Handbook of psychology volume 7 educational psychology

Definitional Issues

457

458

Learning Disabilities

composed of two scales—a Verbal Scale in which the ques-

tions are largely based on language, from which one calcu-

lates a Verbal IQ, and a Performance Scale, from which one

calculates a Performance IQ. It is also possible to calculate a

total or full-scale IQ score. If a discrepancy between IQ and

achievement is used, then which IQ score should be used—the

Verbal, Performance, or full-scale score? It is quite possible to

be categorized as dyslexic on the basis of one IQ score but not

if another IQ score is used. For example, Valtin (1978–1979)

found that it makes a difference which scale is used, the

Verbal or Performance Scale. Children are often classified as

dyslexic when the classification is based on one scale, but they

are not considered dyslexic when the decision is based on the

other. In addition, depending on the type of IQ scores, the dif-

ference between the good and poor readers could be signifi-

cant or not significant, depending on which IQ score was used

in the definition.

There is also empirical evidence that suggests it is not nec-

essary to use the concept of intelligence in defining reading

disabilities. When children with reading disabilities were di-

vided into groups based on their IQ level and compared on a

variety of reading, language, memory, spelling, and phono-

logical tasks, there were no differences between the IQ

groups on the reading-related tasks (Siegel, 1988b). There-

fore, the reading-disabled group was quite homogeneous in

relation to reading-related skills; administering an IQ would

not provide useful information about performance differ-

ences on reading-related tasks.

One typical use of the IQ test is to use the IQ score to mea-

sure the discrepancy between IQ and academic achievement.

If there is a discrepancy, then the individual is said to have a

learning disability. If the individuals are poor readers but

show no discrepancy between their IQ and reading scores,

then they are not considered reading disabled. Fletcher et al.

(1989) maintains that it has not been clearly demonstrated

that children with discrepancies in IQ and achievement have

more specific disabilities than do poor achievers whose IQ

scores were not discrepant. In fact, they contend that there is

relatively little empirical evidence to show that similarly de-

fined children differ on measures other than IQ. In an epi-

demiological study on the influence of various definitions of

learning disabilities on the selection of children, Shaywitz,

Shaywitz, Barnes, and Fletcher (1986) found that although

variations in the use of IQ indexes led to the identification

of different children as learning disabled, there were few dif-

ferences in cognitive abilities. Moreover, there were few

differences among identified LD children with discrepant and

nondiscrepant IQ scores.

A significant number of studies have found no difference

in the reading, spelling, phonological skills, and even reading

comprehension of learning disabled individuals with high

and low IQ scores and no differences between individuals

with dyslexia and poor readers on measures of the processes

most directly related to reading (e.g., N. Ellis & Large, 1987;

Felton & Wood, 1991; Fletcher, Francis, Rourke, Shaywitz, &

Shaywitz, 1992; Fletcher et al., 1994; Friedman & Stevenson,

1988; Gottardo, Stanovich, & Siegel, 1996; Hall, Wilson,

Humphreys, Tinzmann, & Bowyer, 1983; Jiménez-Glez &

Rodrigo-López, 1994; Johnston, Rugg, & Scott, 1987a,

1987b; Jorm, Share, Matthews, & Maclean, 1986; Share,

McGee, McKenzie, Williams, & Silva, 1987; Siegel, 1988b,

1992, 1998; Silva, McGee, & Williams, 1985; Stanovich &

Siegel, 1994; Taylor, Satz, & Friel, 1979; Toth & Siegel,

1994). For example, Siegel (1992) compared two groups of

children who had low reading scores. One group, which was

composed of individuals with dyslexia, had reading scores

that were significantly lower than were those that were pre-

dicted by their IQ scores; the other group, the poor readers,

had low reading scores, but these reading scores were not sig-

nificantly lower than would be predicted by their IQ scores.

