Classification of Learning Disabilities: An Evidence-Based Evaluation
Jack M. Fletcher, University of Texas; G. Reid Lyon, National Institutes of Health; Marcia Barnes, University of Toronto; Karla K. Stuebing, University of Texas; David J. Francis, University of Houston; Richard K. Olson, University of Colorado; Sally E. Shaywitz, Bennett A. Shaywitz, Yale University
Learning Disabilities Summit: Building a Foundation for the Future White Papers
This paper is available in alternative formats: | Download Word | Download pdf |
Pages: | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
EXCLUSION HYPOTHESIS
Most definitions of LD include an exclusion clause, which simply states that LD is not primarily the result of other conditions that can impede learning. These other conditions include mental deficiency; sensory disorders; emotional disturbance; cultural, social, and economic conditions; and inadequate instructional opportunities. Given the role of the exclusion element within definitions of LD, children identified as LD are often identified on the basis of what they are not, rather than what they are. This is unfortunate for three major reasons. First, by placing an emphasis on exclusion, the development of inclusionary characteristics that are linked to assessment and intervention is difficult. To illustrate, the 1977 operationalization of the federal definition suggests that RD be assessed with IQ and achievement tests; it does not specify the domains of reading and ignores the variation in components of reading that may represent differential treatment emphases. Second, an exclusionary definition is a negative definition that adds little to conceptual clarity and clearly constrains understanding the disorder to its fullest extent. Think of the difference in clarity when we identify a child as LD versus identify the child with a reading comprehension disability. Third, many of the conditions that are excluded as potential influences in LD are themselves possible factors that interfere with the development of those cognitive and linguistic skills that lead to the academic deficits that form the basis for LD (Lyon et al., 2001). Parents with reading problems, for example, may find it difficult to establish adequate home literacy practices because of the cumulative effects of their reading difficulties (Wadsworth et al., 2000).
It is reasonable to stipulate that children with mental deficiency and sensory disorders are excluded from classifications of LD. Separate categories exist and their treatment needs are different. This stipulation begs the question of how to differentiate mild mental deficiency and LD (Gresham, MacMillan, & Bocian, 1996), given the weakness of psychometric definitions. Other exclusions are even more difficult to justify. For example, where is the evidence suggesting that RD and MD are different in children who are anxious, depressed, or even psychotic? Recent longitudinal studies suggest that early achievement is causally related to and often precedes the development of behavioral problems, and the interventions that enhance academic achievement prevent behavioral difficulties (Kellam, Rebok, Mayer, Ialongo, & Kalodner, 1994; Onatsu-Arvilommi & Nurmi, 2000). Thornier are exclusions based on social, economic, and cultural disadvantage, and inadequate instruction. This differentiation is based on the presumption that constitutional factors are more relevant for children with LD than environmental factors. In addition to reviewing evidence for exclusion according to environmental factors, we will also review additional evidence on the role of constitutional factors in LD, raising again the question of how well either environmental or constitutional factors distinguish children with LD from low achievement commonly ascribed to environmental factors.
Social, Economic, and Cultural Disadvantage
A variety of factors related to the literacy environment in which a child develops are clearly related to the acquisition of academic skills. When optimal social and economic conditions are not present, the child is at a much higher risk for the development of an academic problem. In reading, a variety of factors have been studied, including print exposure, parental literacy levels, and reading to the child. All these factors are related to the development of reading skills (Adams, 1990) and probably to other academic skills as well. Recent qualitative studies (Hart & Risley, 1999) have provided graphic documentations of the differences in the language environment experienced by advantaged and disadvantaged children. For example, by the age of 5 years, economically advantaged children have vocabularies of approximately 500,000 words, while economically disadvantaged children have vocabularies of approximately 250,000 words (Hart & Risley, 1999). It is widely believed that these types of differences in language development have some (unspecified) effect on brain development, and they are certainly related to the development of proficiency in academic skills.
These types of factors impede oral language development. When oral language development is affected, a variety of language skills are at risk, including those related to the development of word recognition and reading comprehension skills. In a series of longitudinal studies, Whitehurst and Lonigan (1998) provided excellent documentation of the relationship of different oral language skills and the acquisition of reading ability. In evaluating children entering Head Start programs at age 3, Whitehurst and Lonigan (1998) found that skills related to knowledge of the alphabet and word structures were closely tied to reading success in kindergarten, Grade 1, and Grade 2. More general oral language skills that involved vocabulary, language comprehension, and exposure to language through literature and oral reading were related to the development of reading comprehension skills, particularly in Grades 2 and 3.
