Robert E. Slavin

Center for Research on Effective Schooling for Disadvantaged Students The Johns Hopkins University, Baltimore, Maryland

Preventing early reading failure with one-to-one tutoring:
A review of five programs.

Tutoring is the oldest form of instruction. Parents have always provided one-to-one instruction to their children, and learning settings from driving instruction to on-the-job training typically employ one teacher for each learner for at least part of the learner's instruction.

     In elementary and secondary instruction, one-to-one tutoring exists around the sure of group instruction. For example, teachers often work with individual children during seat work periods, recess, study hall, or after school. Parents often hire tutors to work with their children. Tutoring is often used in special education, and sometimes in other remedial programs such as compensatory education.

     The topic of tutoring has come to the fore in recent years because of a renewed focus on students who are at risk of school failure, coupled with a renewed commitment to see that all students learn basic skills in the early grades. In particular, modest effects of traditional U.S. Chapter 1/ Title I pullout programs (Carter, 1984) and the loosening of restrictions on uses of Chapter 1 funds have contributed to a broader range of services being provided under Chapter 1 funding.

     One-to-one tutoring is one option often being considered or implemented. In recent years, increased flexibility in Chapter 1 and other factors have led to the use of tutors with first graders to prevent early reading failure. Advocates of tutoring programs argue that first grade is a critical year for the learning of reading, and reading success in the early grades is an essential basis for success in the later grades. Clay (1979), for example, argues that early intervention for children who have problems learning to read is crucial to children's later success. For students who do. not learn to read in traditional classrooms or with traditional reading programs, one-to-one tutoring is a possible solution to preventing early reading failure.

     Research on Chapter 1 programs suggests that remediation of learning problems after the primary grades is largely ineffective (see Kennedy, Birman & Demaline, 1986). It may be that it is easier to prevent learning problems in the first Place than to attempt to remediate them in the later grades. Considering how much progress the average reader makes in reading between the first and last days of first grade, it is easy to see how students who fail to learn to read during first grade are far behind their peers and will have difficulty catching up.

     The major drawback to tutoring is its cost. Providing tutoring to large numbers of students across the grade span would, of course, be prohibitive. But if in fact early intervention can prevent children from experiencing failure and can help them get off to a successful start in school, the use of this expensive intervention may be cost effective in the long run.

     The importance of understanding the effects of first-grade tutoring goes far beyond the pedagogical and technical issues involved. Edmonds's (1981) statement that every child can learn and Bloom's (1981) assertion to the same effect contributed to a variety of discussions among policy makers about learning as an "entitlement' for all children, on the basis that if every child can learn, then schools have an ethical and perhaps legal responsibility to see that every child does learn. One manifestation of this point of view is a document produced by the Council of Chief State School Officers (1987) that describes model state statutes to entitle every U.S. child not only to an appropriate education but to success in achieving an acceptable level of performance (also see Council of Chief State School Officers, 1989). If success is seen as an entitlement, educators must have methods that produce success for ad non-retarded children regardless of home background, no matter how expensive these methods may be. In any discussion along these lines, one-to-one tutoring for at-risk students is sure to be one element of the strategy to ensure success for all.

     Recently, there is an unprecedented willingness among educators to adopt expensive early intervention programs if they are believed to reliably produce large effects. Examples of this include Project STAR in Tennessee (Word et al., 1990) and Project Prime Time in Indiana (Farr, Quilling, Bessel, & Johnson, 1987), which have implemented substantially reduced class sizes in the early elementary grades. Growing provision of preschool and extended day kindergarten programs and of IBM's Writing to Read computer program are other examples. Recently, many districts have adopted Reading Recovery and Success for All, intensive reading programs with tutoring, as means of preventing early school failure.

     It would be important to know the effectiveness of such programs that are expensive to implement and maintain in school districts. If school districts plan to allocate Chapter 1 funds to expensive programs , the effectiveness of these programs should be of great concern. It is important to know how large the effect of tutoring is (in comparison to plausible alternatives), to what degree effects of tutoring are maintained over time, and which specific tutoring programs and practices produce the largest gains in student reading achievement.

     The purpose of this article is to review the research on the effectiveness of one-to-one tutoring programs to identify what is currently known about the answers to these and other questions.

     Previous reviews of research on tutoring have primarily focused on peer tutoring (e.g., Devin-Sheehan, Feldman, & Allen, 1976; Scruggs & Richter, 1985). The one review that included tutoring by adults primarily focused on applications in special education (Polloway, Cronin, & Patton, 1986). None of these earlier reviews discussed any of the first-grade reading prevention models emphasized here.

     In the present article, we consider the effectiveness of tutorial program from two perspectives: empirical and pragmatic. From the empirical perspective, one can ask questions such as "Does the program work?" and "How strong are its effects?" To answer these questions, we computed effect sizes for each of the five programs. (This is discussed in detail in the section on review methods.) From a pragmatic standpoint, one can ask questions such as "What components of reading are included?" and "Does it matter if the tutors are certified teachers or paraprofessionals?" and "Why are some programs apparently more effective than others?"

     It would also be important to examine the theoretical similarities and differences of these programs regarding the approach taken to learning in general, and reading in specific, and how the relationship between the tutor/student dyad facilitates learning. One aspect of effectiveness of tutorial programs could be explained by appealing to domain-general theories such as Vygotsky (1978) that have been formulated to account for the transmission of knowledge in one-to-one dyads. However, while the Vygotskian perspective has been expired with one program, Reading Recovery (see Clay & Cazden, 1990), theories to account for transmission of knowledge from tutor to student have not been explored in the other programs. Similarly, it would be important to examine the different theories of reading as espoused by advocates of each tutoring program. However, again with the exception of Reading Recovery, the programs do not articulate a theory of reading.

     In what follows, we review five tutoring programs. In the course of describing these programs, we discuss the model of reading to which each program subscribes to and identify the key components of reading found in each program. From reviewing the curriculum of the tutoring programs, we have identified eight components of the reading process that are emphasized in these programs: perceptual analysis of print, knowledge of print conventions, decoding, oral language proficiency, prior knowledge, lexical access, syntactic analysis of sentences, and prose comprehension. We acknowledge that this is by no means a complete list, since key aspects of reading such as phonemic awareness are not included. However, these components were extracted from the programs reviewed. We then discuss which components each program includes. We also consider the nature of the tutors and how the program are implemented. Then we provide effect sizes to qualify the empirical effects of the programs. If one tutoring program appears to be more effective than another, it could be because (a) practical differences in the program lead to different outcomes, (e.g., certified teachers are used in one and not the other), or (b) tutors in one are using more effective methods or curricula than those in the other, or (c) different programs to emphasize to different degrees or reading components that are considered to be central in contemporary theories of reading. In our discussion, we consider these and other explanations.

