ISSUES FACING THE FIELD OF LEARNING DISABILITIES.
Parts of this paper were presented at the Distinguished Lecture series, International Council for Learning Disabilities, Albuquerque, New Mexico, November 1998. Partial funding for this paper comes from the Peloy Endowment Fund, University of California-Riverside.
The number of students classified as having learning disabilities (LD) has increased substantially over the last 20 years. There were 783,000 children identified with LD in 1976, but by 1992-1993 the LD population totaled approximately 2.3 million. These children comprise almost 50% of all placements into special education (U.S. Office of Education, 1994). In addition, approximately another 120,000 students each year are identified as LD, a number equal to all Americans who have contracted AIDS, hepatitis, and tuberculosis in 1995 (Roush, 1995). Based on these figures, one could argue that the classification of children with LD is epidemic. In an article published in Science entitled "Arguing Over Why Johnny Can't Read," Roush (1995) stated "If learning to read and write or do math at expected levels were a disease, then American school children would be in the middle of an epidemic" (p. 1986). This is a costly epidemic considering that public schools spend approximately $8,000 a year to educate an LD student compared with approximately $5,500 for an ordinary student (see also Dillon, 1994).
The popular press has published a plethora of articles on this epidemic. For example, several articles have related that would-be lawyers who are dyslexic are suing over bar exams (see the New York Times, October 23, 1997), dyslexic students are suing colleges over entrance exams (e.g., see the Los Angeles Times, September, 15, 1997), and some articles have suggested that we have gone too far in accommodating the needs of students with LD (see Robert Sternberg's article in the New York Times, August, 25, 1997).
Three news articles are of particular interest. The first appeared in the New York Times, Friday, April 8, 1994, with the caption "A Learning Disabilities Program That Gets Out of Hand." The article described a wealthy New York City real estate family donating in excess of 2 million dollars to a private school (Dalton School) to create a model education program for children with LD. One of the donors had experienced learning problems herself as a young girl and it was not until she was 50 years old that she was diagnosed as having dyslexia. Her contribution to the school supported approximately 14 full- and part-time learning specialists. She also financed research at the school to develop a screening test that would identify students with learning disabilities at an early age. Finally, she gave hundreds of thousands of dollars to a nearby university to evaluate the program and publish scholarly papers.
Of interest in this news story is the relationship between money and outcome. With increased resources, the school's new team of remedial specialists suddenly found increasing numbers of bright little children (IQ between 102-132) with LD. This is noteworthy because approximately 40 percent of the seniors who graduate from this private school attend Ivy League colleges. In one three-year period, during the second month of kindergarten, the school had labeled 77 children (out of 250) at-risk and commenced to give them remedial help. What emerged was a situation in which parents, who just a few months earlier were proud to have their five-year-old accepted at this very prestigious school, were suddenly told that the screening test indicated potential visual-motor problems or sequencing-ability deficits. Private specialists were then hired at $75 to $250 per hour, and by the fourth grade approximately half of the children in the school had received help or remedial training of some sort.
The tragedy is that despite the thousands of dollars given to this school, as well as to the university incorporated to do the research, no one can say with objective certainty that the remedial program helped the students. One of the researchers was quoted as saying, "In the field of education there is this problem with research. People don't think about setting up controls like a science, so a good intention by the benefactor in terms of offering help that she never got turned bad in several ways." In what ways did it go bad? Two ways. First, funding was out of proportion to what was needed. There were more specialists than necessary. Second, there was bad research. No control group was established.
A second scenario was published in Washington Post, February 26, 1991, which raised the issue "Learning Disabled or Pushed Too Fast?" The article addressed the explosion of cases of reading problems among young children in the Washington, DC, area. One school official in the Chevy Chase neighborhood was quoted as stating that "LD identification is epidemic." The specialist attributed the phenomenon to yuppie families who pushed their kids too fast and the schools that required too much academics too soon. The theme reflected within the news clip was that because children between the ages of five and seven are becoming forced to grow up too fast, many are being identified as learning disabled.
A final news clip of interest occurred in Washington Post, Thursday, December 15, 1994, in which the Federal Trade Commission (FTC) indicated that a reading program entitled "Hooked on Phonics" misled consumers. The advertisement suggested that children with reading problems, dyslexia or short attention span would be able to learn without a teacher via phonics training "put" to music. The advertisers, The Gateway Education Program, claimed that the program could easily teach children and adults, including those with LD. Unfortunately, none of Gateway's claims related to helping children with LD has been proven. The program was characterized by some consumers as simple repetition and no more than you would usually do with other sorts of programs. Because the FTC settlement involved Gateway's misleading ads, but did not contend with the company's actions as being fraudulent, there was no consumer redress as part of the settlement. If the company continues to violate this settlement, however, they will face civil fines of $10,000 for each violation. Interestingly, the company declined to disclose their total income from the sales of the reading program. However, the program, which includes color-coded workbooks, audiocassettes, and flash cards at a cost of $230, has sold over one million copies.
These stories in the popular press capture some important issues in the field of LD that need to be addressed. These issues relate to how best to: (a) define students with LD so they are not overidentified, (b) delineate the boundaries of how specific the disability must be, (c) identify effective education treatments, and (d) implement instructional policy that is in the best interest of the student. Although these issues have been discussed in the literature for some time, several deeper conceptual issues lie below the surface. It is my assumption the issues are related to: (a) a weak research foundation for the validity of discrepancy-defined groups; (b) a narrow focus of recent research, thereby excluding work in other areas; (c) a limited instructional research base to answer very practical questions; and (d) implementation of policy that is independent of research. Each of these issues will be explored, followed by suggestions for a redress.
VALIDITY OF DISCREPANCY-DEFINED GROUPS
The first issue to consider is the validity of discrepancy-defined groups (see Aaron, 1997). The salient and distinct feature of individuals classified with LD within the public school system is that their disability in an academic domain such as reading is highly discrepant from their general intellectual competence and various educational or social opportunities. Calculating discrepancy between IQ and achievement on standardized tests is the most common procedure for representing differences between potential and actual performance. Most states and school districts rely on discrepancy-based formulas that include IQ and achievement as the primary procedure to classify children as LD (e.g, Frankenberger & Fronzaglio 1991; Mercer, Jordan, Allsapp, & Mercer, 1996). The procedure can be traced to the use of expectancy formulas that emerged within the reading literature at the turn of the century (e.g., see Monroe, 1932).
The implicit assumption for the inclusion of discrepancy scores in the classification of LD was that children who experience reading, writing and/or math difficulties, unaccompanied by a low IQ, are distinct in cognitive processing from the general "run of the mill" poor, garden-variety, slow, or mildly retarded learners. A major study on the distinctiveness of "discrepancy-defined children" emerged in the 1970s. Specifically, Rutter and Yule (1975) focused on nine-year-old children who were divided into two groups of poor readers. One group encompassed the backward reader, with an IQ around 80. The other was a specific reading retardation group where the average IQ was around 102. The specific reading retardation group was less likely to have organic brain damage or to display neurological abnormalities when compared to backward readers. Backward readers, on the other hand, displayed a variety of motor abnormalities and left-right confusions. Both groups had similar proportions of family members with histories of delays in language development. More important, Rutter and Yule found differential growth curves for specifically retarded readers when compared to the general retardation group. The backward reader displayed a greater growth in reading and less growth in arithmetic than the specific reading retardation group.
