Phonologic and lexical routes to reading

MISSING IMAGE

Material Information

Title:
Phonologic and lexical routes to reading a comparison of impaired, normal, and superior readers
Uncontrolled:
A comparison of impaired, normal and superior readers
Physical Description:
xii, 136 leaves : ill. ; 29 cm.
Language:
English
Creator:
Morris, Mary K., 1957-
Publication Date:

Subjects

Subjects / Keywords:
Reading   ( mesh )
Clinical and Health Psychology thesis Ph.D   ( mesh )
Dissertations, Academic -- Clinical and Health Psychology -- UF   ( mesh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1986.
Bibliography:
Bibliography: leaves 130-135.
Statement of Responsibility:
by Mary K. Morris.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000881987
oclc - 17558907
notis - AEH9866
sobekcm - AA00006109_00001
System ID:
AA00006109:00001

Full Text


















PHONOLOGIC AND LEXICAL ROUTES TO READING:
A COMPARISON OF IMPAIRED, NORMAL, AND SUPERIOR READERS







BY






MARY K. MORRIS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA


1986






























Copyright 1986

by

Mary K. Morris
















ACKNOWLEDGMENTS


Many people have helped me reach this place and to thank all of

them herein is not possible. Special thanks are extended to my

chairperson, Dr. Eileen B. Fennell, and the other members of my

committee, Dr. Russell M. Bauer, Dr. Jacquelin R. Goldman, Dr. Kenneth

M. Heilman, and Dr. Leslie J. Gonzalez-Rothi, for their knowledge,

patience, and emotional support. Dr. Randy Carter and Dr. Roger

Blashfield provided statistical consultation which was both much

needed and much appreciated.

I also express many thanks to Carol Schramke who spent numerous

hours locating, recruiting, and testing the children who participated

in this study. I must also acknowledge those children who donated

their time as well as the staff and teachers of the Alachua County

School Board whose cooperation allowed this work to be done.

Finally, I am grateful to Howard who has provided much love and

encouragement and worked hard to create a climate in which I can

achieve my goals.


-iii-

















TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS....................................................iii

LIST OF TABLES .....................................................vii

ABSTRACT.............. .................................... xi

CHAPTERS

1 INTRODUCTION................. ...................... ..1

Models of Developmental Dyslexia: Historical Overview.....1
Research on Reading Disability Subtypes....................3
Classifications Based on Reading Achievement/Performance...6
Multivariate Statistical Techniques...................6
Clinical-Inferential Techniques.......................8
Surface vs. Deep Dyslexia............................. 13
Reading Development and Reading Routes...................16
Acquired Alexia and Developmental Dyslexia...............18
Statement of Purpose.....................................25

2 METHODS........................... .... ........ ........ 34

Subjects........ ........ ... ..... .... ............. ..... 34
Stimulus Materials......................................35
Task 1: Battery of Adult Reading Function...........35
Task 2: Brigance Diagnostic Inventory of Basic
Skills.. ........ ........... ............ .......36
Apparatus.............. ...... ..... ... ........ 37
General Procedure................ ................... .37
Specific Procedures...................................... 37
Task 1: Battery of Adult Reading Function...........37
Task 2: Brigance Diagnostic Inventory of Basic
Skills............ ....................... .......... 38

3 RESULTS..............................................39

Overview of the Analysis..................................39
Scoring ....................... ................... ...... 42


-iv-










Battery of Adult Reading Function....................42
Brigance Diagnostic Inventory of Basic Skills........43
Reliability.............................................. 43
Descriptive Statistics....................................43
Assessment of Between-Subjects Factors...................46
Battery of Adult Reading Function.....................46
Brigance Diagnostic Inventory of Basic Skills........64
Assessment of Within-Subject Factors.....................67
Assessment of the Contribution of General Scholastic
Ability............................ ........ ..... .72
Battery of Adult Reading Function...................77
Brigance Diagnostic Inventory of Basic Skills........79
Subtype Identification: Multivariate Statistical.........80
Subtype Identification: Clinical-Inferential.............88

4 DISCUSSION.............................. .................92

General Hypothesis 1.................................... 92
Specific Prediction 1.............................. 92
Specific Prediction 2................................ 94
Specific Prediction 3................................94
Additional Findings.................................95
Conclusions............................. .... .........99
General Hypothesis 2.....................................100
Specific Prediction 4............................... 101
Specific Prediction 5...............................105
General Hypothesis 3..................................... 110
Specific Prediction 6............................ 111
Specific Prediction 7.............................. 111
Specific Prediction 8................................111
General Hypothesis 4: Specific Prediction 9.............112
Summary and Conclusions................................. 113

APPENDICES

1 INFORMED CONSENT.........................................117

2 SUBTEST 1: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)............................120

3 SUBTEST 2: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)...........................121

4 SUBTEST 3: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)............................ 122

5 SUBTEST 4: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984).............................123

6 SUBTEST 5: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)...........................124












7 SUBTEST 6: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)............................125

8 APPENDIX A: BATTERY OF ADULT READING FUNCTION
(GONZALEZ-ROTHI ET AL., 1984)...........................126

9 ERROR TYPE CLASSIFICATION CRITERIA .......................128

10 SUBTYPE CLASSIFICATION RULES.......................... 129

BIBLIOGRAPHY....................... ............................... 130

BIOGRAPHICAL SKETCH.....................................................136


-vi-
















LIST OF TABLES


Table Page

1 Percentile Means and Standard Deviations by Reading
Group for "Metropolitan Achievement Test--Reading"
and "Otis-Lennon School Abilities Test"....................44

2 Racial Composition of Reading Groups for "Metropolitan
Achievement Test--Reading" and "Otis-Lennon School
Abilities Test"............................................ 44

3 Means and Standard Deviations by Reading Group for
"Battery of Adult Reading Function": Error
Percentage....... .............. ..... .... ........ .... ..45

4 Means and Standard Deviations by Reading Group for
"Battery of Adult Reading Function": Mean Response
Time ............... ........................ ......... .... ..47

5 Means and Standard Deviations by Reading Group for
"Battery of Adult Reading Function": Cumulative
Percentage of Error Types...................................48

6 Means and Standard Deviations by Reading Group for
"Brigance Diagnostic Inventory of Basic Skills": Error
Percentage and Reading Rate................................48

7 Analysis of Between-Subjects Factors for "Battery of
Adult Reading Function": Significant MANOVAs...............49

8 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Total Performance.................49

9 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 1........................50

10 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 2......................... 50

11 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 3 .........................51

12 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 4........................51


-vii-












13 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 5........................52

14 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Subtest 6........................52

15 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Functors...........................53

16 Univariate ANOVA: Error Percentage for "Battery of
Adult Reading Function": Contentives.......................53

17 Tukey's Studentized Range Test: Error Percentage for
"Battery of Adult Reading Function"........................ 54

18 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Total Performance.................56

19 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 1........................56

20 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 2.........................57

21 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 3......................57

22 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 4.........................58

23 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 5.......................58

24 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Subtest 6.........................59

25 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Functors..........................59

26 Univariate ANOVA: Mean Response Time for "Battery of
Adult Reading Function": Contentives.......................60

27 Tukey's Studentized Range Test: Mean Response Time for
"Battery of Adult Reading Function".........................61

28 Univariate ANOVA: Error Type for "Battery of
Adult Reading Function": Phonological Errors...............62

29 Univariate ANOVA: Error Type for "Battery of
Adult Reading Function": Lexical Errors.....................62


-viii-











30 Univariate ANOVA: Error Type for "Battery of
Adult Reading Function": Other Errors..................... 63

31 Univariate ANOVA: Error Type for "Battery of
Adult Reading Function": Lexicalization Index.............63

32 Tukey's Studentized Range Test: Error Type for
"Battery of Adult Reading Function".......................65

33 Univariate ANOVA: Error Percentage for "Brigance
Diagnostic Inventory of Basic Skills"......................65

34 Univariate ANOVA: Reading Rate for "Brigance
Diagnostic Inventory of Basic Skills"...................... 66

35 Tukey's Studentized Range Test: Reading Rate for
"Brigance Diagnostic Inventory of Basic Skills".............66

36 Repeated Measures ANOVA: Error Percentage for "Battery
of Adult Reading Function": Subtest 1-4,
Within-Subject Effects.................................... 68

37 Repeated Measures ANOVA: Error Percentage for "Battery
of Adult Reading Function": Subtest 5-6,
Within-Subject Effects.....................................68

38 Repeated-Measures ANOVA: Error Percentage for "Battery
of Adult Reading Function": Functors/Contentives,
Within-Subject Effects..................................... 68

39 Duncan's Multiple Range Test: Error Percentage for
"Battery of Adult Reading Function": Subtests 1-4 .........70

40 Duncan's Multiple Range Test: Error Percentage for
"Battery of Adult Reading Function": Subtests 5-6..........70

41 Duncan's Multiple Range Test: Error Percentage for
"Battery of Adult Reading Function": Functors/
Contentives................................................ 71

42 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Subtests 1-4,
Within-Subject Effects.....................................71

43 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Subtests 5-6,
Within-Subject Effects............................... ..73


-ix-











44 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Functors/
Contentives, Within-Subject Effects........................73

45 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Subtests 1-4..........74

46 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Subtests 5-6........... 74

47 Duncan's Multiple Range Test: Mean Response Time for
"Battery of Adult Reading Function": Functors/
Contentives .............................. ....... ....... 75

48 Repeated Measures ANOVA: Error Type for "Battery
of Adult Reading Function": Phonological, Lexical and
Other, Within-Subject Effects..............................75

49 Duncan's Multiple Range Test: Error Type for
"Battery of Adult Reading Function": Phonological,
Lexical, and Other.........................................76

50 Analysis of Covariance: Error Percentage for "Battery
of Adult Reading Function": Subtest 5 with Covariate =
"Otis-Lennon School Abilities Test"........................78

51 Initial Seeds: FASTCLUS Procedure..........................83

52 Cluster Summary: FASTCLUS Procedure .......................85

53 Cluster Means and Standard Deviations, FASTCLUS Procedure...86

54 Cluster Membership by Reading Group........................87

55 Univariate ANOVA: Reading Rate for "Brigance Diagnostic
Inventory of Basic Skills"--Comparison of Statistically
Generated Clusters ..........................................89

56 Clusters Derived by Clinical-Inferential Classification
Rules ...................................................... 91
















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

PHONOLOGIC AND LEXICAL ROUTES TO READING:
A COMPARISON OF IMPAIRED, NORMAL, AND SUPERIOR READERS

BY

MARY K. MORRIS

December, 1986


Chairperson: Eileen B. Fennell, Ph.D.
Major Department: Clinical Psychology


Ninety 12-year-old males were classified as impaired, normal, and

superior readers based on their reading percentiles on the

Metropolitan Achievement Test. Two reading measures were administered

to these children. The Battery of Adult Reading Function (BARF) is

designed to assess the relative integrity of the phonological and

lexical reading routes. It was developed based on a model of acquired

reading disorders and has proven useful in differentiating subtypes of

acquired alexia. Two primary subtypes have been described, resulting

from impairment of one route with relative sparing of the other. The

Brigance Diagnostic Inventory of Basic Skills was utilized to assess

silent reading comprehension and reading rate.

The three reading groups were observed to differ both

quantitatively and qualitatively on the BARF. Analyses were selected

to test the hypothesis that subtypes of developmental reading


-xi-










disorders are analogous to subtypes of acquired reading disorders.

Two classification methods were chosen: (a) cluster analysis and (b)

visual inspection, using a set of classification rules based on a

model of alexia.

Results obtained with these two methods were congruent. Impaired

readers were predominantly characterized by deficits in the

phonological route. Selective deficits in the lexical route with

spared phonological processing were not typical of this group. The

ability to use the phonological route appears to permit normal reading

achievement, even in the absence of well developed lexical skills. It

was also hypothesized that lexical and phonological routes would

differ in speed of processing. While the difference between the

reading rates associated with those two routes was in the expected

direction, it did not achieve significance. These findings are

discussed in the context of a developmental model of reading

acquisition in an attempt to explain why selective deficits in the

lexical route do not result in impaired reading in children.


-xii-
















CHAPTER 1
INTRODUCTION



Models of Developmental Dyslexia: Historical Overview

According to the World Federation of Neurology, children with

specific developmental dyslexia are those who fail to acquire normal

reading proficiency despite conventional instruction, sociocultural

opportunity, average intelligence, and freedom from gross sensory,

emotional or neurological handicap (Critchley, 1970). Although

substantial research attention has been devoted to this population

over several decades, disagreement persists regarding the nature and

etiology of this disorder. Rourke (1985) has recently reviewed the

historical development of models of developmental dyslexia. The

earliest theoretical models viewed developmental dyslexia as a unitary

entity with a single causal deficit. Hinshelwood (1917) described

cases of "otherwise sharp and intelligent" children who could not

learn to read. He attributed this selective impairment of reading

acquisition to "congenital visual word blindness," an inability to

recognize words, resulting from damage to the visual memory center for

words in the left angular gyrus. This model assumed a single type of

reading disability that occurred as an isolated disorder in an

otherwise completely normal child. It also explicitly connected the

congenital disorder with the acquired impairment of reading ability

following brain damage which had been observed in adults.