Review Methods

     This review uses a set of procedures called best evidence synthesis, which combines elements of meta-analysis with those of traditional narrative reviews (Slavin, 1986). Briefly, a best-evidence synthesis requires locating all research on a given topic and discussing the substantive and methodological issues in the research as in a narrative review, A prior criteria for germaneness to the topic at hand and for methodological adequacy are typically applied. Whenever possible, study outcomes are characterized in terms of effect size (ES), the difference between experimental and control means divided by the control group standard deviation. When means or standard deviations are not reported, effect sizes are estimated from F, t, or other statistics (see Glass, McGaw, & Smith, 1981). The numerator of the effect size formula may be adjusted for pretests or covariates by computation of gain scores or use of ANCOVA, but the denominator is always the unadjusted individual level standard deviation of the control group or (if necessary) a pooled standard deviation.

     Inclusion Criteria. Studies were included in the present review if they evaluated one-to-one instruction delivered by adults (certified teachers, paraprofessionals, or volunteers) to students in the first grade who are learning to read for the first time. Studies had to compare tutoring to traditional instruction in elementary schools over periods of at least 4 weeks on measures of objectives pursued equally in experimental and control conditions. This duration requirement did not exclude any studies of first-grade tutoring. The first-grade requirement excluded only three studies (Bausell, Moody, & Walzl, 1972; Fresko & Eisenberg, 1983; and Shaver & Nuhn, 1971), which looked at remedial tutoring in the third grade and higher. Studies of cross-age and same-age peer tutoring (e.g., Cloward, 1967; Greenwood, Delquardi, & Hall, 1989; and von Harrison & Gottfredson, 1986), did not fit this criterion and were not included. All studies ever written in English were included. The only study done outside of the U.S., by Clay (1985), examined only students who were successful in tutoring, not all who received it. This study is described in the section on Reading Recovery. Therefore, this best-evidence synthesis included all methodologically adequate studies of one-to-one tutoring that focused on instruction delivered by adults to first graders. in a complete review of published as well as unpublished studies, a total of 16 studies met the inclusion criteria.

     All of the studies that met the inclusion criteria specified above evaluated a total of five tutoring programs. These programs incorporated instructional materials as well as provision of one-to-one tutors. Some of the major characteristics of these programs are summarized in Table 1.

Table 2 provides additional detail on models of reading used in each program. As is apparent from the Tables, the five programs vary widely in curriculum, integration with classroom instruction, use of certified versus paraprofessional tutors, and other factors not intrinsically related to the one-to-one setting. These programs also differ in their model of reading and the measures used to assess the effectiveness of these programs. As a result, we make no attempt to combine findings across studies in any way. However, we do discuss how different approaches to reading translate into the method used in the tutoring process. Finally, we discuss how ultimately the reading model is tied to the type of assessment each program uses to evaluate its effectiveness, suggesting that curriculum, instruction, and assessment are interrelated (Weade, 1987).

Reading Recovery

     The preventive tutoring program that has received the most attention and use in recent years is Reading Recovery. This program was originally developed by Marie Clay (1985) in New Zealand, and is widely used in that country. In l985, Marie Clay and a colleague, Barbara Watson, spent a year at the Ohio State University. They trained a group of teachers to use the program, and trained several Ohio State faculty members to train others. Since that time, research on Reading Recovery has been conducted at Ohio State, and the program has rapidly expanded in use.

     As applied in the longitudinal studies, Reading Recovery provides one-to-one tutoring to first graders who score in the lowest 20% of their classes on a program-developed diagnostic survey. The tutors are certified teachers who receive training for 2.5 hours per week for an entire academic year. Students are tutored for 30 minutes each day until one of two things happen. If students reach the level of performance of their classmates in the middle reading group, they are "discontinued." if they receive 60 lessons without achieving this level of performance, the students are released from the program but considered "not discontinued."

Model of reading
     In the Reading Recovery program, reading is viewed as a psycholinguistic process in which the reader constructs meaning from print (Clay, 1979; Pinnell, 1985). According to Clay, reading is defined as a "message-gaining, problem solving activity, which increases in power and flexibility the more it is practiced." Clay states that within the "directional constraints of the printer's code, language and visual perception responses are purposefully directed in some integrated way to the problem of extracting meaning from text, in sequence, to yield a meaningful communication, conveying the author's message" (Clay, 1979, p. 6).

     Clay does not specifically address how language and visual perception are coordinated in order to extract meaning from text. Nevertheless, her discussion of reading and components of the Reading Recovery Program suggest that she includes the following components of reading in her model: perceptual analysis, knowledge of print conventions, decoding, oral language proficiency, prior knowledge, inference making, reading strategies, metacognition and error detection, and error correction strategies (see Table 2).

     Clay (1979) describes reading as the "process by which the child can, on the run, extract a sequence of cues from printed texts and relate these, one to another, so that he can understand the precise message of the text." in order to master this process, the child must have good control of oral language, developed perceptual skills, the physiological maturity and experiences that allow the child to coordinate what s/he hears in language and sees in print, and enough hand-eye coordination so s/he can learn the controlled, directional patterns required for reading (Clay, 1979). Expert teachers are assumed to have sufficient implicit knowledge of the processes that they can recognize the source of the child's difficulty.

     From this theory of reading, three major theoretical principles serve as a foundation for the Reading Recovery Program. First, reading is considered a strategic process that takes place in the child's mind. Reading requires the coordination of many strategies and visual information, the integrating of letter-sound relationships, features of print, and the child's own background knowledge. Meaning is never derived just from the print alone, but from the interaction of the reader's unique background and the print. Second, reading and writing are interconnected. Having the child make the connection between reading and writing is essential to literacy development. Third, "children learn to read by reading" (Pinnell, 1989; Pinnell, DeFord, & Lyons, 1988). Children must engage in reading of connected text and should avoid working on isolated skills in order to become proficient in reading. It is only by reading frequently that the child can come to detect regularities and redundancies present in written language.

     These three principles set the foundation for the Reading Recovery program. Children in Reading Recovery spend most of their time engaged in reading and writing activities. There is no systematic presentation of phonics, yet during the reading and writing activities, letter-sound relationships are taught as one of the basic strategies in solving problems. Tutors use a variety of strategies to help students develop "independent, self generating systems for promoting their own literacy" (Pinnell, 1985).

Structure of tutoring

     For the first few tutoring sessions, the teacher and student "roam around the town," reading and writing together in an unstructured, supportive fashion, to build a positive relationship and to give the teacher a broader knowledge of the child. After this, teachers begin to use a structured sequence of activities, as follows (adapted from Pinnell et al., 1988, pp. 10-11).