In contrast to this earlier study, several recent studies suggest that, generally, backward and specific reading retardation groups are not distinct (see Stanovich & Siegel, 1994, for review). No doubt, some of these studies (see reviews by Fletcher et al., 1994; Siegel, 1992; Stanovich, 1991) provide equivocal findings (i.e., some yield null results, others yield small, but statistically significant differences between IQ-based discrepancies and low reading groups). Recent studies correcting some of the flaws in the earlier studies, however, converge on the notion that children classified with LD, specifically in the area of reading, reflect a normal distribution of a multifactorial trait found in reading characteristic of other poor readers (Fletcher et al., 1994; Stanovich & Siegel, 1994). Quite simply, when compared to nondiscrepancy-defined poor achievers, learning disabled-defined groups are more similar in processing difficulties than different (e.g., Shaywitz, B. et al., 1992; Siegel, 1989; Stanovich & Siegel, 1994; Stanovich & Stanovich, 1997). Thus, it is becoming an untenable idea that aptitude-achievement discrepancy tells us anything important about processing mechanisms underlying such areas as reading disabilities (see Aaron, 1997, for a review). This is an interesting situation because such IQ-reading discrepancies are stable (or persistent) in some children across time (Fergusson, Horwood, Caspi, Moffitt, & Silva, 1996; however, see Shaywitz, Escobar, Shaywitz, Fletcher, & Makugh, 1992).
Interestingly, the implications of findings between LD and garden-variety poor readers have not filtered down to actual diagnostic practice in the schools. The majority of research at the state or school district level focuses on computation (identifying more conservative formulas that account for such issues as regression and overidentification, e.g., Reynolds, 1984-1985; Shephard, 1980), not necessarily issues of construct validity (meaningful classification systems). To date, research that focuses on methodology (e.g., comparing formulas in terms of their ability not to overidentify) in many cases operates independent of construct validity issues. Because of this split in research, one of the most serious conceptual challenges to the field is that the cognitive profile of children with LD cannot always be reliably discriminated from generally low-achieving children. No doubt, one can garner several criticisms for the line of research that compares discrepancy and nondiscrepancy groups and finds no differences in performance. These criticisms are not powerful in and of themselves because they indicate only problems in the depth of our knowledge, rather than isolating serious conceptual flaws. These general criticisms will be briefly outlined, however, followed by the more specific concerns.
First, there is a mismatch between conventional wisdom and research outcomes. Most lay people view LD as reflecting some sort of difference between what they can do and what they actually do in certain situations (Swanson & Christie, 1994). While models of LD do not have to match a lay person's understanding, the notion of discrepancy is a frequently reported phenomenon of parents of LD children and the person with LD (Gerber & Reiff, 1991). The weakness of this argument is that people who are generally slow in all academic areas may also operate with the notion, "I have more potential than I am showing."
Second, there has been little empirical evidence available on the validity of a discrepancy score other than between intelligence and reading. A discrepancy in discriminating between groups may have validity in the deficient areas of mathematics, problem solving, writing, and so on.
Third, in contrast to statistical artifacts, we do not know why a discrepancy emerges between IQ and achievement (e.g., reading). In a "psychological" sense, very little research has been directly focused on understanding the determinants of a discrepancy (however, see Sovik, Frostad, & Lie, 1994).
Fourth, most research designed to test the discrepancy model is conceptualized in terms of searching for the association of deficits. Support for the null hypothesis may provide a very insecure basis for theorizing. The finding that differences between poor readers and the learning disabled reader are statistically comparable or that variations of a score are on the same reading continuum is not theoretically compelling. More compelling is the possibility of dividing discrepancy groups into subgroups in terms of those who show disassociation in processes and those who do not. In neuropsychology, for example, there has been a tendency to deemphasize associations and to place greater reliance on disassociation as a source of theoretical insight (Shallice, 1988).
In sum, one could make the argument that the right questions have not been studied. For example, "Do the reading difficulties of the dyslexic stem from problems different from those characterized as poor readers without an IQ discrepancy?" is not the question to ask. Rather, the question is, "Which cognitive processes mediate the emergence of discrepancies and nondiscrepancies in low achievers?"
Although the above arguments are serious, they are not compelling. Direct answers to these questions still call into question current classification procedures for identifying children with potential LD. Attempts to validate the use of potential-achievement discrepancies in the identification of children with LD rest on some very key assumptions. Unless these assumptions are met, classification studies will continue to reflect the artifacts of "difference scores" discussed in the literature of the 1950s and 1960s. To make progress in accurately classifying children as learning disabled based on a discrepancy model, some key assumptions must be met. These assumptions are as follows:
1. There is construct integrity in the measures. The beginning step in capturing the notion of a discrepancy is to define the measures and match them to the construct definition. A test of the construct validity of discrepancy groups stands a greater chance of being assessed if the constructs included in the classification of groups are firmly grounded in theory. Most critically, "there is little reason to believe, and certainly a lot of empirical support to disbelieve, the contention that some arbitrarily weighted function of two variables will properly define a construct" (see Cronbach & Furby, 1970, p. 79). Important criteria for establishing construct validity include the demonstration of convergent and discriminant validity of the measures. Although schools commonly use the WISC-III and standardized achievement tests to determine discrepancy scores, this is not an argument for conceptual integrity. Neither theoretical rationale nor empirical evidence is available to substantiate the claim that IQ tests such as the WISC-III capture the construct of "potential." For example, it is not the case that individuals with comparable IQ scores on the WISC-III have the same potential. In addition, the difference between an intelligence score on the Wechsler and a serious performance deficit on an achievement test in the area of reading is not a valid test of the discrepancy model. In most cases, neither test fits into a theoretical framework of intelligence or reading.
2. There is independence among the measures. Discrepancy scores (or discrepancy-defined groups) are correlated with their component parts, and therefore the discrepancy measure will relate significantly to other variables correlated with the component parts (Cronbach & Furby, 1970). When difference scores are correlated with their component parts, there is a greater-than-chance tendency for them to be correlated with other variables that are associated with those component parts.
An example of the above rule is as follows: We know that reading recognition (word skill accuracy) is highly correlated with measures of phonological awareness. Therefore, when: (a) reading recognition is part of the discrepancy score, and (b) when diagnostic groups are comparable on reading recognition performance, then performance is comparable on processes (phonological awareness) related to reading. Thus, the discrepancy group is little more than a surrogate of the poor reading group. This problem of circularity in findings has been recognized in the literature for some time (Bereiter, 1963; Cronbach & Gleser, 1953). In fact, it can be demonstrated that systematic relationships between component part scores and difference scores exist even when the difference scores are generated randomly (e.g., Wall & Payne, 1973).
3. The direction of outcomes is clearly stated. The direction of the discrepancy must be consequential in performance outcomes. A critical assumption in testing the discrepancy model is that differences in the direction of the profiles are important. The fact that a student has a high reading score but low intelligence score should reflect a different "set of" or "level of" processes when compared to a student with a high IQ score but low reading score. Interestingly, such a basic assumption has not been adequately and directly tested in the literature regarding LD. Thus, a major assumption yet to be tested in variable selection for classification purposes is that the direction of the discrepancy is theoretically consequential (Cronbach & Furby, 1970; Johns, 1981). For example, even though phonological processing is comparable between average-IQ and poor readers, and between low-IQ and poor readers, it is not necessarily the case that low-IQ and high readers have better phonological processing skills (Sparks, 1995). In summary, it appears that the information that goes into a discrepancy must have face validity in terms of the direction.
