Recent developments in reading intervention research: Introduction to the special issue
- First Online:
- Cite this article as:
- Tunmer, W.E. Read Writ (2008) 21: 299. doi:10.1007/s11145-007-9108-4
- 359 Views
Reading intervention refers to instructional approaches and programs designed to either prevent or remediate persistent reading difficulties. Prevention programs typically focus on at risk children with limited amounts of crucial reading-related knowledge, skills and experiences at school entry who often, but not always, come from low-income family backgrounds. Remedial programs target students who are failing to make adequate progress in learning to read.
Three general subtypes of reading difficulties
Reading is the process of extracting and constructing meaning from text. Given that the child’s fundamental task in learning to read is to discover how print maps onto their existing spoken language, the process of learning to derive meaning from print can be adversely affected in one of two ways, or both: the child’s spoken language system may be deficient in various ways, or the process by which print is connected to the child’s spoken language system may be defective. These basic ideas are represented in a model of the proximal causes of reading difficulties that makes two claims; first, that reading may be decomposed into two components, decoding (or more generally, word recognition) and oral language comprehension; and second, that each of these components is necessary, neither being sufficient in itself (Gough & Tunmer, 1986). That is, R = D × C, where R is reading comprehension, D is decoding skill, and C is oral language comprehension. Stated simply, students who have trouble recognizing the words of (age appropriate) text and/or have trouble understanding the language being read, will have trouble understanding the text.
An important feature of the model is that it provides a framework for conceptualizing three broad categories of reading difficulties, each of which appears to require a different intervention strategy. The model predicts that reading comprehension problems can result from weaknesses in recognizing printed words, weaknesses in comprehending spoken language, or both. Students who can understand a text when it is read aloud to them but cannot decode the words even after receiving extensive (evidenced-based) instruction are referred to as dyslexics; students who can read words accurately but have difficulty constructing the meaning of text are described as having specific reading comprehension difficulties (Nation, 2005); and students who have problems in both word recognition and oral language comprehension are described as having a mixed reading disability (Catts & Kamhi, 2005).
The language-based deficits that contribute to developmental reading problems vary across the three subtypes of reading difficulties. Dyslexia is generally associated with problems in the phonological domain, especially deficiencies in phonemic awareness and phonological decoding, both of which are crucial for developing word reading skills (Vellutino & Fletcher, 2005). As predicted by the model, the development of adequate facility in word identification is a necessary (though not sufficient) condition for the development of reading comprehension ability. Growth in the ability to construct meaning from text will therefore be impeded if children fail to develop the phonemically based skills necessary for constructing the detailed orthographic representations required for the automatization of word recognition, or what Ehri (2005) calls sight word knowledge. Systematic reading interventions involving dyslexic children have therefore targeted phonological awareness and decoding skills (Torgesen, 2004, 2005).
The majority of poor readers, however, have mixed reading disability. These children, who are also called “garden variety” poor readers (Gough & Tunmer, 1986), have more widespread language impairments than are typically found among children with dyslexia (Catts & Kamhi, 2005; Tunmer & Chapman, 2007). In addition to phonological processing deficits, children with mixed reading disability have impairments in vocabulary, morphology, syntax, and/or discourse-level processing. The resulting weakness in oral language comprehension places an upper limit on reading comprehension (Hoover & Tunmer, 1993), which would account for research showing that in addition to phonological factors (e.g., phonological awareness), nonphonological oral language factors (e.g., expressive vocabulary, sentence or story recall) are predictive of long-term reading outcomes (Leach, Scarborough, & Rescorla, 2003; Scarborough, 2005). It would also explain why prevention programs for at risk students with mixed deficits focusing mostly on teaching phonemic awareness and phonemically based decoding strategies initially show positive effects on reading achievement (typically word reading) but fail to maintain these positive effects in later grades when reading comprehension measures are used. Torgesen (2004) suggested that this pattern of results occurs “for the simple reason that as reading material becomes more complex (with increasing vocabulary demands and more difficult concepts), the role of broad verbal ability in accounting for reading comprehension difficulties becomes larger” (p. 368). The pattern of results observed in prevention studies is even more pronounced in remedial intervention studies involving older poor readers. From an examination of 14 studies providing interventions to children with moderate to severe word-level reading difficulties, Torgesen (2005) found that reported gains for phonological decoding skills were consistently larger than those for reading comprehension skills (see his Table 27.2, p. 530).