On a variety of reading, spelling, and phonological tasks,

results showed no differences between these two groups in

reading comprehension. In other words, there is no need to

use IQ scores to predict the difference between the individu-

als traditionally called learning disabled and those who have

equally poor achievement but also have lower IQ scores.

These results have also been replicated in a study of adults

with reading disabilities (Siegel, 1998). These findings sug-

gest that there is no need to use IQ tests to determine who is

learning disabled. We need to use only achievement tests. In

addition, IQ scores do not predict who is able to benefit from

remediation (Arnold, Smeltzer, & Barneby, 1981; Kershner,

1990; Lytton, 1967; Van der Wissel & Zegers, 1985;

Vellutino & Scanlon, 1996; Vellutino, Scanlon, & Lyon,

2000). One study (Yule, 1973) found that reading disabled

children with lower IQ scores made more gains than did read-

ing disabled children with higher scores.

Other research shows that reading problems may actually

interfere with the development of language, knowledge, and

vocabulary skills, further complicating the issue of the rela-

tionship between IQ and reading; this is called the Matthew

effect. The Matthew effect refers to the bidirectional relation-

ship between reading and cognitive development. Stanovich

(1986, 1988a, 1988b) has conceptualized the Matthew effect

as “the tendency of reading itself to cause further develop-

ment in other related cognitive abilities”—that is, IQ—such

that the “rich get richer and [the] poor get poorer” (Stanovich,

1986, p. 360). Certain minimum cognitive capabilities must

be present to begin reading; however, after reading com-

mences, the act of reading itself further develops these same

Definitional Issues

459

cognitive capabilities; this relationship of mutual reinforce-

ment is called the Matthew effect. This Matthew effect of a

reciprocal relationship between reading and other cognitive

skills is reflected in performance on an IQ test; consequently,

it undermines the validity of using an IQ-discrepancy-based

criterion because children who read more gain the cognitive

skills and information relevant to the IQ test and thus attain

higher IQ scores. Children with reading problems read less;

therefore, they fail to gain the skills and information neces-

sary for higher scores on the IQ test.

If the Matthew effect as described by Stanovich were

operating, it would be expected that IQ scores would decline

with increasing age because vocabulary and knowledge in-

crease as a result—at least in part—of experiences with print.

If reading disabled children have less experience with print

than do children without reading problems, then the chance

to acquire new knowledge is reduced and the IQ scores are

expected to decrease over time. In a cross-sectional study,

Siegel and Himel (1998) found that the IQ scores—in partic-

ular, vocabulary scores—of older dyslexic children were

significantly lower than were those of younger dyslexic chil-

dren. Similar declines in IQ and vocabulary were not noted

for the normally achieving readers—that is, children who

showed age-appropriate reading skills. However, standard

scores of the children with reading problems compared to

chronologically age-matched children remained relatively

constant with time, so that there was not an overall decline in

skills. Scores on a subtest of the IQ test, Block Design, a test

of visuospatial as opposed to verbal skills, also remained

constant over time. This subtest measures visuospatial skills

more than it does verbal skills, and it is not directly related to

the knowledge required by reading. In addition, younger

children were much more likely to be classified as dyslexic

(as opposed to poor readers) than were older children because

of the so-called decline in IQ scores that resulted from lack of

print exposure. Finally, if intelligence is a measure of some

stable construct of ability or potential and IQ tests measure it,

then these test results should be stable over time. Elliot and

Boeve (1987) found that the variable time has a statistically

significant effect on both children’s WISC-R verbal and per-

formance scores. Therefore, for these students, it is question-

able if the WISC-R is measuring a stable construct.

