Handbook of psychology volume 7 educational psychology

462

Learning Disabilities

with phonological processing, word recognition, or both is

used as the basis of the definition of a reading problem, then

disabled readers appear to have reasonably homogeneous

cognitive profiles and—in particular—deficits in the language

areas (e.g., Fletcher, 1985; Rourke & Finlayson, 1978;

Rourke & Strang, 1978; Siegel & Ryan, 1984, 1988, 1989b;

Strang & Rourke, 1983b). Therefore we and others (e.g.,

Siegel & Heaven, 1986; Siegel & Ryan, 1989a; Vellutino,

1978, 1979) argue that single-word or nonword reading con-

stitute the purest measures of reading and that an operational

definition of a reading disability should be based on nonword

tests to measure phonological skills, single-word tests to

measure word recognition skills, or both.

The classification of children with arithmetic problems is

equally problematic. As Siegel and Ryan (1989a) note, it

is almost impossible to find a group of reading-disabled

children who also do not have severe deficits in arithmetic.

At the same time, a number of investigators (Fletcher, 1985b;

Rourke & Finlayson, 1978; Rourke & Strang, 1978; Siegel &

Feldman, 1983; Siegel & Linder, 1984; Siegel & Ryan, 1984,

1988, 1989a) have found a group of learning-disabled

children with difficulties in arithmetic but with average or

above-average reading scores. Some evidence (e.g., Fletcher,

1985a; Rourke & Finlayson, 1978; Rourke & Strang, 1978;

Siegel & Feldman, 1983; Siegel & Linder, 1984; Siegel &

Ryan, 1984, 1988, 1989a) suggests that these children with

arithmetic deficits but normal reading (word recognition)

have cognitive profiles that are different from those of chil-

dren with reading difficulties. Therefore, it is important that

children designated as arithmetic LD not have problems that

are confounded by difficulties in reading.

SUBTYPES

Given the heterogeneity of LD groups, Siegel (1988b) con-

tends that if all LD children are grouped together, then inac-

curate conclusions may be reached. Evidence in support of

this position has been found by many investigators (e.g.,

Fletcher, 1985; McKinney, Short, & Feagans, 1985;

Rourke & Finlayson, 1978; Siegel & Linder, 1984; Siegel &

Ryan, 1984, 1988, 1989a, 1989b). For example, Siegel and

Ryan (1984, 1988, 1989a, 1989b) and Fletcher (1985) found

differences between specific arithmetic-disabled children

without reading problems and reading-disabled children in

both short-term and working memory. McKinney et al., using

a cognitive battery designed to assess a wide range of linguis-

tic and perceptual abilities, were able to classify 55 first-

and second-grade, school-identified LD children into six

subtypes. They then demonstrated that the three subtypes with

atypical cognitive profiles had poorer academic outcomes

than did the three groups with normal or near-normal profiles.

These differences might not have been evident had these chil-

dren been grouped together and not divided into subtypes.

Because of this heterogeneity of LD groups, considerable

effort has been made to identify specific subgroups of LD

children who share common attributes that distinguish them

from other subtypes. Not only do subtypes exist, but they also

seem to take several forms in terms of achievement patterns,

associated cognitive information-processing abilities, or

both. Furthermore, these subtypes may vary as a function of

etiology and age (e.g., McKinney et al., 1985; Rourke &

Finlayson, 1978; Satz & Morris, 1981; Short, Feagans,

McKinney, & Appelbaum, 1986; Siegel & Heaven, 1986). 

Subtyping Models

Early subtype approaches are based on clinical inferences

that have attempted to reduce complex data sets of subjects

into presumably homogeneous classes largely based upon

a priori considerations and visual inspection techniques. These

methods have been criticized for their inability to manage

simultaneously large quantities of information in an objec-

tive fashion as well as for the subjectivity that results from the

bias of clinical decisions made at various stages during the sub-

type development and subject classification (see Satz & Morris,

1981, and Hooper & Willis, 1989, for a complete review).