Two other essential reading skills that are difficult to remediate in older poor readers are vocabulary and fluency. Foorman, Seals, Anthony, and Pollard-Durodola (2003) were concerned that the low vocabulary sizes of at risk children in third and fourth grade would retard their reading comprehension as they began to confront more semantically complex reading materials across the curriculum in later grades. They therefore investigated the effects of a 20-week vocabulary enrichment program for at risk third and fourth graders and found that although the program significantly improved knowledge of word meanings, the positive outcomes did not transfer to reading comprehension, a result that confirmed the findings of others.
Researchers have also found that although interventions for children with moderate or severe impairments in word reading ability can succeed in closing the gap in reading accuracy and, to some extent, reading comprehension, they generally fail to close the gap in reading fluency, even when the intervention allocated considerable instructional time to reading connected text or to modelling and practicing fluent reading (Torgesen, 2004, 2005; Torgesen, Rashotte, & Alexander, 2001). Torgesen (2005) compared outcomes for reading accuracy and reading fluency in remedial and preventive studies and found consistently large differences in outcomes favouring accuracy gains over fluency gains in remedial studies but little or no gap between accuracy and fluency gains in prevention studies. Torgesen suggested that the intractability of closing the gap in fluency in older poor readers is most likely attributable to the problems these children face in making up for the huge deficits in reading practice they have accumulated as a consequence of failing to learn to read during the first 3 or 4 years of school, an example of negative (poor-get-poorer) Matthew effects in reading (Juel, 1988; Stanovich, 1986; Tunmer, Chapman, & Prochnow, 2003). Such lack of practice and restricted exposure to print would severely impair the development of sight word knowledge in these children (Ehri & McCormick, 1998).
Supporting these claims are findings reported in the first paper in this special issue by Wexler, Vaughn, Edmonds, and Reutebuch. Because students who fail to develop fluency in recognizing the words of text will encounter difficulty in comprehending and thinking about what they are reading, and because fluency problems have been shown to be strongly and consistently associated with reading difficulties in older students, Wexler et al. carried out a synthesis of research on the efficacy of fluency interventions on improving the fluency and comprehension of struggling readers in grades 6 through 12. An extensive search of the literature between 1980 and 2005 yielded a total of only 19 intervention studies that satisfied specific criteria for inclusion in the synthesis, suggesting that research on fluency interventions for secondary students is a largely neglected area, despite its apparent importance. Of the 19 studies included in the synthesis, only six involved the use of treatment/comparison designs, and of these, only four used random assignment of students to treatment and comparison groups. The remaining studies used either single group or single subject designs.
Despite these limitations, two major trends emerged from the synthesis. First, although repeated reading interventions, the most widely used fluency intervention approach, seemed to improve reading rate on practiced passages, the gains in fluency generally had no direct effect on reading comprehension ability. Second, an important preliminary finding was that repeated reading interventions may be no more effective with older struggling readers in increasing reading speed, word recognition accuracy, and reading comprehension than the same amount of non-repetitive reading. The latter finding is consistent with Torgesen’s (2005) suggestion that extensive reading practice is the key to closing the gap in fluency in older poor readers.
Although both dyslexic poor readers and children with mixed reading disability have weaknesses in the phonological domain, the more widespread oral language impairments of the mixed disabled readers further impede the development of their phonemic awareness and phonological decoding skills in at least three ways. First, vocabulary growth during the preschool years plays a major role in the development of preliterate phonological sensitivity by causing lexical representations to become more segmental (Carroll, Snowling, Hulme, & Stevenson, 2003). Because deficiencies in vocabulary growth are accompanied by more poorly specified phonological representations of spoken words, the development of phonemic awareness is likely to be more severely impaired in children with poorly developed vocabulary knowledge at school entry. Second, children with poorly developed vocabulary knowledge will have trouble identifying and assigning appropriate meanings to unknown printed words, especially partially decoded or irregularly spelled words, if the corresponding spoken words are not in their listening vocabulary. Third, deficiencies in syntactic knowledge (i.e., implicit knowledge of rules specifying structural relationships in sentences) will impair the development of word identification skills in children with mixed reading disability by limiting their ability to use sentence context as an aid to identifying partially decoded words, irregularly spelled words, or words containing polyphonic spelling patterns (such as ear as in bear and hear, own as in clown and flown). The use of sentence context to confirm hypotheses about what unfamiliar words might be, based on incomplete information from partial decoding attempts, results in correct word identifications which, in turn, facilitates the development of beginning readers’ word-specific orthographic knowledge from which additional spelling-sound correspondences can be induced (Tunmer & Chapman, 1998, 2006).
