ON BABIES AND BATHWATER: ADDRESSING THE PROBLEMS OF IDENTIFICATION OF LEARNING DISABILITIES

Page created by Charles Bradley
 
CONTINUE READING
ON BABIES AND BATHWATER:
                      ADDRESSING THE PROBLEMS
                          OF IDENTIFICATION
                       OF LEARNING DISABILITIES

                            Thomas E. Scruggs and Margo A. Mastropieri

                    Abstract. In this article, we review problems in identification of
                learning disabilities and consider proposed alternatives to present
                procedures. We argue that no proposed alternative meets all the
                necessary criteria for identification of learning disabilities, and
                that radically altering or eliminating current conceptualizations of
                learning disabilities may amount to “throwing the baby out with
                the bathwater.” We conclude that the major problems of identifi-
                cation of learning disabilities — including overidentification, vari-
                ability, and specificity — can be eliminated by increasing
                specificity and consistency of state criteria and strict adherence to
                identification criteria on the local implementation level. Finally,
                we argue that scarce special education funds should not be
                employed to address the problems of general education.

                        THOMAS E. SCRUGGS, Ph.D., is professor, George Mason University.
                       MARGO A. MASTROPIERI, Ph.D., is professor, George Mason University.

   Since the inclusion of learning disabilities (LD)          gory of learning disabilities be eliminated or signifi-
as a category in the Education for All Handicapped            cantly altered (Aaron, 1997; Algozzine, 1985; Lyon
Children Act (PL 94-142; 1975), concerns have been            et al., 2001).
expressed over appropriate identification procedures.           In this article, we review problems in identifica-
Although far from alone in its search for more appro-         tion of learning disabilities and commonly agreed-
priate definitions (e.g., the American Association on         upon characterizations of learning disabilities and
Mental Retardation has revised its 1921 definition of         consider proposed alternatives to present procedures
mental retardation nine times; see Beirne-Smith,              for identification of learning disabilities. We argue
Ittenbach, & Patton, 1998), the field of learning dis-        that no proposed alternative meets all necessary cri-
abilities has experienced some very specific problems in      teria for identification of learning disabilities. We
identification (Kavale & Forness, 1995; Scruggs &             conclude that the major problems of identification
Mastropieri, 1994-1995). These problems include               of learning disabilities — including overidentifica-
(a) overidentification, (b) variability, (c) specificity,     tion, variability, and specificity — can be eliminated
(d) conceptual considerations, (e) discrepancy issues,        by increasing the specificity of state criteria and
(f) early identification, and (g) local implementation.       strict adherence to identification criteria on the local
As a consequence, it has been suggested that the cate-        implementation level.

                                             Volume 25, Summer 2002   155
PROBLEMS IN IDENTIFICATION                                 a discrepancy, while only 2 of the highest 10 states
Overidentification                                                 employed such a requirement. Such a finding reflects
   Since 1975, the population of individuals identified            the position of Dawes, Faust, and Meehl (1989) that
as having learning disabilities has increased about                more objective procedures in identification may lead to
150% to a level that represents over 50% of all students           more consistent outcomes. More recently, Lester and
with disabilities and over 5% of all students in school,           Kelman (1997) evaluated state identification rates, con-
a percentage far higher than those of, for example, stu-           cluding that demographic and sociopolitical factors
dents with mental retardation or serious emotional dis-            moderately predicted some aspects of state prevalence
turbance (U.S. Department of Education, 2000). This is             of learning disabilities, although these factors were not
frequently the first fact noted in arguments that cur-             related to prevalence of physical disabilities.
rent definitions are problematic (e.g., Fuchs, Fuchs, &            Specificity
Speece, 2002; Lyon et al., 2001).                                    It has been argued that individuals with learning dis-
   These high identification rates may result from im-             abilities can not be reliably distinguished from indi-
precision in federal and state definitions of learning             viduals with generally low achievement (LA) (e.g.,
disabilities (e.g., Mercer et al., 1996). However, re-             Algozzine, 1985; Ysseldyke, Algozzine, Shinn, &
searchers have suggested that a confounding of differ-             McGue, 1982), or, more specifically, that students with
ent high-incidence disabilities has resulted in classify-          reading disabilities can not be distinguished from “gar-
ing individuals as having learning disabilities who                den variety” poor readers (Fletcher et al., 1992;
previously may have been classified as having mental               Fletcher et al., 1994; Spear-Swerling, 1999; Wagner &
retardation (MacMillan, Siperstein, & Gresham, 1996).              Garon, 1999). Shaywitz, Escobar, Shaywitz, Fletcher,
Wong (1996) suggested that the concept of learning                 and Makach (1992) have suggested that “dyslexia”
disabilities may have been overgeneralized, perhaps                simply reflects the lower tail of a normal distribution
reflecting “teachers’ noble goal of teaching as many               of reading ability. Ysseldyke, Richey, and Graden
problem learners as possible, and not restricting in-              (1982) compared students who had been identified as
structional help only to students with learning disabil-           having learning disabilities with students who had not
ities” (p. 8; see also Torgesen, 1999, p. 108; Zigmond,            been identified but had scored lower than the 25th
1993). Regardless of the reason, many consider the                 percentile on achievement tests, and concluded that
prevalence rates for learning disabilities to be “alarm-           “there were no psychometric differences in the per-
ingly high” (Algozzine & Ysseldyke, 1987). According               formances of the two groups of students” (p. 83).
to Wagner and Garon (1999), “the prevalence of this                Citing this and related investigations, Algozzine
disability is likely to be closer to 1-3 percent of school-        (1985) noted that “the learning disabilities category
age children as opposed to recent estimates of 20-30               has outlived its usefulness” (p. 73). Some researchers,
percent” (p. 100).                                                 reanalyzing research in which learning disabilities-low
Variability                                                        achievement comparisons were made, have concluded
  Identification of individuals with learning disabilities         that differences are reliable and substantial (e.g.,
has revealed considerable variability across agencies.             Kavale, Fuchs, & Scruggs, 1994; Fuchs, Fuchs, Mathes,
Finlan (1992) reported that states varied considerably             & Lipsey, 2000). In response, it has been suggested
in identification rates, from a low of 2.10% (Georgia) to          that even if some differences can be identified, the two
a high of 8.66% (Rhode Island). Coutinho (1995) found              groups can benefit from qualitatively similar instruc-
similar variability using more recent data, from a low of          tional approaches, and therefore specific identification
approximately 2% (Wisconsin) to over 7% (Massachu-                 of learning disabilities is unnecessary (Algozzine,
setts) (see also Frankenberger & Fronzaglio, 1991; Lester          Ysseldyke, & McGue, 1995; Lyon, 2001; Spear-
& Kelman,1997). Although there has been some debate                Swerling, 1999).
on whether the variability in identification exceeds               Conceptual Problems
that of other disability areas (Algozzine & Ysseldyke,               It has been proposed that differing and inconsistent
1987; Hallahan, Keller, & Ball, 1986), there is little             identification arises from problems in conceptualizing
doubt that the variability is considerable and may                 the definition of learning disabilities. Kavale and
reflect lack of consistency or precision in identification         Forness (2000) evaluated previous definitions and sug-
procedures. Finlan (1992), for example, identified a sys-          gested, “… the failure to produce a unified definition
tematic relation between identification rate and                   has meant that LD lacks two critical elements: under-
whether or not states employed a specific discrepancy              standing — a clear and unobscured sense of LD — and
requirement. Seven of the lowest 10 states in rates of             explanation — a rational exposition of the reasons
identification employed a specific method of assessing             why a particular student is LD” (p. 240).