4. The comparable performance between groups on various measures is not the same as the relationship among variables within groups. Few studies have been undertaken to examine the interrelationship among cognitive processing variables in predicting LD children's academic performance when compared to nondiscrepancy groups. It seems that the validity of discrepancy definitions is enhanced if it can be shown that the interrelationship among processes that contribute to achievement difficulties differ between discrepancy and nondiscrepancy groups. For example, phonological coding may contribute tremendously to word recognition in the garden-variety poor readers, whereas phonological coding, working memory, and vocabulary may contribute unique variance to the reading process in discrepancy-defined groups. In addition, some cognitive deficits may co-occur in various reading groups, but that does not mean that they all share the same causal link to the reading problems (Swanson & Alexander, 1997).
5. Measures related to the discrepancy scores are only valid if assessed on variables above and beyond their components and correlates. Most researchers recognize the reliability problems with discrepancy scores (e.g., Zimmerman & Williams, 1982), although few recognize that the use of discrepancy scores implies that they accomplish something beyond their component parts. What needs to be tested is whether the discrepancy scores are related to variables that are independent of their component parts. One obvious test is whether students defined by discrepancy scores are more likely to respond favorably to a specific treatment than poorly achieving students without discrepancies. Responsiveness to instruction seems to be a missing test of the majority of studies comparing discrepancy and nondiscrepancy groups. To date, there are no systematic analyses to support the notion that the discrepancy model is a useable construct when it comes to intervention and prognosis of intervention. A recent summary of the intervention literature (Swanson & Hoskyn, 1998; Swanson & Sache-Lee, in press) attempts to address this issue. These syntheses suggest that studies that report mean sample low-average IQs and mean sample low reading scores yield higher outcomes (as measured by the magnitude of effect size) than studies reporting high mean IQs and low mean reading scores. Thus, even though discrepancy and nondiscrepancy groups cannot be clearly separated on cognitive variables related to individual diagnosis, they may be separated in terms of treatment outcome.
The major premise here is that recent LD classification research has confirmed what has been known about the limitations of difference scores for some time and that an adequate testing of the construct has yet to occur. Research would be better served by either (a) meeting the aforementioned assumptions for testing the validity of discrepancy scores and/or (b) choosing an alternative means of assessing discrepancies (e.g., concurrent validity on multiple discrepancy measures; profile analysis; or multivariate linear modeling analogous to growth-curve analysis; Bryk & Raudenbush, 1987; Francis, Fletcher, Stuebing, Davidson, & Thompson, 1991) on those measures closely aligned with theoretical models of measuring potential and actual behavior.
NARROW FOCUS OF RECENT RESEARCH
The second major issue that is crippling the field of LD is the proliferation of recent studies focusing on isolated rather than integrative domains of cognitive processes. This is a paradoxical situation, because commentaries of the field 10 years ago suggested that we needed more specificity in identifying the deficits of children suspected of LD. In an article that focused on a meta-theory of LD about 10 years ago, Swanson (1988) suggested a framework to tie together the multitude of differences being reported between LD and non-LD children. Although most of these differences were artifacts of design (heavy reliance on an IQ- and CA-matched design), the field has been threatened by so many differences that the assumption of specificity was undermined (Stanovich, 1986). As stated by Stanovich (1988), "When researchers went looking for cognitive differences between reading disabled and nondisabled children they found them virtually everywhere" (p. 155).
Overemphasis on a Construct
It appears that there is now an overemphasis on some constructs (also see Pressley & Allington, in press, for a relevant review). An informal count of studies in the last five years of publications that focus on cognitive processes and LD indicates an emphasis on phonological processing (also see Beitchman & Young, 1997). The advantage of focusing on phonological coding is that it gives some specificity to the processing deficits of students with LD. I would not argue that a phonological deficit, such as phonological awareness, cannot act as a critical precursor to reading problems in students with learning disabilities. Rather, I would argue (as have others, Bishop, 1991) that its importance has been greatly exaggerated.
Consider a study by Bishop and Adams (1990) in which they assessed children at four years of age with normal intelligence, but who were experiencing language difficulties ranging from very severe with selective problems in expressive phonology to general impairments in language expression and comprehension. The basic question of interest to the researchers was what language measures best predicted literacy attainment. The authors' hypothesis was that young children with severe expressive phonological disorders would be particularly at risk for reading and spelling difficulties on the grounds that it is difficult for grapheme-phoneme correspondences to occur without normal development in the phonological system. A follow-up at ages 4 1/2, 5 1/2 and 8 1/2 years showed that a child's ability to describe what was happening in a picture, to convey the gist of a story, to produce complex sentences, and to understand and use grammatical inflections, was a better predictor of literacy attainment than expressive phonological competence. At 8 1/2 years, among the high-risk children with normal exposure to reading instruction, no reading difficulties emerged. In addition, some of the sample were above-average readers. The implication of their findings is that children who experience phonological impairments at either four or five years of age do not necessarily go on to have reading difficulties. In fact, Bishop and Adams' study found that some high-risk children became superior readers. These findings, according to Bishop and Adams, coincide with Bradley and Bryan's (1985) earlier study on phonological competence. Phonological competence in the preschool period was a reasonable predictor of children who would become unusually good readers, but the converse was not necessarily true. That is, poor performance on segmentation tasks at four or five years old does not necessarily predict reading problems.
Bishop (1991) considered two possible interpretations of the above findings. First, children with reading problems in the early school grades have low verbal skills (e.g., phonological awareness) as a consequence of limited reading ability. A child who reads learns new words and new information, and this is denied the LD reader who falls behind in a number of language areas. While Bishop (as well as I) agrees with this assumption, it is not necessarily the whole explanation. Bishop and Adams (1990) showed substantial correlations between syntactic and semantic skills measured in children at four years of age, before they even start reading instruction, and their reading ability at 8 1/2 years of age. Thus, some general language processes may directly influence reading acquisition.
Another interpretation is that general language problems in students with LD (e.g., semantic and syntactic deficits) reflect underlying problems in phonological processes. For example, several reviews (e.g., Shankweiler & Crain, 1986; Siegel, 1992) argue that a deficient capacity to form phonetic representations limits the development of semantic and syntactic competence. There seems to be a theoretical compromise of just calling the general language problems in children with LD in the area of reading secondary manifestations of phonological processing. Bishop (1991) humorously suggested that with sufficient creativity on the researcher's part, most tasks that involve a verbal stimulus can be a reflection of phonological processing. Even nonlinguistic deficits may be interpreted as phonological disorders. For example, Jorm, Share, MacLean, and Mathews (1986) found "finger localization to be predictive of specific reading retardation and suggested that this may be a class of phonological processing disability" (p. 52).
There have been some clear attempts to discount high-order deficits as a primary deficit for learning problems in children with LD (e.g., Shankweiler & Crain, 1986; also see Stanovich, 1990; Swanson, 1993, for a relevant review). Information-processing research (see Baddeley, 1986; Gathercole & Baddeley, 1993, for review) has been clear, however, that there are at least two components of an information-processing system that appear to make contributions to language processing, (a) phonological processes and (b) central executive processing (see Gathercole & Baddeley, 1993, for a review). The LD literature clearly shows that there has been considerably less research concerned with the involvement of a central executive processing system than with phonological codes. The lack of focus on executive processing is interesting in light of the number of treatment programs, especially strategy interventions, that make strong assumptions about the existence of a central executive processing system that accounts for a lot of the problems students with learning disabilities experience, especially adolescents (see Meltzer, 1994, for a review).