Given these considerations, a prediction that follows from the classification scheme derived from the Gough and Tunmer (1986) model of reading difficulties is that mixed disabled readers should show greater phonological processing deficits than dyslexic poor readers at the beginning of school and during the early stages of learning to read. In support of this claim, Tunmer and Chapman (2007) found in a longitudinal study of language-related differences between mixed disabled readers and dyslexic poor readers that in addition to expected differences on oral language measures, the mixed disabled readers also showed consistently greater phonological processing deficits than the dyslexic poor readers across a range of phonological processing measures (see Snowling, Gallagher, & Frith, 2003, for similar findings). These findings and those discussed earlier (Leach et al., 2003; Scarborough, 2005; Torgesen, 2005) suggest that for mixed disabled readers, which constitute by far the largest group of poor readers, early intervention programs need to focus on improving these children’s oral language skills as well as their phonological skills (Gersten & Dimino, 2006).
In addition to dyslexic poor readers and mixed disabled readers, the third broad category of disabled readers specified by the Gough and Tunmer (1986) model comprises children with specific reading comprehension difficulties. These poor readers are generally free of phonological processing deficiencies and demonstrate satisfactory alphabetic coding skills, but (like the poor readers with mixed reading disability) show weaknesses in vocabulary, morphology, syntax, discourse-level processing and/or comprehension strategies which, in turn, negatively impact reading comprehension performance (see Nation, 2005, for a review of research). However, the developmental mechanism responsible for the oral language impairments in children with specific reading comprehension difficulties may differ from what causes such impairments in children with a mixed reading disability. For children with mixed reading disability, oral language impairments stem primarily from limited access to linguistic and environmental opportunities during the preschool years, whereas for children with specific reading comprehension difficulties, such impairments appear to be largely a consequence of having substantially less reading and reading-related experience than typically developing readers, which ultimately produces negative Matthew effects in reading (Nation, 2005). As Nation argued, “Poor comprehenders may read less, and learn less from their reading experiences than their peers; therefore impacting on subsequent reading and learning opportunities over time and leading to the formation of weak ‘intellectual habits’...” (p. 264). Intervention programs for children with specific reading comprehension difficulties may therefore need to focus particular attention on motivating these children to increase both the amount and range of their personal reading.
Explicitness and balance in reading intervention programs
An important feature of effective reading interventions is the degree of explicitness with which specific reading-related knowledge, skills and strategies (phonological awareness, alphabetic coding, vocabulary, fluency, comprehension strategies) are taught. For word identification skills, research indicates that compared to normally achieving children, struggling readers are less able to discover letter-sound patterns as a by-product of more general reading, suggesting that these children require more explicit instruction in alphabetic coding (Calfee & Drum, 1986; Torgesen, 2004). In predominantly constructivist approaches to reading intervention like Reading Recovery (Clay, 1993), word analysis activities arise primarily as “mini-lessons” given in response to children’s oral reading errors during text reading (Iversen, Tunmer, & Chapman, 2005; Tunmer & Chapman, 2003, 2004). Although this intervention strategy may be suitable for children who are mildly at risk, children with more severe reading difficulties or higher degrees of risk appear to require a more highly structured, systematic approach that includes teaching word analysis skills outside the context of reading connected text (Foorman, Francis, Fletcher, Schatschneider, & Mehta, 1998; Torgesen et al., 1999; Torgesen et al., 2001b; Tunmer et al., 2003).
In support of this claim are the results of a study by Foorman et al. (1998) of the effects of different methods of beginning reading instruction on the reading growth of at-risk first grade students. They found that the degree of explicitness of instruction in the alphabetic code and related skills was positively associated with amount of improvement in reading, and that more explicit instruction in alphabetic coding resulted in less disparity between at-risk students in reading achievement at the end of the year than less explicit approaches to teaching spelling-sound patterns.
