of scores. A d i f f e r e n t picture emerged for the SDRT i n f e r e n t i a l comprehension variable. The boxes of Subgroups 4M and NM1 d i d not overlap the medians of either. The i n d i c a t i o n was that members of Subgroup 4M tended to comprehend inference better than members of Subgroup NM1. There was greater v a r i a b i l i t y i n the d i s t r i b u t i o n of Subgroup IM than the other subgroups. The amount of overlap for the SDRT Reading Rate variables (see Figure 16) indicated s i m i l a r i t y among the subgroups. Subgroup 4M had the widest range and most v a r i a b i l i t y of scores and Subgroup NM1 had the narrowest range and lea s t v a r i a b i l i t y f or correct responses. Subgroups NM1 and NM2 had the widest and narrowest ranges, respectively, for Total Responses. Subgroup 4M and NM1 had the most v a r i a b i l i t y and IM and NM2 had the l e a s t . Comparison of the two variables showed that members of Subgroup NM1 tended to make more comprehension errors when reading at speed. Subgroup IM d i f f e r e d from Subgroup 4M on the Oral Reading Time va r i a b l e (see Figure 17) as the boxes of t h e i r d i s t r i b u t i o n s did not overlap the medians of either. On the whole, members of Subgroup 4M took longer to read o r a l l y than members of Subgroup IM. 172 F a c t u a l C o m p r e h e n s i o n /5" # c o r r e c t B 1. I n f e r e n t i a l . C o m p r e h e n s i o n # c o r r e c t ;M JMI urn. M o d e l G r o u p s a* 6 a M o d e l G r o u p s F i g u r e 1 5 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n S D R T R e a d i n g C o m p r e h e n s i o n c o m p o n e n t v a r i a b l e s 173 C o r r e c t R e s p o n s e s T o t a l R e s p o n s e s 3» # c o r r e c t 8 .1.. 3^  # o f r- r e s p o n s e s M o d e l G r o u p s T M o d e l G r o u p s F i g u r e 1 6 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n SDRT R e a d i n g R a t e c o m p o n e n t v a r i a b l e s 174 O r a l R e a d i n g T i m e T i m e i n s e c o n d s 6 S B T S i l e n t R e a d i n g T i m e T i m e i n s e c o n d s *4 M o d e l G r o u p s a M o d e l G r o u p s F i g u r e 1 7 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n D u r r e l l R e a d i n g T i m e v a r i a b l e s 175 Subgroup 4M also had the widest range of scores and Subgroup IM the narrowest. However, Subgroup NM2 had the most v a r i a b i l i t y i n these scores and Subgroup IM had the l e a s t . There was no difference between the subgroups on the S i l e n t Reading Variable. The box-plots for the Du r r e l l Comprehension variables are presented i n Figure 18. Subgroup NM1 tended to do better on the Oral Reading Comprehension variable than Subgroup NM2 as t h e i r boxes do not overlap either of t h e i r medians. The scores of Subgroup NM1 also show more v a r i a b i l i t y although NM2 had one outer low score and two extreme low scores whereas Subgroup NM1 only had one extreme low score. The d i s t r i b u t i o n s of Subgroups IM and 4M are a l i k e and more s i m i l a r to Subgroup NM2 than NM1 on t h i s variable. The subgroup d i s t r i b u t i o n s are s i m i l a r on the S i l e n t Reading Comprehension and Listening Comprehension varia b l e s . The majority of members of each c l u s t e r made no Part-Word Repetition Errors so t h e i r median values were zero (Figure 19). Applying the c r i t e r i a for overlap to the error variables, few differences between the subgroups emerged. Subgroup NM1 made more Whole-Word Repetition errors than Subgroup IM and more Whole-Word Omission errors than Subgroup NM2. Subgroup IM made more Whole-Word Addition errors than NM1. Subgroup 4M exhibited the widest range on a l l but Part-Word Omission and Part-Word Substitution and most v a r i a b i l i t y on a l l except Part-Word Repetition and Part-Word Omission. No one subgroup consistently had the narrowest range or exhibited the le a s t v a r i a b i l i t y o v e r a l l . 176 O r a l R e a d i n g C o m p r e h e n s i o n 13 # c o r r e c t I 1 'XI 1 M o d e l G r o u p s S i 1 e n t R e a d i n g C o m p r e h e n s i o n L i s t e n i n g C o m p r e h e n s i o n # c o r r e c t # c o r r e c t 13 JM 4« M l M o d e l G r o u p s i : -r L o. M o d e l G r o u p s F i g u r e 1 8 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s - o n D u r r e l l C o m p r e h e n s i o n v a r i a b l e s 177 P a r t - w o r d R e p e t i t i o n E r r o r s # o f e r r o r s 3 8 W h o l e - w o r d R e p e t i t i o n E r r o r s B B B 8 8 3 BX BX # o f e r r o r s 8 M o d e l G r o u p s IM *rM A / M I I / M I M o d e l G r o u p s P a r t - w o r d Omi s s i o n E r r o r s # o f e r r o r s I I OK X T' —r~ .1 W h o l e - w o r d Omi s s i o n E r r o r s % B # o f e r r o r s M o d e l G r o u p s 3 r .L M o d e l G r o u p s F i g u r e 1 9 . B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n D u r r e l l O r a l R e a d i n g E r r o r v a r i a b l e s 178 P a r t - w o r d S u b s t i t u t i o n E r r o r s II # o f e r r o r s i J . .0 M o d e l G r o u p s W h o l e - w o r d S u b s t i t u t i o n E r r o r s # o f e r r o r s T 8CL T 8 M o d e l G r o u p s ± P a r t - w o r d A d d i t i o n E r r o r s # o f e r r o r s 8 T B B W h o l e - w o r d A d d i t i o n E r r o r s # o f e r r o r s M o d e l G r o u p s 8 /ri HM tfm AM*. M o d e l G r o u p s F i g u r e 1 9 ( c o n t . ) B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n D u r r e l l O r a l R e a d i n g E r r o r v a r i a b l e s 179 Figure 2 0 indicates no difference among the subgroups on Total Errors, Self-Corrected Errors, and Proportion of Self-Corrected Errors. However, the in d i c a t i o n i n Figure 19 that Subgroup 4M exhibited the widest range and most v a r i a b i l i t y on the majority of error variables was borne out by i t s d i s t r i b u t i o n on the Total Errors v a r i a b l e . The box-plots of the a b i l i t y variables are shown i n Figure 21. The box of Subgroup 4M does not overlap the medians of any of the other subgroups for either of the a b i l i t y v a r i a b l e s . This means that the majority of members of Subgroup 4M had higher CCAT and PPVT scores than members of the other three subgroups. The subgroups had s i m i l a r v a r i a b i l i t y for t h e i r CCAT d i s t r i b u t i o n s but Subgroup NM1 had greater v a r i a b i l i t y i n PPVT scores than the others. Summary of Model Subgroups Few differences were observed among the subgroups. On the SDRT I n f e r e n t i a l Comprehension subtest Subgroup 4M scored higher than Subgroup NM1 but no higher than Subgroups IM or NM2. Members of Subgroup 4M were slower o r a l readers than Subgroup IM. Members of t h i s subgroup made more whole-word addition errors but fewer whole-word r e p e t i t i o n errors than members of Subgroup NM1. Members of Subgroup NM2 made fewer whole-word omission errors and were better comprehenders of or a l reading than members of Subgroup NM1. The 180 T o t a l E r r o r s 5*3 # o f e r r o r s r M o d e l G r o u p s S e l f -C o r r e c t e d E r r o r s IS B # s e l f -c o r r e c t e d P r o p o r t i o n S e l f -C o r r e c t e d E r r o r s 5 $ 1 J . P e r c e n t id B B 3 6 T B .1 M o d e l G r o u p s M o d e l G r o u p s F i g u r e 20. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s o n D u r r e l l O r a l R e a d i n g E r r o r v a r i a b l e s : T o t a l E r r o r s , T o t a l S e l f - C o r r e c t e d E r r o r s , a n d P r o p o r t i o n S e l f -C o r r e c t e d E r r o r s . 181 C C A T P P V T in U n i v e r s a l A g e S c o r e So T. njf. I S c a l e d S c o r e 6s* M o d e l G r o u p s G <v I .1 ox T ) 5 p 1H ftt\l tJ^-u M o d e l G r o u p s F i g u r e 21. B o x - a n d - w h i s k e r p l o t s f o r G r o u p D m o d e l s u b g r o u p s ' ~ o n V e r b a l a n d N o n v e r b a l A b i l i t y v a r i a b l e s 182 omission of words could i n t e r f e r e with the continuity of the passage being read and may have contributed to the lower comprehension scores of Subgroup NM1. Overall, Subgroup 4M tended to have the widest range and most v a r i a b i l i t y i n or a l reading errors and SDRT Reading Rate scores. However, Subgroup 4M also had the lea s t v a r i a b i l i t y i n the SDRT Decoding Variables. Members of Subgroup 4M were slower o r a l readers than members of Subgroup IM who tended to have the narrowest range of scores for the decoding and comprehension varia b l e s . Members of Subgroup 4M had higher nonverbal and verbal a b i l i t y than the other subgroups but generally did not perform any better than the other subgroups on the reading t e s t s . Correlations Between Analogy Data and Reading Data In a study that used the People Piece Analogies, Sternberg (1977) found that solution scores correlated with several reference a b i l i t y scores. In the current study two measures of a b i l i t y and the various reading variables were the reference scores. The dependent or c r i t e r i o n variable of most i n t e r e s t i n the componential analysis part of t h i s study i s C r i t e r i o n Variable 1 (CV1). This variable r e f l e c t s quantity and qu a l i t y of performance i n solving analogies. Each student had twenty-four scores for CV1; one f o r each booklet. A single mean solution score was calculated for each student and correlated with the reading and a b i l i t y 183 variables for Group N and each of the four subgroups, 4M, IM, NM1, and NM2, within Group D. In presenting the c o r r e l a t i o n a l r e s u l t s , reading and reading-rel a t e d variables were arranged into s i x groups. These groups contained: (1) the two a b i l i t y variables represented by the CCAT and the PPVT, (2) the four subtests i n the SDRT Decoding component, (3) the two SDRT Reading Rate variables, (4) o r a l and s i l e n t reading time, (5) the SDRT and D u r r e l l comprehension measures, and (6) the o r a l reading errors. Correlations between solution scores (CV1) and reading and related variables for Group N and Subgroups 4M, IM, NM1, and NM2 are shown i n Table 21. Correlations Within Group N I t was predicted that, for Group N, co r r e l a t i o n s between observed sol u t i o n scores (CV1) and time and error scores would be p o s i t i v e and that correlations between observed sol u t i o n scores and a b i l i t y and accuracy scores correlations would be negative (Hypothesis 9.1). In other words, the more p r o f i c i e n t analogical reasoners would read faster, make fewer errors, have higher a b i l i t y scores and display greater accuracy i n decoding and comprehending. The solution scores did not correlate s i g n i f i c a n t l y with eith e r of the a b i l i t y t e s t s . However, there were s i g n i f i c a n t negative correlations with two of the SDRT Decoding variables (PA-Consonant, SA-Blending), two SDRT Reading Rate variables (Accuracy, 184 Table 21. Correlation of observed solution times (CV1) with  a b i l i t y , reading, and related variables Variable CV1 Groups N 4M IM NM1 NM2 A b i l i t y t e s t s CCAT .00 -.15 -.03 -.19 -.15 PPVT .03 .30 .46 .10 .21 Decoding t e s t s (SDRT) PA (consonants) -.47* .09 -.29 -.27 -.08 PA (vowels) -.24 -.03 -.42 -.12 -.15 SA (syllables) -.24 -.11 .31 . 