                                              Learning Disability Quarterly   156
Until the definition of learning disabilities is clearly    that if the condition of learning disabilities is defined as
conceptualized, problems in identification will con-          a lack of academic progress, such a determination can-
tinue (Kavale & Forness, 1995; Kavale, Forness, &             not be made on the preschool level, and that different
Lorsbach, 1991). Part of the difficulty arises from the       characterizations should be applied. Nevertheless, others
lack of agreed-upon positive measures of learning dis-        have suggested that it may be better not to wait for fail-
ability, and the partial reliance on exclusionary criteria    ures to occur prior to implementing interventions. For
— that is, things that learning disabilities can not be,      example, Fletcher and Foorman (1994) described evi-
including mental retardation, emotional disturbance,          dence that learning problems are more easily remediated
or cultural disadvantage (Kavale & Forness, 2000).            in the earlier grade levels, concluding that “the focus
Discrepancy Issues                                            should be on prevention and early intervention for chil-
   The use of IQ-achievement discrepancy criteria in          dren at risk for developing learning difficulties” (p. 187).
identification of learning disabilities, although widely         However, present models of identification that
employed (Schrag, 2000), has been criticized by many          emphasize observable academic failure may lose valuable
for a variety of conceptual and statistical reasons           time for treatment (Hallahan, Kauffman, & Lloyd, 1999).
(Aaron, 1997; Mastropieri, 1987; Scruggs, 1987; Spear-        Lyon et al. (2001) argued that the present IQ-achieve-
Swerling, 1999; Stanovich, 1991). For example, techni-        ment discrepancy criterion necessitates a “wait to fail”
cal problems include the degree of measurement error          model, in that many children can not be reliability iden-
associated with discrepancy methods and the problem           tified as having learning disabilities until third grade
of statistical regression (Cone & Wilson, 1981; Shepard,      because of psychometric limitations in discrepancy cri-
1980). It has been argued that use of IQ tests is not         teria (p. 269). However, grade-level discrepancies (e.g.,
helpful, and that students exhibiting discrepancies do        academic functioning two years below grade level) may
not differ from poor readers with lower IQs on many           be more problematic than IQ-achievement discrepan-
reading, spelling, language, or memory tests (Siegel,         cies, which do not necessarily require long periods of
1999). It has also been suggested that IQ is not predic-      academic failure. Horn and O’Donnell (1985), for exam-
tive of reading achievement (Vellutino, Scanlon, &            ple, found that discrepancy scores, even in the first few
Lyon, 2000; Fletcher et al., 1998), although this posi-       weeks of first grade, were better predictors of future
tion has been challenged (Naglieri, 2001). Further,           learning disabilities classification than was a low-
Stanovich (1991) argued that discrepancy definitions,         achievement criterion. Mastropieri (1988) concluded
at least with respect to reading disability, are threat-      that when children with learning disabilities are identi-
ened by findings that literacy development also devel-        fied and treated early, outcomes are usually positive.
ops the cognitive skills revealed on aptitude measures        Local Implementation
such as IQ tests. Some research suggests that IQ-               One extremely important aspect of the identification
achievement discrepancies are not predictive of aca-          process that is often lost in purely conceptual or tech-
demic growth rates (Lyon, 2001; Vellutino et al., 2000;       nical analyses is the individual or group judgment of
but see Speece & Case, 2001, for a different perspec-         professionals who are part of the referral and eligibility
tive). According to Kavale and Forness (1995), an over-       process. Such judgments can be critical in identifying
riding problem with discrepancy is how it has been            students who might otherwise “fall through the cracks”
applied; that is, that “the prominence of discrepancy in      and lead to false negatives or false positives that may
conceptualizations of LD has resulted in its reification      result from blindly applying quantitative criteria. In
and deification” (p. 162). In some cases, discrepancy         that regard, MacMillan, Gresham, and Bocian (1998)
(underachievement) has come to be seen as equivalent          stated, “the assignment of responsibility to a commit-
to learning disabilities, rather than one possible com-       tee clearly conveys the desire to receive input from var-
ponent of a conceptual understanding of learning dis-         ied perspectives and not rely exclusively on test scores”
abilities. Lyon et al. (2001) reviewed relevant literature    (p. 322). Nevertheless, “actuarial” procedures can result
and concluded, “The IQ-achievement discrepancy,               in much more consistent and systematic identification
when employed as the primary criterion for the identi-        methods, and can be key factors in reducing variability
fication of LD, may well harm more children than it           in identification (Dawes et al., 1989; Ysseldyke,
helps” (p. 266).                                              Algozzine, Richey, & Graden, 1982). The issue was
Early Identification                                          framed by Meehl (1973) as, “When should we use our
  Nearly 30 years ago, Keogh and Becker (1973) alerted        heads instead of the formula?” (p. 81). Gresham (2001)
the field to the importance of early identification and       described the identification process as involving local
treatment and the dangers of misidentification (see also      criteria for school failure, national norms for assess-
Keogh, 1986). In response, Haring et al. (1992) argued        ment, and classification decisions based upon the