It is my opinion that a model that incorporates the resources of a limited, but flexible central executive capacity system, as well as a more specialized system such as phonological coding, is ideally suited to provide a framework for the wide range of academic deficits experienced by students with LD. Both the phonological and executive system play an intricate role in basic cognitive activities, especially in the area of language (e.g., Baddeley, 1986; Gathercole & Baddeley, 1993). As such, they might function as independent cognitive components, but at times highly interactive in such a way that the structure is much more complex than when looked at on the lower level. Thus I would like to reemphasize some assumptions about information processing that have application to research directed at understanding the processing deficits of children with LD. Four common assumptions are considered (see Kail & Bisanz, 1993, for a review).
Assumption 1. Cognitive activity can be explained in terms of processes and representations that intervene between a stimulus and response. We assume that acts of knowing are best understood in terms of mental mechanisms of representations and these mental mechanisms occur in real time and space. As it applies to LD, researchers have attempted to determine what types of information are represented, how much is coded, and how well they are organized.
Assumption 2. There are a small number of elementary processes that may account for the problems experienced by LD students in various cognitive endeavors. A number of information-processing theories are based on the belief that most acts of knowing can be decomposed into some very distinct components. It is debatable at which level the component should be analyzed. For example, in the area of word decoding we may ask if we should analyze the problem at the auditory perception level or if it should be analyzed at the phoneme level or merely at the word recognition level.
Assumption 3. Individual processes operate in concert with other processes. The study of elementary processes is only useful when we understand how elementary processes have been combined into more complex routines. Thus, based on the preceding comments regarding the overemphasis on phonological coding, a critical goal of research in LD is not only to understand the fundamental processes, but to more critically comprehend how these processes are combined and organized to produce performance of different tasks such as problem solving, mathematics, writing, and comprehension. It is assumed that certain high levels of processing may represent organizations of mental processes that are quite distinct from some of the low-level operations.
An example of this assumption can be seen in the area of reading. Recent models of reading, known as connectionist, neural net, or parallel-distributed processing models, are built on the assumption that learning progresses as the student is able to integrate several pieces of information. The key concept in these models is that reading is neither top down nor bottom up in nature; instead, all the relevant processes are simultaneous and both active and interactive (Rumelhart, 1977; Seidenberg, 1992). All components of reading accommodate information to and from each other. The importance of this model is that no one form of knowledge dominates the other; rather, it is the coordination and cooperation of the different components that best capture the learning process. A focus on elementary operations, such as phonological coding, is only useful when elementary processes are combined with other operations that form routines, which, in turn, are combined with other routines to allow for higher-order kinds of operations, such as reading comprehension. Thus, a critical goal of research in the field of LD is to understand how fundamental processes are combined and organized to produce performance on different reading tasks.
Assumption 4. Gains in our understanding of cognitive development occur when the child is involved in some sort of self-modification, direction, instruction, or monitoring. Researchers in the field of LD who are of the information-processing orientation attempt to explain how changes in the cognitive system occur by focusing on internal factors. The mechanisms that enable children to strategically monitor information-processing loads are important in various contexts. That is, no matter what the form of the learning context, the interest is in how the child ultimately codes, stores, and retrieves information. The efficiency of how this is done is understood by looking at the mechanisms of self-modification (see, e.g., Butler & Winne, 1995; Harris, 1991, for a comprehensive review).
How can the above core assumptions be helpful in providing an appropriate focus on processing difficulties in students with LD? Most researchers who investigate the cognitive workings of children with LD agree with the first two assumptions; Assumptions 3 and 4 are troublesome. In fact, some researchers argue that an inability to link isolated lower-order processes (phonological processes) directly to poor achievement, independent of the influence of a general self-monitoring system (Assumption 4), undermines the concept of specificity. This argument is problematic for three reasons.
First, specificity does not need to be isolated to a deficient cognitive process (phonological coding). Rather, a more appropriate conceptualization may be that a cognitive deficit is reflected in a specific cognitive operation, such as manipulating, representing, storing, or allocating mental resources.
Second, differences in higher-order processes, working memory, and metacognition have been found to be independent of general intelligence. That is, individuals with low and high cognitive processing skills, but with average or above-average intelligence, perform differently on reading tasks (Daneman & Carpenter, 1980). Thus, a multitude of low- and high-order processes may operate independently of intelligence (as measured on psychometric tests).
Third, the concept of specificity is not dependent on establishing a direct link between deficient cognitive mechanisms and reading (or any other academic skill). This is because, as stated earlier, reading involves the concurrent execution of several subcomponent processes in an interactive, hierarchical, and/or opportunistic fashion (Frederiksen & Warren, 1987). In other words, reading is not necessarily unidirectional in terms of beginning at the lowest level with phonological processing and proceeding upwards to higher-order processes. The attempt to reduce an LD reader's difficulties to an isolated mechanism is not sufficient if one is to have a complete picture of an LD reader's performance. Although it is essential to identify deficient elementary processes in LD readers, how these processes coalesce into general operations provides a more complete picture of their information-processing ability.
In summary, there has been an overemphasis on investigating the contribution of phonological processing to LD. Although research must continue in this area, a broader perspective is needed in terms of how lower- and higher-order processes are coordinated on tasks such as reading, mathematics, writing and so on.
The third major issue is of a practical nature: identifying effective instruction for students with LD. The practicality of this issue outweighs the aforementioned conceptual issues that have emerged in the field. Although LD as a diagnostic entity involves the largest single category of students receiving special education services (U.S. Department of Education, 1994), a simple question such as "Which intervention works best for students with learning disabilities?" or, more appropriate, "Which intervention works best for which type of LD student?" cannot be directly answered. This is because systematic knowledge about the efficacy of various educational interventions for individuals with LD is sparse (Kavale, 1990). This is unfortunate because policy issues such as "where" instruction should occur for students with LD will not be resolved without a clear understanding of instructional intervention research.
On this issue, my colleagues and I have just completed a comprehensive meta-analysis of experimental instructional (intervention) research with samples of children and adolescents with LD (Swanson, Carson, & Lee, 1996; Swanson & Hoskyn, 1998; Swanson & SacheLee, in press). Before I summarize three important findings, let me address two problems related to synthesizing intervention research. The first is that our knowledge about treatment effectiveness is biased by the publication of positive outcomes. This bias, affectionately called the "the Dodo verdict" (i.e., "everybody has won and all must have a prize"; see Weisz, Weiss, Han, Granger, & Morton, 1995), suggests that all educational programs are beneficial or equally effective. The second is that the majority of studies published on intervention for students with LD are poorly designed, have poor internal validity, fail to indicate if the LD sample is in the normal range of intelligence, and/or fail to provide usable information. Of the approximately 900 studies reviewed that actually reported data, fewer than a fourth met the criteria for inclusion in our synthesis. Thus, approximately 75% of the studies that report intervention outcomes for students with LD have flawed designs (e.g., no control condition). In addition, of those studies that were included in the synthesis, fewer than 5% met high standards in terms of the methodological criteria we have outlined. These criteria are cumulative scores related to internal validity, treatment integrity, adequacy of description of the control condition, breadth of sample description, adequacy of treatment sessions and sampling procedures, and use of reliable dependent measures. Thus, poor methodology must be taken into consideration when interpreting treatment effects. Given that we controlled for methodological artifacts between studies, what did we find?