-1-










Prior to 1950, the only other major theory of developmental

reading disorder was formulated by Orton (1937). Orton also believed

that reading involved visual word images that were stored in a

specific part of the brain. However, rather than attributing reading

problems to deficits in these brain regions, he proposed a lag in the

development of normal left hemisphere dominance for language

abilities. Because of this developmental lag, visual images stored as

mirror images in the nondominant hemisphere were not suppressed by the

dominant hemisphere. This interhemispheric competition interfered

with the perception of normally oriented visual images and resulted in

"strephosymbolia" or "twisted symbols." Orton also attributed speech

and writing problems, stuttering, and general clumsiness to this

developmental lag. Thus, he began to move away from the notion of an

isolated impairment of reading, postulating a relationship between

reading problems and deficits in other nonreading abilities.

While both of these early models assume a unitary disorder based

on neurological dysfunction, the differences between them anticipate

controversial questions in future reading disability research. Does

developmental dyslexia result from a specific deficit or a more

general developmental lag? Are the impairments in reading isolated or

secondary to difficulties in other nonreading abilities? And finally,

are models of acquired reading dysfunction useful in understanding

developmental disorders?

Exploration of nonreading skills became the central focus of

reading disability research after 1950. Investigators began to search

for an underlying nonreading skill deficiency which might cause





-3-


reading impairment. The dominant experimental method was the single

syndrome paradigm (Doehring, 1978). In this paradigm, an experimental

group of disabled readers was compared to a matched control group of

normal readers on one or more measures of nonreading ability.

Statistically significant differences between groups were accepted as

evidence of underlying behavioral causes of reading disability. A

wide variety of potential underlying causes were identified by such

studies. Some examples include disturbance in the development of

visual perception (Bender, 1958), directional confusion (Hermann,

1959), delayed left hemisphere specialization (Satz & van Nostrand,

1970), perceptual deficits (Cruickshank, 1968), and verbal processing

deficits (Vellutino, 1978).

Follow-up studies consistently demonstrated that each of these

unitary deficit hypotheses was inadequate to account for the diversity

among reading disabled children. These failures of replication led to

the gradual realization that the accumulation of contradictory and

inconsistent findings may have been a function of heterogeneity in the

population of interest. Several prominent researchers recommended

that searches for the underlying causes of a unitary dyslexia be

abandoned in favor of attempts to identify different categories of

reading disability (Benton, 1975). During the 1970s, the search for

subtypes dominated research in learning disabilities, in general, and

developmental dyslexia, in particular.

Research on Reading Disability Subtypes

In a recent review of the learning disability subtype literature,

Satz and Morris (1981) describe two different approaches to the










problem of subtype identification. The clinical-inferential approach

utilizes both a priori assumptions about significant group attributes

and visual inspection techniques to reduce large data sets into

presumably homogeneous classes. Groupings have been based on (a)

etiological inferences, (b) performance on neuropsychological/

cognitive measures, and (c) direct measures of reading achievement

and/or performance. These techniques have been frequently criticized

since they tend to reify a priori theoretical biases and cannot

contend with the complexity of large multivariate data sets. However,

valuable insights into significant clinical dimensions of dyslexia are

likely to be gained.

Multivariate statistical methods (e.g., factor analysis and

cluster analysis) have also been employed to create systems of

classification by identifying the hidden substructure of complex

multidimensional data sets. Such data sets commonly reflect

performance on either neuropsychological and/or reading achievement

measures. Studies employing a statistical approach to classification

make no a priori assumptions regarding either number or type of

subgroups in the solution. The use of statistical methods also allows

the accommodation of much larger data sets than can be handled by

visual inspection techniques. However, the validity of subtypes

generated by these techniques has been questioned. These methods will

create clusters from random data and the nature of the subtype

solution is to some extent a function of the clustering algorithm

chosen. Thus, current statistical classification approaches are also

heuristic devices which require external validation.










Classifications based on inferred etiology and neuropsychological

strengths and weaknesses comprise a substantial portion of the subtype

literature. Classifications based on etiology have typically

attempted to clinically identify a subset of reading disabled children

whose difficulties are due to an intrinsic constitutional deficit.

This syndrome has been termed "specific developmental dyslexia"

(Critchley, 1970) or "primary reading retardation" (Rabinovitch,

1968). Rutter (1978) has argued that such classification schemes are

based on a diagnosis by exclusion, involving circular reasoning, with

only presumptive evidence of a constitutional basis. Taylor, Satz,

and Friel (1979) were unable to distinguish a specific developmental

dyslexic subgroup from nondyslexic poor readers along any of several

dimensions including severity of reading disturbance, frequency of

reversal errors, familial reading competency, or personality

functioning.

Classifications based on nonreading abilities have employed both

clinical inferential and multivariate statistical techniques. Since

reading is a highly complex activity, dysfunction in a wide variety of

component cognitive and/or linguistic skills is likely to produce

impairment. Frequently cited studies utilizing a clinical-inferential

approach include Denckla (1972); Kinsbourne and Warrington (196,6); and

Mattis, French, and Rapin (1975). A multivariate statistical approach

based on a battery of neuropsychological measures is also exemplified

in a growing body of literature (Fisk & Rourke, 1979; Fletcher & Satz,

1985; C. S. Johnston, 1986; Lyon, Stewart, & Freedman, 1982; Lyon &

Watson, 1981; Petrauskas & Rourke, 1979; Rourke & Finlayson, 1978).






-6-


Although classifications based on etiology and on nonreading ability

have provided important information about reading disability subtypes,

the present discussion will focus on classifications based on direct

measures of reading achievement and performance. This emphasis

reflects the belief that the search for an underlying mechanism must

begin with a careful analysis of the dysfunctional reading process

itself rather than by examining related cognitive processes.

Classifications Based on Reading Achievement/Performance

Multivariate Statistical Techniques

Doehring and Hoshko (1977) utilized a Q-technique factor analysis

to identify homogeneous subtypes of learning disabled children.

Thirty-one separate tests of reading related skills were administered

to two different groups of children, one with reading deficits (N=34)

and another with mixed learning disabilities (N-31). The Q-technique

clusters subjects (not test variables as in standard factor analysis)

who show similar patterns of performance.

Three subgroups emerged within the reading disabled population.

Type 1 (N-12) demonstrated good performance on both visual-visual and

auditory-visual matching tests and deficits on oral reading tests

(composed of single words and syllables). The authors attributed this

pattern of performance to a linguistic deficit. The profile for Type

2 (N-11) was characterized by normal visual scanning, impaired

auditory-visual letter matching, and impaired oral reading. Doehring

and Hoshko (1977) described these children as phonologically deficient

with particular difficulties in making grapheme-phoneme

associations. Type 3 (N-8) were able to make visual-visual and










auditory-visual matches of single letters, but were unable to perform

similar matches with syllables and words. The group was described as

deficient in intersensory integration.

One of the strengths of the Doehring and Hoshko (1977) study was

the attempt to validate the identified subtypes with an external

criterion. An independent estimate of the reading skills of these

children, based on teacher recommendations for compensatory

educational procedures, was congruent with the hypothesized

deficits. However, as Satz and Morris (1981) point out, the

differences between subgroups on this independent criterion were not

assessed by statistical procedures. In addition, no control group of

normal readers was tested to determine whether the proposed subtypes

were unique to a reading disabled population. Finally, the Q-

technique of factor analysis has been criticized for its inability to

deal with loadings on multiple factors and its insensitivity to

elevation differences when creating subtypes (Satz & Morris, 1981).

Satz and Morris (1981) applied cluster analytic techniques to the

large unselected sample (N-236) of 11-year-old white males who had

participated in the Florida Longitudinal Project (Satz, Taylor, Friel,

& Fletcher, 1978). Cluster analysis is a procedure designed to

facilitate the creation of classification schemes. Several techniques

have been developed to group individuals into homogeneous clusters

based on each subject's performance on a set of clustering

variables. In this study, cluster analysis was used to define a

target population, prior to the search for subtypes.









Nine subgroups emerged, representing different patterns of

reading, spelling, and arithmetic ability. Two of these groups were

characterized by deficits in all three areas (N=89) and were

consequently labeled "learning disabled." Validation of these

subgroups using Peabody Picture Vocabulary Test-IQ (PPVT-IQ),

neuropsychological performance, neurological status, and socioeconomic

status (SES) was also reported. The "learning disabled" groups had

lower PPVT-IQs, poorer performances on both language and perceptual

tests, more "soft" neurological signs, and lower SES. Although this

study illustrates the use of cluster analysis based on achievement

variables it explored subtypes within a broad spectrum of learning

disabilities. Consequently, it does not specifically address the

problem of developmental dyslexia subtypes.

Clinical-Inferential Techniques

Several studies have relied on clinical observation of the

reading process as a basis for subtype classification. Monroe (1932)

identified ten types of errors made by children with reading

difficulty. However, patterns of error type did not differentiate

among reading disabled, mentally retarded, and behavior disordered

subjects. Ingram, Mason, and Blackburn (1970) also used error type to

classify reading disabled subjects into audiophonic, visuospatial, and

mixed subtypes.

Boder's (1973) qualitative analysis of the patterns of reading

and spelling in a sample of dyslexic children has had considerable

influence on the subtype literature. Using this approach, she

identified three subtypes in a sample of 107 children, ages 8 16,










who met the World Federation definition of specific developmental

dyslexia. Subtype 1 was described as "dysphonetic" with selective

impairments in analyzing phonemic properties of language. They

experienced difficulties in both sounding out and blending the

component graphemes of words and syllables, preferring to approach

written material in a more global fashion, based on the visual

recognition of whole words. Sixty-seven percent of Boder's sample

fell into this group. Subtype 2 was labeled "dyseidetic" based on

their deficiencies in discriminating and remembering visual

gestalts. These children read very slowly, relying on phonemic

analysis to sound out each word as if they were encountering it for

the first time. This group comprised 10% of the sample. Finally,

Subtype 3 was described as a mixed dysphonetic-dyseidetic group with

impairments in both phonemic analysis and visual gestalt

discrimination. Twenty-three percent of this population were

classified in the mixed subtype.

Boder (1973) also described a characteristic pattern of errors

associated with each subtype. Dysphonetic readers lacked phonemic

analysis skills and had difficulty blending the component letters and

syllables of a word. Thus, words were often guessed from minimal cues

such as the first or last letter and the length of the word. Word

substitutions were pathognomonic of this group. These substitutions

were often based on visual similarity between the target word and the

response. Occasionally, substitution errors based on a conceptual,

but not visual or phonetic resemblance, occurred. Boder (1973) refers





-10-


to these as "semantic substitution" errors and cites examples such as

"funny" for "laugh," "chicken" for "duck," and "planet" for "moon."

In contrast, dyseidetic children appeared to rely on phonemic

analysis and synthesis, sounding out letters in sequence, to

compensate for their hypothesized deficits in processing visual

gestalts. Errors commonly occurred on irregular words which do not

conform to grapheme to phoneme conversion rules. Responses reflected

the child's attempt to sound out these words by applying conversion

rules.

The mixed dysphonetic-dyseidetic group was described as the most

severely handicapped due to deficits in both processing mechanisms.

Boder characterized these children as nonreaders and described their

errors as bizarre in that they bore little resemblance to the target

on any dimension. Confusion of reversible letters occurred more

frequently in these children but was present to some degree in all

three groups.

Although Boder's (1973) study has been commended for its rich

clinical descriptions of both reading styles and error patterns, her

subtypes remain clinical impressions which require statistical

verification of their validity, reliability, and utility. Although

parallels can be drawn between these groups and those reported by

other investigators (Aaron, 1982; Denckla, 1972; Mattis et al., 1975;

Myklebust, 1968), the globally similar labels used by independent

researchers may actually refer to disparate patterns. Furthermore,

Boder's clinical analysis was not extended to a comparison group of

normal readers in order to directly address the issue of quantitative





-11-


vs. qualitative differences in the reading performance of dyslexics

relative to normal readers.