More recently, with the availability of advanced computer

technology, empirical classification models using applied

descriptive multivariate statistics have been developed. This

approach has involved a search for hidden structure in com-

plex multidimensional data sets generally involving cogni-

tive linguistic skills or direct measures of achievement

or behavior (e.g., Doehring & Hoshko, 1977; Feagans &

Appelbaum, 1986).

These methods also have difficulties. Hooper and Willis

(1989) contend that standards of reliability and validity have

frequently been overlooked or marginally addressed by inves-

tigators using these classification techniques. In addition, they

suggest that “the adequacy and strength of models derived by

empirical classification methods are influenced by many a pri-

ori clinical decisions including those regarding theoretical

orientation, sample selection, and variable selection” (p. 104).

Thus, the appropriate subtyping model remains an open ques-

tion and may depend on the type of research undertaken.

Academic Performance Models

In spite of the difficulties inherent in subtyping models, a num-

ber of investigators (e.g., Fletcher, 1985; Rourke & Finlayson,

Subtypes

463

1978; Rourke & Strang, 1978; Siegel & Feldman, 1983;

Siegel & Linder, 1984; Siegel & Ryan, 1984, 1988, 1989a,

1989b) have suggested that LD students in general can be di-

vided on the basis of their academic achievement as measured

by Wide Range Achievement Test (WRAT) scores in reading,

spelling, and arithmetic. Although each investigator has cre-

ated his or her own classification scheme, two broad cate-

gories of subtype groups have emerged. The first contains

children with at least reading deficits, and the second contains

children with at least arithmetic deficits and normal-to-above-

normal reading scores. Some authors (e.g., Fletcher, 1985;

Rourke & Finlayson, 1978; Rourke & Strang, 1978) have sub-

divided these groups further, basing their divisions on the

presence of other deficits. For example, Siegel and Linder

(1984) and Siegel and Ryan (1984, 1988, 1989a, 1989b) have

used only two academic subtypes: (a) an arithmetic-disabled

group, defined by scores equal to or below the 25th percentile

on the WRAT Arithmetic subtest and scores equal to or above

the 30th percentile on the WRAT Reading subtest and (b) a

reading-disabled group, defined by scores equal to or below

the 25th percentile on the WRAT Reading subtest and no

cutoffs for the other two WRAT subtests. Fletcher (1985b) has

developed the following four subtypes: (a) a reading-spelling-

disabled group, (b) a reading-spelling-arithmetic-disabled

group, (c) an arithmetic-disabled group, and (d) a spelling-

arithmetic-disabled group. According to this categorization,

the reading-spelling-disabled group is defined as consisting of

children with (a) WRAT Reading and Spelling subtest scores

below the 31st percentile, (b) WRATArithmetic subtest scores

above the 30th percentile, and (c) the arithmetic score must be

at least one-half standard deviation above the reading score on

the appropriate WRAT subtest. The reading-spelling-arith-

metic-disabled group is characterized by children with scores

on all three WRAT subtests below the 31st percentile. The

arithmetic-spelling-disabled group contains children who

have (a) WRAT Spelling and Arithmetic subtest scores below

the 31st percentile, (b) WRAT Reading subtest scores above

the 39th percentile, and (c) at least one standard deviation

between their reading and arithmetic scores. The arithmetic-

disabled group consists of children who have (a) WRAT

Reading and Spelling subtest scores above the 39th percentile,

(b) WRATArithmetic subtest scores below the 31st percentile,

and (c) difference of at least one standard deviation between

their reading and arithmetic scores. In contrast, a series of

studies by Rourke and his associates (Rourke & Finlayson,

1978; Rourke & Strang, 1978; Strang & Rourke, 1983a,

1983b) have identified the following three subtypes: (a) a gen-

eral disabled group (reading-spelling-arithmetic-disabled),

(b) a reading-disabled group, and (c) an arithmetic-disabled

group. These investigators defined their reading-spelling-

arithmetic groups as consisting of children with WRAT

subtest scores below the 19th percentile on all three subtests.