     The child rereads familiar books. The child reads again several favorite books that s/he has previously read. The materials are storybooks with natural language rather than controlled vocabulary. Books within a lesson may range from quite easy to more challenging, but the child is generally reading above 90% accuracy. During this time, the child has a chance to gain experience in fluent reading and in using strategies 'on the run" while focusing on the meaning of the text. The teacher interacts with the child during and after the reading, not 'correcting,' but talking with the child about the story and supporting the effective actions the child has taken.

     The teacher analyzes reading using the running record. Each day the teacher takes a running record of a book that was new for the child the previous day. The running record is a procedure similar to miscue analysis (Goodman, Watson, & Burke, 1987). Using a kind of shorthand of checks and other symbols, the teacher records the child's reading behavior during oral reading of the day's selected book. The teacher examines running records closely, analyzing errors and paying particular attention to behavior such as self-correction. In this way, s/he determines the strategies the child is using to gain meaning from text. This assessment provides an ongoing picture of the progress the child makes. While the child is reading, the teacher acts as a neutral observer; the child works independently. The accuracy check tells the teacher whether the text was well selected and introduced the day before.

     The child writes messages and stories and then reads them. Every day the child is invited to compose a message and to write it with the support of the teacher. Writing is considered an integral part of gaining control over messages in print. The process gives the child a chance to closely examine the details of written language in a message that s/he has composed, supported by her/his own language and sense of meaning. Through writing, the child also develops strategies for hearing sounds in words and using visual information to monitor and check her/his reading.

     After the construction of the message, the teacher writes it on a sentence strip and cuts it up for the child to reassemble and read. This activity provides a chance to search, check, and notice visual information. Using plastic letters on a magnetic board, the teacher may take the opportunity to work briefly with the letters to increase the child's familiarity with the names of letters and their use in known words such as the child's name. This work will vary according to the knowledge the child already has.

     The child reads new books. Every day the child is introduced to a new book that s/he will be expected to read without help the next day. Before reading, the teacher talks with the child about the book as they look at the pictures. The teacher helps the child build a frame of meaning prior to reading the text. The purpose of the introduction is not necessarily to introduce new words, but to create understanding in advance of reading so that it will be easier to focus on meaning.

     Every child's program differs. Children do a great deal of reading, but not from a graded sequence. No child reads the same series of books. The small books are carefully selected by the teacher for that child at that time. In writing, children work on their own messages, so they are writing and reading works that are important to them individually, The major difference within and across lessons lies in the teacher's ability to follow each child and to respond in ways that support acceleration and the development of strategies. Strategies may include directional movement, one-to-one matching, self monitoring, cross-checking, using multiple cue sources, and self-correction. The Reading Recovery teacher uses instructional techniques designed to help the child develop and use such strategies.

     The tutoring model in Reading Recovery is separate from the instruction provided in the regular classroom. Most often, Reading Recovery teachers tutor students half time and either teach small groups of Chapter 1 students or teach a regular class the other half. The tutees may thus have the same teacher as their reading teacher and as their tutor, but in general this does not occur.

     Tutor training in Reading Recovery is extensive. During the first year, in addition to teaching a reading clam and tutoring four students, the tutors attend weekly seminars during which they receive training in observational, diagnostic, and assessment techniques and are schooled in the reading philosophy of Marie Clay. The tutors also participate in weekly 'behind the glass' demonstration lessons where they observe actual tutoring sessions behind a one-way mirror and have the opportunity to critique and discuss the lesson.

     Considerable time is spent learning about the reading process and learning how to implement appropriate strategies to meet the needs of individual children. Follow-up service training continues after the first year. Additional training is required of Teacher Leaders who are certified to train Reading Recovery tutors in their areas. Teacher Leaders participate in a 1-year internship at the Ohio State University training center (other states such as New York are establishing regional centers), where they participate in reading and writing seminars and learn to train tutors using the 'behind the glass' technique.

Results

     Research evaluating Reading Recovery in New Zealand (Clay, 1985) focused entirely on the discontinued students (those who were successful in the program), and therefore does not provide a full account of the effectiveness of the intervention. However, the U.S. research has included discontinued and not discontinued students-- all of the students who either graduated from the program or received at least 60 lessons.

     The Ohio State group has conducted two longitudinal studies comparing Reading Recovery to traditional Chapter 1 pullout or in-class methods. The first (pilot) study (Huck & Pinnell, 1986; Pinnell, 1988) of Reading Recovery involved 21 teachers trained by Marie Clay who worked in six inner-city Columbus, Ohio, schools. Each school provided one Reading Recovery class and a matched comparison class. The lowest 20% of students in each class served as the experimental and control group, respectively, Students were pre-tested in September and December, 1984, but the tutoring did not begin until the spring semester, 1985.

     The second longitudinal study (DeFord, Pinnell, Lyons, & Young, 1988; Pinnell, Short, Lyons, & Young, 1986) involved 32 teachers in 12 schools in Columbus. Twelve of these teachers had been tutors in the pilot cohort. In this study, students in the lowest 20% of their classes were randomly assigned to Reading Recovery or control conditions. The research design originally made a distinction between students in the experimental and control groups who had Reading-Recovery-trained versus non-Reading-Recovery-trained teachers in their regular reading program. However, there were no differences on this factor, so the analyses focused on tutored versus untutored children, regardless of who their regular reading teacher was.

 The results at the end of the first implementation year for the two Ohio State studies are summarized in Table 3. Reading Recovery students substantially outperformed control students on almost all measures. The exceptions were tests of letter identification and a word recognition scale, which had apparent ceiling effects in both conditions.

     Each spring for 2 years following the implementation year, all children were assessed on Text Reading Level, an individually a red test in which students are asked to read from books with progressively more difficult content. This measure yields a reading level (e.g., second grade, fire semester).

  The results on this measure, summarized in Table 4, show an interesting statistical paradox. By the criterion of effect size, the effects of Reading Recovery are clearly diminishing each year. By the end of the third grade, the effect size for the pilot cohort has diminished from +.72 to +. 14, and in the second cohort the effect size diminished from +.78 to +.25. On the other hand, the difference in raw units between Reading Recovery and control students remained about the same across all 3 years, hovering around two points in the pilot cohort and three in the second cohort. Is the effect maintaining or not?

     The difference between these two measures is that the standard deviation of the Text Reading Level measure increases each year, making the same raw difference a smaller proportion of the standard deviation. In more substantive terms, the size of the difference may not be diminishing (assuming the measure is an equal interval scale), but the importance of the difference is diminishing. For example, a difference of 3 months on a standardized reading test might be a big difference at the end of the first grade but is a small one at the end of sixth grade.