These results are particularly striking when national evaluations of Head Start programs are examined (Whitehurst & Massetti, in press). These studies have shown that, on average, children who graduate from Head Start programs enter kindergarten knowing one letter of the alphabet. Interviews with teachers showed that they were often discouraged from engaging in activities that promoted understanding of the alphabetic structure of the language because such activities were not viewed as "developmentally appropriate." Nonetheless, when children in Head Start programs were provided with these sorts of activities, higher literacy levels were apparent (Whitehurst & Lonigan, 1998).
What is important in these examples is the illustration that environmental factors influence the development of oral language skills that are known to affect beginning (and later) reading skills. Interventions that address the early development of these skills seem to promote success in reading. Such findings are also apparent in evaluative studies of Title I Programs as well as intervention studies in which alphabetic forms of instruction have been shown to be advantageous for economically disadvantaged children (Foorman et al., 1998; National Reading Panel, 2000). Thus, the mechanisms and practices that promote reading success in advantaged populations appear to be similar to those that promote reading success and failure in disadvantaged populations. There is little evidence that the phenotypic representation of RD varies according to socioeconomic status. Children at all levels of socioeconomic status appear to have reading problems predominantly (but not exclusively) because of word-level difficulties apparent in the beginning stages of reading development (Foorman et al., 1998; Wood, Flowers, Buchsbaum, & Tallal, 1991).
As Kavale (1988) and Lyon (1996) pointed out, the basis for excluding disadvantaged children from the LD category has more to do with how children are served than with empirical evidence demonstrating that characteristics of reading failure are different in LD in economically disadvantaged groups. Indeed, Kavale (1988) suggested that arguments usually point to the fact that "the culturally disadvantaged child is well served by various federally funded title programs, but these are usually mandated under guidelines and revisions different from special education. Specifically, the emphasis is on compensatory education while special education programs function as remedial programs" (p. 195). This has the effect of eliminating economically disadvantaged children from special education services, with the exception of categories related to mental deficiency and emotional disturbance; economically disadvantaged children are disproportionably represented in these special education categories. As Kavale stated,
since culturally disadvantaged children have been shown to exhibit the behavioral characteristics included as primary traits in definitions of LD, it is difficult to determine why the culturally disadvantaged group is categorically excluded from the LD classification. Yet, children from lower SES levels with LD-type behaviors have little chance for receiving LD diagnoses and treatment with an increased likelihood of being labeled retarded in spite of the fact that LD and ED groups are not clearly identifiable as separate entities. (p. 205)
There is little empirical evidence supporting the exclusion of economically disadvantaged children from special education services as a valid classification practice. The exclusion is a policy decision that represents a desire to clearly separate funds dedicated to special education and compensatory education. We have, essentially, a two-tier service delivery system for children with academic difficulties, where advantaged children are designated as LD and served through remedial classes that are questionably effective (Lyon et al., 2001). In contrast, children who qualify for free lunches served under Title I often receive compensatory education programs, which in some studies appear to be effective (Slavin, Karweit, & Madden, 1989). Our concept of LD, however, must not hinge on policy issues. Here there is little basis for distinguishing types of poor achievement according to putative causes, since the phenotypic manifestations seem to be similar across levels of socioeconomic status.
Instruction
Virtually any definition of LD excludes children from consideration if their learning problems are primarily a product of inadequate instruction. Of all the different assumptions in the concept of LD, this assumption is the least frequently examined and perhaps the most important. Some would interpret the exclusion to indicate that children who profit from instruction do not have a biologically based disorder. The functional imaging studies reviewed below suggest that this is hardly the case and that instruction is necessary to establish the neural networks that support reading. Keep in mind that no child is born as a reader; all children are taught to read. Written language is scaffolded upon our natural capacities for developing oral language (Lukatela & Turvey, 1998). It may be that there are differences in brain function that make some children more refractory to intervention than others, but we do not presently have data that would indicate that this is the case.