08 -.33 SA (blending) -.49* . 08 -.55* -.32 .04 Reading Rate (SDRT) Accuracy -.55* -.19 -.44 . 02 -.45 Speed -.50* -.21 .21 . 02 -.44 Reading Time (Durrell) Oral Reading time .66** .21 .23 . 01 -.16 S i l e n t Reading time .64** .22 . 18 . 11 . 08 Comprehension (SDRT) Auditory vocabulary -.01 . 34* . 10 -.06 .27 Factual comprehension -.58** . 00 -.49* -.27 .04 I n f e r e n t i a l comprehension -.50* . 05 -.37 . 24 .04 Comprehension (Durrell) Oral reading -.02 . 11 -.37 . 08 -.54 S i l e n t Reading .16 -.01 -.10 -.36 -.08 Listening Recall -.07 . 07 . 35 -.04 . 10 Oral Reading errors (Durrell) Part-word r e p e t i t i o n -.05 .20 -.06 .24 -.11 Whole-word r e p e t i t i o n .12 . 04 .27 -.21 -.18 Part-word omission -.06 . 04 .34 . 09 .20 Whole-word omission .23 -.15 . 18 . 01 -.10 Part-word su b s t i t u t i o n .01 . 19 .29 -.17 -.05 Whole-word substitution .23 -.03 . 03 -.01 . 01 Part-word addition .09 -.17 .27 . 13 -.10 Whole-word addition .36 . 28 .25 -.08 . 37 Self-corrected errors -.00 .31* . 13 .21 . 02 * = p<.05 **= p<.01 185 Speed) and two SDRT Comprehension variables (Factual, I n f e r e n t i a l ) . These negative correlations indicated that, generally, those who were the fastest, most accurate analogy solvers were better decoders and better comprehenders of d i f f i c u l t as well as easy material. In addition, s i g n i f i c a n t p o s i t i v e c o r r e l a t i o n s between solution scores and Dur r e l l Reading Time variables (Oral, Silent) showed that on the whole, the faster problem solvers were also the fas t e r readers. Therefore, Hypothesis 9.1 cannot be rejected. Lack of s i g n i f i c a n t correlations between solution time and the Dur r e l l comprehension measures may have occurred because of a c e i l i n g e f f e c t for t h i s group. There was no c o r r e l a t i o n between soluti o n time and any of the Dur r e l l Oral reading Errors because there was l i t t l e variance i n t h i s group's error scores. Correlations Within the Model Subgroups Compared to group N, the model subgroups had fewer s i g n i f i c a n t c o r r e l a t i o n s between solution scores and the reading measure (Question 9) . Subgroup NM2 had three s i g n i f i c a n t correlations; Subgroups 4M and IM each had two; and Subgroup NM1 had none. The corr e l a t i o n s seem to imply that those who were more p r o f i c i e n t i n analogical reasoning i n Subgroup 4M sel f - c o r r e c t e d a lower proportion of Dur r e l l o r a l Reading Errors and had lower SDRT Auditory Vocabulary scores; those i n Subgroup IM tended to have higher scores for the decoding variable, SA-Blending and were more 186 accurate factual comprehenders; and those i n Subgroup NM2 were more accurate i n o r a l reading comprehension and did better i n comprehension and speed when the material was r e l a t i v e l y easy. However, these r e s u l t s show l i t t l e consistency among te s t s that measure s i m i l a r s k i l l s and may simply be the r e s u l t of sampling v a r i a b i l i t y . Summary Among a sample of 77 disabled readers, three subtypes were i d e n t i f i e d through c l u s t e r analysis using reading and related v a r i a b l e s . Although the subgroups could be d i f f e r e n t i a t e d on many of the variables a great deal of overlap existed between the c h a r a c t e r i s t i c s of the groups. Componential analysis of analogical reasoning data at the i n d i v i d u a l l e v e l made i t possible to i d e n t i f y the processes and strategies used by some of the normal and disabled readers. Four subgroups were formed by grouping disabled readers who indicated s i m i l a r component and strategy use i n solving analogies. A comparison of disabled readers' strategies with those used by normal readers indicated that the majority solved analogies i n a d i f f e r e n t way. Comparison of reading d i s a b i l i t y c l u s t e r membership with analogy subgroup membership showed no r e l a t i o n s h i p existed. Furthermore, c o r r e l a t i o n a l analysis indicated that for normal readers there i s a r e l a t i o n s h i p between pro f i c i e n c y of analogical reasoning and speed and accuracy i n reading. No s i m i l a r r e l a t i o n s h i p was found for subgroups of disabled readers. 187 CHAPTER VI: DISCUSSION Reading Typology I n i t i a l l y , the purpose of t h i s study was to explore the cognitive processes and strategies used by subtypes of disabled readers i n an analogical reasoning task and to compare these with the processes and strategies used by a group of normal readers. The f i r s t questions asked i n t h i s study concerned the presence of c l e a r l y discernable subtypes of disabled readers i n a n o n c l i n i c a l sample. Methods of subtyping involved the Boder Test of Reading-Sp e l l i n g Patterns - a c l i n i c a l method; and c l u s t e r analysis - a s t a t i s t i c a l method. The Boder t e s t f a i l e d to place the disabled readers into any of the three s p e c i f i c reading d i s a b i l i t y subtypes s p e c i f i e d by the method. However, the s p e l l i n g patterns of these subtypes were found among some members of the reading-disabled group suggesting that milder forms of the subtypes e x i s t . Others were s i m i l a r to Boder's nonspecific reading d i s a b i l i t y subtype. Using a h i e r a r c h i c a l agglomerative technique employing Ward's algorithm for minimum variance with data obtained from the reading variables, the normal reader group (N=20) was c l e a r l y distinguishable from the disabled readers group (N=77). Three subtypes were i d e n t i f i e d among the reading-disabled students. As i n other subtyping studies, three c l u s t e r s of students within Group 188 D were described i n terms of the variables that were used to di s t i n g u i s h them. To some extent, the members of a l l three clu s t e r s had poor decoding s k i l l s , were slow readers, made many or a l reading errors, and even when the material was easy, used i n e f f i c i e n t reading strategies. These cl u s t e r s were characterized by the degree to which they were d e f i c i e n t on these v a r i a b l e s . There was considerable overlap between them. This i s not unique. Most empirical studies have produced subtypes that have shared c h a r a c t e r i s t i c s (e.g. Lyon, Stewart, & Freedman, 1982; McKinney, Short, & Feagans, 1985; Speece, Mckinney, & Appelbaum, 1985). As discussed i n a review of the l i t e r a t u r e , reading typology research has raised quite substantial methodological issues. In seeking to discover i n t r i n s i c differences i n reading s k i l l s as a source of reading d i s a b i l i t y t h i s study has been as successful as other methods. Students were placed i n cl u s t e r s according to s i m i l a r i t i e s i n reading and related s k i l l s that are t y p i c a l l y measured i n school settings. However, i n seeking the underlying cause of each c l u s t e r ' s predominating d e f i c i t , the method was no more successful than any other used i n subtyping research. Componential Analysis Componential analysis was used to determine in d i v i d u a l differences i n information processing among the group of normal readers (Group N) and the group of disabled readers (Group D). 