Unscientific Policies and Practices in LD
Several unscientific practices that are prominent in current policies and practices are discussed in this section. Proposals for changes follow.
Unscientific Vestiges of Early LD: Processing Disorder
Perhaps the most obvious of the unscientific vestiges of early LD (1960-1980) is the emphasis on the processing component in LD definitions and, occasionally, LD classification criteria. Definitions of diagnostic constructs specify key dimensions and other conditions fundamental to its meaning. Classification criteria specify what characteristics must exist in order for persons to be eligible or ineligible for the diagnosis. A healthy diagnostic construct has consistent definitions and classification criteria. An unhealthy diagnostic construct has major inconsistent features across the definition and classification criteria. LD is an unhealthy diagnostic construct because key elements in definition and classification criteria are inconsistent and scientifically unvalidated (note, unvalidated, not necessarily, invalid).
Current LD policy is driven by a definition (IDEA, 1997, 1999) that focuses primary attention on psychological process disorders as the presumptive cause of inefficient learning in several academic domains. LD classification criteria (IDEA, 1997, 1999) do not mention psychological process disorders, instead placing primary attention on a severe discrepancy between academic achievement in specified areas and intellectual ability, a criterion that itself appears to be problematic. Critical elements of the definition and classification criteria are inconsistent. What should change?
If we apply our criteria for change, scientific evidence and outcomes for children, the broad consensus among researchers is that both must change. Neither psychological processes nor severe achievement-ability discrepancy has empirical support.
The names for the psychological processes have changed over the years as noted by Mann nearly 25 years ago, Process training has always made the phoenix look like a bedraggled sparrow. You cannot kill it. It simply bides its time in exile after being dislodged by one of history's periodic attacks upon it and then returns, wearing disguises or carrying new noms de plum, as it were, but consisting of the same old ideas, doing business much in the same old ways. (Mann, 1979, p. 539),
Psychological processes, whether called modality preferences, psycholinguistic strengths-weaknesses, learning channels, hemisphere strengths, learning styles, or the more recent version, cognitive processes, can be used in LD for 3 purposes. In this analysis, I am referring to highly inferential assertions about internal child attributes based on psychometric profiles from standardized tests.
Processes in LD Identification. Some advocates argue that cognitive processes should be used in LD identification. In a few states today low achievers with cognitive processing weaknesses are diagnosed as LD while those without identifiable cognitive weaknesses are deemed ineligible. Many problems exist with this use and a full discussion would take more space than is available. But three well documented problems should pretty much lead to dismissal of cognitive processes as a consideration in LD identification. First, profile differences across cognitive tests are normal and not at all unusual. A high degree of variability across cognitive measures is typical of normal populations. There are many processing measures. Assessment across several measures nearly always identifies what could be regarded as a processing disorder in the majority of normal achievers. Any school or clinical psychologist can find a processing deficit in virtually every child, regardless of the child's school success. Processing deficits cannot, therefore, be a signifying feature of LD if nearly everyone has one or more of them regardless of achievement levels.
Second, difference scores on cognitive or neuropsychological measures are unreliable due to the phenomenon of unreliability of difference scores. Briefly, if there is error in each score on which the difference is based, there is still more error in the difference between the scores. An instructive example was contributed Danielson and Bauer (1978) who pointed out that the reliability of the difference score between IQ and achievement was 0.5, if the correlation between the tests was 0.7, the reliability of the achievement measure was 0.8 and the reliability of the IQ score 0.9. The reliability of difference scores increases as the reliability of each of the scores on which they are based increases and as the correlation between IQ and achievement increases. I note that the internal consistency form of reliability, (an upward bound estimate of test score reliability) is typically used in estimations rather than stability coefficients that add the element of elapsed time. Stability is a more relevant kind of reliability since constancy at least over short periods of time is crucial to making diagnostic and resource allocation decisions.
Third, many of the alleged processes derived through interpretation of published tests do not conform to the test's empirically derived factor structure, making the existence of the alleged process ephemeral. Some might argue that "auditory processing" is an exception; however, is it more effective to administer an instrument like the Woodcock-Johnson Cognitive Battery and perhaps find an auditory processing weakness or to directly measure phonemic awareness and alphabetical principles in the context of reading? The latter is more direct, more closely related to interventions, and less wasteful of time and resources. Greater emphasis on cognitive processing concepts has little potential for improving LD identification.