Aaron (1982) has proposed a model of developmental dyslexia based

on constructs utilized by cognitive psychologists to describe the

reading process. The encoding operation in normal reading is believed

to involve two dissociable information processing strategies,

simultaneous and sequential (Das, Kirby, & Jarman, 1979). The

simultaneous operation is characterized by the parallel processing of

several letters within a word, resulting in the encoding of visual

word images as gestalts. The sequential operation involves a serial

conversion of individual letters to their phonetic equivalents. It is

hypothesized that skilled reading is based on the concurrent

implementation of both encoding operations (Das et al., 1979). Thus,

an underutilization of one or the other of these strategies is

believed to hinder efficient reading. Aaron (1982) attributes

developmental dyslexia to an imbalance in the deployment of these

contrasting operations. This model appears to be congruent with

Boder's (1973) typology in that her dysphonetic group could be

described as sequential-deficient and her dyseidetic group,

simultaneous-deficient.

Attempts to empirically demonstrate the existence of the two

clinical subtypes suggested by this hypothesis have been reported

(Aaron, 1982). This author utilized Boder's (1973) diagnostic

screening inventory to classify dyslexic children into sequential-

deficient and simultaneous-deficient groups. Subjects were presented

with a list of words of graded difficulty and asked to read each





-12-


aloud. Based on their performance on this task, they were asked to

write ten of those words to dictation. Five words which were read

without hesitation and five read with difficulty were presented.

These were presumed to reflect sight vocabulary and unknown

vocabulary, respectively. Subjects were classified into the two

subgroups on the basis of spelling errors on the dictation task.

Sequential-deficient dyslexics were characterized by visual word

substitutions and a tendency to omit, reverse or displace letters

within a word. Simultaneous-deficient dyslexics produced phonetic

misspellings which reflected an application of grapheme-phoneme

correspondences.

After this initial screening, children in each subtype were

matched on the basis of chronological age, mental age, and sex,

resulting in 14 pairs. Another matched group of 14 normal readers

served as a control group. Four measures assumed to reflect either a

sequential or a simultaneous processing strategy were administered:

(a) a memory for faces test (simultaneous), (b) Wechsler Intelligence

Scale for Children-Revised (WISC-R) Digit Span (sequential), (c)

written reproduction of closely paired letter stimuli as either a

single gestalt (simultaneous) or as two discrete letters (sequential),

and (d) written reproduction of individual letters and shapes. On the

final measure, mirror image errors on a delayed reproduction task were

interpreted as evidence for simultaneous processing of the stimulus as

a visual gestalt without regard for left-right orientation.

Performance on this battery supported the existence of two

distinct subgroups of dyslexic children. The simultaneous-deficient





-13-


group was impaired on the memory for faces task relative to the

sequential-deficient and control groups. Conversely, the sequential-

deficient group was impaired on Digit Span and made more mirror image

reversals in their delayed reproductions of letters and shapes.

Finally, the sequential-deficient group also reproduced closely paired

letters as a single fused gestalt more frequently than the other two

groups. Aaron (1982) concludes that these results are congruent with

his hypothesis that dyslexic children are characterized by an

imbalance in encoding strategies which is not characteristic of normal

readers. His findings also support the construct validity of Boder's

(1973) subtypes and the processing strategies that she associated with

them.

Surface vs. Deep Dyslexia

Studies of alexia, an acquired reading disorder secondary to

neurological damage, have also focused on the identification of

subtypes with different clinical presentations. Marshall and Newcombe

(1966, 1973) have described two dissociable symptoms in alexic

patients based on an examination of reading errors. These

investigators propose that reading impairment resulting from left

hemisphere damage reflects a disruption of one of the two processing

routes necessary for normal reading.

One reading route or mechanism, referred to as "visual,"

"lexical," or "whole word," involves visual recognition of the printed

word as a unitary entity. Once recognized, the visual word form is

believed to directly access its semantic representation, without

phonologic mediation. The second route, described as a phonologicc"









or "grapheme-to-phoneme" conversion process, requires the derivation

of a phonemic representation of the printed form by the application of

conversion rules, prior to accessing semantics (Coslett, Gonzalez-

Rothi, & Heilman, 1985).

Disruption of each of these reading mechanisms is associated with

a characteristic pattern of reading deficits. Impairment of the

visual or whole word mechanism has been termed "surface dyslexia"

(Marshall & Newcombe, 1973). These patients have difficulty reading

aloud orthographically irregular words, for which there is no direct

grapheme-to-phoneme correspondence. They read by means of a direct

letter-by-letter phonemic transcoding, resulting in paralexic errors

which are phonologically similar to the target word.

Another group of alexic patients is able to read familiar (high

frequency) words, regardless of their conformity to phonemic

conversion rules. Their errors are primarily visual or semantic in

nature, rather than phonemic substitutions. However, they are unable

to read either nonsense words or function words such as prepositions

and articles. Their reading performance improves with concrete words

or words presented in a semantic context. Marshall and Newcombe

(1973) referred to this syndrome as "deep dyslexia."

Alexic reading patterns reflect both deficits in the processing

system and available compensatory strategies employed to overcome

these deficits. A direct analysis of reading errors is an essential

tool for the identification of both impaired and spared routes in

patients with acquired reading disorders. Four types of errors are

commonly described in the literature on acquired alexia.










Semantic errors are responses which are within the same semantic

category as the target word. Coltheart (1980) describes two types of

semantic errors. In the first type, the response is a synonym of the

target. For example, the target "merry" might be read as "happy."

Coltheart refers to this as a shared feature semantic error. In fact,

this type of semantic error may be divided into two categories. The

first subtype is the superordinate error in which the response is a

semantic superordinate of the stimulus (e.g., robin + "bird"). The

second is the co-ordinate error in which both target and response are

exemplars of a single semantic category (e.g., robin + "sparrow").

Subordinate errors (e.g., dog + "poodle") are logically possible but

rarely encountered. The second category of semantic errors is

characterized by an associative link between target and response

(e.g., merry + "Christmas").

Phonemic errors sound like the target word. These can also be

divided into two types. Some phonemic errors result from an

impairment of the grapheme-to-phoneme conversion process. Others

result from the misapplication of the rules of English orthography.

For example, failure to apply the rule that a terminal "e" lengthens

the preceding vowel would result in the response "cap" for "cape."

Visual errors are responses which are visually similar to the

target, sharing a significant number of graphemes (e.g., chair +

"chain"). Finally, derivational errors are responses which share a

common root morpheme with the target word (e.g., walk + "walking").

Overlap between these categories occurs frequently, particularly for

visual and phonemic errors. This classification problem has often


-15-





-16-


resulted in the introduction of additional categories for errors which

meet the criteria for more than one type.

Reading Development and Reading Routes

Studies of reading development have supported the existence of

two general mechanisms used by children who are learning to read

words. One mechanism is based on spelling-sound correspondence rules

or "phonics." The other mechanism relies on direct knowledge about

the visual word form. "Look-say" teaching techniques capitalize on

this mechanism. The phonic approach is useful in reading

orthographically regular words and nonsense words. It is also the

only method available when the developing reader faces a new and

unfamiliar word. The visual approach cannot be used to read nonsense

words or novel words, but is well suited for reading familiar words

and orthographically irregular words which do not conform to spelling-

sound correspondence rules. The relative importance of these two

mechanisms and the efficacy of teaching techniques emphasizing one or

the other have been controversial topics in educational philosophy.

Rather than opting for one or the other of these extreme

viewpoints, one might assume that both the look and the sound are

important for children learning to read. Frith (1985) has postulated

a three-phase theory of reading acquisition that emphasizes the

predominance of different processing stategies at different stages of

development. The following three stages are hypothesized to occur in

invariant sequence. Each new strategy is assumed to capitalize on the

previous one.






-17-


The logographic phase encompasses the initial phase of reading

acquisition. Reading during this stage is based on logographic skills

which permit the instant recognition of familiar words, probably on

the basis of salient features. Phonologic features are secondary to

recognition. The use of the logographic strategy allows the child to

amass a sizeable sight vocabulary. Reading by the "look-say" method

emphasizes these skills.

At the second phase, an alphabetic strategy is adopted.

Alphabetic skills involve the knowledge and use of grapheme-phoneme

correspondence rules. This is an analytic skill, based on a

systematic, sequential approach. During this stage, the child learns

more than letter-sound correspondence; the more complex and context

sensitive rules of English orthography are acquired during this

phase. The teaching of "phonics" emphasizes the alphabetic strategy.

Finally, the developing reader adopts an orthographic strategy.

This strategy permits the instant analysis of words into orthographic

units without intervening phonological conversion. These orthographic

units are believed to correspond to morphemes and are internally

represented as abstract letter-by-letter strings. Frith (1985)

describes this strategy as both nonvisual and nonphonological. It is

viewed as the product of a merger of instant recognition and

sequential analytic skills, each of which has been predominant at an

earlier phase.

Frith's (1985) model does not address how each of the initially

acquired strategies is continued in skilled reading. The degree to

which each distinct strategy remains accessible and continues to be






-18-


employed is unknown. Clearly, a variety of contextual variables would

play a key role in the pattern of strategy utilization.

Acquired Alexia and Developmental Dyslexia

The literature on both normal reading development and

developmental dyslexia in children provides convergent support for the

existence of two distinct processing mechanisms in reading. Without

empirical demonstration, one cannot assume the identity of processing

routes referred to by different names. However, the apparent

similarity between phonologicc," "sequential," and "alphabetic"

strategies on the one hand, and between "visual," "simultaneous," and

"logographic/orthographic" strategies on the other, is quite

compelling. More importantly, the same set of constructs has proven

useful in understanding the loss of reading competence in adults after

brain injury. Unfortunately, the studies of acquired and

developmental reading disorders have proceeded independently, with

only occasional acknowledgement of congruent findings. However, there

has been some recent discussion concerning the possibility that

developmental dyslexia may exhibit similarities to one or more of the

syndromes of acquired alexia.

Jorm (1979) proposed that developmental dyslexia is similar to

deep dyslexia (Marshall & Newcombe, 1973) in that developmental

dyslexics have a specific impairment of the ability to process written

material via a phonological route. He cites unpublished work by Firth

in 1972 in which groups of dyslexic and normal readers were given a

large battery of tests believed to tap various component skills in the

reading process. According to Jorm (1979), Firth found that tests of





-19-


nonsense word reading, auditory blending, and auditory analysis best

discriminated the two groups. The information presented was

insufficient to evaluate these findings. In particular, the selection

criteria for subjects and the means of assessing visual word form

processing were not described. Jorm (1979) also points out that

dyslexics have been shown to learn the meanings of logographic Chinese

characters with ease (Rozin, Poritsky, & Sotsky, 1971) and are

comparable to normals in their ability to directly associate meanings

with word-like visual forms (Jorm, 1977).

On this basis, Jorm (1979) concludes that the visual processing

route is intact and that developmental dyslexia results from a unitary

deficit in phonological processing. He attributes this deficit to

dysfunction of the left inferior parietal lobule. No evidence that

developmental dyslexics present with other aspects of the deep

dyslexia syndrome, such as semantic, visual, and derivational errors,

impaired reading of functors, or an imageability effect was presented.

In fact, although Boder (1973) makes reference to "semantic

substitution" errors, such errors are rarely reported in clinical

descriptions of developmental dyslexics. However, two recent papers

have supported their occurrence in this population. R. S. Johnston

(1983) presented data from an 18-year-old disabled reader who produced

semantic errors when reading aloud single words. This patient showed

the same pattern of reading errors as the cases of acquired deep

dyslexia reported in the literature (Coltheart, Patterson, & Marshall,

1980). She made semantic, visual, derivational, and function word

substitution errors and was unable to read nonwords. She also showed





-20-


imageability and part of speech effects. However, the proportion of

semantic errors was only 3%. Thus, this patient's reading disturbance

may be more similar to the acquired alexia known as phonological

alexia (Beauvois & Derouesne, 1979; Shallice & Warrington, 1975).

Phonological alexics have a severe impairment of the phonological

reading route, as evidenced by their inability to read nonsense words,

but do not make semantic errors in reading aloud. Temple and Marshall

(1983) have reported a case study of developmental phonological alexia

in a 17-year-old girl of average intelligence. She was impaired at

nonword reading relative to word reading and frequently responded to

nonwords with lexicalizations. A large proportion of her errors were

derivational and visual paralexias. No semantic errors were noted.

More recently, Siegel (1985) presented evidence for semantic

errors in reading single words in a group of six dyslexic children,

ages 7 and 8. These children were unable to read even simple

nonwords. Siegel and Ryan (1984) have demonstrated the ability to

read simple nonwords in normal readers of this age. The majority of

their responses were omissions; however, when responses were given,

they were frequently real words. Lexicalization of nonwords in this

manner has been described in deep dyslexic adults (Marshall &

Newcombe, 1973). Two comparison groups, normal controls and other

dyslexic children, were employed. No semantic substitutions occurred

in either control group. Thus, Siegel (1985) raises the possibility

that only a small subset of dyslexic children will exhibit reading

deficits analogous to those associated with the syndrome of deep

dyslexia.