189 S p e c i f i c a l l y , questions were asked as to which components would be used by disabled and normal readers i n solving schematic picture analogies, which strategy would they prefer, and to what extent they would adhere to the l i n e a r combination r u l e . Additional questions concerned i d e n t i f i c a t i o n of subgroups within the reading disabled sample based upon in d i v i d u a l use of components and strategies, and the rel a t i o n s h i p between membership i n these subgroups and membership i n the three c l u s t e r s . Regression Models: Components and Strategies At the group l e v e l , the re s u l t s of the present study support some of the findings of previous research (Sternberg & Ketron, 1982; Sternberg & R i f k i n , 1979; Wilson, 1980). However, generalizations are made with caution as the sample used i n the present study d i f f e r e d from the samples used i n other studies. In the studies of Sternberg and R i f k i n (1979) and Wilson (198 0) , groups were formed on the basis of age or a b i l i t y and componential analysis was ca r r i e d out and reported at the group l e v e l . Therefore comparisons with previous research are made at the group rather than the in d i v i d u a l l e v e l . For some groups, t h i s study revealed inconsistencies between analyses at the group and in d i v i d u a l l e v e l that allow speculations to be made regarding previous research. The studies of Sternberg and R i f k i n (1979) and Wilson (1980) had shown, at the group l e v e l , that the preferred model used i n 190 solving separable-attribute analogies was generally a modified (no mapping component), self-terminating one (Model 4M) and that t h i s preference became more consistent through childhood to adulthood and increased with l e v e l of a b i l i t y . Only Wilson's low a b i l i t y group (Wilson, 1980) did not show preference for Model 4M but marginally preferred a modified exhaustive model (IM). In the present study, analysis at the i n d i v i d u a l l e v e l showed that, within Group N (n=20), 14 members preferred Model 4M and one member preferred Model IM. The remainder of the group indicated no preference for any of the t h e o r e t i c a l models. Within group D (n=77), 29 members preferred Model 4M, 12 preferred Model IM, and 3 6 indicated no model preference. Within t h i s group of 36, two separate groups could be discerned. One group, l a b e l l e d Subgroup NM1 (No Model 1) , consisted of 18 students who had s i g n i f i c a n t l e v e l s of regression F but very bad f i t of t h e i r data to a model. The other group, l a b e l l e d Subgroup NM2 (No Model 2) consisted of 18 students who had no l e v e l s of s i g n i f i c a n c e i n t h e i r regression equations. Subgroups were formed within Group D, based upon ind i v i d u a l model preference. The data were then collapsed across groups and group regression equations obtained. The preferred group models for Subgroups 4M and IM were stronger than any of t h e i r preferred models at the i n d i v i d u a l l e v e l . Their regression equations showed that they appeared to have used the same t h e o r e t i c a l component processes as Sternberg and R i f k i n ' s (1979) and Wilson's (1980) 191 groups and to have adhered to the l i n e a r combination rule, although Subgroup 4M had the stronger model. The model that best f i t the data for Subgroup NM1 resembled the exhaustive Model IM but had a nonsignificant encoding/response component. The preferred model for Subgroup NM2 was c l e a r l y Model 2 IM but the value of R indicated that the model accounted for only 20 percent of v a r i a b i l i t y i n the data. Data analysis at the i n d i v i d u a l l e v e l did not support the r e s u l t s of analysis at the group l e v e l for these two subgroups. Their i n d i v i d u a l regression equations were unable to show which of the t h e o r e t i c a l components were used and suggested that the l i n e a r combination rule was probably v i o l a t e d . The preferred model of Group N was Model 4M even though six members of Group N did not indicate preference for t h i s model at the i n d i v i d u a l l e v e l . Group N most resembled Wilson's Average A b i l i t y group which consisted of 20 average achieving students. In fact, t h e i r group regression equations were very s i m i l a r . Both had s i g n i f i c a n t regression c o e f f i c i e n t s (p<.05) for the s e l f -terminating inference/application components and small, nonsignificant regression c o e f f i c i e n t s for the self-terminating 2 encoding component. They had lower values for R than any of Sternberg and Rif k i n ' s (1979) groups or of Wilson's (1980) High A b i l i t y group. These regression equation s i m i l a r i t i e s suggest that Wilson's average group may also have contained students who did not prefer Model 4M. This could have caused the regression equation to 192 2 have lower values of R and regression F and the encoding component to have a nonsignificant regression c o e f f i c i e n t . The group regression equation of Subgroup 4M within Group D was s i m i l a r to Sternberg and Rifk i n ' s Grade 4 and Adult samples and s i m i l a r to, but stronger than, Wilson's high achieving group and Sternberg and Ri f k i n ' s Grade 2 and Grade 6 samples. The higher a b i l i t y of Wilson's group suggests that they should have had a stronger model. One explanation may be that Wilson's high achieving group and Sternberg and Ri f k i n ' s Grade 1 and 6 samples contained members who, at the in d i v i d u a l l e v e l , used a model other than 4M. The other Subgroups IM, NM1, and NM2 a l l showed preference for Model IM at the group l e v e l , although NM1 and NM2 had weak models. As Subgroup IM was made up of students who had a l l shown preference for Model IM at the in d i v i d u a l l e v e l , the group model was n a t u r a l l y ' stronger. The members of Wilson's low achieving group, who also showed preference for Model IM, had a group regression equation that, i n comparison, was much weaker than that of Subgroup IM. Both groups had s i g n i f i c a n t regression c o e f f i c i e n t s for the confounded inference/application and encoding/response components but, compared to Wilson's low group, Subgroup IM had a 2 higher R and regression F. In fact, the regression equation of Wilson's low a b i l i t y group bore more resemblance to that of Subgroup NM2. Subgroup NM2 had s i g n i f i c a n t regression c o e f f i c i e n t s for the two confounded 2 components but an R that accounted for only 20 percent of the 193 v a r i a b i l i t y i n scores and a small regression F which was s i g n i f i c a n t at p<.01. This group was made up of students whose in d i v i d u a l regression equations, for most of the models, had few s i g n i f i c a n t regression c o e f f i c i e n t s , and no s i g n i f i c a n t values of regression F even at p<.10 l e v e l of s i g n i f i c a n c e . This suggests that, where preference for a model i s marginal at the group l e v e l , the members may not indicate a preference for any model at the i n d i v i d u a l l e v e l . Component Scores Calculation of component scores for reading-disabled subgroups indicated that Subgroup 4M spent more time executing the encoding component than the inference/application component. This supports the Sternberg and R i f k i n ' s (1979) for a l l age l e v e l s , except Grade 4, and Wilson's (1980) high a b i l i t y group. Subgroup 4M also took longer to execute the self-terminating encoding component than the response component. This also follows the pattern for Sternberg and Ri f k i n ' s Grade 2, Grade 64, and Adult groups, - t h e i r Grade 4 group and a l l of Wilson's groups had response time that was longer than encoding time. Sternberg (1977) showed that more successful adult reasoners took longer to encode s t i m u l i than unsuccessful ones and suggested that there was a trade-off between encoding speed and the speed of performing l a t e r operations. The pattern of component times for Subgroup IM cannot be 194 compared d i r e c t l y with any of Sternberg and Ri f k i n ' s or Wilson's groups as the component scores for t h e i r groups were calculated using t h e i r Model 4M regression equations. Also i t i s not possible to say exactly how much time Subgroup IM spent executing the encoding component as i t i s confounded with the response component i n Model IM. Subgroup IM must have spent less time executing the encoding component than the inference/application component as the encoding time confounded with response time was less than the time taken to execute the inference application component. Thus members of t h i s subgroup did not use the trade-off between encoding speed and the speed of performing l a t e r operations. I f they had, they might have discovered that they did not need to process the components exhaustively. The Relationship Between Analogical Reasoning and Reading Members of Group N were average to good readers and had adequate verbal and nonverbal a b i l i t y . Seventy percent of them showed preference for a self-terminating strategy to solve the schematic picture analogies. Their preferred model at the group l e v e l also indicated a self-terminating strategy which i s consistent with previous research (Sternberg & R i f k i n , 1979; Sternberg & Ketron, 1982; Wilson, 1980). In contrast, only 38 percent of the members of Group D showed preference for a s e l f -terminating strategy and ju s t under h a l f (47%) did not use any of 195 the t h e o r e t i c a l models to solve the schematic picture analogies. This suggested that something i n the nature of t h e i r reading d i s a b i l i t y might be responsible. Examination of the frequency table between c l u s t e r s and analogy subgroups (Table 19) indicated that there was no r e l a t i o n s h i p between membership i n the reading c l u s t e r s and the analogy subgroups. Box-and Whisker plots indicated few differences between the subgroups on the reading and related v a r i a b l e s . In general, Subgroup 4M had the widest range of scores and Subgroup IM had the narrowest, demonstrating that Subgroup IM was probably more homogenous than 4M. Only one trend was consistent with previous research; i t indicated that those students who showed a preference for Model 4M tended to have higher verbal and nonverbal a b i l i t y as measured by the PPVT and CCAT. This suggests that Subgroup 4M were the more reading-disabled subgroup for, despite t h e i r higher a b i l i t y , they did not have better reading s k i l l s . Although the solution scores of Group N did not correlate s i g n i f i c a n t l y with a b i l i t y scores, c o r r e l a t i o n a l analysis of solu t i o n scores with reading variables produced several s i g n i f i c a n t c o r r e l a t i o n s which were generally i n the predicted d i r e c t i o n . In other words, those i n Group N who were more p r o f i c i e n t at solving analogies were generally more p r o f i c i e n t readers i n terms of decoding s k i l l s , speed of reading and comprehension. Indeed, no r e l a t i o n s h i p was c l e a r l y discernable for any of the reading-disabled subgroups. 