     Actually, there is a more complex story on the longitudinal effects of Reading Recovery. The students who succeeded in Reading Recovery, those categorized as discontinued, were performing on average at a level like that of their classes as a whole, and substantially better than the comparison group of low achievers. On the other hand, all of the not-discontinued students (who had at least 60 tutoring sessions but failed to achieve at the level of the rest of their class) were still below the level of their classmates by the third grade, and were substantially lower than the control group. These not-discontinued students represented 27% of the former Reading Recovery students tested in the third grade in the second cohort study (DeFord et al., 1988).

     Effects of Reading Recovery on promotions from grade to grade. Participation in Reading Recovery increased students' chances of being promoted to the second grade in comparison to the control low achievers. Although 31% of comparison students were retained in first grade or assigned to special education, this happened to only 22% of Reading Recovery students (DeFord et al., 1988). However, by the third grade this difference had mostly disappeared. Two years after the children were in the first grade, a total of 59.6% of Reading Recovery children and 57.8% of control children were in the third grade 2 years after first grade. A school district evaluation in Wakeman, Ohio, found that first grade retention's dropped from 24 to 1 in the 3 years after implementation of Reading Recovery (Lyons, Pinnell, DeFord, McCarrier, & Schnug, 1989).

     One additional study compared Reading Recovery to control treatments in first grade. This was a study conducted in four Chicago elementary schools. As in the earlier studies, students were randomly assigned to Reading Recovery or control conditions. Because neither standard deviations nor statistical tests are presented, effect sizes cannot be computed, but program effects in comparison to control students were clearly substantial. Applying standard deviations from the Ohio studies to the same measures used in Chicago yields end-of-first grade effect sizes of approximately +.90 on dictation and text reading level.

     The most recent major study of Reading Recovery conducted by the Ohio State group (Pinnell, Lyons, DeFord, Bryk, & Seltzer, 1991) evaluated the full program in comparison to three alternative programs and a control group in 10 Ohio school districts. The treatments were as follows:

1. Reading Recovery (RR) was implemented as in earlier assessments.

2. Reading Success (RS) was the same as Reading Recovery except that teachers received a 2-week training session in the summer instead of the yearlong, 2 to 3 hours per week training with "behind the glass" demonstration teaching used in Reading Recovery. In comparison to Reading Recovery, this treatment tested the possibility that effects like those for the program as usually implemented could be obtained with far less extensive training, a major stumbling block to widespread diffusion.

3. Direct Instruction Skills Plan (DISP) was an individual tutorial program that tested the possibility that the one-to-one tutoring, not the particulars of the Reading Recovery model, explains the effects of the program. DISP used direct instruction in specific skills such as letter, sound, and word recognition, sequencing, filling in blanks, answering questions, and reading extended text. Teachers were encouraged to design lessons themselves to teach these and other skills.

4. Reading and Writing Group (RWG) was a small group tutorial model caught by teachers who had been trained as Reading Recovery teachers. They used Reading Recovery materials and strategies but were asked to adapt them to the small group setting in their own ways. This treatment essentially tested the effects of the one-to-one tutoring aspect of Reading Recovery, holding curriculum constant.

5. Control group for each treatment was the Chapter 1 pullout program already in existence in each school.

     Four schools (one per treatment) were involved in each district. In each school that already had a Reading Recovery teacher, students were randomly assigned to RR or control (Chapter 1) treatments. In other schools, additional teachers were hired from the district's substitute lists to implement the RS or DISP tutoring models. Trained Reading Recovery teachers were added to schools to implement the Reading and Writing Group (RWG) treatment. Students were randomly assigned to treatment or control classes.

     The treatments were implemented starting early in first grade. Students were then assessed in February, in May, and again in the following October, The results are summarized in Table 5.

     As is clear from Table 5, the effects varied considerably according to measure and time of test administration. The February measures clearly favored the Reading Recovery on all measures and the Reading Success model on the two measures developed as part of the Reading Recovery program, Dictation and Text Reading Level. However, the February measures are biased in favor of the three tutoring models. By February, the tutoring was concluded, and students moved into the Chapter I group program. In contrast, the RWG and Chapter 1 control group programs were yearlong interventions, so measuring effects in February discriminates against them.

     Unfortunately, the only test given in May was the standardized Gates-MacGinitie, which found few effects for any treatment.

     The October follow up provides the best indication of the effects of the four programs. The most positive effects were found for Reading Recovery on Dictation (ES - +.35) and Text Reading Level (ES - +.75). Neither of the other two tutoring methods (RS and DISP) found any positive effects. It is interesting to note that after the full program, it was the Reading and Writing Group (RWG) treatment that had the most positive effects (ES + 29 for Dictation, + 32 for Text Reading Level). This treatment also had the largest positive effects on the May Gates-MacGinitie of all treatments (ES - +.34).

     One important factor may be confounded with the effects of the four programs. The teachers in the two most successful treatments, Reading Recovery and Reading and Writing Group, were experienced Reading Recovery teachers who had a year of Reading Recovery training and at least a year of experience in implementing the program. In contrast, the Reading Success and DISP teachers were hired from the substitute list and may have been considerably less skilled and less experienced.

     At a minimum, the Ohio statewide study provides one more convincing evaluation of Reading Recovery, showing large effects, especially on Text Reading Level, which maintain into the school year following the intervention. The findings suggest that the yearlong training, the particular curriculum and instructional model used, and the one-to-one aspect of the tutoring are all critical to the success of the model, but these conclusions may be tempered by possible differences in teacher quality in the groups that received shorter training (RS) and the alternative tutoring model (DISP).

     A few methodological issues about the Reading Recovery research are worth raising. First, there is an articulation between the Reading Recovery program and the measures used to evaluate the program, suggesting that what is taught is what is measured. The measures used were all individually administered scales designed either by Marie Clay and her associates or by the Ohio State researchers. Five of the measures, Letter Identification, Word Test, Concepts about Print, Writing Vocabulary, and Dictation, make up the Diagnostic Survey, which was developed by Clay. 'The Letter Identification test asks students to identify 54 letters in upper and lower case. The Word Test is a list of high frequency words from the basal reader used in the school district. Concepts about Print asks the students to identify conventions of print and reading. 'The Writing Vocabulary has the children write down as many words as they can, starting with their own name, in 10 minutes. 'The Dictation test assesses children's ability to write down every word in a sentence that is read to them. In scoring this test, children are given credit for every sound correctly represented. The Text Reading Level is the sixth test administered in the Reading Recovery evaluation. This test consists of a series of graded stories that the child reads. A running record of the child's oral reading is taken and then an accuracy level is calculated. These measures correspond to the model of reading in Reading Recovery. As d earlier, the reading model emphasizes oral language, perceptual analysis, concepts of print, reading strategies, and metacognition. All of these aspects are emphasized in the outcome measures. Therefore, children who were tutored in Reading Recovery were also more familiar with the assessment than were the children in the control groups.