Relatedly, Morris, Tyner, and Perney (2000) examined the effectiveness of Early Steps, a first-grade early intervention program that is very similar to Reading Recovery, especially in the emphasis it places on contextual reading and writing. However, a fundamental difference is that Early Steps also includes direct, systematic study of orthographic patterns that is “purposefully isolated from meaningful context so that the child can pay full attention to the patterns being studied” (p. 682). Providing struggling readers with explicit and systematic instruction in word analysis skills outside the context of reading text helps to ensure that these children see the importance of focusing on word-level cues as the most useful source of information in identifying words, as well as helping them to overcome their tendency to rely on sentence context cues to identify unfamiliar words in text rather than using context to supplement word-level information. One of the major distinguishing characteristics of struggling readers is their tendency to rely heavily on sentence context cues to compensate for their deficient alphabetic coding skills (Pressley, 2006). Morris et al. (2000) found that Early Steps was highly effective, especially for those children who were most at risk. Similarly, Hatcher, Hulme, and Ellis (1994) reported that adding systematic training in phonological awareness combined with instruction in the alphabetic principle to a remedial reading program modelled after Reading Recovery was highly effective and produced greater gains in poor readers than either the remedial reading program or phonological awareness training on its own.
Further support for these claims comes from the second paper in the special issue by Ryder, Tunmer, and Greaney. The aim of their study was to determine whether explicit instruction in phonemic awareness and phonemically based decoding skills would be an effective intervention strategy for children with early reading difficulties in a whole language instructional environment where word analysis activities, if any, arise primarily from the child’s responses during text reading and focus mainly on boundary letters (i.e., initial and final letters). Ryder et al. maintain that children experiencing ongoing difficulties in detecting phonemic sequences in words will not be able to fully grasp the alphabetic principle and develop the ability to intuitively perceive redundant patterns between speech and print. A prediction that follows from these considerations is that explicit instruction in phonemic awareness and phonemically based decoding skills should be a more effective intervention strategy for children with early reading difficulties than standard (New Zealand) whole language classroom literacy programs supplemented with remedial reading instruction focusing largely on teaching the use of sentence context cues to identify unfamiliar words in text. Twenty-four 6- and 7-year-old struggling readers were randomly assigned to an intervention or control group. The children in the intervention group were divided into four groups of three students each and received small group remedial reading instruction over a period of 24 weeks that comprised 56 highly sequenced, semi-scripted lessons in phonemic awareness and alphabetic coding skills and book reading activities involving decodable text. In support of the prediction, posttest results showed that the intervention group significantly outperformed the control group on measures of phonemic awareness, pseudoword decoding, context free word recognition, and reading comprehension. Two-year follow-up data indicated that the positive effects of the intervention program were not only maintained but had generalized to word recognition accuracy in connected text.
Another important issue relating to reading intervention programs is the balance that needs to be achieved in the instructional components of each lesson, especially between learning new skills and actually using them. Focusing on explicit and systematic teaching of word analysis skills to struggling readers does not mean adopting a rigid skill-and-drill approach in which word identification skills are taught largely in isolation with little or no connection to actual reading. Although struggling readers should receive explicit instruction in letter-sound patterns outside the context of reading connected text, they should also be taught how and when to use this information during text reading through demonstration, modelling, direct explanation, and guided practice. It cannot be assumed that struggling readers who are successful in acquiring word analysis skills will automatically transfer them when attempting to read connected text (Lyon & Moats, 1997). Rather, struggling readers need to be encouraged to become active problem solvers with regard to graphic information in text. This includes adopting a “set for diversity” (Gaskins et al., 1988) in which they learn to use irregular spelling patterns (which are common in English orthography), polyphonic spelling patterns, and partial decoding attempts to generate alternative pronunciations of target words until one is produced that matches a word in their spoken vocabulary and is appropriate to the sentence context.
Perhaps the most important finding of this research is that supplemental intervention approaches derived from different theoretical perspectives were both effective... Both interventions provided for instruction in key reading skills, balanced with opportunities to apply reading and writing skills in connected text... (p. 179)
Advances in data analysis tools
New statistical tools such as growth curve analysis and hierarchical linear modelling are increasingly being used by reading intervention researchers to examine the effects of various social, cultural, environmental and ecological factors on reading interventions (Lyon & Moats, 1997; Pressley, Graham, & Harris, 2006). Until recently the evaluation of interventions relied almost exclusively on traditional pretest-posttest designs. However, assessing the effects of interventions in this manner fails to provide estimates of intraindividual change, especially in larger scale interventions delivered over an extended period of time. As Lyon and Moats (1997) pointed out, the advantages of growth curve analysis over traditional designs is that it enables the researcher to examine “different forms and different patterns of growth at different stages of development and/or during treatment...[and]...provides the investigator with a new method for addressing the inevitable heterogeneity that will accompany any sample of children, no matter how carefully selected” (p. 584).