-21-


Other authors have argued that developmental dyslexia is

analogous to surface dyslexia as described by Marshall and Newcombe

(1973). Marshall and Newcombe attributed surface dyslexia to a

moderate to severe impairment of the direct route from visual written

forms to semantic representations, combined with a lesser deficit in

knowledge of grapheme-phoneme correspondences. The vast majority of

errors in this group of patients can be described as partial failures

of grapheme-phoneme conversion rules, in particular, context sensitive

rules.

Holmes (1978) studied single word reading in four male dyslexics,

ages 9 13, and reported that the majority of errors made by these

subjects were conversion failures. They had difficulty reading

irregular words and typically produced regularization errors.

Coltheart, Masterson, Byng, Prior, and Riddoch (1983) described

several cases of surface dyslexia, both developmental and acquired.

The two groups were characterized as quite similar in reading

performance. However, all of the developmental surface dyslexics also

made visual errors, suggesting some incompetence of the phonologic

route as well. In contrast, acquired surface dyslexics exhibited

greater variability in the commission of visual errors; some adult

patients never made errors of this type. Finally, Temple (1984)

presented the case of a 13-year-old male with epilepsy, diagnosed at

18 months. His reading performance was consistent with surface

dyslexia in that he had difficulty with irregular words and displayed

homophone confusion. However, unlike the developmental surface






-22-


dyslexics described by Coltheart et al. (1983), this patient made no

visual errors. All of his misreadings were strictly rule governed.

Temple (1985) has proposed an expanded model of the phonological

route and utilized this model to more accurately isolate the nature of

the hypothesized phonological impairment in four developmental

dyslexics. Phonological processing is divided into three main

stages: (a) parser, (b) translator, and (c) blender. The parser

divides the grapheme sequence into a number of segments or chunks.

Chunks may consist of single or multiple graphemes. The translator

assigns each chunk to one of several potentially valid phonological

representations. Finally, these phonologic correspondences are

combined in the blender. Dysfunction at each of these stages would

result in a different type of error. Temple was able to demonstrate

qualitative differences in the performances of the developmental

surface dyslexics which could be explained by the expanded model.

Some cases appeared to represent a pure impairment of a single stage

of phonological processing while others were suggestive of more

widespread impairment.

It appears that initial efforts to apply knowledge about acquired

alexia to the study of developmental dyslexia have focused on

identifying the developmental disorder with one or the other of the

acquired syndromes. An awareness of heterogeneity, evident in the

dyslexia subtype literature, appears to have been replaced by the

unitary deficit model in the papers described above. A notable

exception to this trend is the meticulous analysis of reading

functions in four developmental dyslexics reported by Seymour and






-23-


MacGregor (1984). Based on an information processing model of

reading, specific cognitive-experimental paradigms were selected to

assess the functional integrity of the visual and phonological

processors and of semantic access. Error and reaction time data were

collected.

Findings strongly supported the existence of subtypes. Each of

the four developmental dyslexics exhibited a distinct pattern of

impairment with differential involvement of the hypothesized

processing mechanisms. Case 1 was described as a case of

developmental phonological dyslexia with a primary deficit in the

phonological processor. Knowledge of grapheme-phoneme correspondence

was impaired with consequent inability to read nonwords. Case 2 was

described as developmental morphemic dyslexia based on a primary

defect in the wholistic function of the visual processor. This

subject was unable to handle letter arrays as wholes and was

restricted to letter-by-letter processing compatible with the

phonological processor. Case 3 exhibited impairments of both the

visual and the phonological processor and was thus termed

multicomponent dyslexia. Finally, Case 4 appeared to have a

restricted impairment in the analytic function of the visual

processor. Seymour and MacGregor (1981) refer to this as

developmental visual analytic dyslexia. They propose that dysfunction

at this level of processing affects the speed at which the visual

processor can operate in the analytic mode without preventing the

development of the phonological or the visual wholistic processors.





-24-


Therefore, such dysfunction is primarily manifested in slowed response

times, particularly with nonwords which require segmental parsing.

A recent study in a closely related area provides support for

both the existence of two dissociable processing systems for written

language and the similarity of developmental and acquired disorders.

Roeltgen and Tucker (1986) examined acquired and developmental

agraphia. Acquired agraphia refers to an impairment in the production

of written language secondary to neurological damage. Agraphia may be

secondary to motoric deficits resulting in poor grapheme formation or

may reflect linguistic deficits as evidenced by specific patterns of

spelling disability. Assessment of linguistic agraphia patients has

supported the existence of two distinct spelling systems, one

phonological and one lexical or orthographic (Roeltgen & Heilman,

1985). These systems appear to be analogous to the phonological and

lexical reading mechanisms. Roeltgen and Tucker (1986) analyzed the

spelling performance of 22 adolescent and adult subjects with

developmental agraphia and compared it to that of control subjects and

subjects with acquired agraphia. Both developmental and acquired

graphics could be divided into phonological and lexical groups. More

importantly, the two groups of subjects with a developmental disorder

were almost indistinguishable from the two groups of subjects with an

acquired disorder. The only significant difference involved the

spelling of regular words. Both nonword and regular word spelling

were impaired in the acquired phonological graphics. In

developmental phonological graphics, regular word spelling was

preserved relative to the spelling of nonwords.






-25-


Statement of Purpose

Several lines of evidence have been discussed, all of which

support the existence of strong parallels between developmental and

acquired reading disorders. The notion of two distinct subtypes based

on the relative impairment and preservation of two dissociable

processing mechanisms is also consistent with theories of normal

reading which postulate dual reading routes. Such parallels are not

surprising. If ontogenetic development is assumed to proceed along

clearly delineated paths, then degeneration is likely to proceed along

these same paths which would serve as natural lines of fracture

(Aaron, Baxter, & Lucenti, 1980).

Unfortunately, awareness of these parallels has rarely been

translated into empirical exploration. Direct and systematic

investigation of phonological and lexical reading deficits in

children, with paradigms designed to highlight acquired/developmental

parallels, have been rare. In particular, despite widespread

discussion of the relative merits of phonologic vs. whole word

reading, there has been little effort directed toward developing

clinical measures designed to disentangle these two reading mechanisms

in children. The assessment system of Boder and Jarrico (1982)

represents an initial step in this direction. However, their system

is not based on either a theory of the reading process or a model of

the neurological basis of reading.

Fortunately, such a measure has been developed for adults.

Gonzalez-Rothi, Coslett, and Heilman (1984) have developed a clinical

tool, the Battery of Adult Reading Function, to identify the integrity






-26-


of the lexical and phonological routes to reading in patients with

acquired alexia. The direct analysis of reading errors is an

essential component of this measure. Dissociation of these two

reading processes was demonstrated using this battery in a patient

with mixed transcortical aphasia (Coslett et al., 1985). Their

findings suggested that the lexical and phonologic conversion systems

were both functionally and anatomically distinct.

The Battery of Adult Reading Function contains six subtests and

two appendices. Subtests 1 4 contain 30 words each, to be read

aloud by the subject. All word lists are balanced for both word

frequency (Kucera & Francis, 1967) and number of graphemes per word.

Each subtest contains the following word type: (a) Subtest 1--

phonologically possible nonwords, (b) Subtest 2--regular words, (c)

Subtest 3--rule governed words, and (d) Subtest 4--irregular words.

Subtests 5 and 6 assess silent reading. Subtest 5 consists of 2

practice plates and 18 test plates each containing one picture and two

printed words. The words are homophonic but not homographic

(identical sound but different spelling) and the subject is asked to

point to the word that goes with the picture. Target words and foils

are matched for word frequency and number of graphemes. Subtest 6

contains 2 practice and 18 test plates, each containing one word and

three pictures. Each word is a nonword which is homophonic to a real

English word. The subject's task is to point to the picture which

depicts that real word. Words depicted by the other pictures begin

with the same phoneme as the target word and are again matched for

word frequency and graphemes.






-27-


Appendix A assesses the subject's ability to read functors.

Sixty words are presented to be read aloud; 30 are contentives (nouns,

verbs, and adjectives) and the remainder are functors (prepositions

and articles) matched for frequency and graphemes. Appendix B tests

written production by dictating individually the word pairs used in

Subtest 5 within the context of a sentence.

On Subtests 1 4, the subject's oral production is scored both

for accuracy and specific error type. The following error types are

described in the battery manual:

1. semantic--a semantic associate of the target (ocean

for fish),

2. phonologic--phonologically similar to the target (fine

for phone),

3. visual--visually similar to target; no more than 30%

of graphemes different (cat for eat),

4. visual/phonological--visually and phonologically

similar (fin for fun),

5. derivational--words sharing a common root morpheme

(jump for jumping),

6. other--any not included above. (Gonzalez-Rothi et

al., 1984, p. 6)

Subtests 5 and 6 and Appendix A are scored for error rate. The

writing test included in Appendix B is scored for both error rate and

error type (homophonic/nonhomograph or other).

Gonzalez-Rothi et al. (1984) make the following predictions

regarding differential reading performance by subjects relying on the

lexical or the phonological route. Subjects employing a lexical route





-28-


should read aloud regular, rule governed, and irregular words better

than nonwords. Errors should be predominantly visual or semantic.

Performance on Subtest 5 should be superior to Subtest 6 since 5 is

based on semantic guidance while 6 requires phonologic transcoding.

Finally, on Appendix A, particular difficulty reading functors

relative to contentives is predicted.

Conversely, the use of a phonological route should lead to

difficulty in reading irregular words relative to nonwords, regular

words, and possibly rule governed words. Errors should be primarily

phonological and performance on Subtest 6 should be superior to

Subtest 5. Reading of functors in Appendix A should be equivalent or

superior to the reading of contentives.

This study employed the Battery of Adult Reading Function

(Gonzalez-Rothi et al., 1984) to assess the integrity of the

phonological and lexical reading mechanisms in a group of reading-

impaired children. This measure has proven useful in distinguishing

subtypes of acquired alexia and has been designed to reflect a sixth

grade reading level. Thus, it should be appropriate with older

children. Use of a measure from the acquired alexia literature

maximized the comparability of subtypes which emerged. Utilization of

the Battery of Adult Reading Function also allowed the direct

examination of both oral and silent reading, unconfounded by spelling

and writing. Previous studies (Aaron, 1982; Boder, 1973) have equated

input and output mechanisms by focusing on spelling errors in their

classification schemes. However, reading and spelling, although

closely related, utilize different neurological mechanisms (Frith,

1983; Heilman & Rothi, 1982).







-29-


The performance of an impaired reading group on this battery was

compared to the performance of age matched groups of normal and

superior readers. The majority of studies of reading disability

subtypes have not included these controls. Their inclusion made it

possible to determine whether these same subtypes occur in unimpaired

readers or are idiosyncratic to the impaired group. The assumption

that utilization of both routes is necessary for skilled reading was

also evaluated by examining these unimpaired groups.

A timed passage comprehension task, the Brigance Diagnostic

Inventory of Basic Skills (Brigance, 1977) was also administered.

This measure permitted an assessment of both silent reading

comprehension and reading rate. The two hypothesized reading routes

should differ in processing speed, i.e., lexical processing should

occur more rapidly than phonological processing. Reading rates of

lexical and phonological readers, classified by the Battery of Adult

Reading Function, were compared. The Brigance Diagnostic Inventory of

Basic Skills also provided an independent assessment of reading

comprehension in these subjects.

Finally, the application of an acquired alexia model to

developmental dyslexia has the advantage of facilitating an

understanding of the hypothesized neurological basis of this

disorder. Localization of brain lesions in alexic patients has

provided valuable information regarding the anatomic substrate for

each of the hypothetical processing mechanisms (Heilman & Rothi,

1982). The proposed neurological etiology for developmental dyslexia

has proven difficult to demonstrate. It is likely that this






-30-


difficulty results from the nature of the neuropathology, which may be

at a microstructural level. Galaburda and his colleagues (Galaburda &

Kemper, 1979; Galaburda, Sherman, Rosen, Aboitiz, & Geschwind, 1985)

have reported cytoarchitectonic abnormalities in several developmental

dyslexics who have come to autopsy. These abnormalities were in the

region of the left angular gyrus. The identification of parallels

between acquired and developmental reading impairments may provide

important clues as to the location of such subtle abnormalities.

The following general hypotheses were proposed:

1. Impaired, normal, and superior readers will exhibit different

patterns of performance on the Battery of Adult Reading

Function (BARF).

2. Subtypes based on pattern of performance on the BARF will be

identifiable only within the impaired group.

3. Impaired, normal, and superior readers will exhibit different

patterns of performance on the Brigance Diagnostic Inventory

of Basic Skills (Brigance).

4. Performance on the BARF will be related to reading rate on the

Brigance.

The following specific predictions were tested:

1. Impaired readers will make more errors and have slower

response times on BARF as compared to normal and superior

readers.