     It also appears that bias in favor of the kinds of skills taught in the program is most likely at the low levels of the Text Reading Level measure, where assessments focus on concepts of print, using pictures and Patterns to guess story content, and other skills specifically taught in Reading Recovery. The finding of particularly large effects on Text Reading Level (in contrast to other measures) was especially pronounced in the Ohio statewide study (Pinnell et al., 1991).

     Secondly, Reading Recovery has a policy of not serving students who have already been retained in first grade and students identified for special education. One of the reports (Pinnell et al., 1986) implies that some students originally selected for tutoring failed to make adequate progress in early tutoring sessions and were excused from tutoring (and therefore excluded from the evaluation). Any of these practices might have influenced the Reading Recovery sample by excluding the very lowest achievers.

     These criticisms aside, the effects of Reading Recovery are impressive at the end of the implementation year, and the effects are maintained for at least 2 years. In addition, the Ohio State researchers have studied implementation Issues that affect the quality of the program. For example, Lyons (1991) studied the effects of duration of training on Reading Recovery teachers. Teachers who had a 2-week inservice program were compared to teachers who amended a yearlong training program. The results showed that students who had teachers who received more extensive training out performed students who had teachers in the 2-week program on Text Reading level.

     In another study, Handerhan (1990) conducted a socialistic analysis of teachers and children in Reading . Reading Recovery tutoring sessions were videotaped and sessions of four of the most and least successful teachers (based on what was accomplished with the student) were analyzed. Handerhan (1990) found that across tutors there was consistency in how they structured the lessons regarding language, materials, and procedural techniques. However, more successful tutors showed greater variability in the strategies they used and the less successful tutors engaged more in presenting letters and words as discrete skills without reading for meaning. This study is important because it documents the variability in instruction during tutoring as well as identifying what behaviors are necessary to be a successful tutor helping children learn to read. The rapidly expanding use of Reading Recovery throughout the U.S, (see Lyons, Pinnell, DeFord, McCarrier, & Schnug, 1989) shows that the program is practical to use.

Success for All
     Success for All (Madden et al., 1991; Slavin, Madden, Karweit, Dolan, & Wasik, 1992; Slavin, Madden, Karweit, Livermon, & Dolan, 1990) is a comprehensive schoolwide restructuring program that is designed primarily for schools serving large numbers of disadvantaged students. Its main intention is to see that ail children are successful in basic skills, particularly reading, the just time they are taught. One major element of Success for All is one-to-one tutoring by certified teachers for students in Grades 1-3 who are having difficulties learning to read. The program includes many other elements, such as a beginning trading program, preschool and kindergarten programs, and family support services. However, for low-achieving first graders, who receive most of the tutoring services, the Success for All program can be seen primarily as a preventive tutoring program.

Model of reading
     The Success for All tutoring program is based on research that "points to the need to have students learn to read in meaningful contexts and at the same time have a systematic presentation of word attack skills" (Slavin et al., 1992). Its underlying philosophy is that there is certain regularity to language, and that direct presentation of phonics is viewed as a helpful strategy which children can use to figure out words. Children also need to build a strong sight vocabulary that will help in identifying words that are not decodable. Along with the systematic presentation of phonics, children engage in reading meaningful connected text. The Success for All program emphasizes that reading is a strategic process that takes place in the student's mind and that these strategies should be taught directly.

     Unlike Reading Recovery, Success for All does not articulate a complete theory of be reading. However, an underlying model of reading can be seen in the structure and content of the program. There are four components that drive the Success for All tutoring program. First, children learn to read by reading meaningful text. Reading skills are not acquired by children learning isolated, unconnected information about print. Second, phonics needs to be taught systematically as a strategy for cracking the reading code. Children engage in reading stories that are meaningful and interesting, yet have a phonetically controlled vocabulary, Third, children need to be caught the relationship between reading words and comprehending what they read. Mere word recognition is not reading. The emphasis on comprehension is directly related to the fourth component, the emphasis on children's need to be caught strategies to help them become successful readers. Children who have problems learning to read often do not know how to effectively use metacognitive strategies to help them read. Through direct instruction, children are taught when, how, and why they should use strategies.

     In summary, Success for All emphasizes the following components in its model of reading: perceptual analysis, decoding, prior knowledge, oral language proficiency, inference making, reading strategies, metacognition and error detection, and error correction strategies.

Structure of tutoring
     The tutoring model used in Success for All is different in many ways from that used in Reading Recovery. One difference is that in Success for All, the tutoring model is completely integrated with the reading program. The tutor's most important responsibility is to make sure that the student is making adequate progress on the specific skills and concepts being taught in the reading class.

     Another difference is that in Success for All, first graders receive tutoring as long as they need it. Although most students receive tutoring for part of a year, some receive it all year and then continue to be tutored into the second grade. The commitment in Success for All is to see that every child succeeds, that no child is retained or assigned to special education except under extreme circumstances.

     First graders are initially selected into tutoring in Success for All on the basis of individually administered informal reading inventories given in September. Afte2r that, however, students are assessed every 8 weeks in terms of their progress through the reading curriculum. On the basis of these 8-week assessments, students who are doing well may be rotated out of tutoring as other students are rotated into tutoring. The amount of tutoring received by a given student may vary from 8 weeks to the entire year or more.

     Students receive tutoring every day for 20 minutes This time is usually scheduled during an hour long social studies/science block, so that tutoring represents additional time in reading.

     The tutors are certified teachers recruited in the same way as other teachers. Each tutor teaches a 90 minute reading class each day (to reduce class size for reading) and then spends the rest of the day tutoring three children per hour. Because the tutors teach a reading class, they are fully aware of what the reading program is; if a child is struggling with Lesson 37, the tutor knows exactly what is required for success in Lesson 37 because he or she has taught it.

     In many cases, tutors work with students who are also in their morning reading class. When scheduling does not allow this, the student's reading teacher fails out a tutor/teacher communication form that indicates what lesson the student is working on in class and the teacher's assessment of the specific problems the student is having with that lesson. The tutor uses this information to plan the tutoring session. This communication ensures coordination between the classroom instruction and tutoring.

     The tutors receive 2 days of training (along with all other beginning reading teachers) to learn to teach the Success for All beginning reading program (described below), and then they receive 4 additional days of training on assessment and on tutoring itself. Tutors are observed weekly by the program facilitator and given direct feedback on the sessions.

     A strong emphasis is placed on teaching comprehension strategies. The tutor's goal is to get the students to read fluently, and to understand what they read. Tutors are trained to explicitly teach metacognitive strategies to help students monitor their comprehension. For example, a tutor will teach a student to Stop at the end of each page and ask, "Did I understand what I just read?" The students learn to check their own comprehension and to go back and reread what they did not understand.