In addition to being useful in assessing the effects of interventions, the analysis of growth curves (which yields slopes and intercepts as variables) provides the basis for a recently developed approach for identifying students at risk for reading failure called dual discrepancy, where both the student’s performance level and growth rate are below that of classroom peers (Fuchs, Fuchs, McMaster, & Otaiba, 2003). The approach involves two steps. A screening tool is first used to determine if students are performing below grade-level expectations. An analysis of individual growth curves is then used to measure rate of improvement in response to instruction (which may be supplemental instruction in a multi-tiered system; Denton & Mathes, 2003). To illustrate the logic of this approach, Fuchs et al. (2003) gave the example of an endocrinologist considering the possibility of underlying pathology in a child’s physical growth. Not only is the child’s height measured at one point in time, but the child’s growth trajectory is measured in response to an adequately nurturing environment. If both measures are below normal, the child is considered a likely candidate for special treatment.
Nested models such as hierarchical linear modelling have provided researchers with the means for examining interactions between interventions and child, classroom, and school variables. The models have also enabled researchers to examine contextual effects in risk determination, where individual risk for reading difficulties is nested within classroom and within school levels of risk. An example of the latter is provided by the third paper in the special issue by Foorman, York, Santi, and Francis. The primary purpose of their study was to examine contextual effects on predicting risk for reading difficulties in first and second grade. The study addressed three questions. First, which is a better predictor of risk in second grade, the student’s pretest score in first grade or the student’s score in first grade combined with the pretest mean of the student’s first grade classroom? Second, are student outcomes influenced by the particular pairing of year one to year two teachers who instructed students as they moved from first grade to second grade? Third, does the administration format of the early reading assessment used in the study (paper, paper plus desktop computer with internet connection, or handheld device plus desktop computer) or level of teacher support in making use of the assessment data (web mentoring or no web mentoring) influence outcomes?
To explore these issues 171 schools (with a total of 542 first grade classrooms) were selected to participate in the study using a stratified random sampling procedure. The assessment administered by the classroom teachers was the Texas Primary Reading Inventory. Although no effects were found for either the administration format of the assessment or level of teacher support in making use of the assessment data, an analysis of predictor and outcome scores for each pretest-posttest pairing of teachers revealed that student achievement gains varied as a function of the particular pairing of teachers a student had when moving from first to second grade. The results further indicated that a combination of student pretest and mean of pretest classroom was a better predictor than student pretest alone. A closer examination of interaction effects for specific reading measures indicated that students with high word reading scores in first grade tended to have higher word reading scores in second grade if they were in a high-scoring classroom, whereas students with low fluency scores in first grade tended to have higher fluency scores in second grade if they were in high-scoring classrooms. An important implication of the latter finding is that it may not be guided reading alone that promotes fluency but rather guided reading in the context of faster reading models.
Demise of IQ-achievement discrepancy approach to identifying reading disability
As suggested earlier, the difficulties that intervention researchers have experienced in attempting to help struggling readers close the gap in fluency, vocabulary and reading comprehension are most likely the result of negative Matthew effects in reading. Because they receive considerably less practice in reading, struggling readers are not able to take full advantage of the reciprocally facilitating relationships between reading achievement and other aspects of development, including component skills of reading itself (Stanovich, 1986; Tunmer & Chapman, 1996; Tunmer & Hoover, 1993). These developmental spinoffs include vocabulary growth, ability to comprehend more syntactically complex sentences, development of richer and more elaborate knowledge bases, and greater reading practice opportunities for building fluency and facilitating implicit learning of letter-sound patterns (Gough & Hillinger, 1980; Share, 1995), all of which facilitate further growth in reading and language by enabling children to cope with more difficult materials.
One of the clear implications of the negative Matthew effects associated with each of the three broad categories of poor readers (dyslexic, mixed, specific reading comprehension difficulties) specified by the Gough and Tunmer (1986) model of reading difficulties is the need for early intervention. Historically the notion of “unexpected underachievement” has been the central defining feature of the reading disability construct. Children are identified as reading disabled/dyslexic when factors that would be expected to cause problems in all areas of learning, not just reading, are excluded (e.g., mental retardation, emotional and social difficulties, attentional problems, limited access to linguistic and environmental opportunities, inadequate or inappropriate classroom instruction). In actual practice, however, reading disability is normally defined as a discrepancy between reading achievement and intellectual potential as measured by standardized intelligence tests. Most exclusionary factors are generally ignored, especially experiential and instructional deficits. As Vellutino and Denckla (1991) pointed out, “virtually all of the research available has failed to evaluate or adequately control for the environmental and/or educational deficits that may cause a reading disorder” (p. 603).