                                             Volume 25, Summer 2002   157
degree of presumed “profitability” to the student,                   Consideration of such issues has led researchers in
rather than established criteria.                                  two different directions over the years (Lester &
   Recent research has suggested that “public school               Kelman, 1997). On one hand are those who have sug-
practices for diagnosing children with learning disabili-          gested that improved precision of identification proce-
ties bear little resemblance to what is prescribed in fed-         dures will lead to a resolution of these issues, and
eral and state regulations” (MacMillan et al., 1998,               support treatment of those who “truly” have learning
p. 323). Gottlieb, Alter, Gottlieb, and Wishner (1994)             disabilities. Reynolds (1984), for example, argued:
reported that urban students identified with learning                   The tremendous disparities in measurement mod-
disabilities had an average IQ of 81.4, concluding, “these              els adopted in the various states in their respective
children today are classified as learning disabled, when                learning disability guidelines and the varying levels
in fact most are not” (p. 463). MacMillan, Gresham,                     of expertise are obvious, major factors contributing
Siperstein, and Bocian (1998), using a California sample,               to the difference in the proportion of children
found that fewer than half of the students identified                   served as LD in the various states. Lack of a specific
as having learning disabilities met state requirements. In              definition, improper or lack of application of the
an earlier investigation, MacMillan et al. (1996) identi-               severe discrepancy criterion, and the failure to
fied students with IQ scores lower than 75, and as low as               develop appropriate mathematical models … are
58, who had been classified as having learning disabili-                the primary, certainly interrelated causes of these
ties. McLeskey and Waldron (1990) examined the                          disparities. (p. 455; see also Keogh, 1994)
records of 1,742 Indiana students referred for evaluation            On the other hand, some have suggested that the
for learning disabilities eligibility. They reported that,         concept of learning disabilities has little or no practical
although students identified as having learning disabili-          utility and should be abandoned: “It is time to quit
ties (52%) differed markedly from students referred but            viewing eligibility decision making as a technical prob-
not identified (48%), over one third of students identi-           lem. It means putting an end to efforts to try to find
fied as having learning disabilities did not meet existing         better ways of defining concepts and conditions that
state criteria.                                                    cannot be defined and may not exist” (Algozzine &
   Although variability may exist across and within                Korinek, 1985, pp. 392-393).
states, evidence to date suggests that a very substantial            Nevertheless, any accounting of learning disabilities,
source of overidentification of learning disabilities is           positive or negative, must take into account the existence
the application (or misapplication) of state criteria at           of descriptive reports dating back to the 19th century or
the local level. These data suggest that identification            earlier (see Anderson & Meier-Hedde, 2001; Berlin, 1887;
frequency could be reduced by as much as one third or              Dejerine, 1892; Hinshelwood, 1896; Kussmaul, 1877), the
more simply by consistent and systematic application               proliferation of professional and advocacy organizations,
of state criteria at the local level. Further research             such as the Council for Learning Disabilities, the Division
could provide more evidence on the reasons for this                for Learning Disabilities of the Council for Exceptional
overidentification and the steps that could be taken to            Children, and the Learning Disabilities Association of
reduce it.                                                         America, and the thousands of service providers who not
                                                                   only testify to the existence of learning disabilities but
            CHARACTERIZATIONS                                      also advocate for improved services (see Scruggs &
          OF LEARNING DISABILITIES                                 Mastropieri, 1988). Although measurement issues are far
  Problems in identification, based to some degree on              from resolved, a relatively consistent set of observations
conceptual inconsistencies, have led some to argue that            has been reported over the years. Wong (1996) suggested,
learning disabilities does not exist as a viable condition.             … those very characteristics observed by parents,
Some have proposed that learning disabilities is, in fact,              educators, psychologists, and medical professionals
socially or organizationally constructed to serve the                   about children with learning disabilities in 1963
interests of certain elements of society rather than chil-              are the very same characteristics that we see today
dren (Christensen, 1999; Coles, 1987; Skrtic, 1999;                     in children, adolescents, and adults with learning
Sleeter, 1995; see Kavale & Forness, 1998, for a review).               disabilities! (p. 22)
Others have suggested that students characterized as                 As an example, Mastropieri (2001) described the case
having learning disabilities are not distinguishable               of “Andrew,” a middle third-grader in a middle-class,
from other poor readers or low achievers not character-            suburban school. On a tape-recording of his reading of
ized as having learning disabilities (Aaron, 1997;                 a grade-level reader, he reads a painful six words per
Fletcher et al., 1994; Shaywitz, Holahan, & Shaywitz,              minute. On an early second-grade text, he reads eight
1992; Ysseldyke, Algozzine, Shinn, & McGue, 1982),                 words per minute. Psychoeducational testing reveals
and that perhaps the concept is not useful.                        considerable variability: a Full-Scale IQ of 104, but a