     Each tutoring session is structured, but the tutor is constantly diagnosing and assessing the individual needs of each student and tailoring the sessions to fit the student's specific problem. If a student is having difficulty with fluency, the tutor will have the student do repeated reading aloud of a story. With similar materials, a tutor may work with another child on comprehension monitoring.

     A typical tutoring session begins with the student reading out loud a familiar story that he or she has read before in tutoring and in the reading class. This is followed by a 1-minute drill of letter sounds to give the student the opportunity to practice the letter sounds taught in class. The major portion of the tutoring session is spent on reading aloud 'shared stories' that correspond to the beginning reading lessons. The shared stories are interesting, predictable stories that have phonemically controlled vocabulary in large type and other elements of the story in small type. The teacher reads aloud the small-type sections to provide a context for the large type portions read by the students. The tutor works with the student to sound out the phonetically regular words, asks comprehension questions about the whole story, and has the student reread passages out loud to gain fluency. Writing activities are also incorporated into the reading activities.

     As noted, the tutoring model is closely integrated with the reading program (Slavin, Madden, Karweit, Livermon, & Dolan, 1990), in which students are regrouped according in their reading levels. Use of tutors as reading teachers allows schools to reduce class size to about 15 students who are all at one level, so there are never multiple reading groups in any reading class. This allows teachers to spend the entire class period actively teaching reading, as it removes the need for the follow-up or seat work activities typical of classes with multiple reading groups. The beginning reading program emphasizes reading to students, engaging students in discussions of story structure, and developing oral language skills. Students begin using shared stories, as described earlier. As letter sounds and sound blending strategies are taught, students can apply them in their books. Students do a great deal of partner reading and pair practice activities, and writing is caught along with reading.

     There is a high degree of structure in the beginning reading program, which is helpful in integrating the classroom instruction with the tutoring session. Expectations for each lesson are clear, so the teacher and tutor can know that they are working on the same objectives. Integration is also facilitated by the use of brief tutor/teacher communication forms, on which each can tell the other about particular successes or problems a child is experiencing.

     Success for All is currently being evaluated in several schools in several school districts in six states. Evaluations most relevant to the tutoring aspect of the program relate to low achievers in two Baltimore schools that have had adequate funding to provide a high level of tutoring services for several years. Abbottston Elementary, the original pilot school, has implemented Success for All for 4 years. City Springs Elementary is a fully funded site whose implementation began a year after Abbottston. Each school was matched with a similar comparison school, and then students were individually matched on standardized reading measures. The student bodies at both Baltimore schools are almost entirely African American. Seventy-six percent of Abbottston's students qualify for free lunch. City Springs serves the most disadvantaged student body in the district; all its children come from housing projects, and 96% receive free lunch. Both are Chapter I schoolwide projects. Each May, students are individually assessed on scales from the Woodcock Language Proficiency Battery (1984) and the Durrell Analysis of Reading Difficulty (1980).

   The results for the students in Grades 1-3 who scored in the lowest 25% on the pretests are summarized in Table 6. The amount of tutoring received by these students varied depending on their needs; almost all received some tutoring, but in some cases they received 8 weeks, while some second or third graders may have received more than a year of daily tutoring.

     The results shown in Table 6 indicate powerful effects of the combination of tutoring, curricular changes, and family support services used in Success for All. At both schools in all years, first-grade low achievers have better than their matched counterparts in control schools (mean effect size - + 1. 15). Second graders who started in Success for All in the first grade or earlier also scored substantially better than control students (mean effect size - +.82), as did third graders in the program for 3 years (ES - + 1. 16). These second- and third-year effects should not be compared with the second- and third-year effects of Reading Recovery; the Reading Recovery data relate to the lasting effect of a first-grade intervention, while those for Success for All relate to the effects of continuing intervention. Although effect sizes stayed at approximately the same level in second and third grades as in first, this is an indication of 2 growing effect. Because standard deviations increase each year, a constant effect Size means a growing difference between experimental and control groups in grade equivalencies or raw scores.

     In addition to effects on reading achievement, all three schools substantially reduced assignments of students to special education for learning problems and essentially eliminated retention's (Slavin, Madden, Karweit, Livermon, & Dolan, 1992).

     As with Reading Recovery, there are methodological limitations to research on Success for All that may affect the results. First, because only one school was involved in each comparison, school effects could account for part of the observed differences. Lack of random assignment of schools or students also could have affected the results.

     The effects of Success for All were positive for the lowest achieving quarter of students involved as well as for the other students in the school (Slavin, Madden, Karweit, Dolan, & Wasik, 1990; Slavin, Madden, Karweit, Livermon, & Dolan, 1990; Madden et al., 1991, 1992). However, the effects for the higher achieving students must be ascribed to the curriculum and other program elements, as few of them received any tutoring. Also it is important to note that schools using Success for All without extra resources for tutoring also obtained very positive results, although not as positive as those for the fully funded schools (see Slavin et al., 1992). These schools used their existing Chapter I funds to provide some tutoring (almost all to first graders), but could not sustain the amount of tutoring provided to Abbottston and City Springs low achievers. A school in Philadelphia used a modified version of Success for All to work with limited English-proficient (LEP) Cambodian students, and also found very positive outcomes for these students and for non-LEP students (Slavin & Yampolsky, 1991). The evaluation of Success for All shows the potential power of a tutoring program that is integrated with a structured reading program. Evaluations of additional years will be needed to determine whether the program's goal of success for every child is realistic. Follow-up studies are needed to determine the validity of the program's assumption that success through the elementary grades will have long-term consequences, but the data collected to date clearly demonstrate the program's effectiveness when used at the beginning of students' school careers.

     Like Reading Recovery, Success for All articulates instruction with assessment. The Word Identification, Letter Identification, and Word Attack subtests of the Woodcock assess letter and word knowledge and phonics skills. The Durrell Oral Reading test asks children to read passages in a limited amount of time and answer comprehension questions. Given the emphasis on decoding, oral language, perceptual analysis, reading strategies, and prose comprehension, these measures correspond with the model of reading in the Success for All program.

     Like children in Reading Recovery, children in Success for All may be more familiar with the items that they are being evaluated on than are the controls because of the relationship between instruction and assessment.

     Finally, unlike the verification of Reading Recovery, the fidelity of program implementation in Success for All has not been documented. It would be important to know whether the model of reading is being appropriately implemented and if tutoring sessions look similar across sites. Also, if there is no consistency of delivery of instruction across tutoring sessions, what do the effect sizes mean? Qualitative implementation data need to be collected in order to validate the consistency in delivery of instruction and to determine how this translates into increased reading performance.