Another problem with the inadequate instruction exclusion is that it presumes that the field has a good understanding of what constitutes adequate instruction. At the time the federal definition was adopted, this was not case. Recent consensus reports (National Reading Panel, 2000; Snow et al., 1998) make it clear that we do know a lot about teaching children to read. Given what we know, consideration of the students' response to well designed and well implemented early intervention as well as remediation programs may need to become part of the definition of LD. Why should the complex identification criteria and expensive due process procedures of special education be used before an attempt is made to provide a powerful intervention early in the child's development? A child's failure to respond to intervention may be the best way to operationalize the notion of adequate instruction. While a child's failure to respond to appropriate instruction is a very strong indication of a disability, the cognitive problems associated with their LD parallel those exhibited by children who do not respond to inadequate instruction. The two types of children are equally disabled and there is no evidence that there are differences at a neurological level, prior to intervention or in terms of their intervention needs, that would make them different. For children with mental deficiency, sensory disorders, and emotional disturbance, there are other classifications in IDEA that can lead to services. For the child who is deemed culturally, economically, or socially disadvantaged, compensation education programs are available. What is there for the child who develops academic difficulties because of poor instruction? Excluding children on the basis of inadequate instruction does not seem a reasonable practice.
Constitutional Factors
Approaching the exclusion hypothesis from the perspective of classification research shows little evidence supporting exclusions based on emotional disturbance; social, cultural, and economic disadvantage; or inadequate instruction. This reflects the difficulties of differentiating forms of low achievement that are presumably specific or unexpected from those than can be attributed to other causes, where low achievement is expected. Related to this hypothesis is another source of data that is frequently invoked in explaining unexpected low achievement. That is the notion that unexpected LD is due to constitutional factors that are intrinsic to the child. In the current federal definition of LD, the intrinsic/neurological component is implicit in the use of terms like "basic psychological processes." In other contemporary non-federal definitions the concept is explicitly stated. For example, the definition of LD proposed by the National Joint Committee on Learning Disabilities (1988) states: "these disorders [LD] are intrinsic to the individual, presumed to be due to central nervous system damage, and may occur across the life span" (p.1).
Neurobiological factors do not represent formal classification hypotheses in the sense that they are used to identify students with LD. They do represent components that can be tested for validity purposes. If children with unexpected low achievement differ from children in whom achievement is expected on constitutional factors, then this might support the hypothesis that expected and unexpected low achievement should be differentiated.
It has long been assumed that neurobiological factors were the basis of LD, reflecting its conceptual origins in the notion of organically based behavior disorders (Doris, 1993; Rutter, 1982; Satz & Fletcher, 1980). Neurobiological (constitutional) dysfunction was inferred from what was then known about the linguistic, cognitive, academic, and behavioral characteristics of adults with documented brain injury or lesion. As the field progressed, definitions of LD continued to attribute the disabilities in learning to intrinsic (neurological) rather than extrinsic (e.g., environmental, instructional) causes, even though there was no objective way to adequately assess the presence of putative brain damage or dysfunction. These assumptions of constitutional etiology were buttressed by associations of a variety of indirect indices of neurological dysfunction and LD. These indirect indices included observations of perceptual-motor problems (i.e., difficulty copying geometric figures), paraclassical or "soft" neurological signs (e.g., gross motor clumsiness, fine motor incoordination), and anomalies on electrophysiological measures, such as an electroencephalogram (Taylor & Fletcher, 1983). Even at the time, the lack of specificity of these observations with either LD or neurological integrity was widely acknowledged (Satz & Fletcher, 1980). Nevertheless, the neurobiological deficits were presumed to be selective rather than diffuse, resulting in specific difficulties processing linguistic, visual, and motor information critical to academic learning without concomitant loss of general intellectual functions.
Over the past two decades, some evidence, varying widely in methodological quality, has been obtained from investigations designed to identify, more directly, the neurological basis for LD and particularly RD. For example, data derived from postmortem studies performed on dyslexic adults and structural neuroimaging studies with children and adults have indicated that some individuals with RD are characterized by differences in the size of specific brain structures (e.g., planum temporale) and in the presence of specific neuroanatomical anomalies (e.g., ectopias) (Filipek, 1996; Galaburda, 1993). Structural imaging studies reliably show that people with RD have a smaller left hemisphere, or less asymmetric hemispheres. Both the autopsy and structural imaging studies have been confounded by subject selection problems, failure to account for comorbid neurological disease (e.g., seizures) and other variables (e.g., handedness). Interpretation of the structural imaging studies has been impeded by the use of different neuroimaging methods and data analytic techniques, as well as difficulties replicating the findings of these studies (Filipek, 1996; Shaywitz et al., 2000).