196 Factors Underlying Strategy Choice Sternberg and Ketron (1982) were unsuccessful i n t h e i r attempts to t r a i n u n i v e r s i t y students to use a strategy other than a self-terminating one. They attributed t h i s strong compulsion to the e f f i c a c y and e f f i c i e n c y of the strategy i n terms of speed and accuracy, ease of use and maintenance, and smaller memory load. Yet 12 students who made up Subgroup IM chose to use an exhaustive strategy that was less e f f i c i e n t than a self-terminating one and 36 students rejected a l l Sternberg's (1977) t h e o r e t i c a l models i n favour of some other, less appropriate strategy. The p o s s i b i l i t y exists that strategy use i s d i r e c t l y related to i n t e l l e c t u a l a b i l i t y . Subgroup 4M had higher verbal and nonverbal scores than the other three subgroups but members of the other three subgroups also had scores that were above the PPVT and CCAT means of 100, as did some of those members of Group N who did not appear to choose a model. This would appear to rule out i n t e l l e c t u a l a b i l i t y as the only factor i n strategy choice and raises the p o s s i b i l i t y that a factor other than i n t e l l e c t u a l a b i l i t y was also involved. What Subgroups IM, NM1, and NM2, as well as Wilson's (1980) low a b i l i t y group have i n common i s low achievement. Wilson's a b i l i t y groups were chosen according to t h e i r achievement l e v e l s , on the assumption that achievement indicates a b i l i t y . The low achievement of learning-disabled students i s , by d e f i n i t i o n , not 197 i n d i c a t i v e of low i n t e l l i g e n c e and Wilson's low a b i l i t y sample may have contained some students with higher a b i l i t y than t h e i r achievement l e v e l s would suggest. I t may be that there are factors at work which in t e r a c t with a b i l i t y to influence not only academic achievement but the extent to which strategy choice i s e f f i c i e n t and appropriate. I t may be that for Group D the factor or factors underlying strategy choice were related i n some way to the c h a r a c t e r i s t i c s exhibited by disabled readers that were not d i r e c t l y measured i n t h i s study, such as short-term memory or metacognition. In addition, the factors may be influenced by motivational and/or personality c h a r a c t e r i s t i c s . Metacognition i s regarded as a conscious p l a n f u l , evaluative, decision-making process (Anderson, 1975; Brown, 1980; F l a v e l l , 1978) . The students i n t h i s study appeared to be using metacognitive strategies i n solving analogies as a l l of them were eager to share t h e i r strategies with the examiner and had to be encouraged to wait u n t i l the t e s t i n g sessions were over. This suggests that many of the students did not lack metacognitive strategies, but that t h e i r strategies were i n e f f i c i e n t and inappropriate. Nor does lack of motivation appear to have influenced strategy choice as the examiner noted that a l l students were interested and enthusiastic i n t h e i r approach to the analogies t e s t . K o l l i g i a n and Sternberg (1987) suggested a l i n k might exi s t 198 between i n e f f i c i e n t strategy use and i n f l e x i b l e response s t y l e , giving as an example the i n e f f i c i e n t use of speed. Subgroup IM, whose members chose a less e f f i c i e n t strategy which required more processing than a self-terminating one, nevertheless completed the analogies i n the shortest time. They had twice the error rate of Subgroup 4M and appeared to have s a c r i f i c e d accuracy for speed. 2 . Wilson (1980) had also suggested that a low R i n the regression equation might be caused by a v i o l a t i o n of the l i n e a r additive rule or by a missing component such as response speed. She reasoned that the errors i n her normal and low a b i l i t y group might be caused by impulsive responding i n which accuracy was s a c r i f i c e d f or speed. Re s t r i c t i o n s of working memory have been implicated i n reading d i s a b i l i t i e s (Jorm, 1983; Torgesen, 1977) and i n e f f i c i e n t strategy use ( K o l l i g i a n & Sternberg, 1987). Members of Subgroup NM1 took the most time to solve analogies and had the same error rate as Subgroups IM and NM2. Individual analysis indicated that members of t h i s group had probably not processed components i n a s e r i a l fashion. I t i s possible that members of t h i s group may have attempted to use an exhaustive strategy and found that, because of d e f i c i e n t working memory, t h i s strategy could not be maintained for the more d i f f i c u l t analogies. They may have then resorted to a h o l i s t i c strategy that caused them to take longer than the other groups and make more errors. In h i s t r i a r c h theory of i n t e l l i g e n c e , Sternberg (1987) theorized that strategies are under the control of the executive 199 or metacomponents. However, K o l l i g i a n and Sternberg (1987) suggested that, where d e f i c i e n t working memory i s unable to hold metacomponential information, i n e f f i c i e n t or incomplete strategies are used. I t i s therefore possible to speculate that one c h a r a c t e r i s t i c of members of Subgroup NM1 may have been a d e f i c i t i n working memory that allowed inappropriate strategies to be used. There appears to have been no r e l a t i o n s h i p between the amount of processing required i n the solution of the analogies and the time taken by Subgroup NM2. Although members of t h i s group completed the t o t a l number of analogies at the same rate as Group 4M, they made twice as many errors. This suggests that t h e i r strategy might be associated with impulsive reasoning i n which accuracy was s a c r i f i c e d for speed, or i t might be associated with some other factor such as guessing. Limitations of the Study The l i m i t a t i o n s of t h i s study concern the extent to which the r e s u l t s can be compared with previous research or generalized to other populations. Comparability and g e n e r a l i z a b i l i t y depend upon the nature of the sample and the methodology employed. The sample of reading-disabled students used i n t h i s study did not resemble the c l i n i c a l samples used by many researchers (e.g. Doehring et a l . , 1981; Petruskas & Rourke, 1979), the s c h o o l - i d e n t i f i e d samples of the longitudinal studies (e.g., Lyon & Watson, 1981; Mckinney, 200 1984), or students from high SES neighbourhoods with IQ scores of 90 and above (e.g., V e l l u t i n o , 1979, 1980). Some of the reading-disabled students i n Group D resembled c l i n i c a l subjects i n that t h e i r reading d i f f i c u l t i e s had been recognized and t h e i r educational programs modified. The remainder were receiving i n s t r u c t i o n i n the regular program; some were recognized by t h e i r teachers as achieving below the l e v e l of t h e i r peers; others were not. The sample did not contain a l l possible c l i n i c a l subjects or below-average readers i n the Grade 5 population because some students did not take part i n the screening phase of the study eith e r through absence, f a i l u r e to complete the te s t s , or lack of parental permission. Permission to take part i n the second phase of the study was not given for some students who met the c r i t e r i a for membership i n Group D. No reason was asked for but i n two cases parents reported that they f e l t t h e i r c h i l d had been subjected to enough t e s t i n g . In three other cases parents indicated that t h e i r c h i l d refused to take part i n a study which emphasized reading because t h i s was an a c t i v i t y the caused the c h i l d a great deal of stress and anxiety. The constraints of time and resources make i t necessary to screen large populations with group administered t e s t s . The SDRT was chosen for the present study because of i t s diagnostic a b i l i t y and the CCAT was chosen to measure nonverbal a b i l i t y because i t had been normed i n Canada. There are students who do not perform well 201 on group administered t e s t s so that i t i s possible some disabled readers were d i s q u a l i f i e d by the i n t e l l i g e n c e c r i t e r i a and others included who would not have been i d e n t i f i e d as disabled readers had i n d i v i d u a l l y administered reading t e s t s been used. Conclusion Reading typology provided a way of grouping disabled readers into s p e c i f i c subtypes according to c l i n i c a l p r o f i l e s , reading performance, or patterns of cognitive development. This makes i t possible to give a single diagnostic l a b e l to a l l the students who belong to a p a r t i c u l a r subtype. However, the reading typology l i t e r a t u r e has raised some quite substantial methodological issues, chief of which i s the f a i l u r e to look at the student at the in d i v i d u a l l e v e l . The c h a r a c t e r i s t i c s which are t y p i c a l of the group are not necessarily the same as those which are s p e c i f i c to the i n d i v i d u a l . The unique contribution of t h i s study i s that i t has offered a t o t a l l y d i f f e r e n t way of looking at reading d i s a b i l i t i e s . I t has used a t r a d i t i o n a l subtyping approach combined with a c l i n i c a l experimental technique. Thus subtypes were i d e n t i f i e d using a conventional method, then members of each subtype were examined one by one using a task that has been proven as a good predictor of a b i l i t y . This i s a technique that has implications for programming i n terms of devising a remedial program t a i l o r e d not only for a 202 group but for an i n d i v i d u a l . Not only did the method provide another way of looking at