Prevention of Learning Disabilities
     Prevention of Learning Disabilities is a program developed by the Learning Disorders Unit of the New York University Medical Center that identifies first and second graders who are at risk for school failure and provides intensive instruction before they begin to fall behind in basic skills, Students involved in the program are screened in first grade using an instrument (SEARCH) that focuses primarily on neurological indicators of learning disabilities and on perceptual and general immaturity. Using diagnostic information from SEARCH, first graders are given lessons either individually or in small groups that attempt to strengthen their areas of perceptual weakness. The instructional interventions, called TEACH, are designed primarily to build perceptual skills, such as recognition, discrimination, copying, and recall, and are administered by certified teachers in 30-minute sessions three to five times per week.

Model of reading
     Unlike Reading Recovery and Success for All, the Prevention of Learning Disabilities tutoring program is based on a physiological view of teaming and learning disorders. As to reading itself, Silver and Hagin's (1990) model is based on the assumption that reading is a "complex process that must be analyzed according to component skills in order to understand the learning difficulty." However, these authors take a very atomistic view of reading and teaching reading. In compartmentalizing these reading skills, the goals are to identify those components with which the child is having difficulty and to teach to those specific skills.

     Silver and Hagin propose that children need to have four skills in order to read: pre reading skills, word attack skills, comprehension, and study skills, Pre-reading skills include the visual discrimination of letters, recognition of symbols in their correct orientation, the ability to organize symbols into groups, and several auditory skills. Word attack skills involve not only the use of phonics to figure out words, but also the identification of whole words using visual cues such as letter combinations. Comprehension involves having a rich vocabulary, being able to select the right meaning of a word, and making inferences. Study skills are described as the tools for acquiring information. These skills enable children to locate and select relevant elements within a sequence and organize the content of the text for later recall based on the goal of reading.

     Although Silver and Hagin proposed these components of reading, not all of these aspects are directly taught in the program. There is considerable emphasis on matching, copying, and recalling individual letters and words, and little emphasis on reading for meaning. Phonics are not systematically presented, but instead letter-sounds are reinforced. In total, the Prevention of Learning Disabilities program includes in its model the following components of reading: perceptual analysis of print, decoding, and oral language proficiency.

Structure of tutoring
     No coordination with the regular reading program is apparent in program descriptions. Children come to tutoring for 30 minutes, 3 to 5 days a week. Tutors, who are certified teachers, work on perceptual skills such as recognition, discrimination, copying, and recalling of information. There is no emphasis on reading connected text and no systematic presentation of phonics.

Results
     An evaluation of Prevention of Learning Disabilities was conducted in inner-city New York City classrooms (Hagin, Silver, Beecher, 1978; Silver & Hagin, 1979). Students were randomly assigned to experimental or control classes, and those in the experimental group received TEACH instruction for 2 years. Table 7 summarizes the findings. On reading measures as well as on perception measures, the experimental students performed substantially better than controls. In the same study, Silver and Hagin (1979) found that students who had a full year of TEACH performed better than those who had only a half year.

     In a similar study, Silver, Hagin, and Beecher (1981) found that third graders who received the TEACH intervention in first and second grade showed significantly greater performance in oral reading, word identification, and word attack skills (a measure that the ability to sound out words) when compared to a no treatment control group.

     Arnold et al. (1977) replicated the Prevention of Learning Disabilities program in inner-city and middle class schools in Columbus, Ohio. Using SEARCH, 86 first graders were identified as being at-risk for reading problems and were assigned to one of three groups: the TEACH intervention group, a group who received academic tutoring from a teacher, and a no-treatment control group. Students in the TEACH and regular tutoring group received tutoring for 30-minute sessions twice a week. Table 7 summarizes the findings. At the end of one year, the effects for both the TEACH intervention and the regular tutoring were minimal on the WRAT achievement test. However, at the end of the second year of the intervention, students in the TEACH group performed significantly better than the students in the regular tutoring and the no-Treatment control group on the WRAT.

     A more recent study by Mantzicopoulos, Morrison, Stone, and Setrakian (1990) found few effects for the TEACH intervention. In this study, first graders who were identified as at risk for reading failure by the SEARCH screen were assigned to three groups: a TEACH group, a phonics tutoring group, and a no-contact control group. In the phonics tutoring group, students were given phonics instruction, were drilled in phonics, and read phonemically regular books. This is in contrast to the TEACH group, which worked on visual-auditory discrimination activities. In both the TFACH and phonics tutoring groups, students received one-to-one tutoring for 30-minute sessions twice a week. The findings are summarized in Table 7.

     On reading measures and perceptual measures, students in the TEACH group did not perform any differently than the phonics tutoring group or the no-contact controls. Not surprisingly, the phonics tutoring group did show some significant improvement in word attack skills, compared to the no-contact control.

     Mantzicopoulos et al. (1990) suggest that one mason for the disappointing effects of TEACH was that the high attrition rate of their students produced a skewed simple distribution. Attrition, of course, is a factor in working with at-risk populations.

     As in Reading Recovery and Success for All, the measures used to assess this program are consistent with the model of reading and the instruction delivered in tutoring. The Word Identification and Word Attack subtests of the Woodcock and the Gates-MacGinitie Vocabulary test assess letter, letter-sound knowledge, and word knowledge. The WRAT (oral reading), SRA comprehension, and the Gates-MacGinitie Comprehension scales assess reading connected text and comprehension, although, these are multiple-choice tests and do not assess on line reading as does the Text Reading Level of Reading Recovery and the Durrell Informal Reading Inventory. Students in Prevention of Learning Disabilities are provided with minimal instruction in reading and answering comprehension.

     A final concern about the findings from Prevention of Learning Disabilities is that, as with Success for All there is no information about program implementation. There is no way of determining if the program implemented by Mantzicopoulos et at. (1990) was the same as Silver and Hagin's program or if variability in program implementation produced these different results. Qualitative implementation data need to be collected in order to determine if there is consistency across tutoring sessions and if what is proposed in the model of reading is being carried out in instruction.

Wallach Tutoring Program
     The Wallach Tutorial Program (Wallach & Wallach, 1976) is, like Reading Recovery and Success for All, based on the idea that students who fail to learn to read in first grade are seriously at risk, and that carefully structured tutoring intervention can prevent reading failure. in this model, students receive 30 minutes of tutoring per day for a year. Unlike Reading Recovery and Success for All, the Wallach model uses paraprofessionals as tutors. The tutoring is to students who score below the 40th percentile on a standardized reading test.