An important consequence of the discrepancy-based assessment procedure is that children with reading disability are not normally identified until after they have been exposed to reading instruction for 2–3 years, and often longer. This “wait-to-fail” approach to identification is antithetical to early intervention and the prevention of negative Matthew effects in reading (Fuchs & Fuchs, 2006). Even more damaging to the discrepancy-based approach is a considerable amount of research indicating that groups of poor readers formed on the basis of the presence or absence of IQ-achievement discrepancies do not reliably differ in long-term prognosis, response to intervention, or the cognitive subskills (e.g., phonemic awareness, phonological decoding) that underlie the development of word recognition (Fletcher, Denton, & Francis, 2005; Fletcher et al., 1994; Francis et al., 2005; Fuchs & Young, 2006; Hatcher & Hulme, 1999; Stanovich, 1991; Stanovich & Siegel, 1994; Stuebing et al. 2002; Vellutino, Scanlon, & Lyon, 2000). These findings suggest that IQ tests are largely irrelevant to the identification of reading disability.
Further support for the claim that non-discrepancy-defined (i.e., low IQ) poor readers and discrepancy-defined poor readers (i.e., those with IQs in the average to above average range) do not acquire reading skills in a fundamentally different manner comes from the fourth paper in the special issue by Goetz et al. The major aim of their study was to determine whether poor readers with more severe intellectual impairments benefit from essentially the same type of intervention approach involving explicit instruction in phonological awareness and phonemically based decoding strategies that has been found to be effective with poor readers who have mild or no cognitive impairment. As Goetz et al. note, the prevailing view among most researchers has been that children with intellectual disability are primarily logographic readers who rely on visual cues to recognise words, in which case they should be taught to read using a “Look and Say” approach in which words come to be recognized as whole word visual patterns. However, this way of learning to recognize words is very developmentally limiting because it is not generative; it does not provide a means for identifying words not seen before.
Goetz et al. carried out a short-term intervention study in which 15 children with down syndrome (DS) attending mainstream schools received instruction in phoneme segmentation and phoneme blending skills, letter-sound relationships that included explicit links between phonological awareness and print, and book reading activities. The intervention built on two existing programs, but included additional activities thought to be helpful to children with severe learning disabilities, such as speech-based work focusing on the articulation of target sounds in isolation and in combination with other sounds. One group of DS children (n = 8) received two 8-week blocks of training, and the second group (n = 7) received one block that began after the first group had competed the first block of training. Assessments for both groups occurred immediately prior to the intervention, after each block of training, and five months after the intervention had finished. Results indicated that the first group of DS children significantly outperformed the second group on measures of letter-sound knowledge and word recognition following eight weeks of intervention, that the second group began to make progress once they received the intervention, that both groups improved in phonological awareness and made accelerated progress in reading during the period of the intervention, and that the positive effects of the intervention were maintained at the five-month follow-up testing. These findings suggest that children with intellectual disability can benefit from a phonics-based approach to reading intervention.
In addition to problems related to IQ, another major shortcoming of the discrepancy-based procedure is the associated definition-by-exclusion approach for identifying reading disability in which it is typically assumed that poor achievement reflects disability rather than poor or inadequate teaching when the latter is more often than not the primary contributing factor (Fuchs & Fuchs, 2006; Tunmer & Chapman, 1996; Vellutino et al. 1996; Vellutino, Scanlon, Small & Fanuele, 2006). As Calfee (1983) pointed out over two decades ago, “the bureaucratic procedures used at the present to label a child as having a specific reading disability make it possible and even likely that many youngsters, perhaps a preponderance of those so designated, reflect an instructional dysfunction rather than a constitutional shortcoming of the child” (p. 26). More recently, Vellutino et al. (2006) argued that because discrepancy-based approaches “do not control for the child’s preschool and educational history, they do not adequately distinguish between reading difficulties caused primarily by experiential and instructional deficits and reading difficulties caused primarily by biologically based deficits in reading-related cognitive abilities” (p. 157). As a consequence, the number of children classified as reading disabled is highly inflated.