First, the instructional areas we know most about in terms of treatment are in the domains of reading (e.g., word recognition, reading comprehension). The least researched domains, and by implication the areas we know least about, are in the areas of intelligence, creativity (e.g., divergent and convergent thinking), problem solving, mathematics, and language. Second, LD participants in these intervention studies exhibited no deficit that would suggest that one process is more resistant to treatment than another. This raises the question of "What's all this fuss about phonological awareness?" Finally, only 30% of the total number of instructional components (e.g., appropriate sequencing, reminders to use strategies, small group instruction) of the total we coded contributed significant variance to treatment outcomes. Further, none of these critical components in isolation predicted outcomes. The implication is that a small number of components operating as a interactive instructional core significantly predict outcomes.
Given the number of studies rejected for inclusion in the above synthesis because of serious problems in internal validity, more comprehensive intervention research is needed. Further, given the bias for publishing positive outcomes, it is reasonable to suggest that we increase the veracity of the studies we do have that show positive outcomes. As found in the above synthesis, few intervention studies provide measures of treatment integrity and even fewer represent followups or independent replications. Perhaps the lack of followup and replication is a policy and/or researcher consensus problem. That is, intervention researchers are leery about their work being replicated, the same way teachers do not like the idea of some body verifying their claims that reading program X worked with student A, but not with student B (see Pressley & Harris, 1994, for a review).
A common defense when intervention outcomes differ from what is found in the mainstream is to suggest that the replication was done poorly and that the particular version of the intervention created by the investigators did not represent well the intervention as studied previously. Sometimes this defense is appropriate, but at other times replication failure should be taken seriously (Pressley & Harris, 1994). Perhaps a change is needed in the consensus on the notion of independent replication. Following are some solutions to this issue (also see Pressley & Harris, 1994).
Solution 1. One way to know more about what works with students with learning disabilities would be to replicate intervention studies independently. Independence needs careful documentation, such as a research independence among the authors and the funding source. A simple policy for an editor of a research journal would be to publish only intervention studies that include an independent replication. The cost of such a policy in terms of editorship would be staggering. However, the outcome would be some sort of truth in advertising about the importance of particular interventions with students with learning disabilities.
Solution 2. Another policy is that consumers should learn to value research that has stood the test of peer review and replication. There are examples in which many research studies are cited by teachers without having stood the test of science. No doubt, part of the problem here relates to the sophistication of teachers as consumers. They may fail to adequately identify a credible study. What is a credible study? A credible empirical study, quite simply, is one that is methodologically sound. Credibility is a function of the study's internal validity characteristics (Pressley & Harris, 1994).
Solution 3. Emphasize the technology of transfer after an accumulation of replicated intervention studies. Some researchers believe that replicating complex cognitive social conditions is almost out of the question. However, it could also be argued that it might not be the rigor of intervention research, but rather the generalizability of effective procedures that is at issue. Thus, a rigorous approach to intervention research should be adopted, as well as a crafted set of procedures to enhance transfer. This technology of transfer should include an agreed-upon set of "markers" (e.g., age range, specificity of disability).
The final issue that affects the field is inflexible policies that view problems as either/or versus policies that are complementary. This is a broad statement. Therefore, the discussion will be limited to two areas: the politics of reading instruction and "full inclusion."
Whole Word vs. Phonics
There is some polarization in teaching reading, via phonics or whole word, to students with LD (see Adams & Bruck, 1993; Foorman, 1995; Lerner, Cousin, & Richeck, 1992; Pressley & Rankin, 1994, for a review of these issues). However, I assume that most teachers of LD students incorporate some aspect of word segmentation as well as the reading of authentic texts in their remediation program. Phonics instruction has a strong tradition in the remediation of nonreaders (Dearborn, 1929; also see Gray, 1940, for review of the value of phonics), and nonreaders have been found to benefit from both phonics and whole word instruction (Gray, 1940). Thus, remediation (to use terms by Whitehurst & Lonigan, 1998) reflects two domains: inside-out skills (phonological awareness, letter knowledge) and outside-in skills (e.g., language, conceptual knowledge). However, there has been some polarization related to funding (Pressley & Allington, in press). For example, the National Institute of Child Health and Human Development for the past 10 years has spent millions of dollars trying to decipher some of the features of LD and seems to have traced the aspects of the condition to a deficit in phonological awareness or the inability to code words into individually assigned units. Although a few of these studies have questionable internal validity (Troia, 1999), from these data a case is made for a revision in teaching methods nationwide whereby current "context-based" reading instruction is replaced with highly structured explicit and intensive instruction in phonological rules and other applications to print. The implication of teaching whole language without due attention to decoding places students with LD at greater risk than students in more traditional programs. Children who cannot discover the alphabetic principle independently are denied explicit instruction on the regularity conventions of letter strings.
What Does the Intervention Research Say?
We recently conducted a meta-analysis of instructional research with samples of children and adolescents with LD in the domains of real word recognition and reading comprehension (Swanson, 1999). The results indicated that a combined instructional model that includes components of both strategy and direct instruction positively influences reading comprehension performance and that direct instruction improves word recognition. More importantly, our results suggest that an emphasis on segmenting information, such as in phonics instruction, is not a sufficient condition to bolster real word recognition (also see Adams, 1990, for a review). In fact, regression modeling within our synthesis suggested that careful sequencing of materials and information across experiments partials out the effects that segmenting information (such as in a phonics program) has on word recognition. Although there has been some debate about whether the emphasis of intervention in sight word reading should be directed at the whole word level, and/or at the analytic level (isolated sounds), and/or at the synthesis level (sound units), our synthesis suggests that this debate might be better addressed, at least at the sight word level, when systematic comparisons are made between treatments that involve segmentation with an instructional core (e.g., careful sequencing, feedback, and daily review).
Much has been written on the issue of full inclusion for students with LD (e.g., Cruickshank, 1977; Klinger, Vaughn, Schumm, & Shay, 1998; Manset & Semmel, 1997; Marston, 1996; Roberts & Mather, 1995). Some of the most recent research suggests that children with LD prefer pull-out to inclusion classes (Klinger et al., 1998), that inclusive programs are less effective than pull-out programs for some LD students (Manset & Semmel, 1997), and that teacher satisfaction is greater for combined (pull-out & regular education)models than inclusion-only models (Marston, 1996). Regardless of these findings for maintaining pull-out programs, there is a trend to leave students with LD in the regular classroom. Some of the arguments justifying this trend suffer from the Lake Wobegon effect. In Garrison Keillor's fictional tales about Lake Wobegon, "all children are above average." When this idea is transferred to the field of LD, there are those who suggest that special education has great things to offer children with LD but that it has been offering them in the wrong place (see Fuchs & Fuchs, 1994, for a comprehensive review of various arguments). The full inclusion argument is that special education as we now conceptualize it in terms of a pull-out program is conceptually flawed; therefore, what is unique about instruction for students with disabilities needs to occur in the regular classroom. No doubt, there has been a trend, for financial reasons, to leave students with learning disabilities in the regular classroom (see Shanker, 1995, for discussion of this issue). Some of this policy is also fueled by the lack of conclusive research on outcomes when such children are serviced in pull-out programs (e.g., resource room) or regular classroom (Fuchs & Fuchs, 1993; Martson, 1991).