The reading-spelling-disabled group consisted of children

with (a) WRAT Arithmetic subtest scores at least 1.8 years

higher than their WRAT Reading and Spelling subtest scores

and (b) WRAT Reading and Spelling subtest scores below

the 15th percentile. The arithmetic-disabled group contained

children whose WRAT Reading and Spelling subtest scores

were at least 2 years above the WRAT Arithmetic subtest

scores.

Rourke and colleagues have developed subtypes of LD

children based on patterns of academic performance (Fisk

& Rourke, 1979, 1983; Ozols & Rourke, 1985; Porter &

Rourke, 1985; Petrauskas & Rourke, 1979; Rourke &

Finlayson, 1978; Rourke & Strang, 1978; Rourke & Fisk,

1981; Strang & Rourke, 1983, 1985a, 1985b; Sweeney &

Rourke, 1978). Depending on the particular investigation,

percentile cutoff scores of 20, 25, or 30 on the WRAT have

been used to define a particular LD group.

In one study, Rourke and Finlayson (1978) investigated

the performance of three groups of LD children on a neu-

ropsychological battery. The subtypes were formed on the

basis of their WRAT scores in reading, spelling, and arith-

metic. The first subtype exhibited uniformly deficient perfor-

mance in all three academic areas. They were found to be

superior to a specifically arithmetic-disabled group on

measures of visual-perceptual and visuospatial abilities. In

the second subtype, children were relatively good at arith-

metic as compared to their reading and spelling, but all areas

were below average. Superior visuospatial abilities and Per-

formance IQ scores were characteristic of this subtype. The

third subtype was composed of children whose reading and

spelling scores were average or above, but whose arithmetic

score was at least 2 years below that. This group exhibited

Verbal IQ scores that were higher than their Performance

scores. Their performance on measures of visual-perceptual-

organizational skills was somewhat deficient. Their per-

formance on measures of verbal and auditory-perceptual

abilities was superior to that of the first two groups, who were

reading disabled and low achievers in arithmetic. Rourke and

Strang (1978) found further deficiencies in this third subtype

on complex psychomotor measures and differences in hand

superiority on the Tactual Performance Test. Difficulties in

visuospatial orientation, including right-left problems, and

impaired bilateral tactile-perceptual abilities, including finger

agnosia, were also characteristic of this group.

Strang and Rourke (1983a, 1983b) found that the specific

arithmetic-disabled children made significantly more errors

than did the reading and arithmetic disabled children on the

Halstead category test. The arithmetic-disabled children had

464

Learning Disabilities

lower scores on those subtests, which “require a substantial

degree of ‘higher order’ visual-spatial analysis” (Strang &

Rourke, 1985b, p. 173). An analysis of the arithmetic errors

made by the third group (specifically arithmetic disabled) on

the WRAT subtest indicated a large number and variety of

errors. The quality of errors, as expected, changed somewhat

with age. The most prevalent types of mechanical errors were

identified as follows: (a) spatial organization, (b) visual de-

tail, (c) procedural errors, (d) failure to shift psychological

set, (e) graphomotor, (f) memory, and (g) judgment and rea-

soning, indicating deficits in visuospatial and perceptual-

motor abilities.

Ozols and Rourke (1985) compared the performance of

two groups of LD children to a group of average-achieving

children of the same age on four tasks—two verbal and

two nonverbal. The groups were identified as follows: (a) a

language-disorder group, who exhibited relatively well-

developed visuospatial abilities but poor auditory-perceptual

and language-related abilities (WRAT subtest percentile

scores in reading, spelling, and arithmetic were all below 25);

(b) a spatial-disorder group, who exhibited relatively well-

developed auditory-perceptual and language-related abilities

but poor visuospatial abilities (WRAT arithmetic subtest

score was the only percentile score below 25). The language-

disorder group performed significantly more poorly on the

verbal tasks than did the controls, and the spatial-disorder

group performed significantly more poorly than did the con-

trols on the nonverbal tasks.