Vellutino and colleagues (Vellutino et al., 1996; Vellutino, Scanlon, & Jaccard, 2003) investigated the extent to which experiential/instructional deficits are primary causes of early and protracted reading difficulties, especially for children who have acquired such limited amounts of crucial reading-related knowledge, skills, and experiences from home and preschool that they are unable to acquire basic literacy skills by means of regular classroom instruction without additional support. Vellutino and colleagues carried out a longitudinal study in which a large sample of children was tracked from the beginning of kindergarten to the end of third grade. The children who had significant reading problems by the middle of the first grade (approximately 9% of their sample) were provided with one-to-one remedial instruction during the second semester of first grade. Vellutino et al. (1996) found that 67% of these children (the “readily remediated poor readers”) were within the normal range of reading achievement following the remediation. In addition, these children performed significantly better than the “difficult-to-remediate poor readers” (i.e., those who did not respond to the intervention) on measures of phonological processing administered prior to the intervention. Vellutino and colleagues concluded from these results that most children with early reading difficulties suffer from experiential and instructional deficits, and that the truly disabled readers (1.5% of the sample) are those children with relatively severe phonological processing deficits who, as a consequence, do not respond to either regular classroom instruction or intensive, short-term intervention efforts.
Response to intervention (RTI)
Emerging from these findings and those of others is the response-to-intervention (RTI) approach to preventing and identifying reading disability (Deschler, Mellard, Tollefson, & Byrd, 2005; Fuchs & Fuchs, 2006). Researchers concerned with reading disability/dyslexia have concentrated their efforts on answering three key questions: What is it? What causes it? What can be done about it? Although the focus of this special issue is on the latter question, the answers to these questions are highly interrelated and cannot be pursued in isolation from one another. A researcher’s conceptualization of what skilled reading is and how it is acquired will greatly influence how they define reading disability, what they think causes problems in learning to read, and what they believe are the most effective intervention strategies for helping students to overcome persistent literacy learning difficulties. RTI simultaneously addresses all three questions by defining reading disability as the inability of otherwise typically developing children (i.e., those who satisfy standard exclusionary criteria) to respond adequately to high quality instruction/intervention because of an impairment in the phonological processing skills required to learn to read. The RTI model includes procedures for identifying reading disability, for closely monitoring progress in acquiring the phonemically based skills and strategies known to be causally related to early reading development, and for implementing research-based secondary and tertiary interventions for children with persistent literacy learning problems.
RTI operationalizes unexpected underachievement in terms of both low performance on reading and reading-related measures and poor response to high quality instruction. This is the dual discrepancy assessment procedure described earlier, a procedure that provides the basis for the early identification of students at risk for reading failure. In this approach intervention serves as the “test stimulus” and rate of growth (i.e., degree of responsiveness to intervention) serves as the “test performance” in identifying reading disability (Fuchs & Fuchs, 2006, p. 95). RTI uses evidence-based instruction and continuous progress monitoring across multiple tiers (usually three) to provide early intervention for children at risk for reading failure and to develop a more reliable procedure for identifying students with reading disability.
The first tier of RTI models typically involves “enhanced classroom instruction” (Denton & Mathes, 2003, p. 233) where literacy teaching in the earliest years of school addresses the individual needs of all of the children in the classroom, especially those experiencing early literacy difficulties. The second tier normally involves more explicit and extended (small group) instruction for children whose rates of progress in the first tier identify them as at risk for reading difficulties and in need of supplemental instruction (i.e., secondary intervention). Children who continue to progress at a very slow rate after the provision of second-tier supplementary instruction are placed in more intensive third-tier interventions (e.g., daily one-to-one tutoring) of longer duration (see Denton & Mathis, 2003, for a more detailed discussion of the three-tier model). Continuous monitoring of individual student progress is used in each of the three tiers to determine whether a child no longer needs supplemental instruction, needs continuing support at the existing level, or is eligible for a higher level of support. A variety of multi-tiered, RTI models are currently being investigated (Al Otaiba & Fuchs, 2006; Denton, Fletcher, Anthony, & Francis, 2006; McMaster, Fuchs, Fuchs, & Compton, 2005; O’Connor, Fulme, Harty, & Bell, 2005; Vaughn, Linan-Thompson, & Hickman, 2003; Vellutino et al., 2006).
An important aim of the RTI model is to increase the accuracy of selecting children who are truly in need of secondary intervention (i.e., “true positives”) to ensure that the most vulnerable children receive supplemental instruction, thus preventing the development of more significant reading problems (Fuchs & Fuchs, 2006). Two types of errors can lower the degree of accuracy in identifying at risk students, “false positives” and “false negatives”. Selecting students for intense services who are not in need of them results in false positives, which undermines the effectiveness of RTI models by inflating the number of at-risk children and stressing available resources for providing secondary interventions. False negatives occur when children score above the cut off scores on predictive measures but later develop reading difficulties. A high number of false negatives diminishes the preventive aspect of RTI models by depriving at-risk children of the additional support that they require.