2. Normal and superior readers will not have significantly

different error rates on the BARF.






-31-


3. Superior readers will have more rapid response times on the

BARF relative to normal readers.

4. Within the impaired group, three subgroups will be

identifiable based on their patterns of performance on the

BARF.

A. A phonological reading group will be characterized by the

following pattern of responding:

1) Increased errors and/or response time on Subtest 4

relative to Subtest 1.

2) The production of neologistic responses on Subtest 1.

3) Increased errors and/or response time on Subtest 5

relative to Subtest 6.

4) Error rate and response time for functors in Appendix

A equal to or less than for contentives.

5) A predominance of phonological and visual/phonological

errors.

B. A lexical reading group will be characterized by the

following:

1) Increased errors and/or response time on Subtest 1

relative to Subtest 4.

2) The production of lexical (real word) responses on

Subtest 1.

3) Increased errors and/or response time on Subtest 6

relative to Subtest 5.

4) Error rate and response time for functors in Appendix

A greater than for contentives.






-32-


5) A predominance of visual, derivational, and semantic

errors.

C. A mixed deficit reading group will be characterized by the

following:

1) Increased errors and/or response time on both Subtest

1 and Subtest 4.

2) The production of lexical (real word) responses on

Subtest 1.

3) Increased errors and/or response time on both Subtest

5 and Subtest 6.

4) Error rate and response time for functors in Appendix

A equal to contentives.

5) The presence of all error types.

5. Analogous subtypes within normal and superior readers will not

be observed.

6. Impaired readers will make more errors and have slower reading

rates on the Brigance as compared to normal and superior

readers.

7. Normal and superior readers will not have significantly

different error rates on the Brigance.

8. Superior readers will have more rapid reading rates than

normals on the Brigance.

9. Those subjects whose performance on the BARF suggests a

lexical reading route will have more rapid reading rates than






-33-



those subjects whose BARF performance suggests a phonological

reading route.















CHAPTER 2
METHODS


Subjects

Ninety subjects were selected from the population of 12-year-old

males enrolled in the Alachua County school system. They were chosen

on the basis of their reading scores on the Metropolitan Achievement

Test (1978), standardly administered by this school district. This

measure is self-administered and involves reading brief stories and

answering multiple choice questions about them. Thus, it is best

described as a test of silent reading comprehension. Three groups

were defined, each composed of 30 subjects. Impaired readers had

reading percentile ranks which placed them at least 2 years below

their current grade level. Normal readers had reading scores within 1

year of current grade level. Superior readers had attained scores

which placed them at least 2 years above current grade level. All

subjects had obtained a standard score of 80 or above on the Otis-

Lennon School Abilities Test (1979), also standardly administered by

the school district. This measure is also a self-administered

multiple choice test. It is designed to assess a wide variety of

cognitive abilities related to scholastic aptitude. However, its

method of administration and verbal format suggest that reading

comprehension skills are essential to performance. Subjects were

identified from school records. Parents were contacted by


-34-






-35-


telephone and interviewed regarding their child's developmental

history. Children with gross sensory, emotional or neurological

handicaps were not chosen to participate. Parents who agreed to

participate were mailed an informed consent form (see Appendix 1)

which was to be returned to their child's teacher.

Stimulus Materials

Task 1: Battery of Adult Reading Function

All reading subtests of the Battery of Adult Reading Function

(Gonzalez-Rothi et al., 1984), i.e., Subtests 1 6 and Appendix A

were administered. Subtests 1 4 contain 30 words each, to be read

aloud by the subject. Each word was printed in lower case letters on

a 3 X 5 card to allow individual presentation. Word lists in each

subtest are balanced for both word frequency (Kucera & Francis, 1967)

and number of graphemes per word. Each subtest contains the following

word type: (a) Subtest 1--phonologically possible nonwords (see

Appendix 2), (b) Subtest 2--regular words (see Appendix 3), (c)

Subtest 3--rule governed words (see Appendix 4), and (d) Subtest 4--

irregular words (see Appendix 5).

Subtests 5 and 6 assess silent reading. Subtest 5 consists of 2

practice plates and 18 test plates, each containing one picture and

two printed words. The words are homophonic but not homographic

(i.e., identical pronunciation but different spelling) and the

subject is asked to point to the word that goes with the picture.

Target words and foils are matched for word frequency and number of

graphemes (see Appendix 6). Subtest 6 contains 2 practice and 18 test

plates, each containing one word and three pictures. Each word is a





-36-


nonword which is homophonic to a real English word. The subject's

task is to point to the picture that depicts that real word. Words

depicted by the other pictures begin with the same phoneme as the

target word and are again matched for word frequency and number of

graphemes (see Appendix 7).

Appendix A assesses the subject's ability to read functors.

Sixty words are presented to be read aloud; 30 are contentives (nouns,

verbs, and adjectives) and the remainder are functors (articles and

prepositions), matched for frequency and number of graphemes (see

Appendix 8). Each word was printed in lower case letters on a 3 X 5

card. Words were presented individually in a manner identical to

Subtests 1 4.

Task 2: Brigance Diagnostic Inventory of Basic Skills

Reading rate and comprehension were assessed using the Brigance

Diagnostic Inventory of Basic Skills (Brigance, 1977). This measure

consists of paragraphs of approximately 100 words, equated for units

of information, which the subject is asked to read silently. Passage

reading is timed in order to calculate a reading rate in words per

minute. Comprehension of each paragraph is measured by the subject's

accuracy in responding to five multiple choice questions. The five

questions assess comprehension on a literal level, focusing on clearly

stated factors or details, sequence, cause and effect, vocabulary, and

main theme. Paragraphs are graded in difficulty from the primer to

the eleventh grade level.






-37-


Apparatus

Oral reading responses on Subtests 1 4 and Appendix A were

audiotaped using a Sony Walkman Professional Stereo Cassette-Corder,

Model WM-D6, augmented by a Realistic Omnidirectional Electret Tie

Clip Microphone, Model 33-1056. Recordings were made on Maxell

XLII-90 Epitaxial cassettes. Reaction times and reading rate were

recorded on a Micronta LCD Quartz Chronograph, Model 63-5009A,

accurate to 1/100 of a second.

General Procedure

Testing sessions were scheduled at the subject's school after

parental informed consent was returned. Each subject completed the

measures in a single testing session, lasting approximately 1 hour.

Subjects were tested individually in a quiet room. Order of

administration of the two measures was counterbalanced within each

reading ability group. Order of subtest presentation on the Battery

of Adult Reading Function was standardized for all subjects. Subtests

1 6 were administered in sequence, followed by Appendix A. Only the

first response produced for each stimulus was scored.

Specific Procedures

Task 1: Battery for Adult Reading Function

In Subtests 1 4 and Appendix A, the subject was presented with

a 3 X 5 card on which a single word was printed in lower case letters

and instructed to read it aloud. On Subtest 1, subjects were informed

that the stimuli were nonwords and asked to pronounce them as they

thought they should be pronounced. If a real word was produced, the

instructions were repeated. Subsequent real word responses were not






-38-


followed by reminders. The examiner phonetically transcribed each

response and scored its accuracy. Reaction times for each stimulus

word were obtained using a manually operated stopwatch. An audiotape

of each subject's responses was recorded in order to permit an

assessment of inter-rater reliability both for response accuracy and

classification of error type. In Subtest 5, the subject was presented

with the standard test plate and instructed to "Point to the word that

goes with the picture." In Subtest 6, the subject was told that,

"Although this isn't a real word, it sounds like a real word that goes

with one of these pictures. Point to the picture that goes with the

word." Response choice and reaction time were recorded for Subtests 5

and 6.

Task 2: Brigance Diagnostic Inventory of Basic Skills

Each subject was presented with a typed paragraph commensurate

with his reading grade level. He was instructed to, "Read this

paragraph silently as quickly and carefully as you can. I will be

asking you some questions about it when you're done. Begin when I say

'Go' and let me know when you are finished." Reading time was

recorded by manually operated stopwatch. Immediately after completing

the passage, each subject was given five comprehension questions,

typed on a single sheet, and asked to "Circle the best answer for each

question." Four possible responses appeared for each item.
















CHAPTER 3
RESULTS


Overview of the Analysis

This chapter will begin with a description of scoring procedures

used with both experimental tasks, the BARF and the Brigance, and a

discussion of the interrater reliability associated with these

procedures. Descriptive statistics will then be presented to

characterize each reading group in terms of both classification

variables and dependent measures. Discussion of five separate

analyses will follow. These analyses were designed to assess (a) the

contribution of between-subjects factors to performance on the BARF

and the Brigance, (b) the contribution of within-subject factors to

performance on the BARF, (c) the contribution of general scholastic

ability to performance on both tasks, (d) the presence of subtypes

within this population using cluster analytic techniques, and (e) the

presence of subtypes within this population using clinical-inferential

techniques.

Analysis of between-subjects factors was performed using

multivariate analysis of variance (MANOVA), with error percentage and

mean response time as the dependent variables. Two between-subjects

factors, Reading Group and Race, were included in the model. Follow-

up univariate analyses of variance (ANOVA) were performed for each

dependent variable. Tukey's Studentized Range Test was used for


-39-






-40-


posttest comparison of group means. Separate analyses were performed

for each of nine pairs of error percentages and mean response times on

the BARF and for error percentage and reading rate on the Brigance.

Thus, each pair of dependent variables was treated as an independent

task in this analysis. Univariate ANOVAs were also used to analyze

the contribution of between-subjects factors to variance in four

separate error types. Reading Group and Race were again included in

the model and Tukey's test was used for comparison of group means.

Analysis of within-subject factors utilized a multivariate

repeated measures ANOVA. This analysis allowed an assessment of

performance differences across subtests of the BARF and of Subtest x

Reading Group interactions. Subtests were treated as repeated

measures rather than independent tasks. Planned comparisons were made

across theoretically meaningful groups of subtests and were performed

separately for the three types of dependent variables: (a) error

percentages, (b) mean response times, and (c) error types. Duncan's

Multiple Range Test was used for these comparisons. Race was no

longer included in the model since between-subjects analysis had not

revealed a significant effect for this factor.

The contribution of general scholastic ability to performance on

both experimental tasks was explored with an analysis of covariance.

Performance on the Otis-Lennon School Abilities Test was observed to

differ across reading groups and was designated the covariate in this

analysis. Each of the MANOVAs and ANOVAs from the between-subjects

analysis was repeated with the effects of general scholastic ability

held constant.






-41-


The three analyses described above involve multiple F tests and

multiple posttest comparisons of means performed on the same data

set. Caution must be exercised in this situation since the

performance of multiple tests increases the probability of making a

Type I error. Type I error is defined as the rejection of the null

hypothesis when it in fact reflects the true state of affairs in the

population (Winer, 1971). In order to reduce the probability of this

type of error, the alpha level was set conservatively at .01 for all

the significance tests performed in this study.

A cluster analytic technique termed FASTCLUS (SAS Institute,

Inc., 1985) was employed to search for subtypes in this population.

This program is based on a K-means iterative partitioning technique

and is useful when specific subtypes are expected based on an explicit

theoretical model. Mean profiles of expected clusters can be

specified as "seeds" and the program clusters observations around

these profiles. Seed profiles were created using a small subset of

dependent variables believed to be most predictive of reading

mechanism. Five clusters were predicted containing (a) superior

readers, (b) normal readers, (c) lexical readers, (d) phonological

readers, and (e) readers with deficits in both reading routes. The

external validity of the cluster solution was assessed using

univariate ANOVA followed by Duncan's tests to compare the Brigance

reading rates of the derived clusters.

Finally, a set of classification rules (see Appendix 10) was

generated based on the theoretical model of reading disability

subtypes developed by Gonzalez-Rothi et al. (1984). Two subtypes were











defined: (a) phonological readers and (b) lexical readers. Subjects

who did not meet the criteria for either subtype were placed in an

unclassified group. Visual inspection of individual profiles was used

to classify each subject. Brigance reading rates were compared across

these subgroups as well.

Scoring

Battery of Adult Reading Function

Error percentages were computed for Subtests 1 6 and for both

Functors and Contentives in Appendix A. Mean response times were

computed for each of these tasks, based only on correct trials. The

total error rate and total mean response time across all tasks were

also calculated. Each error on Subtests 1 4 was classified by type

based on the taped recording of each subject's responses. Errors

could be classified as (a) semantic, (b) phonological, (c) visual, (d)

visual-phonological, (e) derivational, or (f) other, according to the

criteria described in Appendix 9. The percentage of total errors

which fell into each category was computed across all subtests.

Ultimately, these six categories were consolidated to form three

independent error types. "Phonological" errors included all

phonological and visual-phonological errors. "Lexical" errors

included all semantic, visual, and derivational errors. Finally,

errors which did not meet the criteria for any error type were

maintained in the category referred to as "Other." Finally, a

lexicalization index was computed. This index reflected the

percentage of responses on Subtest 1 which were real words.