                                              Learning Disability Quarterly   158
reading standard score of 72, with similar deficits in          Forness, 1995). Mastropieri’s (2001) case of a student
math and spelling. His vocabulary is above average; his         with average abilities in intelligence and general lan-
listening comprehension and verbal expression scores            guage skills, who nevertheless exhibited very consider-
are almost exactly average. For a “free writing” activity,      able difficulty in math and literacy tasks, provides an
he drew a picture that seems to be a comet, and wrote           example of this unexpected underachievement. That
below the picture the single sentence, “It vush the oue         this level of achievement was not expected from an
wun” [it was the only one]. His teachers describe him           evaluation of the student’s intellectual or linguistic
as a good-natured student of good verbal ability, but           functioning is integrally linked to a general under-
having significant problems in reading and math con-            standing of learning disabilities.
cepts. As such, Andrew provides a classic example of a             Learning disabilities are multifaceted in that they
student who exhibits many of the characteristics con-           describe a variety of manifestations in areas of academic
sidered “typical” of learning disabilities.                     and/or psychological functioning. Although most stu-
   Kavale and Reese (1991) surveyed 547 teachers of stu-        dents identified as having learning disabilities exhibit
dents with learning disabilities in the state of Iowa (aver-    problems in the literacy areas of reading, spelling, and
age learning disabilities teaching experience 8.5 years,        written language, for which considerable co-morbidity
13 years overall), and found substantial agreement on           has been identified (Beitchman, Cantwell, Forness,
the nature of learning disabilities. More than 80% of the       Kavale, & Kauffman, 1998), specific problems in mathe-
teachers agreed that the condition of learning disabilities     matics are also often identified (American Psychiatric
is associated with the following statements: (a) a dis-         Association, 2000; Baroody & Ginsburg, 1991; Fleischner,
crepancy between ability and achievement; (b) learning          1994; Kosc, 1981; McLeskey & Waldron, 1991). These
strengths as well as learning weaknesses; (c) academic          problems may (Scruggs & Mastropieri, 1986) or may not
strengths as well as academic weaknesses; (d) a process-        (Rourke, 1989; Strang & Rourke, 1983) be associated with
ing deficit that interferes with learning ability; (e) aver-    problems in reading or spelling. In addition, deficits in
age or above intelligence; (f) a need for special materials     specific cognitive processes such as memory (Cooney &
and instructional techniques; and (g) the ability to learn      Swanson, 1987; Siegel & Ryan, 1988; Swanson, 1993a) or
at a different rate than individuals with mental retarda-       attention (Hallahan & Cottone, 1997) may underlie prob-
tion (pp. 146-147).                                             lems in more than one academic area. Even among stu-
   These statements correspond closely with the three           dents with reading disabilities, it is less clear whether the
components of learning disabilities conceptualizations          frequently observed deficits in phonological awareness
over the past 30 years, as described by Keogh (1994):           (Torgesen, 1999) are a foundational element of all read-
“1) unexpected low achievement relative to aptitude or          ing disorders, responsible also for identified deficits in
ability; 2) deficits and uneven profiles in specific per-       reading comprehension (Mastropieri & Scruggs, 1997;
ceptual or cognitive processes; and 3) evidence of              Sternberg, 1999).
within-child, presumably causal, neurological condi-               This multifaceted nature of learning disabilities was
tion(s)” (p. 16). Keogh concluded, “although they may           demonstrated in a multivariate analysis of a large data
have validity on a construct level, these definitional cri-     set (Kavale & Nye, 1985-1986, 1991), which revealed
teria present serious problems of measurement when              learning disabilities to be most consistently related to
making diagnostic decisions or assigning individuals to         deficits in reading and math, as related in turn to deficits
classes” (p. 16).                                               in linguistic (e.g., semantic, syntactic, phonological),
   Wong (1996) has described many of these commonly             neuropsychological (e.g., selective attention, memory,
reported features of learning disabilities, including the       cognitive style), and social/behavioral functioning (e.g.,
following.                                                      interpersonal perception, intrapersonal perception).
   Learning disabilities involve unexpected or unan-            Such results suggest that learning disabilities, at least
ticipated underachievement. This unexpected nature              as commonly understood and evaluated, are multifac-
of learning disabilities is a commonly cited character-         eted rather than unitary in nature. In order to under-
istic (Keogh, 1994). It is not completely separate from         stand the different features observed in students with
other considerations, since the unexpected or unantic-          learning disabilities, attempts have been made to iden-
ipated nature of learning disabilities doubtless has            tify “subtypes” (e.g., Lyon, 1985; McKinney, Short, &
much to do with the concept of deficits in cognitive            Feagans, 1985). However, general agreement has not
processing in spite of normal intelligence. Learning            been reached on the nature of such learning disabilities
problems associated with such factors as low intelli-           subtypes (Kavale & Forness, 1987).
gence, sensory impairments, or cultural disadvantage,              Learning disabilities are associated with intraindi-
in contrast, are “expected,” and provide a direct con-          vidual differences or deficits, and uneven profiles in
ceptual contrast with learning disabilities (Kavale &           specific perceptual or cognitive processes. Although

                                               Volume 25, Summer 2002   159
researchers have observed deficits in a large number of           ties definitions. Further, measures of cognitive pro-
domains, individuals with learning disabilities are fre-          cessing, including attention, memory, and linguistic
quently observed to demonstrate relative patterns of              processes, although not necessarily directly related to
strengths and weaknesses (Keogh, 1994). These were                neuropsychological dysfunction, have revealed pro-
described by Kirk (1971) as intraindividual differences,          cessing deficits in students with learning disabilities.
and by Gallagher (1966) as “imbalances.” Although not
necessarily revealed in large group summaries of sub-                        ALTERNATIVE PROCEDURES FOR
type scores (e.g., WISC profiles; Kavale & Forness,                                IDENTIFICATION
1984), researchers and practitioners have frequently              Proposed Alternatives
noted relative strengths and weaknesses within indi-                Clearly, the field of learning disabilities is facing a
vidual patterns of functioning (Wong, 1996).                      crisis. On the one hand are those who have concluded
  Learning disabilities are associated with within-               that problems in identification have critically compro-
child, presumably neurological conditions and, as such,           mised the concept of learning disabilities, to the extent
are not primarily due to sensory, motor, or intellectual          that it should be abandoned entirely. On the other
deficits, or cultural disadvantage. This characteristic is        hand is the considerable volume of research and
included in many definitions of learning disabilities             descriptive evidence that, while acknowledging the
(Kavale & Forness, 2000), and has contributed both to             problems in identification, nevertheless supports the
conceptual models of learning disabilities and to the             concept of learning disabilities. Thirty years ago,
most severe criticisms of the concept of learning dis-            McCarthy (1971) wrote:
abilities. Since cognitive deficits of known neurological
                                                                       The most important decision you will make is that
origin have been difficult to measure directly, problems
                                                                       of definition — because your definition will dictate
of reliability and validity of psychological “process”
                                                                       for you the terminology to be used in your pro-
measures called the issue of neurological functioning
                                                                       gram, the prevalence figure, your selection criteria,
into question (e.g., Larsen & Hammill, 1975). Mercer,
                                                                       the characteristics of your population, and the
Forgnone, and Wolking (1976) suggested, “due to the
                                                                       appropriate remedial procedures. (p. 14)
vague nature of the concept of a process disability, a
                                                                    Today, issues of definition, identification, and assess-
description of it for the purpose of analyzing definitions
                                                                  ment are more important than ever in the continuance
of LD is subject to criticism” (p. 378).
                                                                  of the field of learning disabilities. For these reasons,
  Such analysis, however, was helpful in redirecting
                                                                  alternatives to the present identification procedures
previous assumptions that learning disabilities gener-
                                                                  have been proposed, several of which contain overlap-
ally reflect visual-perceptual processing difficulties,
                                                                  ping considerations. Alternative identification proce-
and require perceptual-motor training (Kavale &
Mattson, 1983). Nevertheless, modern techniques,                  dures for learning disabilities include the following.
including functional magnetic resonance imaging                     Double-deficit criteria. Wolf and Bowers (1999) pro-
(fMRI) techniques, have suggested that a neuropsy-                posed criteria based on deficits on (a) phonological
chological basis for learning disabilities does exist             analysis tasks and (b) rapid continuous naming of digits
(Hynd, Clinton, & Hiemenz, 1999), and that “both                  and letters. These criteria have also been seen to dis-
anatomical and physiological signatures of dyslexia               criminate between students with reading disabilities and
exist,” for example, in decreased activation of the               normally achieving readers in the early grades. Allor
left temporoparietal and superior temporal cortex dur-            (2002) reviewed 16 studies of phonemic awareness and
ing phonological processing (Rumsey, 1996, p. 72).                rapid naming conducted between 1990 and 1997, and
Research from autopsy studies (Galaburda, 1991), posi-            concluded that performance on each of these two tasks
ton emission tomography (PET) scans (Wood, Flowers,               contributes uniquely to development of word reading,
Buchsbaum, & Tallal, 1991), and genetic studies                   and that continued efforts to evaluate the double-deficit
(Olson, 1999) has tended to support these findings,               hypothesis are warranted. However, more recently,
leading Hynd et al. (1999) to suggest, “learning dis-             Ackerman, Holloway, Youngdahl, and Dykman (2001),
abilities are most appropriately viewed from a neu-               testing a sample of 101 elementary school children with
ropsychological perspective” (p. 60). Although such a             and without learning disabilities, found that these chil-
characteristic may or may not be a useful component               dren differed on a number of tasks in addition to the
of a definition of learning disabilities (Spear-Swerling,         double-deficit criteria, including orthographic tasks,
1999), the consideration that learning disabilities is            attention, arithmetic achievement, and most WISC-III
associated with neuropsychological dysfunction —                  factors. Further, students with a double deficit were no
which may be manifest as disorders of cognitive                   more limited in reading and spelling than peers with a
processes — has been prominent in learning disabili-              single deficit in phonological analysis.