This topic is investigated in the final two papers of the special issue. The study by Fuchs, Compton, Fuchs, Bryant, and Davis examined three key questions about the RTI method for preventing and identifying reading disabilities. As noted earlier, RTI models typically involve three tiers. Regular classroom instruction represents the primary intervention, small-group tutoring for students demonstrating unsatisfactory progress in the primary intervention represents the more intensive secondary intervention, and intensive special education for students who respond poorly to the secondary intervention represents the tertiary intervention. The three questions addressed in the study focused on RTI’s pivotal secondary intervention. The questions were: Who should be selected for secondary intervention? What are the effects of research-based secondary intervention on the reading development of the participating at-risk students? How should “non-responsiveness” to secondary intervention be defined?
To investigate these questions Fuchs et al. analyzed data from a two-year longitudinal study that started at the beginning of first grade and involved the six lowest performing children from each of 42 classes, giving a total of 252 students. Regarding the first question, Fuchs et al. examined the use of CART, a classification tree analysis that develops a set of multivariate decision rules with cut-scores and builds decision trees to sort categorical variables (in this case, reading disability vs. non-reading disability) using a set of predictors. They found that a CART model developed in earlier work for predicting children at risk for reading disability (see Compton, Fuchs, Fuchs, & Bryant, 2006) could be improved by adding a cognitive ability measure (performance IQ) to the prediction model that reduced the number of false positives. Interestingly, this finding is consistent with results obtained in a study by Tunmer, Herriman, and Nesdale (1988), who found that preliterate children with low levels of phonological awareness at school entry but above-average levels of decentration ability showed significantly greater improvement in phonological awareness and reading achievement during first grade than children with similarly low levels of phonological awareness but below-average levels of decentration ability at school entry.
Regarding the second question, Fuchs et al. found that not only did the students in their tutoring program (which occurred during the spring semester of first grade and included activities relating to improving sight word recognition, decoding, and fluency) significantly outperform controls on both a progress monitoring measure and several standardized reading measures, but that the positive effects of the secondary intervention were maintained throughout second grade. Regarding the third question, the researchers explored four promising approaches for defining “non-responsiveness” and found that each of these RTI methods resulted in different (but overlapping) groups of children being designated as having reading disability, indicating the need for further research.
The aim of the final paper of the special issue by Vellutino, Scanlon, Zhang, and Schatschneider was to determine whether measures of response to kindergarten and first grade intervention would provide a more effective procedure for identifying children at risk for long-term reading difficulties than kindergarten screening measures, measures of intelligence, or measures of reading-related cognitive abilities. Entry-level kindergarten children (n = 1,373) from 28 classrooms in nine schools from five school districts were classified as “at risk” or “not at risk” on the basis of a letter identification test. Half of the at-risk children were randomly assigned to a project-based intervention condition where they received supplementary intervention two days each week (30 min per day) in small groups until the end of the kindergarten year. The intervention program included activities designed to facilitate the development of basic literacy skills (phonological awareness, knowledge of letter sounds, letter-sound decoding, and sight word identification) as well as ample opportunities to apply these skills in authentic reading and writing. All at-risk children were again assessed at the beginning of first grade and assigned to a “continued-risk” group or a “no-longer-at-risk” group based on a composite measure of word-level skills. A subset of the children in the continued-risk group received project-based intervention (daily one-to-one tutoring) throughout first grade focusing mostly on the acquisition of phonological skills.
Vellutino et al. tracked the literacy development of children in all groups until the end of third grade. The results of the study indicated that of the total sample of at-risk children identified at the beginning of kindergarten who received intervention in kindergarten or both in kindergarten and first grade (about 30% of the cohort), 84% became independent readers. Of particular interest, an examination of various models for identifying kindergarten children who may be at-risk for later reading difficulties showed that RTI-based measures of incremental growth (i.e., growth curve analysis) more effectively and more consistently identified such children than did psychometric measures. These findings and those reported by Fuchs et al. provide further evidence that RTI-based procedures are more effective in identifying genuine at-risk-students (i.e., true positives) and not selecting those not in need (i.e., false positives) than standard psychometric screening procedures.