-42-






-43-


Brigance Diagnostic Inventory of Basic Skills

Error percentage, based on five multiple choice questions, and

reading rate in words per minute were computed for each subject.

Reliability

Interrater reliability for judgments of response accuracy and

error type classification on the BARF was assessed utilizing taped

recordings of subject responses. Nine subjects, constituting 10% of

the total sample, were randomly selected to assess reliability of

response accuracy judgments. Two independent raters achieved 97%

agreement in rating responses as correct or incorrect. Percent

agreement on individual subtests ranged from 91% on Subtest 1 to 99%

on Appendix A. A second sample of protocols was selected to assess

the reliability of error type classification judgments. Three

subjects were randomly selected from each reading group to form a

subset of nine subjects or 10% of the total sample. Selection from

each group was employed to maximize the likelihood that a wide variety

of error types would be represented. Two independent raters achieved

82% agreement in classifying errors according to the criteria

presented in Appendix 9.

Descriptive Statistics

Percentile means and standard deviations for the Metropolitan

Achievement Test (1978) and the Otis-Lennon School Abilities Test

(1979) are presented for each reading group in Table 1. The racial

composition of each reading group is displayed in Table 2. Table 3

contains means and standard deviations of all error percentages on the

BARF by reading group. These values are not suggestive of floor or






-44-


Table 1
Percentile Means and Standard Deviations
"Metropolitan Achievement Test--Reading"
Abilities Test"


by Reading Group for
and "Otis-Lennon School


Superior Normal Impaired


Metropolitan X = 85.47 51.07 20.00
S.D. = 9.36 5.67 8.76

Otis-Lennon X 77.13 59.10 34.77
S.D. 18.12 20.00 15.52


Table 2
Racial Composition of Reading Groups for "Metropolitan Achievement
Test--Reading" and "Otis-Lennon School Abilities Test"


Superior Normal Impaired


Black 7% 33% 57%

White 93% 67% 43%






-45-


Table 3
Means and Standard
Reading Function":


Deviations by Reading Group for "Battery of Adult
Error Percentage


Superior Normal Impaired


Total


Subtest 1


Subtest 2


Subtest 3


Subtest 4


Subtest 5


Subtest 6


Appendix A:
Functors

Appendix A:
Contentives


x -
S.D. -

S.D.

S.D. -

X =
S.D. =

X =-
S.D. -
X =
S.D. -

X
S.D. -

X -
S.D. -

X
S.D. -

X -
S.D.


13.33
4.71

33.47
13.89

4.57
5.35

8.27
6.40

22.60
11.01

14.70
9.34

17.07
8.05

4.80
8.51

4.67
11.94


20.67
8.58

39.27
19.73

9.17
8.30

19.53
11.61

40.17
12.83

20.33
11.28

23.20
14.10

7.57
5.78

4.97
6.33


34.00
10.53

60.77
16.29

21.30
15.28

35.67
14.14

59.30
12.05

25.47
15.12

38.70
11.68

18.93
13.10

11.60
9.01






-46-


ceiling effects for accuracy in any of the groups. Comparable data

for mean response times on the BARF appear in Table 4. Table 5

contains the means and standard deviations of error types within each

reading group. These values represent the percentage of total

errors. Finally, means and standard deviations for error percentage

and reading rate on the Brigance appear in Table 6.

Assessment of Between-Subjects Factors

Battery of Adult Reading Function

Multivariate Analysis of Variance (MANOVA). A two-factor

(Reading Group, Race) MANOVA was performed on nine separate pairs of

error percentages and mean response times (Total Performance, Subtests

1 6, Appendix A: Functors, and Appendix A: Contentives).

Significant MANOVAs appear in Table 7.

Analysis of Variance: Error percentage. Error percentages were

analyzed using univariate ANOVA with two between-subjects factors

(Reading Group, Race). The results of these analyses revealed highly

significant main effects of Reading Group for the following tasks:

(a) Total Performance, (b) Subtests 1 4 and 6, and (c) Appendix A:

Functors. The effect of Reading Group approached significance for

Subtest 5 and Appendix A: Contentives. No other significant main

effects or interactions were found (see Tables 8 16). Thus,

impaired readers made significantly more errors than normal and

superior readers on all BARF subtests except Subtest 5 and Appendix

A: Contentives. Posttest comparison of group means using Tukey's

Studentized Range Test appear in Table 17. Significant differences

between all three groups were demonstrated for (a) Total Performance,






-47-


Table 4
Means and Standard Deviations by Reading Group for "Battery of Adult
Reading Function": Mean Response Time


Superior Normal Impaired
secss.) (sees.) secss.)


Total X 1.13 1.21 1.69
S.D. 0.27 0.20 0.65

Subtest 1 X 1.27 1.33 2.21
S.D. 0.34 0.37 1.37

Subtest 2 X 0.84 0.86 1.19
S.D. = 0.17 0.17 0.65

Subtest 3 X 0.85 0.88 1.26
S.D. 0.16 0.16 0.67

Subtest 4 X 0.91 0.93 1.11
S.D. 0.21 0.17 0.33

Subtest 5 X 1.41 1.71 2.36
S.D. 0.41 0.57 1.02

Subtest 6 X 2.18 2.35 3.41
S.D. 0.88 0.84 1.70

Appendix A: X 0.82 0.83 1.02
Functors S.D. 0.17 0.12 0.26

Appendix A: X = 0.78 0.79 0.96
Contentives S.D. 0.16 0.12 0.24





-48-


Table 5
Means and Standard
Reading Function":


Deviations by Reading Group for "Battery of Adult
Cumulative Percentage of Error Types


Error Type


Visual/
Phono- Phono- Deriva-
Semantic logic Visual logical tional Other


Superior X = 0.23 0.37 5.10 92.30 1.90 0.10
S.D. = 0.90 1.13 5.23 6.31 2.81 0.55

Normal X 0.20 0.67 6.53 88.80 3.33 0.47
S.D. 0.61 1.58 4.99 7.09 4.10 1.01

Impaired X = 0.30 0.77 15.70 77.97 3.37 1.53
S.D. = 0.65 1.25 10.91 13.12 2.66 1.87









Table 6
Means and Standard Deviations by Reading.Group for "Brigance
Diagnostic Inventory of Basic Skills": Error Percentage and Reading
Rate


Superior Normal Impaired


Error % X 17.33 29.33 27.33
S.D. 16.39 25.59 27.53

Reading Rate X 187.03 148.93 112.63
(words per S.D. 54.47 38.01 41.84
minutes)






-149-


Table 7
Analysis of Between-Subjects Factors for "Battery of Adult Reading
Function": Significant MANOVAs


Total Performance

Subtest 1

Subtest 2

Subtest 3

Subtest 4

Subtest 5

Subtest 6

Appendix A:
Functors

Appendix A:
Contentives


READING

READING

READING

READING

READING

READING

READING


GROUP

GROUP

GROUP

GROUP

GROUP

GROUP

GROUP


READING GROUP


READING GROUP


F(4,164) 13.96

F(4,164) 9.39

F(4,164) 6.97

F(4,164) 14.29

F(4,164) 16.28

F(4,164) 4.91

F(4,164) 11.37


F(4,164) 6.85


F(4,164) = 4.70


Table 8
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Total Performance


Source DF SS F PR > F


GROUP 2 3863.33 27.30 0.0001

RACE 1 5.15 0.07 0.7879

GROUP*RACE 2 20.19 0.14 0.8673


p = .0001

p = .0001

p = .0001

p = .0001

p = .0001

p = .0009

p = .0001


p = .0001


p = .0013






-50-


Table 9
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Subtest 1


Source DF SS F PR > F


GROUP 2 8977.74 15.75 0.0001

RACE 1 113.52 0.40 0.5297

GROUP*RACE 2 458.54 0.80 0.4509


Table 10
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Subtest 2


Source DF SS F PR > F


GROUP 2 3067.16 13.49 0.0001

RACE 1 7.82 0.07 0.7937

GROUP*RACE 2 49.41 0.22 0.8051






-51-


Table 11
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Subtest 3


Source DF SS F PR > F


GROUP 2 7107.45 27.74 0.0001

RACE 1 129.40 1.01 0.3178

GROUP*RACE 2 18.30 0.07 0.9311


Table 12
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Subtest 4


Source DF SS F PR > F


GROUP 2 9295.02 31.42 0.0001

RACE 1 32.92 0.22 0.6383

GROUP*RACE 2 74.25 0.25 0.7786






-52-


Table 13
Univariate ANOVA: Error Percentage
Function": Subtest 5


for "Battery of Adult Reading


Source DF SS F PR > F


GROUP 2 693.06 2.42 0.0953

RACE 1 16.41 0.11 0.7359

GROUP*RACE 2 508.20 1.77 0.1762


Table 14
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Subtest 6


Source DF SS F PR > F


GROUP 2 4481.92 16.39 0.0001

RACE 1 8.82 0.06 0.8001

GROUP*RACE 2 117.47 0.43 0.6522






-53-


Table 15
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Functors


Source DF SS F PR > F


GROUP 2 2217.57 11.63 0.0001

RACE 1 4.92 0.05 0.8208

GROUP*RACE 2 32.69 0.17 0.8427









Table 16
Univariate ANOVA: Error Percentage for "Battery of Adult Reading
Function": Contentives


Source DF SS F PR > F


GROUP 2 797.97 4.40 0.0152

RACE 1 10.23 0.11 0.7379

GROUP*RACE 2 34.96 0.19 0.8250











Table 17
Tukey's Studentized Range Test: Error Percentage for "Battery of
Adult Reading Function"


Reading Group


Impaired Normal Superior


Total Performance
Error % 34.00 20.67 13.33

Subtest 1
Error % 60.77 39.27 33.47

Subtest 2
Error % 21.30 9.17 4.57

Subtest 3
Error % 35.67 19.53 8.27

Subtest 4
Error % 59.30 40.17 22.60

Subtest 6
Error % 38.70 23.20 17.07

Appendix A: Functors
Error % 18.93 7.57 4.80






-55-


(b) Subtest 3, and (c) Subtest 4. On Subtests 1, 2, and 6 and

Appendix A: Functors, impaired readers made more errors than normal

and superior readers. The error percentages of normal and superior

readers were not significantly different.

Analysis of Variance: Mean response time. Mean response times

were analyzed using univariate ANOVA with two between-subjects factors

(Reading Group, Race). The results of these analyses revealed highly

significant main effects of Reading Group for the following tasks:

(a) Total Performance; (b) Subtests 1-3, 5, and 6; (c) Appendix A:

Functors; and (d) Appendix A: Contentives. The effect of Reading

Group approached significance for Subtest 4. No other significant

main effects or interactions were found (see Tables 18 26). Thus,

impaired readers had significantly slower response times than normal

and superior readers on all BARF subtests except Subtest 4. Posttest

comparisons of group means using Tukey's Studentized Range Test appear

in Table 27. On all subtests, impaired readers performed more slowly

than both normal and superior readers who did not differ from each

other.