                                             Learning Disability Quarterly    160
Phonological process core difference model. Stanovich          further suggested that these measures could discrimi-
(1988) proposed that differences in phonological pro-             nate among subtypes of learning disabilities (e.g., non-
cesses could discriminate between “dyslexic” and                  verbal learning disabilities).
“garden-variety poor readers,” particularly in reading               Assessment of cognitive processing. In recent years,
achievement at early grade levels. Generally supporting           researchers have proposed the use of tests of cognitive
this idea, Torgesen and Wagner (1998) have proposed the           processing that could be useful in providing direct evi-
use of tests of phonological awareness, rapid automatic           dence of relevance to identification of learning disabili-
naming, and verbal short-term memory in identification            ties. For example, Naglieri (2001) described the Cognitive
of reading disabilities at the early grade levels (Wagner,        Assessment System (Naglieri & Das, 1997), and Swanson
Torgesen, & Rashotte, 1999). Similarly, Fletcher et al.           (1993b) described the Swanson Cognitive Processing Test
(1998), arguing against the utility of discrepancy models         as a dynamic assessment measure. Either of these meas-
of learning disabilities, suggested that learning disabilities    ures is suited to examining individual differences in
be considered for any student failing to reach the 25th           information processing, and as such could be potentially
percentile for specific reading skills, including core read-      useful in identifying learning disabilities.
ing process measures. They noted that this consideration             Operational interpretation. Kavale and Forness
would result in higher numbers identified, but may have           (1995, 2001) suggested an operational interpretation
greater relevance than present policy-based decisions,            that would permit a number of concepts to be given
which in their view merely fulfill a “gate-keeping func-          meaning by describing their individual operational def-
tion” (p. 199), that is, to keep numbers down.                    initions. They proposed a five-level process of identifi-
   Chronological age definitions. As have other research-         cation that included:
ers, Siegel (1989) suggested that IQ is not relevant to              1. underachievement/discrepancy as a necessary but
the concept of learning disabilities. Specifically, she                  not sufficient first criterion;
noted the bidirectional influence of IQ and achieve-                 2. pervasive deficits in basic skills, focusing on the
ment, arguing that lack of achievement may in time                       four major academic areas of reading, writing, lan-
exert a negative influence on IQ. Instead, Siegel (1989)                 guage, and math;
suggested that students be identified as having learning             3. deficits in learning efficiency, including measures
disabilities if they score below age expectancy in achieve-              of strategy use and rate of learning;
ment (particularly, reading of pseudowords) and are not              4. psychological process deficits in areas including
mentally retarded. Lyon and Fletcher (2001) suggested                    attention, memory, linguistic processing, and
that discrepancy criteria be abandoned and replaced by                   metacognition;
comparing achievement in academic areas with age and                 5. the exclusion of alternative causes of learning failure,
grade levels.                                                            such as mental retardation, sensory impairment,
   Bayesian procedures. Alley, Deshler, and Warner                       emotional disturbance, or inadequate instruction.
(1979) described an approach to identification based              Students would be identified as having learning disabil-
upon earlier models proposed by Wissink, Kass, and                ities only when all five operational criteria are met.
Ferrell (1975). Bayes’ formula combines prior and cur-               Failure to respond to validated treatment proto-
rent information to determine the probability of learn-           cols, or dual discrepancy criteria. Berninger and
ing disabilities. Regular class teachers are asked to             Abbott (1994) argued that sufficient “opportunities to
complete a checklist for students experiencing learning           learn” are often assumed rather than demonstrated,
problems. Each factor on the checklist is assigned a              and as a result identification of learning disabilities
numerical weight, depending on its odds of defining               may not discriminate between constitutional and
learning disability. Finally, a formula is used that com-         experiential deficits. They proposed the development
bines probability data. For example, although reading             of validated protocols of which treatment approaches
decoding had a probability of only .21 of prediction              work best for which learning characteristics (based
of learning disabilities, decoding + reading recogni-             upon static assessment of learning in 11 domains).
tion + detecting spelling errors + problems with math             When this knowledge is available, learning disabilities
algorithms resulted in a .96 probability of prediction of         can be diagnosed on the basis of treatment nonre-
learning disabilities. Further testing was done to verify         sponding; that is, failure to make expected gains on
these considerations.                                             treatment protocols demonstrated to achieve positive
   Neuropsychological assessment. Rourke (1993) de-               results for similar students.
scribed a battery of neuropsychological assessments                  Fuchs and Fuchs (1998; see also Speece & Case, 2001)
(e.g., tactile-perceptual tests, visual-perceptual tests,         have suggested that deficits in level and slope of learn-
auditory-perceptual and language-related tests) that he           ing (as evaluated by curriculum-based measurement
argued could be used to identify learning disabilities. He        [CBM] procedures) would represent a dynamic means