Training Cognitive Processes. Cognitive processes are sometimes advocated as targets for intervention. For example, Naglieri claims that training in "planning" improves mathematical computation performance with children with planning weaknesses (Naglieri & Das, 1997). The evidence is weak or disconfirmatory (e.g., Kroesbergen, Van Luit, & Naglieri, 2003; Naglieri & Johnson, 2000). Many similar claims have been made over the last 40 or more years, nearly always with the same result.
The pattern goes like this. Authors of tests measuring the "process du jure" claim positive evidence for gains in the processes and, occasionally, in achievement. Once these claims reach a certain level of prominence, other scholars subject the claims to further scrutiny, usually with better research methods, typically resulting in rejection of the claims and dismissal of yet another processing "innovation." As Mann (1979) noted, however, there always are new process test authors and new claims. Thorough reviews of the training of cognitive processes produced unequivocal conclusions across the last 2 generations of LD researchers (e. g., Hammill & Larsen, 1974; Kavale & Forness, 1999; Mann, 1979). Vaughn and Linan-Thompson (2003) stated these conclusions succinctly and well, "Although it may be accurate that many students with LD have underlying neurological and/or processing disorders, researchers and educators have been singularly unsuccessful at reliably identifying these difficulties and designing specific treatments to remediate them. (p. 141) "However, it is important to note that despite lack of support for process identification and treatment models, they continue to persist." P. 141).
Matching Teaching Methodology to Cognitive Process Strengths. A very attractive and old idea is to match teaching methodology to cognitive strengths. In the research literature the matching process is called "aptitude by treatment" interaction. Cognitive processes (aptitudes) are matched to treatments (teaching methods). The matching is justified by asserting that children with achievement problems will learn best if the method of teaching capitalizes on their processing strengths. The potential forms of matching are as varied as the cognitive processes themselves, but the most common match is to pair cognitive process (e.g., visual-auditory modality, simultaneous-successive processes, right-left hemisphere) to reading methodology with the first of the named process pairs assigned to whole language-whole word methods and the second to phonetic methods.
No substantial evidence supports these pairings, although they continue to be popular (Kavale & Forness, 1999). As Vaughn & Linan-Thompson (2003, p. 142) commented, "There is no empirical support for the use of modality-matched instruction or learning styles as a means to enhance outcomes for students with LD." Cronbach (1957), one of the originators of the concept of aptitude by treatment interactions, disavowed the construct in 1975, "Once we attend to interactions, we enter a hall of mirrors that extends to infinity." (Cronbach, 1975, p. 119).
The most common LD definition that focuses on LD as a disorder in psychological processes has misguided the field into emphases on concepts, assessments, and treatments that do not work. Cognitive processes are not useful as a basis for LD identification, as targets for interventions, or in the design of treatments. It is time to change to LD identification models that are more reliable, valid, and treatment relevant.
Unscientific Vestiges of Early LD: Severe Discrepancy
The Bureau of Education for the Handicapped faced a dilemma in 1975. Congress mandated that they establish methods to limit LD prevalence and provide guidance to states on LD identification (Federal Register, 1976, 1977). If these criteria were not published in final form by January 1, 1978, what amounted to a 2% cap on LD prevalence would become permanent.
Although an LD definition that emphasized underlying process disorders was established in Federal requirements, research developments in the early to mid-1970s precluded using processing deficits as the basis for classification criteria. A series of scathing articles were published in the early to mid 1970s that severely criticized the use of processing assessment and LD treatments based on overcoming processing deficits (Hammill & Larsen, 1974, 1978; Mann, 1979; Newcomer, Larsen, & Hammill, 1975; Ysseldyke, 1973).
A proposed set of regulations to establish LD classification criteria emphasizing a discrepancy between academic achievement and intellectual ability was published in the Federal Register, November 29, 1976. The most controversial feature was the expectancy formula that would precisely define the nature and magnitude of the discrepancy. No mention of processing factors appeared in the proposed LD regulations or in the revisions eventually adopted. Sharp debate ensued and nearly 1000 mostly critical letters were submitted to the BEH ("Coming Face to Face," 1978; Danielson & Bauer, 1978; Federal Register, December 29, 1977). The field was virtually united in opposition to the expectancy formula due to technical and conceptual issues; however, relatively little discussion of the intellectual ability-achievement discrepancy criterion per se occurred at this time.