Siegel and Ryan (1984) used performance on achievement

tests to subdivide LD children into more heterogeneous

groups. Three LD groups were created and compared to a

control group of normally achieving children. The LD groups

were (a) reading-disabled children who had low scores on

the WRAT Reading test, a test of word recognition skills;

(b) children who were arithmetic written work disabled who

had low scores on the WRAT Arithmetic test, a measure of

computational arithmetic skills; and (c) children with atten-

tion deficit disorder (ADD; hyperactive) but with no other

learning disabilities. Typically, the reading-disabled children

had significantly below-average ability to understand the

syntactic and morphological aspects of language and—at the

youngest ages—memory for sentences. None of the other LD

groups showed deficits in these areas. The reading-disabled

children had a significant deficit in reading and spelling non-

words and in recognizing the visual form of a spoken sound.

None of the other LD children had deficits in these areas,

with the exception of the youngest, specifically arithmetic-

disabled children, who had some difficulties with these

phonological skills in spite of their normal word-recognition

skills. This finding may be a result of the definition of

arithmetic disability, which at the youngest ages involves

memory skills in addition to computational skills. Thus, these

younger arithmetic-disabled children may represent the more

severely disabled children compared to older children who

have a specific arithmetic disability. The children with ADD

and no other difficulties with achievement did have signifi-

cantly lower scores on a reading comprehension test (which

is in reality a memory test and may require attentional skills),

but these children showed no other deficits.

All of the reading-disabled children had a severe deficit

involving phonological skills. They were quite homogeneous

in this respect. In most cases there was no overlap in the

scores of the reading-disabled and normally achieving chil-

dren of the same chronological age. Furthermore, for the

reading and spelling of nonwords, the reading-disabled chil-

dren performed more poorly than did younger controls

matched for reading level, indicating a very serious deficit.

The children with an arithmetic writing disability and the

children with a reading disability had significantly lower

scores on a short-term memory task involving visually pre-

sented letters, but Siegel and Linder (1984) found that for

auditorily presented letters, the arithmetic-disability group

performed in the normal range, indicating that the deficit of

this latter group was limited to short-term memory for visu-

ally presented information.

Fletcher (1985) used a verbal and nonverbal memory task

to test the hypothesis that subgroups of disabled learners

show different performance patterns. All tasks were based on

selective reminding procedures (i.e., in subsequent presenta-

tions, items are repeated only if they were not recalled on the

previous trial) to determine what inferences could be drawn

concerning storage and retrieval skills in these subgroups.

Control subjects had to obtain percentile scores above 39 on

the Reading, Spelling, and Arithmetic subtests of the WRAT

and show no history of achievement difficulties. Disabled

learners were placed into one of four groups depending on

their pattern of WRAT scores. The groups were (a) reading-

spelling-disabled (R-S), (b) reading-spelling-arithmetic-

disabled (R-S-A), (c) spelling-arithmetic-disabled (S-A), and

(d) arithmetic disabled (A).

The findings indicated that not all LD children showed

similar patterns of memory deficiencies. Children with read-

ing problems showed only retrieval difficulties on verbal

tasks, whereas children with arithmetic problems demon-

strated both storage and retrieval problems on nonverbal

tasks. Children with the R-S pattern had poorer verbal skills,

whereas children with the arithmetic pattern showed poorer

nonverbal skills. The R-S-A pattern had not been previously

studied, but children demonstrating this pattern had relatively

poor performance on the tasks used in this study.

Subtypes

465

Some studies were carried out in an attempt to determine

the relative diagnostic power of a number of tests in dis-

criminating between dyslexic children and those with other

learning disabilities. In one study by Rudel and Denckla

(1976), the performance of developmental dyslexic individu-

als was tested using visual-verbal tasks, such as naming pic-

tured objects, letters, and numbers. A 2-year or greater lag in

reading performance on a test of oral reading skill led to a clas-

sification of dyslexic. All participants were tested on rapid au-

tomatized naming tasks. Stimuli included colors, numerals,

use objects (e.g., comb), and high-frequency lowercase letters.