                                                 Volume 25, Summer 2002   161
for establishing lower academic functioning. Resistance             ment nonresponder, due to the considerable variation
to instruction could be evaluated by slopes on probes of            both between and within treatment nonresponders.
academic skills, including reading and math. Speece and               Recently, representatives of the National Center for
Case (2001) implemented these procedures on K-2 stu-                Learning Disabilities, the National Association of
dents in reading, and identified samples that were dif-             School Psychologists, and the International Reading
ferent in some respects (e.g., lower in age and mean IQ)            Association proposed a “multitiered” approach, apply-
from students identified according to IQ-achievement                ing CBM procedures to identify and serve students
discrepancy formulas. Al-Otaiba (2000) suggested,                   with learning disabilities (Horowitz, Lichtenstein, &
     given widespread and converging evidence that                  Roller, 2002). Specifically, Horowitz et al. recom-
     phonological processing deficits often lead to                 mended that students at risk for academic failure be
     learning disabilities, unresponsiveness to effective           provided with a “research-based, general education
     treatment protocols can provide an alternative to              intervention” (p. 2) for 8-10 weeks, monitored with
     discrepancy-based formulas currently used to iden-             CBM. Students who do not display meaningful gains,
     tify students with learning disabilities. (p. 12; see          as measured by slope and level of learning during this
     also Berninger & Abbot, 1994; Fuchs & Fuchs,                   intervention period, are considered candidates for spe-
     1998; Vellutino et al., 1996)                                  cial education. This process is intended to eliminate
   Referring to unsatisfactory level of performance in              the use of IQ-achievement discrepancies. It is recom-
addition to inadequate rate of growth as a “dual discrep-           mended that “at risk” status be determined by a 20th
ancy” (p. 34), Fuchs et al. (2002) described in detail the          percentile cutoff on standardized, norm-referenced
“treatment validity model” and how it could be used in              academic measures, and that “meaningful gains” be
the identification process. For example, Phase I assess-            determined by failure to achieve growth at the 20th
ment determines whether the general education class-                percentile on CBM measures.
room environment is sufficiently supportive to warrant                Horowitz et al. suggested that there is “sufficient evi-
decision making for individual students. If not, inter-             dence” (p. 3) to suggest that these procedures may be
vention at the classroom level is indicated to increase the         implemented appropriately in school settings, with
student academic growth rate to a level comparable with             substantial impact: they would not change numbers of
the school, district, or nation. If classroom instruction is        children served as LD, although the procedures would
considered appropriate, Phase II assessment identifies              reduce referrals to special education by almost one half
students with dual discrepancies, that is, students func-           (p. 3). However, direct empirical tests of this model
tioning dramatically below peers in level and slope of              appear to be lacking. Published research applications
performance on, for example, reading rate. Phase III
                                                                    would be useful in determining the ultimate usefulness
assessment determines whether adaptations in the gen-
                                                                    of the model as a replacement for current procedures.
eral education classroom can produce acceptable learn-
ing for individual students. If these adaptations are not           Evaluating Alternative Approaches
successful, Phase IV determines, through CBM proce-                    Although these and other identification procedures
dures, whether learning disability classification and spe-          have been proposed, none has achieved general accept-
cial education placement is effective for a given student.          ance. However, given the overall characteristics of the
If not, labeling and placement in special education is not          nature of learning disabilities, as described previously,
considered justifiable.                                             it may be possible to generate criteria that need to be
   Based on a recent review of 23 research reports on the           met in any valid, generally accepted identification pro-
characteristics of students who are unresponsive to early           cedure. These could include the following:
literacy intervention, Al-Otaiba and Fuchs (in press)                  • Does the identification procedure address the multi-
found that most unresponsive students were character-                    faceted nature of learning disabilities? That is, can it
ized by deficits in phonological awareness. Other char-                  be applied to disabilities in reading comprehension
acteristics identified less consistently included deficits in            as well as decoding skill areas, and other possible
phonological retrieval or encoding, verbal ability, behav-               areas, including, for example, math, writing, and
ior problems, or developmental delays. Al-Otaiba and                     spelling; or memory, attention, and study/organiza-
Fuchs suggested that future research address a common                    tional skills? If learning disabilities is considered to
definition of “treatment unresponsiveness” (e.g., reading                be multifaceted, measures of a single area of func-
fluency below 40 words per minute), and that more                        tioning (e.g., phonological core processes) will be
attention be given to characteristics such as phonologi-                 limited in identification of all learning disabilities.
cal memory and low IQ, and to the training and fidelity                • Can the procedure be applied across the age spectrum
of treatment implementation of trainers. Finally, they                   of students with learning disabilities? That is,
highlighted the difficulty in describing a “typical” treat-              although early identification is of great importance,