Ultimately, services may be denied to individuals who

have not been administered an IQ test or who do not achieve

a certain score on the IQ test (e.g., see Padget, Knight, &

Sawyer, 1996). IQ scores are significantly correlated with

socioeconomic status—that is, children from lower socioeco-

nomic backgrounds are likely to have lower IQ scores

probably because of a relative lack of experience with the vo-

cabulary and knowledge measured by the IQ test. Children

from low socioeconomic backgrounds are more likely to be

classified as poor readers than are dyslexics even though

these children have the same degree of reading difficulty as

dyslexics (Siegel & Himel, 1998); this means that the use of

the discrepancy definition discriminates against children

from low-SES backgrounds. Some argue that we should use

IQ test scores because IQ is correlated with achievement

(e.g., Wong, 1996). As Tunmer (1989) has noted, parental

income is correlated with reading achievement (the correla-

tions are almost identical in magnitude to those between

reading and IQ); why not use a discrepancy between parental

income and reading achievement as a measure of dyslexia? A

significant discrepancy would be needed for the individual to

get remedial help. The social consequences and unfairness of

this suggestion are obvious, but the principle is the same as

using a discrepancy between IQ and achievement.

The requirement for the use of the IQ score provides us

with the dilemma of determining an IQ score that is neces-

sary for an individual to achieve to be considered learning

disabled. Some studies use 80, others 85, others 90, and some

even use 100. For example, although they recognize that

there are problems with the use of IQ in the identification of

learning disabilities, Padget et al. (1996) argue that 

Clinical judgment may be needed when considering cases in

which the Verbal IQ is below 90. Two cases in which this may

be particularly important are older students who previously

achieved a Verbal IQ above 90 and students who have low sub-

test scores only on the subtests that require significant auditory,

sequential memory skills to perform the task. In most other cases

when students score below these guidelines it strongly suggests

that reading problems may be related to factors other than, or in

addition to, dyslexia. (p. 59).

However, they present no evidence for this assertion. An

opposing conclusion has been reached by G. Reid Lyon

(1995) of the U.S. National Institute of Child Health and

Human Development as follows:

The assumption that a discrepancy between achievement and

aptitude (typically assessed using intelligence tests) is a clear

diagnostic marker for learning disabilities or can be considered a

pathognomonic sign is at best premature, and at worst invalid.

(p. 5121)

Often, it is recommended that instead of the IQ test, mea-

sures of listening comprehension be used presumably because

this will assess an individual’s level of cognitive processing

(Aaron, 1991; Padget et al., 1996). Listening comprehension

measures typically consist of the examiner’s reading aloud a

passage to the individual and then asking him or her questions

460

Learning Disabilities

about the text. This type of task places heavy demands on

memory, and if an individual fails to answer a question cor-

rectly, the examiner does not know whether the individual did

not understand the passage or has merely forgotten the an-

swer. Ability to remember what has been heard also depends

on one’s background knowledge of the material in the text.

Unlike when one reads text in which the material is available,

one cannot refer back to the material in a listening compre-

hension task. Therefore, the time has come to abandon listen-

ing comprehension as an alternative to the IQ test.

Another assumption of many LD definitions—including

US Public Law 94-142, Ontario’s Bill 82, and the NJCLD’s

definition—is that there must be a discrepancy between IQ

and achievement. In other words, the child’s achievement is

not commensurate with his or her ability or intelligence (IQ). 

The second assumption of the discrepancy definition is

that measures of intelligence and measures of achievement

are independent. This assumption has been questioned by

some investigators (see Lyon, 1987, for a complete discus-

sion), but I maintain that it is necessary if a discrepancy defi-

nition is to be meaningful. If one accepts the argument that

intelligence is not orthogonal to achievement, then there

would be no reason to expect a discrepancy. Therefore, a dis-

crepancy definition is logical if and only if the presence of a

learning disability does not affect IQ test scores but does

affect achievement test scores. Then children with LD would

have a discrepancy between the scores on their IQ tests and

the scores on their achievement tests, whereas normally

achieving children and those with other disabilities will have

scores on these tests that are similar (not discrepant).

A number of investigators (Green, 1974; Hopkins &

Stanley, 1981; Siegel, 1988a, 1988b) have suggested that

this assumption of independence is questionable. Green

(1974) argues, for example, that comparisons of scores for

ability and achievement are meaningful only to the extent to

which unique elements are measured.