Analysis of Variance: Error type. Four error type variables

were analyzed using univariate ANOVA with two between-subjects factors

(Reading Group, Race): (a) Phonological errors, (b) Lexical errors,

(c) Other errors, and (d) the Lexicalization Index. Results revealed

highly significant main effects of Reading Group for Phonological,

Lexical, and Other errors (see Tables 28 30). The Lexicalization

Index did not significantly differ as a function of Reading Group (see

Table 31). No other significant main effects or interactions were






-56-


Table 18
Univariate
Function":


ANOVA: Mean Response Time for
Total Performance


"Battery of Adult Reading


Source DF SS F PR > F


GROUP 2 3.74 10.22 0.0001

RACE 1 0.02 0.12 0.7309

GROUP*RACE 2 0.03 0.07 0.9319









Table 19
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 1


Source DF SS F PR > F


GROUP 2 11.79 8.50 0.0004

RACE 1 0.40 0.58 0.4494

GROUP*RACE 2 2.86 2.06 0.1334






-57-


Table 20
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 2


Source DF SS F PR > F


GROUP 2 1.67 5.17 0.0077

RACE 1 0.05 0.29 0.5902

GROUP*RACE 2 0.27 0.84 0.4343


Table 21
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 3


Source DF SS F PR > F


GROUP 2 2.31 6.71 0.0020

RACE 1 0.02 0.12 0.7336

GROUP*RACE 2 0.06 0.19 0.8290






-58-


Table 22
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 4


Source DF SS F PR > F


GROUP 2 0.52 4.29 0.0168

RACE 1 0.00 0.00 0.9545

GROUP*RACE 2 0.03 0.25 0.7785









Table 23
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 5


Source DF SS F PR > F


GROUP 2 8.34 7.93 0.0007

RACE 1 0.00 0.01 0.9208

GROUP*RACE 2 0.09 0.09 0.9174






-59-


Table 24
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Subtest 6


Source DF SS F PR > F


GROUP 2 19.16 6.39 0.0026

RACE 1 0.00 0.00 0.9628

GROUP*RACE 2 0.81 0.27 0.7631









Table 25
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Functors


Source DF SS F PR > F


GROUP 2 0.57 7.57 0.0009

RACE 1 0.17 0.45 0.5054

GROUP*RACE 2 0.07 0.93 0.3977






-60-


Table 26
Univariate ANOVA: Mean Response Time for "Battery of Adult Reading
Function": Contentives


Source DF SS F PR > F


GROUP 2 0.43 6.88 0.0017

RACE 1 0.00 0.02 0.8819

GROUP*RACE 2 0.09 1.39 0.2556





-61-


Table 27
Tukey's Studentized Range Test: Mean Response Time for "Battery of
Adult Reading Function"


Reading Group


Impaired Normal Superior


Total Performance
RT (sees.) 1.69 1.21 1.13

Subtest 1
RT (sees.) 2.21 1.33 1.27

Subtest 2
RT (sees.) 1.19 0.86 0.84

Subtest 3
RT (sees.) 1.26 0.88 0.85

Subtest 5
RT (sees.) 2.36 1.71 1.41

Subtest 6
RT (sees.) 3.41 2.35 2.19

Appendix A: Functors
RT secss.) 1.02 0.83 0.82

Appendix A: Contentives
RT (sees.) 0.95 0.79 0.77





-62-


Table 28
Univariate ANOVA: Error Type for "Battery of Adult Reading
Function": Phonological Errors


Source DF SS F PR > F


GROUP 2 2181.94 12.64 0.0001

RACE 1 2.90 0.03 0.8549

GROUP*RACE 2 34.59 0.20 0.8188


Table 29
Univariate ANOVA: Error Type for "Battery of Adult Reading
Function": Lexical Errors


Source DF SS F PR > F


GROUP 2 1774.14 11.58 0.0001

RACE 1 3.79 0.05 0.8246

GROUP*RACE 2 32.03 0.21 0.8117





-63-


Table 30
Univariate ANOVA: Error Type for "Battery of Adult Reading
Function": Other Errors


Source DF SS F PR > F


GROUP 2 21.09 6.36 0.0027

RACE 1 0.06 0.04 0.8512

GROUP*RACE 2 0.14 0.04 0.9593


Table 31
Univariate ANOVA: Error Type for "Battery of Adult Reading
Function": Lexicalization Index


Source DF SS F PR > F


GROUP 2 950.97 2.03 0.1377

RACE 1 0.03 0.00 0.9917

GROUP*RACE 2 141.72 0.30 0.7397





-64-


found. Posttest comparisons of group means using Tukey's Studentized

Range Test appear in Table 32. Impaired readers made significantly

fewer Phonological errors than normal and superior readers who did not

differ from each other. Impaired readers made significantly more

Lexical and Other errors than normal and superior readers, who again

were not significantly different.

Brigance Diagnostic Inventory of Basic Skills

Multivariate Analysis of Variance (MANOVA). A two-factor

(Reading Group, Race) MANOVA was performed with error percentage and

reading rate as the dependent variables. The main effect of Reading

Group was highly significant (F(4,164) 6.10, p=.0001). No other

multivariate analyses were significant.

Analysis of Variance: Error percentage. Error percentage was

analyzed using univariate ANOVA with two between-subjects factors

(Reading Group, Race). No significant main effects or interactions

were found (see Table 33). Thus, error percentage on the Brigance was

unrelated to either reading group membership or race.

Analysis of Variance: Reading rate. Reading rates were

analyzed using univariate ANOVA with two between-subjects factors

(Reading Group, Race). The results of this analysis revealed a

highly significant main effect of Reading Group. No other

significant main effects or interactions were found (see Table 34).

Posttest comparisons of group means using Tukey's Studentized Range

Test appear in Table 35. Superior readers had significantly faster

reading rates than normals who had significantly faster rates than

impaired readers.





-65-


Table 32
Tukey's Studentized Range Test: Error Type for "Battery of Adult
Reading Function"


Reading Group


Impaired Normal Superior


Phonological
% Total Errors 78.73 89.47 92.67

Lexical
% Total Errors 19.73 10.07 7.23

Other
% Total Errors 1.53 0.47 0.10


Table 33
Univariate ANOVA: Error Percentage for "Brigance Diagnostic Inventory
of Basic Skills"


Source DF SS F PR > F


GROUP 2 847.73 0.74 0.4794

RACE 1 367.23 0.64 0.4250

GROUP*RACE 2 32.97 0.03 0.9716





-66-


Table 34
Univariate ANOVA:
Basic Skills"


Reading Rate for "Brigance Diagnostic Inventory of


Source DF SS F PR > F


GROUP 2 46770.70 11.04 0.0001

RACE 1 621.95 0.29 0.5894

GROUP*RACE 2 461.84 0.11 0.8969













Table 35
Tukey's Studentized Range Test: Reading Rate for "Brigance Diagnostic
Inventory of Basic Skills"


Reading Group


Impaired Normal Superior


Reading Rate
(words per minute) 112.63 148.93 187.03





-67-


Assessment of Within-Subject Factors

In this analysis, BARF subtests were treated as repeated measures

rather than as independent tasks. Repeated measures ANOVAs with one

between-subjects factor (Reading Group) and one within-subject factor

(Subtest) were performed. Race was dropped from the model due to the

nonsignificant findings in the between-subjects analysis. Three sets

of planned comparisons were made within theoretically related groups

of subtests. Comparisons were made between (a) Subtests 1 4, (b)

Subtests 5 and 6, and (c) Appendix A: Functors and Appendix A:

Contentives. Separate analyses of error percentage, mean response

time, and error type data will be presented. Since the between-

subjects effects of Reading Group were discussed above, only within-

subject main effects and interactions will be reported here.

Error percentage. Repeated measures ANOVA on Subtests 1 4

produced a significant main effect of Subtest, as well as a

significant Reading Group x Subtest interaction (see Table 36).

Results of Duncan's Multiple Range Test, performed on subtest means by

reading group, appear in Table 39. In both normal and impaired

readers, fewest errors occurred on Subtest 2 (regular words),

followed by Subtest 3 (rule governed words), with highest error rates

on Subtests 1 (nonwords) and 4 (irregular words). Significant

differences were revealed between all subtests except 1 and 4.

However, superior readers poorest performance was on Subtest 1

(nonwords) followed by 4 (irregular), 3 (rule governed), and





-68-


Table 36
Repeated Measures ANOVA: Error Percentage for "Battery of Adult
Reading Function": Subtests 1-4, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 3 66370.72 328.88 0.0

SUBTEST*GROUP 6 3542.86 8.78 0.0001


Table 37
Repeated Measures ANOVA: Error Percentage for "Battery of Adult
Reading Function": Subtests 5-6, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 1 1705.09 15.30 0.0002

SUBTEST*GROUP 2 1129.01 5.07 0.0083


Table 38
Repeated Measures ANOVA: Error Percentage for "Battery of Adult
Reading Function": Functors/Contentives, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 1 506.69 17.17 0.0001

SUBTEST*GROUP 2 401.64 6.80 0.0018





-69-


2 (regular). The difference between Subtests 3 and 2 was not

significant.

Analysis of Subtests 5 and 6 showed a significant main effect of

Subtest as well as a significant Reading Group x Subtest interaction

(see Table 37). Table 40 displays the results of Duncan's test on

subtest means, by reading group. Error rates of normal and superior

readers on Subtests 5 and 6 were not significantly different.

Impaired readers made significantly more errors on Subtest 6, which

requires a phonological reading route, compared to Subtest 5 which

requires lexical reading.

Finally, a significant Subtest effect and Subtest x Group

interaction were also shown in the analysis of Functors and

Contentives (see Table 38). Results of Duncan's test again showed

that normal and superior readers perform similarly with these two word

classes while impaired readers have significantly more difficulty with

functors.

Mean response time. Repeated measures ANOVA on Subtests 1 4

produced a significant main effect of Subtest and a significant

Reading Group x Subtest interaction (see Table 42). Results of

Duncan's Multiple Range Test, performed on subtest means by Reading

Group, appear in Table 45. In all three reading groups, significantly

slower response times occurred on Subtest 1 (nonwords). Response

times to Subtests 2 (regular words), 3 (rule governed words), and 4

(irregular words) were not significantly different. However, impaired

readers differed from the other two groups by a nonsignificant

tendency to response more quickly to Subtest 4 (irregular words).





-70-


Table 39
Duncan's Multiple Range Test: Error Percentage for "Battery of Adult
Reading Function": Subtests 1-4


Subtest


Group 1 2 3 4


Impaired Error % 60.77 21.30 35.67 59.30

Normal Error % 39.27 9.17 19.53 40.17

Superior Error % 33.47 4.57 8.27 22.60


Table 40
Duncan's Multiple Range Test: Error Percentage for "Battery of Adult
Reading Function": Subtests 5-6


Subtest


Group 5 6


Impaired Error % 25.47 38.70

Normal Error % 20.33 23.20

Superior Error % 14.70 17.07





-71-


Table 41
Duncan's Multiple Range Test: Error Percentage for "Battery of Adult
Reading Function": Functors/Contentives


Subtest


Group Functors Contentives


Impaired Error % 18.93 11.60

Normal Error % 7.57 4.97

Superior Error % 4.80 4.67










Table 42
Repeated Measures ANOVA: Mean Response Time for "Battery of Adult
Reading Function": Subtests 1-4, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 3 26.21 66.87 0.0001

SUBTEST*GROUP 6 5.91 7.54 0.0001





-72-


Normal and superior readers tended to respond more quickly on Subtest

2 (regular words).

Analysis of Subtests 5 and 6 showed a significant main effect of

Subtest (see Table 43). Table 46 displays the results of Duncan's

test on subtest means, by reading group. In all readers, slower

response times occurred on Subtest 6, which requires phonological

reading, compared to Subtest 5 which requires the lexical route.

A significant main effect of Subtest was also shown in the

analysis of Functors and Contentives (see Table 44). Results of

Duncan's test showed that all subjects had slower response times to

Functors, as compared to Contentives.

Error type. Repeated measures ANOVA was also used to explore

within-subject effects on the occurrence of three types of errors:

(a) Phonological errors, (b) Lexical errors, and (c) Other errors.

The analysis was performed with Reading Group as the between-subjects

factor and Error Type as the within-subject factor. A significant

main effect for Error Type and a significant Reading Group x Error

Type interaction was found (see Table 48). The results of Duncan's

Multiple Range Test appear in Table 49. Phonological errors were the

most frequent error type, followed by Lexical errors, with Other

errors the least frequent in all reading groups. However, both

Lexical and Other errors made up a larger percentage of the total

errors made by impaired readers.

Assessment of the Contribution of General Scholastic Ability

Inspection of group means revealed substantial differences

between reading groups on the Otis-Lennon School Abilities Test






-73-


Table 43
Repeated Measures ANOVA: Mean Response Time for "Battery of Adult
Reading Function": Subtests 5-6, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 1 30.68 53.26 0.0001

SUBTEST*GROUP 2 1.32 1.15 0.3217


Table 44
Repeated Measures ANOVA: Mean Response Time for "Battery of Adult
Reading Function": Functors/Contentives, Within-Subject Effects


Source DF SS F PR > F


SUBTEST 1 0.10 30.14 0.0001

SUBTEST*GROUP 2 0.01 1.46 0.2367






-74-


Table 45
Duncan's Multiple Range Test: Mean Response Time for "Battery of
Adult Reading Function": Subtests 1-4


Subtest


Group 1 2 3 4


Impaired RT (sees.) 2.21 1.19 1.26 1.11

Normal RT (sees.) 1.33 0.86 0.88 0.93

Superior RT secss.) 1.27 0.84 0.85 0.91










Table 46
Duncan's Multiple Range Test: Mean Response Time for "Battery of
Adult Reading Function": Subtests 5-6


Subtest


5 6


RT (sees.) 1.83 2.65





-75-


Table 47
Duncan's Multiple Range Test: Mean Response Time for "Battery of
Adult Reading Function": Functors/Contentives


Subtest


Functors Contentives


RT (secs.) 0.89 0.84












Table 48
Repeated Measures ANOVA: Error Type for "Battery of Adult Reading
Function": Phonological, Lexical and Other for Within-Subject Effects


Source DF SS F PR > F


ETYPE 2 394272.96 2469.04 0.0

ETYPE*GROUP 4 5806.31 18.18 0.0001






-76-


Table 49
Duncan's Multiple Range Test: Error Type for "Battery for Adult
Reading Function": Phonological, Lexical, and Other


Error Type


Group Phonological Lexical Other


Impaired
ET (% Total E) 78.74 19.73 1.53

Normal
ET (% Total E) 89.46 10.07 0.47

Superior
ET (% Total E) 92.67 7.23 0.10





-77-


(Otis-Lennon) (1979). These differences in general scholastic ability

might account for the significant group differences in performance on

the BARF and the Brigance. In order to rule out this possibility,

score on the Otis-Lennon was designated the covariate in an analysis

of covariance. Each of the MANOVAs and ANOVAs from the between-

subjects analysis was repeated, including Otis-Lennon score as a

between-subjects factor. Error percentage and mean response time were

the dependent variables in the MANOVAs. Follow-up ANOVAs were

performed for each dependent variable in isolation. Univariate ANOVAs

were performed for each error type. Follow-up comparisons of least

squared means tested the hypothesis of no Reading Group or Race

differences, with Otis-Lennon score held constant at its mean value.