                                               Learning Disability Quarterly   162
can the procedures also be used (or modified) to          Addressing Problems in Identification
     identify students at other age levels? Procedures            To date, there have been considerable criticisms of
     that focus solely on early identification of learn-       identification of learning disabilities. Several proposals
     ing disabilities either assume that all students can      for modifying identification procedures have been
     be identified early and remediated, or that no            made, none of which is backed by sufficient research
     cases of learning disabilities will emerge after the      to document the likely consequences of implementa-
     primary grades.                                           tion. Additional research should be conducted to
  • Can the procedure be applied with known meas-              determine which, if any, of the proposed alternatives
     ures with demonstrated technical adequacy? That is,       result in improved identification of learning disabili-
     have variability problems due to reliability, valid-      ties over present methods. Before wide implementa-
     ity, or administration problems been addressed            tion, these procedures should be shown to be superior
     adequately?                                               to present methods in conceptual, technical, and
  • Will the procedure reduce overidentification of learn-     implementation domains.
     ing disabilities from present procedures?                    It should be remembered, however, that any modifi-
  • Will the procedure reduce inappropriate variability in     cation in identification must address effectively the
     identification rates across state and local educa-        aspects of identification that are presently most widely
     tional authorities?                                       criticized. A review of the literature critical of current
  • Will the procedure be more likely than current pro-        learning disabilities identification reveals that these
     cedures to identify students who meet present con-        problems include (a) overidentification, (b) variability,
     ceptualizations of learning disabilities, and be less     (c) specificity, (d) conceptual considerations, (e) dis-
     likely to produce false positives (identifying stu-       crepancy issues, (f) early identification, and (g) local
     dents who do not “really” have learning disabili-         implementation. Following is a set of suggestions that
                                                               may be employed to address all of these concerns.
     ties) or false negatives (failing to identify students
                                                                  1. Overidentification. Overidentification is often as-
     who “really” do have learning disabilities)?
                                                               sumed because the proportions of students identified
  Although many concerns have been expressed
                                                               as having learning disabilities have been consistently
regarding existing practice, no alternative to date has
                                                               growing, and because the present percentage of identi-
gained general acceptance. Clearly, any modification
                                                               fied students (over 5%) seems higher to many than
in present identification procedures must address the
                                                               reasonable for a “disability” status. Research in local
criticisms that have been raised previously. For exam-
                                                               and state identification procedures reveals that many
ple, any substantial study of identification procedures
                                                               students — perhaps as many as one third to one half
must address the issue of across-state and within-state
                                                               — are identified as having learning disabilities without
variability in identification rates. However, that does
                                                               meeting state criteria. One reason is that school per-
not necessarily mean that there is no reason for actual
                                                               sonnel realize that many students require special
prevalence rates to vary, and that future procedures           attention, which in many cases can only be provided
must lower variability. Characteristics of residents can       by referral to special education. Another commonly
vary by jurisdiction, and these characteristics can lead       described reason for overidentification is that other
to lower vs. higher prevalence of specific conditions.         categories of exceptionality — particularly mental
For example, the mean percentage of low-birthweight            retardation and emotional disturbance — are seen to
children is 7.03%, but individual percentages by state         be more stigmatizing than learning disabilities.
range from 4.7 to 15.4%. Such an observation does              Therefore, students are identified as having learning
not in itself suggest that procedures for identification       disabilities so that they can be provided with special
of low birthweight should be modified (Lester &                services without use of a more stigmatizing label
Kelman, 1997).                                                 (MacMillan & Siperstein, 2001). Overidentification,
  Nevertheless, variability in identification must be          then, can be significantly reduced not by provision of
understood far better than it is today, so that judg-          a new definition on the federal level, but by requiring
ments can be made on whether such variability is the           local educational authorities to employ strict adher-
result of variability in the characteristics of state popu-    ence to state definitional criteria.
lations or the result of inconsistently applied identifi-         If this is done, however, some provision must be
cation procedures. A true and complete understanding           made for other students who need assistance but do
of the observed variability in identification can lead         not meet learning disabilities criteria. For students who
directly to proposals for policies to reduce such vari-        are simply low achievers, or those who were formerly
ability, or to a better understanding of existing vari-        referred to as “slow learners” (e.g., IQs from 70-85),
ability in learning disabilities.                              schools should provide, and states should support,

                                              Volume 25, Summer 2002   163
some form of additional assistance. Since this could                 5. Discrepancy issues. Lyon et al. (2001) argued that
include over 15% of the school population, scarce spe-            “no definitional element of LD has generated as much
cial education funds should not be used to support                controversy as the use of IQ-achievement discrepancy
these learners. Better training for general education             in the identification of students with LD” (p. 265). And,
teachers, emphasizing educational methods for low-                in fact, some individuals will never accept the idea of
achieving students, could also do much to reduce refer-           discrepancy as a criterion for identification (e.g., Aaron,
rals for learning disabilities. Finally, for students who         1997). Nevertheless, it remains clear that discrepancy is
are identified as having learning disabilities because            a most objective indicator of learning disabilities, and
more appropriate categories are thought to be stigma-             that its elimination would result in overidentification
tizing, perhaps these other categories could be                   and variability at even higher levels than those at pres-
described differently, using terminology with more                ent. Further, it is difficult to understand contemporary
neutral connotations.                                             characterizations of learning disabilities without some
   2. Variability. If local implementation procedures             notion of intraindividual differences between ability
become more consistent, variability will also likely              and achievement (Mastropieri, 2001). More careful
decrease. States can also do much to reduce variability by        application of discrepancy criteria, in conjunction with
employing more consistent and specific criteria (perhaps          carefully documented exclusion criteria, can greatly
encouraged by the U.S. Department of Education). For              improve present identification practices. For example,
example, all states could employ a specific discrepancy           the five-level process recommended by Kavale and
formula (or other comparable criteria), and discrepancy           Forness (2001) includes specific discrepancy criteria in
formulas could be made more consistent. Since regres-             addition to documentation of pervasive basic skills
sion formulas are generally considered technically supe-          deficits, deficits in learning efficiency, psychological
rior (Kavale & Forness, 2000), states could specify that          process deficits, and exclusion of alternative causes of
regression formulas be employed, and provide specific             learning failure. Such a procedure, when carefully
input on the types of measures that are acceptable.               implemented, could potentially address many concerns
   3. Specificity. Present evidence suggests that earlier         with present identification practices.
concerns that learning disabilities could not be distin-             6. Early identification. Students with learning disabili-
guished reliably from general low achievement were                ties who are identified and treated early have a brighter
overstated. Nevertheless, it is important to establish            future than students with learning disabilities who are
that students identified as having learning disabilities          not identified early. However, it is possible that the
are reliably different from general low achievers, and            present relatively lower identification rates during the
that there is reason to infer a deficit in psychological          primary years are to a large extent the consequence of
processing that results in a disability to learn. This can        reluctance to categorize at early ages rather than limita-
be accomplished by strict adherence to explicit state             tions of present psychometric measures. Early identifi-
criteria and careful exclusion of alternative explana-            cation can be increased through state and federal
tions of discrepant functioning, including an inability           initiatives that encourage such practices and emphasize
to learn normally in spite of adequate learning oppor-            that students who are identified early are more likely to
tunities, cultural disadvantage, emotional disturbance,           be successfully treated. Nevertheless, it does not follow
mental retardation, or sensory impairments.                       that “the idea that special education funds can be used
   4. Conceptual considerations. Conceptual considera-            for early identification and prevention is critical” (Lyon
tions can also be addressed by careful adherence to               et al., 2001, p. 281). Special education funds, never fully
state criteria and application of the understanding that          provided by the federal government, are too limited to
IQ-achievement discrepancy is not equivalent to learn-            be employed in general education prevention efforts.
ing disabilities, but in fact is only one consideration           Such efforts — in the form of improved instruction and
used as evidence of the existence of learning disabili-           careful attention to the lower 25% in achievement —
ties. Adequate previous opportunities to learn should             can do much to limit referral to special education, but
be carefully evaluated. Prereferral interventions should          should be provided through general education funds.
be implemented in the general education classroom,                Limited special education funds should be reserved for
and shown to be inadequate, before students are iden-             support of intensive, high-quality interventions for stu-
tified. If all students identified as having learning dis-        dents meeting strict standards of disability.
abilities demonstrate average general abilities, but                 7. Local implementation. As has been argued, improve-
significant problems in areas relevant to academic                ments in local implementation of state and federal cri-
functioning in spite of adequate and even intensified             teria in the identification of learning disabilities may be
general education, many of the conceptual concerns                the key to addressing other problem areas. If state crite-
will have been met.                                               ria for learning disabilities were carefully implemented