members of previously i d e n t i f i e d groups at the i n d i v i d u a l l e v e l but i t provided an additional method of subtyping through the analogical reasoning task. I t allowed subtyping to be approached through a d i f f e r e n t avenue and provided information about mental processing and strategy use that could not be obtained by conventional d i f f e r e n t i a l and information processing methodologies. This study has shown that the majority of a group of reading-disabled students did not use the same processes and strategies i n analogical reasoning as a group of normal readers. The subgroup of reading-disabled students who used the most e f f i c i e n t and appropriate strategy tended to have higher verbal and nonverbal a b i l i t y . However, a b i l i t y alone could not account for the use of i n e f f i c i e n t and inappropriate strategies nor was i t able to explain the lack of r e l a t i o n s h i p between analogical reasoning components and strategies and reading s k i l l s . This lack of r e l a t i o n s h i p between the two systems i s perhaps the most su r p r i s i n g and paradoxical finding of the study as there are many precedents i n the l i t e r a t u r e that l i n k neuropsychological processes to reading d i s a b i l i t i e s . I t may be that what was being measured through the reading variables and analogical reasoning variables impinged i n a d i f f e r e n t way on the reading a b i l i t y of the reading-disabled children. In other words, reading disabled childr e n may have struggled with c e r t a i n of the reading tasks and 203 then have gone on to solve the analogies i n a unique way. This, i n turn, may be explained by the nature of the tasks. Sternberg and Spear (1983) regard decoding and reading comprehension as bottom-up processing and analogical reasoning as top-down processing. Content-driven or bottom-up processing i s dependent upon the content of the material being processed whereas top-down processing i s content free and dependent upon the c h a r a c t e r i s t i c s of the processor. Thus the reading s k i l l s being measured were large l y dependent upon the l e v e l of d i f f i c u l t y of the words to be decoded i n each reading task whereas the analogical reasoning task was r e l a t i v e l y novel and the processes and strategies used were more dependent upon the inherent c h a r a c t e r i s t i c of the student. Another possible explanation suggests that the subtypes formed on the basis of strengths and weaknesses i n reading s k i l l s may have represented measures at a macro l e v e l whereas the componential analysis may have represented measures of processing at a micro l e v e l , rather as a telescope reveals the gross anatomy of the wing of a b i r d and the microscope reveals minute d e t a i l s of the structure of an i n d i v i d u a l feather. In addition, component processing may have been so s p e c i f i c to the i n d i v i d u a l that differences i n processing between indivi d u a l s were buried i n a l l the reading d i s a b i l i t y subtypes. This implies the existence of subtypes within subtypes. 204 Suggestions for Future Research This study was unique i n that i t combined a t r a d i t i o n a l subtyping methodology with the method of componential analysis and looked at students at the group and i n d i v i d u a l l e v e l . However, missing from the study was a c l i n i c a l sample of more severely disabled readers. Most subtyping research has been c a r r i e d out with more severely disabled readers, therefore i t i s suggested that the methodology used here be extended to a c l i n i c a l sample. Research has indicated that a self-terminating strategy i s most e f f i c i e n t and the one most l i k e l y to be used i n solving schematic picture analogies. However, the majority of reading-disabled students did not choose the most e f f i c i e n t strategy; some chose an exhaustive strategy and others chose strategies that contravened the assumption of l i n e a r i t y and v i o l a t e d the additive factor r u l e . No r e l a t i o n s h i p was established between membership in a reading d i s a b i l i t y c l u s t e r and strategy choice i n analogical reasoning. I t has been suggested that t h i s lack of r e l a t i o n s h i p may be associated with the bottom-up versus top-down nature of the two tasks. Perhaps the use of verbal analogies could provide content more i n keeping with a reading task and thereby reveal a r e l a t i o n s h i p that picture analogies were not able to do. K o l l i g i a n and Sternberg (1987) have suggested that d e f i c i e n c i e s i n executive or metacomponents are not involved i n reading d i s a b i l i t i e s as t h i s would a f f e c t a l l aspects of i n t e l l e c t u a l functioning. However, i n 205 t h i s study, the majority of students appeared to be using metacognitive processes that were found to be linked to i n e f f i c i e n t and inappropriate strategies. This l i n k may be mediated by a deficiency i n some area that also i n t e r f e r e s with the reading process. K o l l i g i a n and Sternberg 1987) theorized that d e f i c i e n t working memory was the l i n k between metacomponential functioning and reading d i s a b i l i t i e s . They suggested that d e f i c i e n t working memory i s unable to hold metacomponential information and allows i n e f f i c i e n t or inappropriate strategies to be used. This i s another area which could u s e f u l l y be explored. REFERENCES Anderson, T. H. (1980). Study strategies and adjunct aids. In R. J. Spiro, B. B. Bruce, & W. F. Brewer (Eds.), Theoretical issues  i n reading comprehension (pp. 484-502). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Applebee, A. M. (1971). Research i n reading retardation: Two c r i t i c a l problems. Journal of Child Psychiatry and A l l i e d  D i s c i p l i n e s . 12, 91-113. Bannatyne, A. (1974) . A note on the recategorization of WISC scaled scores. Journal of Learning D i s a b i l i t i e s . 7, 272-274. Benton, A. L. (1978). Some conclusions about dyslexia. In A. L. Benton & D. Pearl (Eds.), Dyslexia: An appraisal of current  knowledge (pp. 451-476). New York: Oxford University Press. Blachman, B. A. (1983). Are we assessing the l i n g u i s t i c factors c r i t i c a l i n early reading? Annals of Dyslexia, 33, 91-109. Boder, E. (1971). Developmental dyslexia: P r e v a i l i n g diagnostic concepts and a new diagnostic approach. In H. R. Myklebust (Ed.), Progress i n learning d i s a b i l i t i e s (Vol. 2, pp. 293-321). New York: Grune & Stratton. Boder, E. (1973). Developmental dyslexia: A diagnostic approach based on three a t y p i c a l reading patterns. Developmental Medicine  and C h i l d Neurology, 15, 663-687. Boder, E., & J a r r i c o , S. (1982). The Boder Test of Reading Sp e l l i n g  Patterns. New York: Grune & Stratton. Borkowski, J . G., Johnston, M. B., & Reid, M. K. (1987). Metacognition, motivation, and controlled performance. In S. Ceci (Ed.), Handbook of cognitive, s o c i a l , and neurological aspects  of learning d i s a b i l i t i e s (Vol I I , pp. 147-174). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Bradley, L. & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connection. Nature, 301, 419-421. Bradley, L. & Bryant, P. E. (1985). Rhyme and reason i n reading and  s p e l l i n g . Ann Arbor, MI: University of Michigan Press. Brady, S., & Fowler, A. E. (1988). Phonological precursors to reading a c q u i s i t i o n . In R. L. Masland & M. W. Masland (Eds.), Prevention of reading f a i l u r e (pp. 204-215). Parton, MD: York Press. 207 Brady, S., Shankweiler, D., & Mann, V. (1983). Speech perception and memory coding i n r e l a t i o n to reading a b i l i t y . Journal of  Experimental C h i l d Psychology, 35, 345-367. Brown, A. L. (1980) Metacognitive development and reading. In R. J . Spiro, B. B. Bruce, & W. F. Brewer (eds.), Theoretical issues  i n reading comprehension (pp. 456-481). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Calfee, R. C. (1977). Assessment of independent reading s k i l l s : Basic research and p r a c t i c a l applications. In A. S. Reber & H.S. Scarborough (Eds.), Towards a psychology of reading (pp. 289-323). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Calfee, R. C., & Drum, P. A. (1978). Learning to read: Theory, research, and practice. Curriculum Inquiry, 8, 183-249. Calfee, R. C., & Spector, J . E. (1981). Separable processes i n reading. In P. Satz & R. Morris (Eds.), Neuropsychological and  cognitive processes i n reading (pp. 3-29). New York: Academic Press. Denckla, M. B. (1977). Minimal brain dysfunction and dyslexia: Beyond diagnosis by exclusion. In M. E. Blaw, J . Rapin, & M. Kinsbourne (Eds.), Topics i n c h i l d neurology (pp. 243-261). New York: Spectrum. Denckla, M. B., Rudel, R. G. (1976). Rapid automatized naming (R.A.N.): Dyslexia d i f f e r e n t i a t e d from other learning d i s a b i l i t i e s . Neuropsychologica. 14. 471-479. Doehring, D. G. (1976). Ac q u i s i t i o n of rapid reading responses. Monographs of the Society for Research i n Child Development, 41, (#165). Doehring, D. G., & Hoshko, I. M. (1977). C l a s s i f i c a t i o n of reading problems by the Q-technique of factor analysis. Cortex, 13, 281-294. Doehring, D. G. , Hoshko, I. M. , & Bryans, B. N. (1979). S t a t i s t i c a l c l a s s i f i c a t i o n of children with reading problems. Journal of  C l i n i c a l Neuropsychology, 1, 5-16. Doehring, D. G., T r i t e s , R. L., Patel, P. G., & Fiedorowicz, C. A. M. (1981). Reading D i s a b i l i t i e s . New York: Academic Press. Dorman, C. (1987). Reading d i s a b i l i t y subtypes i n neurologically-impaired students. Annals of Dyslexia, 37, 166-188. Downing, J . , & Leong, C. K. (1982). Psychology of reading. New York: MacMillan. 