Hopkins and Stanley (1981) examined the overlapping

variance in a well-constructed intelligence test (the Lorge-

Thorndike) and in two subtests (the reading and the arith-

metic) of a well-constructed achievement test (the Iowa Test

of Basic Skills). On average, 47% of the variance was found

to overlap. This finding suggests that when one ability test and

one achievement test are used, about 50% of the time the same

concept is being measured; this clearly violates the assump-

tion of independence among concepts measured by the ability

and achievement tests. It can only be hoped that the other 50%

of the variance is tapping something different that can provide

insight to the true differences revealed by the comparison.

Discrepancy definitions also have been questioned on sta-

tistical grounds. Reynolds (1984–1985, 1985) notes that many

discrepancy models are based on grade equivalents. He points

out that these models have problems in two areas. First, nei-

ther age nor grade-equivalent scores provide adequate mathe-

matical properties (cannot be added, subtracted, multiplied, or

divided) for use in discrepancy analysis. Second, the amount

of retardation reflected by 2 years below grade level changes

with increasing grade level. A much greater level of retarda-

tion is reflected by a 2-year deficit at Grade 2 than at Grade 7,

and even less retardation would be reflected at Grade 11.

Therefore, a much greater deficit would be needed for a child

in Grade 3 to be identified as LD than for one in Grade 10.

In summary, the weight of the evidence leads us to con-

clude that IQ scores are irrelevant to the definitions of learn-

ing disabilities.

Specificity

There are two types of specificity: The first involves the degree

to which the individual’s problem is specific to one or more

cognitive areas, and the second involves the issue of whether

all children with learning disabilities should be considered one

homogeneous group.Anumber of authors (Hall & Humphreys,

1982; Stanovich, 1986, 1988a; Swanson, 1988b, 1989) have

maintained that one of the essential concepts of a learning dis-

ability is its specificity—that is, a learning disability is pre-

sumed to be caused by a neurological inefficiency that affects

a narrow group of subskills of cognitive processes; this affects

a specific domain of academic skills but leaves intellectually

ability intact. In other words, a learning disability reflects a

cognitive deficit possibly due to a neurological dysfunction

that is comparatively specific to a particular domain (e.g.,

reading or arithmetic). Swanson (1988a) has suggested that

these specific deficits must not stray too far into other cogni-

tive domains, or the concept of a specific learning disability

will blend with other more generalized conditions (e.g., men-

tal retardation). As well, Stanovich (1986) maintains that def-

initions of dyslexia must rest on an assumption of specificity.

He contends that dyslexia results from a brain or cognitive

deficit that is reasonably specific to reading.

Siegel (1988a) and Bryant and Brown (1985) argue that

this type of specificity is unrealistic. Siegel (1988a) maintains

that if children have problems in working memory, this con-

dition could affect a variety of academic tasks—especially in

areas such as reading, spelling, and arithmetic. Brown and

Bryant (1985) suggest that if a child has a severe language

problem, this condition could influence a large number of cog-

nitive areas: reading, writing, speaking, listening, or any com-

bination. Moreover, Siegel (1988a) contends that implicit in

the specificity assumption is the assumption that domains such

as reading and arithmetic are entirely independent cognitive

Definitional Issues

461

processes. This assumption is invalid because working mem-

ory and recognizing and labeling abstract symbols are

involved in both reading and arithmetic skills, and a child

who has difficulty with these cognitive processes is likely to

have problems with tasks involving such skills (Siegel,

1988a).