                                             Learning Disability Quarterly   164
by local school districts in all cases, many if not most        funds not be used for this purpose. Federal and state
of the criticisms of identification of learning disabilities    general education funds could provide for low-
would be addressed. On the other hand, as long as local         achieving students in need of academic assistance,
practices remain subjective and idiosyncratic, any              while special education funds can continue to be
number of changes in definition or federal policy will          reserved for students identified as having disabilities.
be unsuccessful.                                                As numbers of students identified as having learning
                                                                disabilities diminish, federal support — not fully pro-
         SUMMARY AND CONCLUSION                                 vided since the inception of IDEA — can pay a more
   We have reviewed criticisms of present practices in          equitable share of the costs of special education.
identification of learning disabilities. Much criticism
involves concerns with overidentification, variability
in identification, and specificity in reliably discriminat-
                                                                                         REFERENCES
ing learning disabilities from general low achievement.         Aaron, P. G. (1997). The impending demise of the discrepancy
In addition, the definition of learning disabilities has          formula. Review of Educational Research, 67, 461-502.
been challenged on conceptual grounds, particularly             Ackerman, P. T., Holloway, C. A., Youngdahl, P. L., & Dykman,
with respect to discrepancy criteria, which, it has been          R. A. (2001). The double-deficit of reading disability does not fit
suggested, cannot provide for early identification.               all. Learning Disabilities Research & Practice, 16, 152-160.
                                                                Algozzine, R. (1985). Low achiever differentiation: Where’s the
Finally, local implementation practices have been                 beef? Exceptional Children, 52, 72-75.
described as subjective and arbitrary.                          Algozzine, R., & Korinek, L. (1985). Where is special education for
   In response to these concerns, several alternative             students with high prevalence handicaps going? Exceptional
methods for identification have been proposed. These              Children, 51, 388-394.
include deficits in level and slope of academic per-            Algozzine, R., & Ysseldyke, J. E. (1987). In defense of different
                                                                  numbers. Remedial and Special Education, 8, 53-56.
formance as assessed by curriculum-based measure-               Algozzine, R., Ysseldyke, J. E., & McGue, M. (1995). Differentiating
ment, deficits in phonological core processes, and                low-achieving students: Thoughts on setting the record straight.
failure to respond to validated treatment protocols.              Learning Disabilities Research & Practice, 10, 140-144.
Each of these procedures has addressed some aspects of          Alley, G. R., Deshler, D. D., & Warner, M. M. (1979). Identifica-
                                                                  tion of learning disabled adolescents: A Bayesian approach.
learning disabilities, but to date none has been demon-
                                                                  Learning Disability Quarterly, 2, 86-83.
strated to provide results superior to present proce-           Allor, J. H. (2002). The relationships of phonemic awareness and
dures. It was suggested that any alternative approach             rapid naming to reading development. Learning Disability
must address conceptual, technical, and implementa-               Quarterly, 25, 47-57.
tion aspects of the identification process, across age          Al-Otaiba, S. D. (2000). Children who do not respond to early liter-
                                                                  acy intervention: A longitudinal study across kindergarten and
levels and skill areas, and that it must identify individ-
                                                                  first grade. Unpublished doctoral dissertation, Vanderbilt
uals who share characteristics commonly associated                University.
with learning disabilities.                                     Al-Otaiba, S. D., & Fuchs, D. (in press). Characteristics of children
   Finally, we suggest that radically altering or elimi-          who are unresponsive to early literacy intervention: A review of
nating the concept of learning disabilities because of            the literature. Remedial and Special Education.
                                                                American Psychiatric Association. (2000). Diagnostic and statistical
problems with current identification procedures
                                                                  manual of mental disorders (4th ed.). Text Revision. Washington,
amounts to “throwing the baby out with the bathwa-                DC: Author.
ter.” We have proposed that many of the commonly                Anderson, P. L., & Meier-Hedde, R. (2001). Early case reports of
heard criticisms of identification of learning disabili-          dyslexia in the United States and Europe. Journal of Learning
ties can be addressed by careful attention to the local           Disabilities, 34, 9-21.
                                                                Baroody, A. J., & Ginsburg, H. P. (1991). A cognitive approach to
implementation aspect of identification. Such prac-
                                                                  assessing the mathematical difficulties of children labeled
tices could reduce numbers of students identified by              “learning disabled.” In H. L. Swanson (Ed.), Handbook on the
one third or more, and could increase the homogene-               assessment of learning disabilities: Theory, research and practice
ity of identified populations. Although these practices           (pp. 177-228). Austin, TX: PRO-ED.
would be more valid, and more within the spirit of              Beirne-Smith, M., Ittenbach, R. F., & Patton, J. R. (1998). Mental
                                                                  retardation (5th ed.). Columbus, OH: Merrill.
IDEA legislation, general education programs should
                                                                Beitchman, J. H., Cantwell, D. P., Forness, S. R., Kavale, K. A., &
also be implemented for the students who would not                Kauffman, J. M. (1998). Practice parameters for the assessment
be served as having learning disabilities, but who                and treatment of children and adolescents with leanguage and
nonetheless need additional assistance in acquiring               learning disorders. Journal of the American Academy of Child and
important academic skills. Such programming would                 Adolescent Psychiatry, 37 (Supplement), 46S-62.S
                                                                Berlin, R. (1887). Eine besondre art der wortblindheit: Dyslexia [A spe-
be very beneficial for general education students, and            cial type of wordblindness: Dyslexia]. Wiesbaden: Bergmann.
would likely reduce referrals to special education.             Berninger, V. W., & Abbott, R. D. (1994). Redefining learning dis-
However, it is important that scarce special education            abilities: Moving beyond aptitude — achievement discrepancies

                                               Volume 25, Summer 2002     165
You can also read