208 Dunn, L. M. , & Dunn, L. M. , (1981). Peabody Picture Vocabulary Test - Revised. C i r c l e Pines, MN: American Guidance Service. D u r r e l l , D. D., & Catterson, J . H. (1980). D u r r e l l Analysis of  Reading; D i f f i c u l t y . New York: Harcourt, Brace, & Janovich. E h r i , L. (1979). L i n g u i s t i c insight: Threshold of reading a c q u i s i t i o n . In G. E. MacKinnon & T. G. Waller (Eds.), Reading  research: Advances i n theory and practice (Vol. 2, pp. 63-114). New York: Academic Press. E l l i s , A. W. (1985). The cognitive neuropsychology of developmental (and acquired) dyslexia: A c r i t i c a l survey. Cognitive  Neuropsychology, 2, 169-205. E l l i s , N. (1981). Visual and name coding i n dyslexic children. Psychological Research. 43., 102-218. E v e r i t t , B. (1980). Cluster analysis. London: Heinemann Educational Books. Feagans, L., & Appelbaum, M. I. (1986). V a l i d a t i o n of language subtypes i n learning disabled children. Journal of Educational  Psychology, 78., 358-364. Feagans, L., & McKinney, J . D. (1981). Patterns of exceptionality across domains i n learning-disabled children. Journal of Applied  Developmental Psychology, 1, 313-328. Fisk, J . L., & Rourke, B. P. (1979). I d e n t i f i c a t i o n of subtypes of learning-disabled children at three age l e v e l s : A neuropsychological, multivariate approach. Journal of C l i n i c a l  Neuropsychology, 1, 289-310. F l e i s s , J . L., Lawlor, W., Platman, S. R., & Fieve, R. R. (1971). On the inverted factor analysis for generating of typologies. Journal of Abnormal Psychology, 77, 127-132. Fox, B., & Routh, D. K. (1980). Phonemic analysis and severe reading d i s a b i l i t y i n children. Journal of Psycholinguistic  Research. 9, 115-119. Fox, B., & Routh, D. K. (1983). Reading d i s a b i l i t y , phonemic analysis, and dysphonetic s p e l l i n g : A follow-up study. Journal  of C l i n i c a l C hild Psychology, 12, 28-32. Fox, B., & Routh, D. K. (1984). Phonemic analysis and synthesis as word attack s k i l l : Revisited. Journal of Educational Psychology, 76, 1059-64. 209 Frederiksen, J . R. (1978). Assessment of perceptual, decoding and l e x i c a l s k i l l s and t h e i r r e l a t i o n to reading pr o f i c i e n c y . In A. M. Lesgold, J . W. Pellegrino, S. Fokkema, & R. Glasser (Eds.), Cognitive psychology and i n s t r u c t i o n (pp. 152-169). New York: Plenum Press. Gadow, K. D., & Swanson, H. L. (1986). Assessing drug e f f e c t s on academic performance. In K. D. Gadow (Ed.), Methodological issues  i n human psychopharmacology (pp. 247-279). Greenwich, CT: JAI Press. Godfrey, J . J . , Syndal-Lasky, A. K., M i l l a j , K. K., & Knox, C. M. (1981). Performance of dyslexic children on speech perception t e s t s . Journal of Experimental Child Psychology. 32, 401-424. Goodman, K. S. (1968). The psycholinguistic nature of the reading process. Detroit, OH: Wayne State University Press. Gordon, H. W. (1984). Dyslexia. In R. E. Tarter & G. Goldstein (Eds.), Advances i n c l i n i c a l neuropsychology (Vol 2, pp. 181-205). New York: Plenum Press. Harris, A. J . (1981). How many kinds of reading d i s a b i l i t y are there? Journal of Learning D i s a b i l i t i e s , 15, 456-460. Hynd, G. , & Cohen, M. (1983). Dyslexia: Neuropsychological theory,  research and c l i n i c a l d i f f e r e n t i a t i o n . New York: Grune & Stratton. Ingram, T. T. S., Mason, A. W., & Blackburn, I. (1970). A retrospective study of 82 children with reading d i s a b i l i t y . Developmental Medicine and Child Neurology. 12., 271-281. Jackson, M., & McLelland, J . (1979). Processing determinants of reading speed. Journal of Experimental Psychology: General, 108, 151-181. Jansky, J . , & DeHirsch, K. (1972). Preventing reading f a i l u r e :  Prediction, diagnosis, and intervention. New York: Harper & Row. Jastak, J . , & Jastak, S. (1978). Wide Range Achievement Test. Wilmington, DE: Jastak Associates. Johnson, D., & Myklebust, H. (1967). Learning d i s a b i l i t i e s : Educational p r i n c i p l e s and practices. New York: Grune & Stratton. Jorm, A. F. (1983). S p e c i f i c reading retardation and working memory: A review. B r i t i s h Journal of Psychology, 74., 311-342. 210 Jorm, A. F., & Share, D. L. (1983). An i n v i t e d a r t i c l e : Phonological reading and reading a c q u i s i t i o n . Applied  Psycholinguistics, 4, 103-147. Karlsen, B., Madden, E., & Gardner, E. F. (1976). Stanford Diagnostic Reading Test (Brown Level). New York: Harcourt, Brace, & Janovich. Kavale, K. A., & Forness, S. R. (1987). The f a r side of heterogeneity: A c r i t i c a l analysis of subtyping research i n learning d i s a b i l i t i e s . Journal of Learning D i s a b i l i t i e s , 20, 374-382. Kaufman, A. S. (1975). Factor analysis of the WISC-R at 11 age le v e l s between 6 1/2 and 16 1/2 years. Journal of Consulting and  C l i n i c a l Psychology, 43, 135-147. Kinsbourne, M., & Warrington, E. K. (1963). Developmental factors i n reading and writing backwardness. B r i t i s h Journal of  Psychology, 54, 145-156. K o l l i g i a n , J . , & Sternberg, R. J . (1987). Intelligence, information processing, and s p e c i f i c learning d i s a b i l i t i e s : A t r i a r c h synthesis. Journal of Learning D i s a b i l i t i e s , 20, 8-17. LaBerge, D., & Samuels, S. J . (1974). Toward a theory of automatic information processing i n reading. Cognitive Psychology, 6, 193-323. L a i , C. (1982). Hier a r c h i c a l grouping analysis with optional  contiguity constraint: Program manual. Computing Services, University of B r i t i s h Columbia. Lawson, M. L. (1984). Being executive about metacognition. In R. J . Kirby (Ed.), Cognitive strategies and educational performance (pp. 89-109). Orlando, FL: Academic Press. Liberman, I. Y., & Shankweiler, D. (1985). Phonology and the problems of learning to read and write. Remedial and Special  Education. 6, 8-17. Liberman, I. Y., Mann, V. A., Shankweiler, D., & Werfelman, M. (1980). Children's memory for recurring l i n g u i s t i c and n o n l i g u i s t i c material i n r e l a t i o n to reading a b i l i t y . Cortex, $ 367. Liberman, I. Y., Shankweiler, D., Fischer, F. W., & Carter, B. (1974). E x p l i c i t s y l l a b l e and phoneme segmentation i n the young c h i l d . Journal of Experimental Child Psychology, 8, 1-12. 211 Liberman, I. Y., Shankweiler, D., Liberman, A. M., Fowler, C. , & Fischer, F. W. (1977). Phonetic segmentation and recoding i n the beginning reader. In A.S. Reber & H. S. Scarborough (Eds.), Towards a psychology of reading (pp. 207-225) . H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Lovett, M. W. (1984). The search for subtypes of s p e c i f i c reading d i s a b i l i t y : Reflections from a cognitive perspective. Annals of  Dyslexia, 34, 155-178. Lyon, R., & Watson, B. (1981). Empirically derived subgroups of learning-disabled readers: Diagnostic c h a r a c t e r i s t i c s . Journal  of Learning D i s a b i l i t i e s , 14, 256-261. Lyon, R., Stewart, N., & Freedman, D. (1982). Neuropsychological c h a r a c t e r i s t i c s of empirically derived subgroups of learning disabled readers. Journal of C l i n i c a l Neuropsychology, 4, 343-365. Lyon, R., Rietta, S, Watson, B., Porch, B., & Rhodes, J . (1981). Selected l i n g u i s t i c and perceptual a b i l i t i e s of empirically derived subgroups of learning-disabled readers. Journal of School  Psychology, 19, 152-166. Malatesha, R. N., & Dougan, D. R. (1982). C l i n i c a l subtypes of developmental dyslexia: Resolution of an i r r e s o l u t e problem. In R. N. Malatesha, & P. G. Aaron (Eds.), Reading disorders:  V a r i e t i e s and treatments (pp. 69-92). New York: Academic Press. Massaro, D. W. (1975). Visual features, perceptual storage, and processing time i n reading. In D. W. Massaro (Ed.), Understanding  language: An information-processing analysis of speech perception (pp. 207-239). New York: Academic Press. Mattis, S. (1978). Dyslexia syndromes: A working hypothesis that r e a l l y works. In A. L. Benton & D. Pearl (Eds.), Dyslexia: An  appraisal of current knowledge (pp. 43-67). New York: Oxford University Press. Mattis, S., French, J . H. , & Rapin, I. (1975). Dyslexia i n children and young adults: Three independent neuropsychological syndromes. Developmental Medicine and Child Neurology, 17, 150-163. McCarthy, D. (1972). McCarthy Scale of Children's A b i l i t i e s . New York: Psychological Corporation. McKinney, J . D. (1984). The search for subtypes of s p e c i f i c learning d i s a b i l i t y . Journal of Learning D i s a b i l i t i e s , 17, 43-50. 212 McKinney, J . D., & Forman, S. G. (1982). Classroom behaviour patterns of EMH, LD, and EH students. Journal of School Psychology, 20, 271-289. McKinney, J . D., & Speece, D. L. (1986). Academic consequences and lon g i t u d i n a l s t a b i l i t y of behavioral subtypes of learning-disabled children. Journal of Educational Psychology, 78, 365-372. McKinney, J . D., McClure, S., & Feagans, L. (1982). Classroom behaviour patterns of learning-disabled and nonlearning disabled children. Learning D i s a b i l i t y Quarterly, 5, 45-52. Morais, J . , Cary, L., Algeria, J . , & Bertelson, P. (1979). Does awareness of speech as a sequence of phonemes a r i s e spontaneously? Cognition, 7, 323-331. Morris, R. , B l a s h f i e l d , R. , & Satz, P. (1981). Neuropsychology and c l u s t e r analysis: Potentials and problems. Journal of C l i n i c a l  Neuropsychology. 3, 70-100. Morey, L. C., B l a s h f i e l d , R., & Skinner, H. A. (1983). A comparison of c l u s t e r analysis techniques within a sequential v a l i d a t i o n framework. Multivariate Behavioral Research, 18, 309-329. Nelson, H. E., & Warrington, E. K. (1980). An in v e s t i g a t i o n of memory functions i n dyslexic children. B r i t i s h Journal of  Psychology, 71, 487-503. Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood C l i f f s , NJ: Prentice-Hall. Olson, R. K., K l i e g l , R., Davidson, B. J . , & Folz, G. (1985). Individual and developmental differences i n reading d i s a b i l i t y . In C. E. Mackinnon & T. G. Waller (Eds.), Reading research:  Advances i n theory and practice (Vol. 4, pp. 1-63). New York: Academic Press. Palinscar, A. S. (1986). Metacognitive strategy i n s t r u c t i o n . Exceptional Children, 53, 118-124. Palinscar, A. S., & Brown, A. L. (1984) The r e c i p r o c a l teaching of comprehension fostering and comprehension monitoring a c t i v i t i e s . Cognition and Instruction, 1, 117-175. P e r f e t t i , C , & Lesgold, A. (1977). Discourse comprehension and sources of i n d i v i d u a l differences. In M. Just & P. Carpenter (Eds.), Cognitive processes i n comprehension (pp. 141-183). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. 213 P e r f e t t i , C., & Lesgold, A. (1979). Coding and comprehension i n s k i l l e d reading and implications for reading i n s t r u c t i o n . In L. B. Resnick & P. A. Weaver (Eds.), Theory and practice for early  reading (Vol. 1, pp. 57-84). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. P e r f e t t i , C., Goldman, S., & Hogaboam, T. Reading s k i l l and the i d e n t i f i c a t i o n of words i n discourse context. Memory and  Cognition, 7, 273-282. Petruskas, R. J . , & Rourke, B. P. (1979). I d e n t i f i c a t i o n of subtypes of retarded readers: A neuropsychological multivariate approach. Journal of C l i n i c a l Neuropsychology. 1, 17-37. Randawa, B. S., Hunt, D., & Rawlyk, S. A. (1974). F a c t o r i a l structure, v a l i d i t y , effectiveness and e f f i c i e n c y of the Canadian Cognitive A b i l i t i e s Test. Alberta Journal of Educational  Research. 20., 208-215. Reeve, R. A., & Brown, A. L. (1985). Metacognition reconsidered: implications for intervention research. Journal of Abnormal Child  Psychology, 13, 343-356. Rispens, J . (1982). Reading disorders as information-processing disorders. In R. N. Malatesha, & P. G. Aaron (Eds.), Reading  disorders: V a r i e t i e s and treatments (pp. 177-197). New York: Academic Press. Rourke, B. P. (1985). Overview of L.D. subtypes. In B. P. Rourke (Ed.), Neuropsychology of learning d i s a b i l i t i e s : Essentials of  subtype analysis (pp. 3-14). New York: G u i l f o r d Press. Rumelhart, D. E. (1977). Toward an i n t e r a c t i v e model of reading. In S. Dornic (Ed.), Attention and Performance (Vol. VI, pp. 45-67). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Rutter, M. (1978). Prevalence and types of dyslexia. In A. L. Benton & D. Pearl (Eds.), Dyslexia: An appraisal of current  knowledge (pp. 4-28). New York: Oxford University Press. Rutter M., & Yule, W. (1975). The concept of s p e c i f i c reading retardation. Journal of Child Psychiatry. 16, 181-197. Samuels, S. J . , & Eisenberg, P. (1981). A framework for understanding the reading process. In F. J . Pirozzolo, & M. C. Wittrock (Eds.), Neuropsychology and cognitive processes i n  reading (pp. 31-67). New York: Academic Press 214 Samuels, S. J . , & LaBerge, D. (1983). A theory of automaticity i n reading: Looking back: A retrospective model. In L. M. Gentile, M. L. Kamil, & J . S. Blanchard (Eds.), Reading research r e v i s i t e d (pp. 39-55). Columbus, OH: M e r r i l l . Samuels, S. J . , & M i l l e r , N. L. (1985). F a i l u r e to f i n d attention differences between learning disabled and normal children on classroom and laboratory tasks. Exceptional Children, 5, 358-375. Satz, P., & Morris, R. (1981). Learning d i s a b i l i t y subtypes: A review. In F. J . Pirozzolo & M. C. Wittrock (Eds.), Neuropsychology and cognitive processes i n reading (pp.109-141). New York: Academic Press. Scarborough, H.S. (1984) . Continuity between childhood dyslexia and adult reading. B r i t i s h Journal of Psychology. 75, 329-348. Seidenberg, M. S., Bruck, M., Fornarolo, G., & Backman, J . (1985). Word recognition processes of poor and disabled readers: Do they necessarily d i f f e r ? Applied Psycholinguistics, 6, 161-180. Shankweiler, D., & Liberman, I. Y. (1976). Exploring the r e l a t i o n s between reading and speech. In R. M. Knights & D. S. Bakker (Eds.), Neuropsychology of learning disorders: Theoretical  approaches (pp.297-313). Share, D. L., McGee, R., McKenzie, D., Williams, S., & S i l v a , P. A. (1987). Further evidence r e l a t i n g to the d i s t i n c t i o n between s p e c i f i c reading retardation and general reading backwardness. B r i t i s h Journal of Developmental Psychology, 5, 35-44. Smith, F. (1971). Understanding reading: A p s y c h o l i n g u i s t i c  analysis of reading and learning to read. New York: Holt, Rinehart & Winston. Sneath, P. H. E., & Sokal, R. R. (1973). Numerical taxonomy. San Francisco, CA: W.H. Freeman. Spear, L. C , & Sternberg, R. J . (1987). An information-processing framework for understanding reading d i s a b i l i t y . In S. Ceci (Ed.), Handbook of cognitive, s o c i a l , and neurological aspects of  learning d i s a b i l i t i e s (Vol I I , pp. 2-30). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Speece, D. L. (1987). Information processing subtypes of learning disabled readers. Learning D i s a b i l i t y Research. 2, 91-102. Speece, D. L., McKinney, J . D., & Appelbaum, M. I. (1985). C l a s s i f i c a t i o n and v a l i d a t i o n of behavioral subtypes of learning disabled children. Journal of Educational Psychology, 77, 67-77. 215 Sporn, E. L. (1981). Is dyslexia a middle class d i s a b i l i t y ? A c r i t i q u e of some d e f i n i t i o n s (Doctoral Dissertation, Yeshiva University, New York). Dissertation abstracts International. 42, 4380A. Spreen, 0., & Haaf, R. G. (1986). Empirically derived learning d i s a b i l i t y subtypes: A r e p l i c a t i o n attempt and longitudinal patterns over 15 years. Journal of Learning; D i s a b i l i t i e s , 19., 170-180. Stanovich, K. E. (1982a). Individual differences i n the cognitive processes of reading I: Word decoding. Journal of Learning  D i s a b i l i t i e s , 15, 485-493. Stanovich, K. E. (1982b). Individual differences i n the cognitive processes of reading I I : Text-level processing. Journal of  Learning D i s a b i l i t i e s , 15, 549-554. Stanovich, K. E. (1986). Cognitive processes and the reading problems of learning disabled children: Evaluating the assumption of s p e c i f i c i t y . In J . K. Torgesen & B. Y. L. Wong (Eds.), Psychological and educational perspectives on learning  d i s a b i l i t i e s (pp.87-131). New York: Academic Press. Stanovich, K. E. (1988). Explaining the differences between the dyslexic and the garden-variety poor readers: the phonological-core v a r i a b l e difference model. Journal of Learning D i s a b i l i t i e s , 21. 590-604, 612. Stanovich, K. E., Cunningham, A. E., & Freeman, D. J . (1984). The r e l a t i o n s h i p between early reading a c q u i s i t i o n and word decoding with and without context: A longitudinal study of f i r s t - g r a d e children. Journal of Educational Psychology, 76, 668-677. Sternberg, R. J . (1977). Intelligence, information processing and  analogical reasoning: The componential analysis of human  a b i l i t i e s . H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Sternberg, R. J . (1978). Intelligence research at the interface between d i f f e r e n t i a l and cognitive psychology: Prospects and proposals. Intelligence . 2, 195-222. Sternberg, R. J . (1980a). The development of l i n e a r s y l l o g i s t i c reasoning. Journal of Experimental Child Psychology. 29, 342-356. Sternberg, R. J . (1980b). Representation and process i n l i n e a r s y l l o g i s t i c reasoning. Journal of Experimental Psychology: General, 109. 119-159. Sternberg, R. J . (1981). Intelligence and nonentrenchment. Journal  of Psychology. 73, 1-16. 216 Sternberg, R. J . (1985). Beyond 10: A t r i a r c h theory of human  i n t e l l i g e n c e . New York: Cambridge University Press. Sternberg, R. J . (1986). Toward a u n i f i e d theory of human reasoning. Intelligence , 10, 281-314. Sternberg, R. J . , & Davidson, J . E. (1983). Insight i n the g i f t e d . Educational Psychologist, 18, 51-57. Sternberg, R. J . , & Ketron, J . L. (1982). Selection and implementation of strategies i n reasoning by analogy. Journal of  Educational Psychology, 74, 399-413. Sternberg, R. J . , & Nigro, G. (1980). Developmental patterns i n the solutio n of verbal analogies. Chi1d Deve1opment, 51, 27-38. Sternberg, R. J . , & R i f k i n , B. (1979). The development of analogical reasoning processes. Journal of Experimental Child  Psychology, 27, 195-232. Sternberg, R. J . , & Spear, L. C. (1985). A t r i a r c h theory of mental retardation. In N. E l l i s & N. Bray (Eds.), International review  of research i n mental retardation (Vol. 13, pp.301-326). New York: Academic Press. Sternberg, R. J . , & Suben, J . (1986). The s o c i a l i z a t i o n of i n t e l l i g e n c e . In M. Perlmutter (Ed.), Perspectives on  i n t e l l e c t u a l development: Minnesota symposia on c h i l d psychology (Vol.19, pp.201-235). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. Sternberg, S. (1969). The discovery of processing stages: Extensions of Donder's method. Acta Psychologica, 30, 276-315. Stevens, R. J . (1988). E f f e c t s of strategy t r a i n i n g on the i d e n t i f i c a t i o n of the main idea of expository passages. Journal  of Educational Psychology, 80, 21-26. Swanson, H. L. (1987a). Information-processing theory and learning d i s a b i l i t i e s : an overview. Journal of Learning D i s a b i l i t i e s . 20, 3-7. Swanson, H. L. (1987b). Information processing theory and learning d i s a b i l i t i e s : A commentary and future perspective. Journal of  Learning D i s a b i l i t i e s , 20, 155-166. Swanson, H. L. (1988). Memory subtypes i n learning disabled readers. Learning D i s a b i l i t y Quarterly, 11, 342-357. 