The substantive issue, however, is whether teachers in the regular classroom effectively accommodate such students (Division for LD, 1993; Gersten, Walker, & Darch, 1988; Kauffman, 1993; Kearney & Durand, 1992; NEA, 1994). NFO Research Incorporated of Greenwich, Connecticut, upon instigation of the National Center for Learning Disabilities (NCLD), conducted a survey among parents and teachers who were NCLD members (no doubt a biased selection), as well as members of the Learning Disabilities Association of America. Approximately 1,500 individuals of an original sampling of 5,000 responded. No doubt, the survey is tainted with some artifacts of methodology. However, the survey (reported in Orton Dyslexia Society newsletter, 1995) showed that parents of children with LD had little confidence in the training of regular classroom teachers for teaching such children. Of the teachers sampled, half agreed with the parents' sentiment. Only 53% of the regular classroom teachers felt that they were adequately trained to teach youngsters with LD compared to their colleagues in special education. By comparison, 98% of the parents of children with LD were convinced that general educators were not trained to teach their children.
In addition to these issues, it is evident that regular classroom teachers are responsible for growing numbers of diverse learners (see Carnine, 1994, Table 1). In reference to my own state of California, this increase results largely from the growth of the number of children living in poverty and those who come from nonEnglish-language homes. The full inclusion movement places increasing numbers of special students in general education classrooms and also puts more demands on regular classroom teachers. Diverse learners, such as students with LD and children from limited English-speaking backgrounds, are reported to demonstrate a variety of problems in the areas of memory, language, strategies, vocabulary, and more. Yet, a number of individuals are demanding, or at least expecting, that these students perform at some particular standard. Although no regular classroom teacher would deny that there is tremendous diversity with respect to the type of students within their classroom, especially in the area of achievement, within the field of LD we may be negatively amplifying the impact of such diversity by placing unrealistic demands on the regular classroom teacher.
In our earlier analysis of the intervention literature (Swanson & Hoskyn, 1998), we narrowly defined full inclusion in the context of our syntheses as students with LD who receive full-time education (with or without treatment) in a general (regular) education classroom. Our results were that larger effect sizes occurred in resource rooms than in other settings. However, only four studies (out of 180), with a total sample size of 151 participants, provided information on effect sizes related to the regular classroom. Of these four studies, two were confounded by using different schools to represent treatment and control conditions. In summary, our results showed that interventions that took place in the pull-out classroom yielded larger effect sizes than other settings.
This conclusion calls for a qualification about how many intervention studies actually reported where treatment occurred. The more substantive issue, however, is whether teachers in the regular classroom can effectively accommodate such children. We have no data on this issue in our synthesis. There was not enough information in the articles to code whether treatment was better carried out by the regular classroom teacher than the special education teacher. In addition, generalizations on treatment outcomes as a function of setting in our synthesis data are constrained by the lack of direct contrasts between special vs. regular education settings within a study. Further, the research question "Which setting parameters of an instructional environment best maximize potential and productivity (however defined) in a general setting?" is not directly addressed in our synthesis.
A conservative interpretation of our intervention studies supports a full continuum of services (high effect sizes were found in both placements). This is a generous interpretation on two accounts: (a) The trend in the data supports pull-out programs, and (b) few studies on intervention occur in the regular classroom setting. Unless it can be shown that all children with LD do not benefit from pull-out programs (our synthesis suggests they do) compared to integrated classrooms, special education placements should be left in place.
Possible solutions to the present problem are difficult to find (Kauffman, 1994), because research in special education seldom informs policy (Keogh, 1994). In contrast to the other issues I have discussed, which are embedded in research, the issues related to full inclusion are politically, philosophically, and/or "affectively" driven. It is clear that many LD students can be educated in full integration classrooms, while others cannot. Unfortunately, some people jump on the "inclusion bandwagon" with no regard for research data or the misinterpretation of such data. More troublesome is that some states are moving toward full inclusion models despite the lack of well-designed supportive research and sufficient information on individual differences in response to instruction (severe vs. moderate LD). Some representatives for the policy have made full inclusion a civil rights issue. No doubt, separation or segregation in a democratic society is repugnant (un-American), whereas integration is in line with our democratic principles; however, the subtle concepts underlying the full inclusion movement as discussed in educational circles have little application to civil rights issues (denying certain people access who have equitable skills versus teaching skills to people so access is more equitable).
Unless it can be shown that all children with LD do not benefit from pull-out programs when compared to an integrated classroom, the full continuum of services should be left in place (also see Council for Learning Disabilities, 1993, position statement). Therefore, the main focus of this discussion is on minimizing errors. That is, the major premise guiding policy for children with LD should be concerned with minimizing errors in terms of instruction. A few dependent measures of error can be inferred from various achievement and social measures of students functioning in pull-out and regular classroom settings. Unfortunately, it is sometimes difficult to discern whether the primary source of errors (e.g., weak gains in achievement) in these dependent measures originates in the environment, which may offer the wrong educational stimulant to the child (whole language instruction), and/or in the mind (cognition) of the policy makers (e.g., teachers, principals, school district officials, etc.), or both. In an attempt to engage in complementary reasoning (neither committing to full inclusion nor to pull-out programs), I suggest five ways of handling errors. While some of these are tongue-in-cheek, they do capture the sentiments of some advocates on either side of the fence.
1. Develop a foolproof system for assessment and intervention in the early grades. That is, teachers and administrators anticipate by kindergarten, first, or second grade that a certain number of children will manifest LD; therefore, provisions are made for a sufficient number of options so that the errors will not result in a poor fit between environment and child needs. This fool-proof approach has various drawbacks, however. First, there is no limit to what a fool will do when put in charge of this system (refer to the first news story earlier in this article). Second, every educational system and option can be switched off and on with various changes in philosophy and attitudes.
2. Select those teachers and school psychologists who make few mistakes. The difficulty here is that we have no studies on what constitutes teaching expertise (i.e., expert knowledge) in the field of LD or how such expertise differs from the expertise of regular classroom teachers in terms of effective instruction of LD students. Further, there are no studies, to my knowledge, identifying school psychologists who use selection procedures of identifying children with LD that can be judged as possessing a high degree of construct validity.
3. Intensify training programs so teachers will make few mistakes. Unfortunately, professional training at a university or public school level does not provide sufficient protection against slips, lapses, and mistakes. Perhaps asking educators to monitor their own behavior may reduce the number of poor decisions, but most errors, I suspect, are made without intention. Further, even when teachers are doing their best, errors emerge.
4. Influence regular teachers by giving them incentives for wanted and unwanted behavior in dealing with individual differences (or diversity) of children within their classrooms. Poor decisions related to students with LD are usually not violations of contingency rules. Moreover, the resulting performance of a child who has not adequately learned new information is more punishment from a teacher's perspective than can ever be provided by an incentive system.
5. Change the environment in which policy is developed. Elicit the desired behavior from those who are making policy by providing the conditions for effective communication. If communication fails, use governmental procedures to remove such groups of individuals from office (analogous to voting all the Democrats out of the House of Representatives).