The final LD classification criteria without the expectancy formula were published just prior to the Congressional deadline (Federal Register, December 29, 1977; 34 C.F.R. 300.540-544). As noted previously, the LD classification criteria specified a severe discrepancy between ability and achievement in 1 or more of 7 achievement areas along with exclusion of other plausible causes. Absent was a discussion of, or guidance on, how the severe discrepancy would be determined or the magnitude of the discrepancy that would be "severe." As shown later, consistent discrepancy determination methods do not exist across the states or, in many instances, within states.
The subtle, but crucial change in what LD meant, from process disorder to unexplained low achievement (Federal Register, 1977) did not receive a lot of discussion in the literature in the late 1970s. An exception was an editorial in the Journal of Learning Disabilities, pointing out that the nearly unanimous outcry against the expectancy formula resulted in overlooking the conceptual shift in LD from psychological processes to unexplained underachievement (Senf, 1978).
The significant shift to severe discrepancy was an uneasy compromise that solved a political problem in the late 1970s. Little research was done on the possible consequences of the severe discrepancy and what was done indicated potentially significant problems with alternative determination methods, reliability, validity, and control of prevalence (Danielson & Bauer, 1978). Severe discrepancy was arguably a necessary expedient to solve a problem of concern to Congress in 1975 that had not been resolved by the LD field, how to identify LD and how to control prevalence. We struggle for answers to that question today, particularly since the solution adopted 25 years ago is fundamentally flawed from a basic measurement perspective (Shepard, 1980, 1983; Shepard & Smith 1983; Shepard, Smith, & Vojir, 1983).
Limited theoretical support for the severe discrepancy classification criterion existed in the mid-1970s (Rutter & Yule, 1975) who reported two groups of children with low reading achievement that varied in response to intervention, a group with ability consistent with achievement and another with ability higher than achievement. Subsequent research has largely disconfirmed the two group analysis (e.g., Fletcher at al., 2002; Francis, Shaywitz, Stuebing, Shaywitz, & Fletcher, 1996; Shaywitz, Escobar, Shaywitz, Fletcher, & Makush, 1992). Convincing data disconfirming the Shaywitz et al. conclusions have not been offered over the past decade.
"This study allowed us to investigate the commonly held belief that dyslexia is a discrete diagnostic entity. Our data do not support this notion. Rather, they suggest that dyslexia occurs along a continuum that blends imperceptibly with normal reading ability. These results indicate that no distinct cutoff point exists to distinguish children with dyslexia clearly from children with normal reading ability; rather, the dyslexic children simply represent the lower portion of the continuum of reading capabilities." (p. 148)
Several good treatments of the inadequacies of the severe discrepancy classification criteria have appeared in the last decade (e.g., Fletcher et al., 2002; Gresham, 2002; Stage, Abbott, Jenkins, and Berninger, 2003; Steubing et al., 2002; Vellutino, Scanlon, & Lyon, 2000) and a number of learned societies or panels have advocated abandonment of the severe discrepancy (Bradley, Danielson, & Hallahan, 2002; Donovan & Cross, 2002; Learning Disabilities Roundtable, 2002; President's Commission on Excellence in Special Education, 2002). We see no reason to repeat these arguments or the efforts to refute them (e.g., Kavale, 2002; Scruggs & Mastropieri, 2002). At best, severe discrepancy is seen by its advocates as one marker of LD, often without specifically delineating other markers beyond the exclusion factors. We believe that the balance of the evidence shows that the severe discrepancy classification criteria are (a) unreliable (particularly in the sense of stability), (b) invalid (poor readers with higher IQs do not differ on relevant variables from those with IQs commensurate with reading levels), (c) easily undermined in practice by giving multiple tests, finding a score that is discrepant and ignoring disconfirming evidence, and (d) harmful because the severe discrepancy delays treatment from kindergarten or first grade when the symptoms of reading disability are first manifested to 3rd or 4th grade when reading problems are more severe, intervention more complex, and the school curriculum shifts to "reading to learn."
In order to avoid harm to children and to move toward LD classification criteria that are more reliable and valid we need to eliminate the severe discrepancy requirement. It does more harm than good.
Previous Page | Next Page
(Introduction) | (Alternatives)