More recently, research using different types of functional neuroimaging methods to measure brain activation in response to visual, linguistic, and reading tasks among skilled and unskilled readers indicates systematic and selective brain activity in several left hemisphere neural systems subserved by the basal surface of the temporal lobe, the middle temporal gyrus, the temporoparietal region, and the inferior frontal region. Converging evidence from a range of functional imaging methods used in studies with both good and poor readers indicate that a network of brain areas is involved in the ability to recognize words accurately, and that adults and children with RD manifest different patterns of activation in these areas when compared with skilled readers (Shaywitz et al., 2000).
A critical question that has been raised by the functional neuroimaging studies of those with LD in reading is whether the patterns seen in these individuals with RD are compensatory in nature ("compensatory hypothesis") or reflect the failure of the environment and/or instruction to impact the brain in a manner necessary to form the neural networks that support word recognition. Thus, the pattern in RD children may be similar to that seen in a young child who has not learned to read and may change by virtue of development, instruction, or even intervention ("normalization hypothesis"). Given this possibility, functional neuroimaging studies may provide an example of how brain and environment interact in forming neural networks for complex behaviors. Such studies are feasible and investigations that combine neuroimaging and reading intervention studies are currently being completed.
Figure 8. Individual activation maps from a 10-year-old child who was experiencing serious difficulties in learning to read, before and after an intense phonologically-based intervention. Activation maps were obtained using a pseudoword rhyme-matching task. The child showed dramatic improvement in phonological decoding skills after 8 weeks (80 hours) of enrollment in the program, analogous to that reported by Torgesen et al. (2001). Note the dramatic increase in the activation of the left temporoparietal regions.
Figure 8 provides an example from a pilot study in which functional neuroimaging studies were performed using magnetic source imaging while a child read words. The imaging studies occurred before and after approximately 60 hours of intense intervention (over 8 weeks) in which the child, who was 10 years old with severe RD, showed significant improvement in word reading ability into the average range. The top part of the figure shows the standard brain activation pattern characterized by activity predominantly in the temporoparietal regions of the right hemisphere. After intervention, the pattern shifts to predominant activation involving the homologous areas of the left hemisphere, an activation pattern typical of non-disabled readers. Thus, these results are more consistent with the normalization hypothesis than the compensatory hypothesis.
The preliminary data from these types of studies suggest different conceptualizations of the role of constitutional factors in LD. The view that is emerging suggests that neural systems develop and are deployed for specific behaviors through the interaction of brain and environment (including instruction) as opposed to representing fixed properties of the nervous system that inherently limit learning potential. As such, the concept of LD retains the optimism that was intended with its inception.
This interaction perspective is also supported by genetic studies of individuals with RD. It has long been known that reading problems reoccur across family generations, with a risk in the offspring of a parent with RD 8 times higher than in the general population. Multiple genes are most likely involved, with similar modes of transmission in dyslexic and non-dyslexic families. Linkage studies implicate markers on chromosomes 1, 2, 6, and 15. However, genetic factors account for only about half of the variability in reading skills, which means that the environment has a significant influence on reading outcomes. This also suggests that what is inherited is a susceptibility for RD that may manifest itself given specific interactions, or lack thereof, with the environment. For example, parents who read poorly may be less likely to read to their children. As such, the quality of reading instruction provided in the school may be more critical for children when there is a family history of poor reading giving rise to limited environmental-instructional interactions in the home (Olson, Forsberg, Gayan, & DeFries, 1999; Pennington, 1999; Wadsworth et al., 2000).
Conclusions: Exclusionary Criteria
There is little evidence that children excluded from LD classifications due to emotional disturbance; social, economic, and cultural disadvantage; or instructional history are meaningfully different from those included as LD. In particular, none of these criteria provide robust differentiations of expected and unexpected low achievement. The notion that expected and unexpected low achievement reflects variation in cognitive and behavioral correlates, prognosis, response to instruction, or even a broad range of neurobiological factors, does not have strong validity. This does not mean that the concept of LD is not valid or that the exclusions should not be used, particularly since many children can be served under other categories in IDEA or other approaches to providing services (e.g., compensatory education). There may well be needs outside the academic area that are better addressed through identification for other categories or programs. Exclusions due to inadequate instruction are not justifiable as lack of instruction can essentially cause LD. The exclusions must be seen as policy-based determinations to facilitate service delivery and avoid commingling of facts, not as classification factors that have strong validity.