Perhaps none of these suggestions in isolation, even if elaborated upon, can reduce errors in poor policy decisions related to the educational environment for students with LD. However, we do need to agree on our goals for educating students with LD. The aim of education is to make good and productive citizens of all children in a general or "shared" society. The specific aim of meeting the needs of LD children is to prepare them for citizenship in general society. This is done by providing an effective array of options. Sometimes, in the debate on full inclusion we find that our aim is not well focused.
In summary, four vexing issues in the field of LD have been presented. First, there are conceptual problems related to the validity of discrepancy-defined groups. Recent research has confirmed what has been known about the limitations of difference (discrepancy) scores for some time and that an adequate testing of the construct of discrepancy has yet to be found. Second, there has been an overemphasis on investigating the contribution of phonological processing to LD. Although research must continue in this area, a broader perspective is needed in terms of how lower-and higher-ordered processes are coordinated on tasks such as reading, mathematics, writing and so on. Third, several striking methodological problems emerge in the majority of intervention studies, which limits answering several practical questions. Finally, sometimes policy is implemented without a clear focus and an adequate research base. No doubt biased, the interpretations of these issues do reflect some suggestions for remediation.
Aaron, P. G. (1997). The impending demise of the discrepancy formula. Review of Education Research, 67(4), 461-502.
Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press.
Adams, M. J., & Bruck, M. (1993). Word recognition: The interface of educational policies and scientific research. Reading and Writing: An Interdisciplinary Journal, 5, 113-139.
Baddeley, A. D. (1986). Working memory. London: Oxford University Press.
Beitchman, J. H., & Young, A. R. (1997). Learning disorders with a special emphasis on reading disorders: A review of the past 10 years. Journal of the American Academy of Child & Adolescent Psychiatry, 36(8), 1020-1032.
Bereiter, C. (1963). Some persisting dilemmas in the measurement of change. In C. W. Harris (Ed.), Problems in measuring change (pp. 3-20). Madison: University of Wisconsin Press.
Bishop, D.V.M. (1991). Developmental reading disabilities: The role of phonological processing has been overestimated. Mind & Language, 6, 97-101.
Bishop, D.V.M. & Adams, C. (1990). A prospective study of the relationship between specific language impairment, phonological disorder, and reading retardation. Journal of Child Psychology and Psychiatry, 31, 1027-1050.
Bradley, L., & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. Ann Arbor: University of Michigan Press.
Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to assessing change. Psychological Bulletin, 101, 147-158.
Butler, D. L., & Winne, P. H. (1995). Feedback & self-regulated learning: A theoretical synthesis. Review of Educational Research, 65, 245-281.
Carnine, D. (1994). Diverse learners and prevailing, emerging and research based educational approaches and their tools. School Psychology Review, 23, 341-350.
Council for Learning Disabilities. (1993). Position statement and concerns about the "Full Inclusion" of students with learning disabilities in regular education classrooms. Learning Disability Quarterly, 16, 126.
Cronbach, L., & Furby, L. (1970). How we should measure "change" -- or should we? Psychological Bulletin, 74, 68-80.
Cronbach, L. J., & Gleser, G. C. (1953). Assessing similarity between profiles. Psychological Bulletin, 74, 68-80.
Cruickshank, W. M. (1977). Least restrictive placement: Administrative wishful thinking. Journal of Learning Disabilities, 10, 193-194.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450-466.
Dearborn, W. F. (1929). Teaching the nonreader. Elementary School Journal, 30, 266-270.
Dillon, S. (1994, April 7). A class apart: Special education soaks up New York's school resources. New York Times.
Division for Learning Disabilities of the Council for Exceptional Children. (1993). Inclusion: What does it mean for students with learning disabilities? The DLD Times, 10(3), 4i.
Fergusson, D. M., Horwood, L.J., Caspi, A., Moffitt, T.E., & Silva, P.A. (1996). The (artefactual) remission of reading disability: Psychometric lessons in the study of stability and change in behavioral development. Developmental Psychology, 32, 132-140.
Fletcher, J., Shaywitz, S. E., Shankweiler, D. P., Katz, L., Liberman, I., Stuebing, K., Francis, D., Fowler, A., & Shaywitz, B. A. (1994). Cognitive profiles of reading disability: Comparisons of discrepancy and low achievement definitions. Journal of Educational Psychology, 86, 6-23.
Foorman, B. (1995). Research on "The great debate" code-oriented versus whole language approaches to reading. School Psychology Review, 24, 376-392.
Francis, D. J., Fletcher, J. M., Stuebing, K. K., Davidson, K. C., & Thompson, N. M. (1991). Analysis of change: Modeling individual growth. Journal of Consulting and Clinical Psychology 59, 27-37.
Frankenberger, W., & Fronzaglio, K. (1991). A review of states' criteria and procedures for identifying children with learning disabilities. Journal of Learning Disabilities, 24, 495-500.
Frederiksen, J. R., & Warren, B. M. (1987). A cognitive framework for developing expertise in reading. In R. Glaser (Ed.), Advances in instructional psychology Vol. 3 (pp. 1-39). Hillsdale, NJ: Erlbaum.
Fuchs, D., & Fuchs, L. S. (1994). Inclusive schools movement and the radicalization of special education reform. Exceptional Children, 60, 294-309.
Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and language. Hillsdale, NJ: Erlbaum.
Gerber, P. I., & Reiff, H. B. (1991). Speaking for themselves: Ethnographic interviews with adults with learning disabilities. Ann Arbor: University of Michigan Press.
Gersten, R., Walker, H., & Darch, C. (1988). Relationship between teachers' effectiveness and their tolerance for handicapped students. Exceptional Children, 54, 433-438.
Gray, W. S. (1940). Reading. Review of Educational Research, 10, 79-106.
Harris, K. R. (1991). Developing self-regulated learners: The role of private speech and self-instructions. Educational Psychologist, 25, 35-49.
Johns, G. (1981). Difference score measures of organizational behavior variables: A critique. Organizational Behavior and Human Performance, 27, 44-463.
Jorm, A., Share, D., Maclean, R., & Matthews, R. (1986). Cognitive factors at school entry predictive of specific reading retardation and general reading awkwardness: A research note. Journal of Child Psychology and Psychiatry, 27, 45-54.
Kail, R., & Bisanz, J. (1993). The information processing perspective on cognitive development in childhood and adolescence. In R. Sternberg & C. A. Berg (Eds.), Intellectual development, Vol 1 (pp. 229-260). New York: Cambridge.
Kauffman, J. M. (1993). How we might achieve the radical reform of special education. Exceptional Children, 60, 6-16.
Kavale, K. A. (1990). Variances and verities in learning disability interventions. In T. E. Scruggs & B.Y.L. Wong (Eds.), Intervention research in learning disabilities (pp. 3-33). New York: Springer-Verlag.
Kearney, C. A., & Durand, V. M. (1992). How prepared are our teachers for mainstreamed classroom settings? A survey of post-secondary schools of education in New York State. Exceptional Children, 59, 6-11.
Keogh, B. (1988). Improving services for problem learners: Rethinking and restructuring. Journal of Learning Disabilities, 21(1), 19-22.
Klinger, J., Vaughn S., Schumm, J., & Shay, C. (1998). Inclusion or pull-out: Which do student prefer? Journal of Learning Disabilities, 31 (2), 148-158.