The second type of specificity involves questions about the

degree to which all children with LD have the same problems

or whether there are subtypes. On the basis of what is now

known, the concept of a generalized homogeneous group la-

beled LD children should be abandoned; the child with LD

should be considered as part of a smaller, more clearly de-

fined subtype (Siegel, 1988a). As well, other theorists (e.g.,

Bateman, 1968–1969; Benton & Pearl, 1978; Boder, 1973;

Kinsbourne & Warrington, 1963; Rourke, 1983, 1985) have

contended that differences within the population of children

with learning disabilities may reflect the existence of distin-

guishable subtypes. In other words, not all learning-disabled

children have the same types of disabilities, and independent

subtypes include distinctive characteristics and antecedent

conditions that consistently predict specific patterns of learn-

ing difficulties. Therefore, failure to differentiate among

types of learning disabilities can lead to inaccurate conclu-

sions. For example, Siegel and Ryan (1984, 1988) found that

children with reading disabilities have difficulty processing

certain aspects of syntax, whereas specific children with

specific arithmetic disabilities (and no reading disability) do

not. Differences between the same two groups have been

found in working memory (Fletcher, 1985; Siegel & Linder,

1984; Siegel & Ryan, 1989a). Swanson (1988b) found

that children with a reading disability may be characterized

by different patterns of memory dysfunction. These differ-

ences are reflected on measures of achievement in reading

and arithmetic. If all these children had been considered to-

gether as a homogeneous group, these differences might have

been obscured.

A Resolution of Definitional Issues

Clearly the field of special education continues to have prob-

lems defining and classifying children with learning disabili-

ties. Current discrepancy definitions are problematic and

should be reconsidered because they cannot be justified in

light of their illogical nature. But where does that leave us

and where can we go from there?

One way in which meaning is given to a concept is by

defining it operationally. Specifically, an operational defini-

tion explains a concept solely in terms of the operations used

to produce and measure it. Recognizing that there are prob-

lems inherent in operational definitions (e.g., the meaning of

the concept is restricted to the narrowly described operations

used for measuring it), I suggest that learning disabilities (a)

need to have an IQ threshold because I recognize that the

field is not ready to accept Siegel’s (1989a, 1989b) challenge

to the use of IQ tests; (b) should refer to a significant diffi-

culty in a school-related area; and (c) should exclude only

severe emotional disorders, second-language background,

sensory disabilities, and specific neurological deficits.

Evidence (Siegel, Levey, & Ferris, 1985; Siegel & Ryan,

1989b) suggests that the type of operational definition used for

the concept may influence the outcome and conclusions of the

study. Siegel and Ryan (1989b) have argued that the actual de-

finition used for a reading disability can make a difference in

the conclusions that are drawn about information-processing

characteristics of the children and whether there are reading-

disability subtypes. In one study, poor readers (all with IQ

scores equal to or above 80) were divided into four groups as

follows: (a) those with deficits in phonological processing—

inadequate phonological skills based on the reading of pseu-

dowords, (b) those with word-recognition deficits, (c) those

with comprehension-only deficits—inadequate reading com-

prehension skills but adequate word-recognition skills, and

(d) those with rate-only deficits—slow reading speed but ade-

quate word-recognition skills. When each disabled group was

compared to an age-matched normally achieving group, dis-

tinct cognitive differences appeared. For example, children

with a phonological deficit or a word-recognition deficit had

scores that were significantly lower than were those of normal

readers on short-term memory tasks but not on language

tasks. The readers with rate-only deficits had cognitive pro-

files similar to those of normally achieving children. There-

fore, the children with word-recognition problems are

probably the ones with language deficits and those with only a

reading comprehension problem probably do not have lan-

guage problems.

In addition, there was approximately a 25% overlap be-

tween poor comprehenders and poor readers; therefore, had a

reading comprehension test been used to define the learning-

disabled group and word recognition not been used as a con-

trol, the reading-disabled group would have consisted of some

children with word-recognition problems and some without.

This leads to the question How should a reading disability

be defined? Stanovich (1986, 1989) has suggested that the

core deficits in a reading disability are problems in phonolog-

ical processing. Although reading is more than simply decod-

ing and recognizing words (one has to remember what was

read, put it into context, etc.), unfortunately currently there

are no accurate tests to measure these variables (see Siegel &

Heaven, 1986, for a complete discussion of this issue).

Further empirical evidence suggests that when a difficulty