217 Sweeney, J . E., & Rourke, B. P. (1985). S p e l l i n g d i s a b i l i t y subtypes. In B. P. Rourke (Ed.), Neuropsychology of learning  d i s a b i l i t i e s : E s s e n t i a l of subtype analysis (pp.147-166). New York: G u i l f o r d Press. T a l l a l , P. (1980). Language and reading: Some perceptual pre r e q u i s i t e s . B u l l e t i n of the Orton Society, 30, 170-78. T a l l a l , P., & Stark, R. E. (1982). Perceptual/motor p r o f i l e s of reading impaired childr e n with or without concomitant o r a l language d e f i c i t s . Annals of Dyslexia, 32, 163-176. Taylor, H. G., Fletcher, J . M., & Satz, P. (1982). Component processes i n reading d i s a b i l i t i e s : neuropsychological in v e s t i g a t i o n of d i s t i n c t reading s u b s k i l l d e f i c i t s . In R. N. Malatesha, & P. G. Aaron (Eds.), Reading disorders: V a r i e t i e s  and treatments (pp.121-147). New York: Academic Press. Taylor, H. G. , Satz, P., & F r i e l , J. (1979). Developmental dyslexia i n r e l a t i o n to other childhood reading disorders: Significance and c l i n i c a l u t i l i t y . Reading Research Quarterly, 15, 84-101. Thorndike, R. L., Hagen, E., Lorge, I., & Wright, E.N. (1974). Canadian Cognitive a b i l i t i e s Test (Multilevel e d i t i o n ) . Toronto: Thomas Nelson & Sons (Canada) Limited. Torgesen, J . (1977). Memorization processes i n reading-disabled c h i l d r e n . Journal of Educational Psychology, 79, 571-578. Torgesen, J . (1978-1979). performance of reading-disabled children on s e r i a l memory tasks. Reading Research Quarterly, 14, 57-87. Tukey, J . W. (1977). Exploratory data analysis. Reading, MS: Addison Wesley Publishing Company. Ve l l u t i n o , F. R. (1979) . Dyslexia: Theory and research. Cambridge, MS: MIT Press. V e l l u t i n o , F. R. (1980). Dyslexia: Perceptual deficiency or perceptual i n e f f i c i e n c y ? In J . F. Kavanagh & R. L. Venezky (Eds.), Orthography, reading, and dyslexia (pp.251-270). Baltimore, MY: University Park Press. Venezky, R. L., & Massaro, D. W. (1979). The r o l e of orthographic r e g u l a r i t y i n word recognition. In L. B. Resnick & P. A. Weaver (Eds.), Theory and pr a c t i c e of early reading (Vol. 1, pp.85-107). H i l l s d a l e , NJ: Lawrence Erlbaum Associates. 218 Wagner, R. K., & Sternberg, R. J . (1984). Alt e r n a t i v e conceptions of i n t e l l i g e n c e and t h e i r implications f o r education. Review of  Educational Research, 54., 179-223. Waites, L. (1968). World Federation of Neurology: Research group  on developmental dyslexia and world i l l i t e r a c y . Report of proceedings, p.22. Ward, J . (1983). Hier a r c h i c a l grouping to optimize an objective function. Journal of American S t a t i s t i c a l Association. 58, 236-244. Wechsler, D. (1949). Manual for the Wechsler i n t e l l i g e n c e scale for  children. New York: Psychological Corporation. Wechsler, D. (1955). Manual for the adult i n t e l l i g e n c e scale. New York: Psychological Corporation. Wechsler, D. (1974). Manual for the Wechsler i n t e l l i g e n c e scale  for c h i l d r e n - revised. New York: Psychological Corporation. Wilson, N. L. (1980). Individual differences i n processes and  strategies i n analogical reasoning. Unpublished master's thesis. University of B r i t i s h Columbia. Wong, B. Y. L. (1979). Increasing retention of main ideas through questioning strategies. Learning D i s a b i l i t y Quarterly. 2, 42-47. Wong, B. Y. L. (1986). Metacognition and spec i a l education: A review of a view. Journal of Special Education, 20, 9-29. Wong, B. Y. L., & Jones, W. (1982). Increasing metacomprehension i n learning-disabled and normally achieving students through self-questioning t r a i n i n g . Learning D i s a b i l i t y Quarterly, 5, 228-240. Wong, B. Y. L., & Wong, R. (1986). Study behaviour as a function of metacognitive knowledge about c r i t i c a l task varia b l e s : An inves t i g a t i o n of above average, average, and learning-disabled readers. Learning D i s a b i l i t y Research. 1, 101-111. Wright, B. D. (1977). Solving measurement problems with the Rasch model. Journal of Educational Measurement, 14, 97-116. 219 Appendix A. Letter of consent 220 Name of School Board. Re: Research project of Margaret Potter Dear Parent: A research project e n t i t l e d "Processes and strategies used by normal and disabled readers i n analogical reasoning" i s being c a r r i e d out by Margaret Potter. Margaret, a teacher within our system, i s currently attending the University of B r i t i s h Columbia. This study forms the basis of the doctoral d i s s e r t a t i o n required for a Doctor of Education degree. The purpose of t h i s study i s to discover how s k i l l e d and u n s k i l l e d readers solve analogies. This information w i l l add to what i s known about u n s k i l l e d or disabled readers and provide a better understanding of why some children have d i f f i c u l t y learning to read. Permission i s requested for your c h i l d to take part i n the study. The t o t a l amount of time required w i l l be approximately two hours. A picture vocabulary tests and two reading t e s t s w i l l be i n d i v i d u a l l y administered i n one session and a series of picture analogies w i l l be given to small groups of students i n a second session. The i d e n t i t y of each student w i l l be kept c o n f i d e n t i a l . Each student w i l l be assigned a number and only the number w i l l be used i n recording the data. A summary of the c h i l d ' s t e s t r e s u l t s w i l l be a v a i l a b l e i f you request them. Parents who wish to share the information with school personnel should indicate t h i s by giving permission for the re s u l t s to be sent d i r e c t l y to the school. When the study has been completed a l l t e s t booklets w i l l be destroyed. You or your c h i l d have the ri g h t to refuse to p a r t i c i p a t e or withdraw from the study at any time. Such r e f u s a l or withdrawal w i l l not influence your c h i l d ' s standing i n class i n any way. Further explanation of any of these procedures w i l l be furnished upon request. Please indicate to your c h i l d ' s teacher that further information i s required and the request w i l l be passed on. 221 I consent/do not consent to my c h i l d ' s p a r t i c i p a t i o n i n t h i s study: Signature of parent or guardian: I consent/do not consent to my c h i l d ' s r e s u l t s being passed on to his/her teacher Signature of parent or guardian: Please sign t h i s form and return i t to the school i n the sealed envelope provided. 222 Appendix B. Screening c h e c k l i s t for teachers 223 SCREENING CHECKLIST FOR TEACHERS Name of Student: Grade: Name of Teacher: School: 1. Education has been i n English since Grade 1? Yes ... No ... 2. V i s i o n (with glasses) i s normal? Yes ...No ... 3. Hearing i s normal? Yes ...No ... 4. Has attendance i n the l a s t school year been at le a s t 75 percent? Yes ...No ... 5. Behaviour problems might prevent t h i s student from completing the requirements of t h i s study? Yes ...No ... 6. Reading i n comparison with peers i s : Above average ... Average Below average ... 224 Appendix C. Schematic Picture Analogy Booklet 225 226 227 228 229 Appendix D. Cards used to i n s t r u c t students i n solving Schematic Picture Analogies 230 231 Appendix E. Table l i s t i n g c l u s t e r membership, Boder s p e l l i n g patterns, and CV1 and CV2 model preference for Group D and Group N 232 Student Cluster Boder s p e l l i n g Model for Model for Number Subgroup CV1 CV2 Dl 1 nonspecific 4M 4M D2 2 mixed 4M NM1 D3 3 nonspecific NM2 4M D4 3 dysphonetic 4M 4M D5 2 dysphonetic 4M 4M D6 3 nonspecific IM IM D7 2 nonspecific NM2 4M D8 2 nonspecific NM1 NM1 D9 3 nonspecific 4M 4M D10 3 nonspecific NM1 4M D l l 2 dysphonetic NM2 4M D12 3 nonspecific IM IM D13 3 nonspecific NM2 NM2 D14 1 dyseidetic 4M 4M D15 1 nonspecific IM 4M D16 1 nonspecific 4M NM2 D17 3 nonspecific NM2 IM D18 2 dysphonetic NM1 NM2 D19 3 nonspecific 4M 4M D20 3 dysphonetic NM2 NM2 D21 1 dyseidetic NM1 NM1 D22 3 nonspecific 4M 4M D2 3 3 nonspecific 4M NM2 D24 3 nonspecific 4M NM2 D2 5 1 mixed 4M NM2 D2 6 2 dysphonetic NM1 4M D27 3 nonspecific NM2 NM2 D28 1 nonspecific NM2 4M D29 2 nonspecific NM2 NM2 D30 3 nonspecific IM NM1 D31 3 nonspecific 4M 4M D32 2 dysphonetic NM2 NM2 D33 2 dysphonetic IM NM2 D34 1 dyseidetic NM1 NM2 D35 1 dyseidetic 4M 4M D3 6 2 dysphonetic 4M 4M D3 7 1 mixed NM2 NM2 D3 8 1 nonspecific NM2 4M D3 9 1 dysphonetic IM NM2 D40 3 nonspecific NM2 NM2 D41 3 nonspecific NM1 NM1 D42 1 nonspecific NM1 4M 233 Student Cluster Boder s p e l l i n g Model for Model for Number Subgroup CV1 CV2 D43 3 nonspecific 4M 4M D44 3 nonspecific IM NM2 D45 1 nonspecific NM1 4M D46 3 nonspecific NM2 NM2 D47 3 nonspecific NM1 4M D48 1 nonspecific IM NM2 D49 1 nonspecific 4M 4M D50 1 nonspecific 4M 4M D51 3 nonspecific 4M 4M D52 3 nonspecific NM1 4M D53 3 nonspecific NM2 NM2 D54 3 nonspecific IM IM D55 1 dyseidetic NM1 NM2 D56 1 dysphonetic NM1 NM2 D57 2 dyseidetic 4M 4M D58 1 nonspecific 4M 4M D59 2 nonspecific 4M 4M D60 1 nonspecific 4M 4M D61 1 nonspecific 4M 4M D62 2 dyseidetic 4M 4m D63 2 dysphonetic 4M 4M D64 3 dyseidetic 4M 4M D65 1 nonspecific IM NM2 D66 1 nonspecific NM2 4M D67 2 nonspecific IM NM2 D68 2 dysphonetic NM2 NM2 D69 3 nonspecific NM1 NM2 D70 3 nonspecific IM NM2 D71 3 nonspecific 4M 4M D72 3 mixed NM1 NM2 D73 1 dysphonetic nM2 NM2 D74 2 nonspecific 4M 4m D75 1 nonspecific NM1 4M D76 2 nonspecific NM1 NM2 D77 1 nonspecific NM1 NM2 Nl NM2 NM2 N2 4M 4M N3 4M NM2 N4 IM NM2 N5 NM2 NM2 N6 4M 4M N7 4M 4M N8 NM2 4M N9 4M 4M 234 Student Cluster Boder s p e l l i n g Model for Model for Number Subgroup CV1 CV2 N10 NM1 IM N i l 4M NM2 N12 4M 4M N13 4M 4M N14 4M 4M N15 4M 4M N16 4M 4M N17 4M 4M N18 4M NM2 N19 4M 4M N2 0 NM1 NM2 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0100946/manifest

Comment

Related Items