The naming speed on all four tasks for normally functioning

individuals was faster than for nondyslexic LD participants,

and nondyslexic participants were significantly faster than

dyslexic subjects.

In this study, the experimental children were termed learn-

ing disabled, but no reference was made to how they were so

designated. The authors stated that division of participants

into dyslexic and nondyslexic groups was done post hoc by

determining the difference between reading age and mental

age. Reading grade level was based on a test of oral reading

skill; no mention was made of what oral reading test was used.

The problems associated with such tests have already been

discussed. The lack of information on the measures used in

this study raises questions about the heterogeneity and sever-

ity of the problems demonstrated by the group being studied.

In a second study (Denckla, Rudel, & Broman, 1981), a

series of tasks was administered: a rapid automatized naming

task, the Oldfield-Wingfield Pictured Object Naming Test,

visual-Braille letter learning (a paired associate learning

task), visual-temporal spatial matching (a same or different

response was required of the participant after he or she

watched a series of light flashes emitted from a single sta-

tionary point source in a black box), and finally the silent

detection of repeated target symbols (a pencil-and-paper

version of the visual matching task paradigm). The dyslexic

group had a high percentage of dysphasic errors and pro-

longed times on repeated naming compared to the non-

dyslexic LD children.

In this study, the participants were defined as dyslexic on

the basis of a discrepancy between the WISC-R score and the

score on the Gray Oral Reading Test. The same problems as

in the earlier study emerge. The latter test is an oral reading

comprehension test that cannot be regarded as a pure test of

reading skill. Failure to do well on this test could be the result

of poor decoding skills, poor attention, poor memory, or poor

comprehension. The result is a heterogeneous sample with

findings that have limited external validity. 

Another attempt at subtyping on the basis of achievement

test scores is described by Satz and Morris (1981). Although

their subtyping scheme resembles Rourke’s, it is flawed

by the use of grade-level retardation on the WRAT subtests

for classification. This scheme is particularly problematic

with the many older children in the study.

Rather than finding significant heterogeneity on language

and memory functioning, these studies found homogeneity

within the population of disabled readers. It is important to

note that the studies that have found homogeneity within the

reading-disabled population have used retardation in word-

recognition skills to define the reading-disabled population.

In one very early study on subtyping that was unsuccessful, it

was found that all reading-disabled children had difficulty

with sound blending and short-term memory (Naidoo, 1972).

As discussed later in this chapter, studies that have used other

definitions (reading comprehension) do not find the same de-

gree of homogeneity. As further evidence of this point, Siegel

and Ryan (1989b) found that when a definition of reading

disability was used that involved poor word-recognition

or phonics skills, reading-disabled children had significant

problems with phonological processing and understanding of

syntax. However, children with low reading comprehension

scores but good phonics, word recognition skills, or both did

not show these problems. In summary, with achievement test

definitions such as those used by Rourke and associates,

Fletcher, and Siegel and colleagues, fairly homogeneous

groups of reading-disabled children have emerged.

Regardless of the exact criteria used, this method of sub-

typing had identified groups of LD children whose arithmetic

difficulties are not confounded by deficits in reading (word

recognition). The emergence of a specific arithmetic-disabled

subgroup has permitted investigators to clarify some of the

characteristics that distinguish this group from other LD

children with reading deficits.

Subtypes Within the Reading Disability Group

A number of investigators have assumed that a reading dis-

ability is not a single disorder but rather represents a group of

more specific subtypes. Some of these conceptualizations are

outlined in the following discussion. However, what appears

to be heterogeneity is really not; because of definitional inad-

equacies, heterogeneity emerges. Several schemes have been

used as subtyping systems. These schemes involve (a) the use

of achievement tests to classify all LD children with the

implicit assumption that all the children with a reading

disability will show a reasonable similarity in cognitive

performance and be different from those without a reading

disability (as was described earlier), (b) the use of patterns

of responses on reading tests to classify subgroups of dis-

abled readers, (c) the use of neuropsychological measures to