Battery of Adult Reading Function

Multivariate Analysis of Variance (MANOVA). A three-factor

(Otis-Lennon, Reading Group, Race) MANOVA was performed on nine

separate pairs of error percentages and mean response times (Total

Performance, Subtests 1 6, Appendix A: Functors, and Appendix A:

Contentives). Only one of these MANOVAs was significant. For Subtest

5, the main effect of Otis-Lennon score was highly significant

(F(2,82) 8.44, p-.0005). No other multivariate analyses were

significant.

Analysis of Variance: Error percentage. Error percentage was

analyzed using univariate ANOVA with three between-subjects factors

(Otis-Lennon, Reading Group, Race). The only significant main effect

was for Otis-Lennon score on Subtest 5 which requires a lexical

reading route (see Table 50). Results of the least squared means






-78-


Table 50
Analysis of Covariance: Error Percentage for "Battery of Adult
Reading Function": Subtest 5 with Covariate = "Otis-Lennon School
Abilities Test"


Source DF SS F PR > F


OTIS

GROUP

RACE

GROUP*RACE


2009.39

41.04

20.40

546.40


16.62

0.17

0.17

2.26


0.0001

0.8442

0.6822

0.1107


Prob > t

i/j

1

2

3

1 Normal readers


Least Squares Means

HO: LSMEAN(i) -


1



0.8759

0.6014

2 = Impaired


2

0.8759



0.5720

readers


LSMEAN(j)

3

0.6014

0.5720



3 = Superior readers





-79-


procedure showed that the differences between reading groups on this

measure could be accounted for by group differences in performance on

the Otis-Lennon (see Table 50).

Analysis of Variance: Mean response time. Mean response times

were analyzed using univariate ANOVA with three between-subjects

factors (Otis-Lennon, Reading Group, Race). No significant main

effects or interactions involving the Otis-Lennon factor were found in

the analysis of response times. The significant main effect of

Reading Group was maintained.

Analysis of Variance: Error type. Four error type variables

were analyzed using univariate ANOVA with three between-subjects

factors (Otis-Lennon, Reading Group, Race): (a) Phonological errors,

(b) Lexical errors, (c) Other errors, and (d) the Lexicalization

Index. Otis-Lennon score was not involved in any significant main

effects or interactions for any of the four error types. The

significant main effect of Reading Group for Phonological, Lexical,

and Other errors was maintained.

Brigance Diagnostic Inventory of Basic Skills

Multivariate Analysis of Variance (MANOVA). A three-factor

(Otis-Lennon, Reading Group, Race) MANOVA was performed with error

percentage and reading rate as the dependent variables. Results of

this analysis revealed no significant main effects or interactions

involving the Otis-Lennon factor.

Analysis of Variance: Error percentage. Error percentage was

analyzed using univariate ANOVA with three between-subjects factors






-80-


(Otis-Lennon, Reading Group, Race). No significant main effects or

interactions involving Otis-Lennon score occurred.

Analysis of Variance: Reading rate. Reading rates were analyzed

using univariate ANOVA with three between-subjects factors (Otis-

Lennon, Reading Group, Race). No significant main effects or

interactions involving Otis-Lennon score occurred. The significant

main effect of Reading Group was maintained.

Subtype Identification: Multivariate Statistical

The FASTCLUS program (SAS Institute, Inc., 1985) was chosen to

search for subtypes in this population. This technique performs a

disjoint cluster analysis on the basis of Euclidean distances computed

from one or more quantitative variables. Observations are divided

such that every observation belongs to one and only one cluster. This

procedure is based on a method referred to as nearest centroid

sorting. A set of point called cluster seeds is chosen to serve as

initial guesses about the mean cluster profiles. Each observation is

assigned to the nearest seed to form temporary clusters. The seeds

are then replaced by the means of these temporary clusters and the

matching process is repeated. This sequence is repeated until no

further changes occur in the cluster solution. This method is most

appropriate when specific subtypes are predicted based on an explicit

theoretical model.

It was necessary to select a small subset of the many dependent

variables assessed in this study to be entered into the clustering

algorithm. Initially, six dependent variables were chosen, based on






-81-


their presumed predictive utility in differentiating phonological and

lexical reading mechanisms. Two of these variables werb newly created

for this analysis. The first was termed HOMO and was defined as

Error % on Subtest 5 minus Error % on Subtest 6. This variable

reflected the difference in performance on a measure of lexical

reading (Subtest 5) relative to a measure of phonological reading

(Subtest 6). The second new variable was referred to as FCDIFF and

was defined as Error % on Functors minus Error % on Contentives. Four

other previously discussed dependent variables were selected: (a)

Error % on Subtest 1 (EP1), (b) Error % on Subtest 4 (EP4), (c) the

Lexicalization Index (WORD1), and (d) Lexical errors (LEXERR).

The initial set of six variables was examined further to assess

their appropriateness for this analysis. The frequency distribution

of each variable was visually inspected. Only one variable, LEXERR

appeared to have a markedly deviant distribution. Secondly,

intercorrelation matrices were generated, separately for each reading

group, to determine the amount of redundancy contained in this set of

variables. In both impaired and normal readers, EP1, EP4, and LEXERR

were highly correlated. In superior readers, EP1 and EP4 were highly

correlated.

A decision was made to drop LEXERR from the set of clustering

variables. This removed the only variable with a deviant frequency

distribution, eliminating the need to standardize the variables before

clustering. One pair of highly correlated variables remained, EP1 and

EP4. However, neither of these was removed from the set due to their

presumed importance for subtype identification.






-82-


Five clusters were predicted containing (a) Superior readers, (b)

Normal readers, (c) Lexical readers, (d) Phonological readers, and (e)

readers with deficits in both reading routes. Mean profiles for each

of these predicted clusters across the five clustering variables

appear in Table 51. These seeds were generated based on the BARF

profiles predicted for phonological and lexical readers (Gonzalez-

Rothi et al., 1984). Superior and Normal readers should (a) perform

equally on Subtests 1 and 4, (b) have a relatively low percentage of

lexicalizations, and (c) perform equally on Subtests 5 and 6 and on

Functors and Contentives. These two groups will differ only

quantitatively, with Superior readers making fewer errors. Lexical

and Phonological readers will make more errors than Normal readers but

will not differ quantitatively from each other. Lexical readers will

(a) make more errors on Subtest 1 relative to Subtest 4, (b) have a

larger percentage of lexicalizations, (c) make more errors on Subtest

6 than Subtest 5, and (d) make more errors on Functors than on

Contentives. Phonological readers will (a) make more errors on

Subtest 4 than Subtest 1, (b) have a relatively low percentage of

lexicalizations, (c) make more errors on Subtest 5 than Subtest 6, and

(d) perform equally on Functors and Contentives. Finally, mixed

deficits readers will make more errors than all other clusters. These

readers will (a) perform equally on Subtests 1 and 4, (b) have a

relatively high percentage of lexicalizations, and (c) perform equally

on Subtests 5 and 6 and on Functors and Contentives.





-83-


Table 51
Initial Seeds: FASTCLUS Procedure


Cluster Name EP1(%) EP4(%) WORD1(%) HOMO(%) FCDIFF(%)


1 Superior 10 10 25 0 0

2 Normal 20 20 25 0 0

3 Lexical 60 30 50 -20 20

4 Phonological 30 60 25 20 0

5 Mixed Deficits 60 60 50 0 0





-84-


A summary of the cluster solution appears in Table 52. The

following frequencies were obtained: (a) Superior 9, (b) Normal =

31, (c) Lexical 13, (d) Phonological 17, and (e) Mixed Deficits =

20. Although the distance between cluster means (or centroids) was

considerably larger than the standard deviation of distances within a

cluster, each cluster had at least one member that was farther away

from its mean than the centroid distance.

Cluster means and standard deviations appear in Table 53.

Although in most cases, the general shape and relative elevation of

the seed profile was maintained, some significant changes were noted:

1. Superior readers had a much smaller percentage of

lexicalizations than Normal readers.

2. The profile of Lexical readers no longer showed differential

performance on Subtests 1 and 4 and the percentage of

lexicalizations was much lower than predicted.

3. Mixed Deficits readers made more errors on Subtest 6 relative

to Subtest 5 and had greater difficulty with Functors relative

to Contentives than was predicted.

4. The performance of Lexical readers differed quantitatively

from that of Phonological readers, with more errors occurring

in the Lexical group.

Finally, the distribution of reading groups placed in each

cluster appears in Table 54. The majority of Superior readers were

placed in either the Superior or Normal cluster. However, a small

number of Superior readers appeared in the Phonological cluster. Over






-85-


Table 52
Cluster Summary: FASTCLUS Procedure


Max. Distance
RMS from Seed to Nearest Centroid
Cluster Name Freq. S.D.* Observation Cluster Distance**


1 Superior 9 8.72 31.03 2 29.74

2 Normal 31 10.44 44.43 4 25.65

3 Lexical 13 10.58 36.31 5 31.58

4 Phonological 17 10.59 34.97 2 25.65

5 Mixed
Deficits 20 11.84 38.08 3 31.58


The root-mean-square distance between observations in the cluster.

** The distance between the means of the current cluster and the
nearest other cluster.






-86-


Table 53
Cluster


Means and Standard Deviations, FASTCLUS Procedure


Cluster Name EP1(%) EP4(%) WORD1(%) HOMO(%) FCDIFF(%)


1 Superior
X 23.00 17.67 4.89 -5.33 4.33
S.D. 10.01 8.46 5.97 8.17 10.30

2 Normal
X = 34.03 25.65 30.84 -5.16 -0.71
S.D. 13.50 9.73 12.55 8.03 6.79

3 Lexical
X 52.85 53.62 26.31 -28.62 6.00
S.D. = 8.15 12.96 11.95 12.70 4.64

4 Phonological
X 36.00 43.47 30.53 12.53 4.12
S.D. 13.27 10.10 12.60 9.77 5.37

5 Mixed Deficits
X 72.20 63.60 38.60 -9.35 6.85
S.D. 10.45 7.49 16.33 12.52 10.57





-87-


Table 54
Cluster Membership by Reading Group


Cluster 1: Superior 9

Reading Group Superior 7
Normal 2


Cluster 2: Normal 31

Reading Group Superior 18
Normal = 11
Impaired 2


Cluster 3: Lexical 13

Reading Group Superior 1
Normal 4
Impaired 8


Cluster 4: Phonological 17

Reading Group Superior 4
Normal = 9
Impaired 4


Cluster 5: Mixed Deficits 20

Reading Group Normal 4
Impaired 16





-88-


half of Normal readers did not appear in either the Superior or Normal

cluster and almost one-third of Normal readers were classified as

Phonological readers. Approximately 50% of impaired readers were

placed in the Mixed Deficits cluster. Impaired readers were also

represented in both the Lexical and Phonological clusters. The ratio

of Lexical to Phonological impaired readers was 2:1. The two impaired

readers with the highest reading achievement scores appeared in the

Normal cluster.

Brigance reading rate was employed to assess the external

validity of the clusters with a variable that did not enter into their

derivation. Lexical reading is believed to be based on simultaneous

processing, while phonological decoding of words requires sequential

processing. Thus, phonological and lexical reading routes should

differ in processing speed, i.e., lexical processing should occur more

rapidly than phonological processing.

Univariate ANOVA with one between-subjects factor (Cluster) was

used to analyze Brigance reading rates. A highly significant main

effect of Cluster emerged (see Table 55). Posttest comparisons of

Cluster means using Duncan's Multiple Range Test also appear in Table

55. Although Lexical readers read more rapidly than Phonological

readers, this difference was not significant. However, the reading

rate of Lexical readers was significantly faster than that of Mixed

Deficits readers.

Subtype Identification: Clinical-Inferential

Visual inspection techniques were used to classify subjects into

lexical and phonological readers, based on a set of classification