Lerner, J. W., Cousin, P. T., & Richeck, M. (1992). Critical issues in learning disabilities: Whole language learning. Learning Disabilities Research & Practice, 7, 226-230.
Manset, G., & Semmel, M. (1997). Are inclusive programs for students with mild disabilities effective? A comparative review of model programs. Journal of Special Education, 31, 155-180.
Marston, D. (1996). A comparison of inclusion only, pull-out only, and combined service models. Journal of Special Education, 30(2),121-132.
Meltzer, L. (Ed.). (1994). Strategy assessment and instruction for students with learning disabilities: From theory to practice. Austin, TX: PRO-ED.
Mercer, C., Jordan, L., Allsopp, D. H., & Mercer, A. R. (1996). Learning disability definitions and criteria used by state education departments. Learning Disability Quarterly, 19(4), 217-232.
Monroe, M. (1932). Children who cannot read. Chicago: University of Chicago Press.
National Education Association. (1994). NEA policy on appropriate inclusion. Washington, DC: Author.
Orton Dyslexia Society. (1994, Fall). ODS position statement on inclusion [Special issue]. Perspectives on Inclusion, 20(4), 2-3.
Pressley, M., & Allington, R. (in press) What should reading instructional research be the research of? Issues in Education.
Pressley, M., & Harris, K. R. (1994). Increasing the quality of educational intervention research. Educational Psychology Review, 6, 191-208.
Pressley, M., & Rankin, J. (1994). More about whole language methods of reading instruction for students at risk for early reading failure. Learning Disabilities Research & Practice, 9, 157-168.
Reynolds, C. R. (1984-1985). Critical measurement issues in learning disabilities. Journal of Special Education, 18, 453-476.
Roberts, R., & Mather, N. (1995). The return of students with learning disabilities to regular classrooms: A sellout? Learning Disabilities Research & Practice, 10, 46-58.
Rousch, W. (1995). Arguing over why Johnny can't read. Science, 267, 1896-1898.
Rumelhart, D. E. (1977). Toward an interactive model of reading. In S. Dornic (Ed.), Attention and performance, Vol. 6 (pp. 573603). Hillsdale, NJ: Erlbaum.
Rutter. M., & Yule, W. (1975). The concept of specific reading retardation. Journal of Child Psychology and Psychiatry, 16, 181-197.
Seidenberg, M. S. (1992). Dyslexia in a computational model of word recognition in reading. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp 243-274). Hillsdale, NJ: Erlbaum.
Shallice, T. (1988). From neuropsychology to mental structures. Cambridge: Cambridge University Press.
Shanker, A. (1995, September 19). Where we stand: A rush to inclusion. New York Times, The Week in Review, p. 9.
Shankweiler, D., & Crain, S. (1986). Language mechanisms and reading disorder: A modular approach. Cognition, 24, 139-168.
Shaywitz, S. E., Escobar, M. D., Shaywitz, B. A., Fletcher, J. M., & Makugh, R. (1992). Evidence that dyslexia may represent the lower tail of a normal distribution of reading ability. The New England Journal of Medicine, 326, 145-150.
Shaywitz, B. A., Fletcher, J. M., Holahan, J. M., & Shaywitz, S. E. (1992). Discrepancy compared to low achievement definitions of reading disability: Results from the Connecticut Longitudinal Study. Journal of Learning Disabilities, 25, 639-648.
Shepard, L. A. (1980). An evaluation of the regression discrepancy method for identifying children with learning disabilities. Journal of Special Education, 14, 79-91.
Siegel, L. S. (1989). IQ is irrelevant to the definition of learning disabilities. Journal of Learning Disabilities, 22, 469-479.
Siegel, L. S. (1992). An evaluation of the discrepancy definition of dyslexia. Journal of Learning Disabilities, 25, 618-629.
Sovik, N., Frostad, P., & Lie, A. (1994) Can discrepancy between IQ and basic skills be explained by learning strategies? British Journal of Educational Psychology, 64, 389-405.
Sparks, R. L. (1995). Phonological awareness in hyperlexic children. Reading and Writing, 7, 217-235.
Stanovich, K. (1986). Matthew effect in reading: Some consequences of individual differences in the acquisition of literacy. Reading Research Quarterly, 21, 360-407.
Stanovich, K. E. (1988). The right and wrong places to look for the cognitive locus of reading disability. Annals of Dyslexia, 38, 154-177.
Stanovich, K. E. (1990). Concepts in developmental theories of reading skill: Cognitive resources, automaticity, and modularity. Developmental Review, 10, 72-100.
Stanovich, K. E. (1991). Discrepancy definitions of reading disability: Has intelligence led us astray? Reading Research Quarterly, 26, 7-29.
Stanovich, K., & Siegel, L.S. (1994). Phenotypic performance profile of children with reading disabilities: A regression-based test of the phonological-core variable-difference model. Journal of Educational Psychology, 86, 24-53.
Stanovich, K. E., & Stanovich, P. (1997). Further thoughts on aptitude/achievement discrepancy. Educational Psychology in Practice, 13(1), 3-8.
Swanson, H. L. (1988). Toward a metatheory of learning disabilities. Journal of Learning Disabilities, 21, 196-209.
Swanson, H. L. (in press). A comprehensive synthesis of reading intervention research for students with learning disabilities. Journal of Learning Disabilities.
Swanson, H. L., & Christie, L. (1994). Implicit notions about learning disabilities: Some directions for definitions. Learning Disabilities Research & Practice, 9, 244-254.
Swanson, H. L., & Hoskyn, M. (1998). A comprehensive synthesis of experimental intervention for students with learning disabilities. Review of Educational Research, 68, 276-321.
Swanson, H. L., & Sache-Lee (in press). A synthesis of single-subject intervention research for students with learning disabilities. Journal of Learning Disabilities.
Troia, G. A. (in press). Phonological awareness intervention research: A critical review of the experimental methodology. Reading Research Quarterly.
Wall, T. D., & Payne, R. (1973). Are deficiency scores deficient? Journal of Applied Psychology, 58, 322-326.
Whitehurst, G. J., & Lonigan, C. J. (1998). Child development and emergent literacy. Child Development, 69(3), 848-872.
Yule, W., Rutter, M., Berger, M., & Thompson, J. (1974). Over-and underachievement in reading distribution in the general population. British Journal of Educational Psychology, 44, 1-12.
Zimmerman, D. W., & Williams, R. H. (1982). Gain scores in research can be highly reliable. Journal of Educational Measurement, 19, 149-154.
Requests for reprints should be addressed to: H. Lee Swanson, School of Education, University of California, Riverside, CA 82521.
H. LEE SWANSON, PH.D., is professor of Educational Psychology and Special Education, University of California-Riverside.
|Printer friendly Cite/link Email Feedback|
|Author:||Swanson, H. Lee|
|Publication:||Learning Disability Quarterly|
|Date:||Jan 1, 2000|
|Previous Article:||PREVALENCE AND EXPRESSION OF DEPRESSIVE SYMPTOMATOLOGY IN STUDENTS WITH AND WITHOUT LEARNING DISABILITIES.|
|Next Article:||LINKS AMONG SOCIAL STATUS, SERVICE DELIVERY MODE, AND SERVICE DELIVERY PREFERENCE IN LD, LOW-ACHIEVING, AND NORMALLY ACHIEVING ELEMENTARY-AGED...|