Model of reading
     According to Wallach and Wallach (1976), reading is a skill that "can be broken down into component parts; that these parts can be arranged in a cumulative, hierarchical manner such that learning of the latter parts builds systematically upon what has been learned already' (p. 56) and can be best learned by "systematically cumulating the mastery of component subskills' (p. 77). Wallach and Wallach propose that in acquiring these subskills, children must first establish competence 'in the recognition and manipulation of sounds,' and then acquire skill "in the use of the alphabetic code and in blending." Finally, they need to effectively apply 'these competencies in reading printed material."

     Because of this skills-mastery approach to reading, phonics are systematically presented in this program. Unlike Success for All, the phonics lessons are not coordinated with emphasis on reading connected text. Instead, reading comes after the letter sounds have been learned. Also, little consideration is given to metacognitive strategies. From Wallach and Wallach's point of view, the goal is to teach the students skills that they need to be readers. No attention is given to finding out the kinds of strategies the students are using and teaching new, more successful strategies.

     Much of this emphasis on the relationship of sound-symbol is a response to a finding in an earlier study'(Wallach, Wallach, Dozier, & Kaplan, 1977), which indicated that at the end of kindergarten, most of a sample of disadvantaged students but few middle-class students had difficulty recognizing phonemes in words read to them, such as knowing that man but not house starts with the mmm sound. Wallach and Wallach argue that disadvantaged students need to be explicitly taught letter sounds so they can serve as a foundation for acquiring other skills necessary for learning to read.

     In total, Wallach and Wallach includes only the following components of reading in their model: perceptual analysis, decoding, and oral language proficiency. They apparently believe that once the code has been cracked, oral language processes take over.

Structure of tutoring
     The Wallach and Wallach program has three parts. For about 10 weeks, children are taught to recognize initial phonemes in words read to them, to recognize letters, and to associate letters and phonemes. In the second stage, students spend 2 to 3 weeks learning to sound out and blend easy words. For the remainder of the year, the children learn to apply their skills to classroom reading materials. Thus, the Wallach model begins as a completely separate tutoring program (like Reading Recovery) his later begins to integrate tutoring with classroom instruction (like Success for All).

Two studies have evaluated the Wallach model. The results of these studies are summarized in Table 8.

     The first evaluation was a field test in two inner city Chicago schools (Wallach & Wallach, 1976). First graders who were identified at the beginning of the school year as low in "academic readiness" were randomly assigned to either tutoring or a no-treatment control.

     On the Spache Word Recognition Scale, the tutored students scored 5 months higher than the control (G.E. 1.8 vs. 1.3) with an effect size of +.64. On the Spache Consonant Sounds Test, the tutored students also outperformed the control group, with an effect size of +.66. On the Spache Reading Passage scales, there were apparent differences favoring the tutored students but these were obscured by a floor effect on the test (which does not measure below 2 grade equivalent of 1.6).

     A second study (Dorval, Wallach, & Wallach, 1978) evaluated the program in rural Roanoke Rapids, North Carolina. Students who received the tutoring were compared to similar students in the same school the previous year, to similar students in a comparison school who received the services of a full time reading aide in their regular reading class, and to other students in the same comparison school who received neither tutoring nor aides. At the end of the year, students took the group administered CTBS and were individually on the Spache Word Recognition and Reading Passages scales. The various control groups did not differ from one another, so they can be pooled.

     On the Spache Word Recognition Scale, the tutored students scored 8 months higher than controls grade equivalent 2.3 vs. 1.5). On the Spache Reading Passages, the tutored students were reading at a median grade equivalent of 1.8, while control students were at a median of 1.6., but again a floor effect may account for this small difference. On the CTBS, tutored students scored at the 56th percentile, comparison students at die 34th, for an effect size of +.75.

     Like the other programs, the measures used in assessing the Wallach and Wallach program match the model of reading and instruction. The Spache Word Recognition Test and the Spache Consonant Sound Test correspond to the emphasis on perceptual analysis and decoding. The CTBS also has sections that require word identification, analysis, and comprehension.

     Again, there are no implementation data on the Wallach tutoring program. It would be important to know if the different effect sizes found in Wallach and Wallach (1976) compared to Dorval, Wallach, and Wallach (1978) are the result of implementation differences or other program factors. The differences found on the Spache Word Recognition test and Spache Reading Passages in the two studies is large enough that differences in program implementation need to be considered.

Programmed Tutorial Reading
     Programmed Tutorial Reading is a highly structured tutoring program used with first graders who are in the lowest quartile on standardized reading tests. The program was originally developed by Douglas Ellson at Indiana University. The tutors for the program are paid paraprofessionals, volunteers, or parents. Students are tutored 15 minutes per day as a supplement to regular classroom instruction.

Model of reading
     Ellson and his colleagues describe reading as " a complex activity that, at minimum, includes oral or sight reading, phonics, and comprehension" (Ellson, Barber, Engle, & Kampwerth, 1965, P. 80). They describe a hierarchy in the important components of teaching reading in which sight reading has priority, followed by a phonetic analysis and synthesis, which they describe as not being necessary at the start, and, finally, comprehension of the meaning of visually presented words of sentences.

     Of all the programs described, Programmed Tutorial Reading is the least explicit regarding its model of reading is partly due to the fact that Ellson and his colleagues were primarily interested in testing the structure of this individualized programmed instruction and intended to extend this model to other content areas such as math. A the only components of reading that appear to be part of this model are perceptual analysis and decoding. They present no clear explanation or indication of how comprehension is taught.

     The curriculum in Programmed Tutorial Reading is designed on the principles of operand conditioning. Students proceed by mastering small, sequential steps in the reading process and are reinforced for correct responses. The primary emphasis is on acquiring sight words. Phonics is also systematically presented in the context of acquiring words. There is no emphasis on reading words in connected text with the goal of teaming new words and teaming to comprehend what is read. Mastering individual components is expected to build a repertoire of behaviors that coordinate into reading. How this occurs is not explicitly stated.

Structure of tutoring
      Students are cycled through a sequence of lessons on sight-reading, comprehension, and word analysis, which is repeated many times. Tutors are trained in specific strategies to present items, reinforce students for correct responses, and route students through the materials according to their responses. There is no coordination between tutoring and classroom instruction.

Results
      Several studies have evaluated Programmed Tutorial Reading, but only three of these have compared the program to control groups over meaningful time periods with non retarded populations. Table 9 summarizes the results of these studies.

     Ellson, Harris, and Barber (1968) evaluated Programmed Tutorial Reading of two duration's, over 2 full school year. Students were assigned to one of four tutored groups: Programmed Tutorial Reading for 15 minutes per day, Programmed Tutorial Reading for 30 minutes per day, an alternative tutoring program called Directed Tutoring for 15 minutes per day, and Directed Tutoring for 30 minutes per day. Then, a matched student was identified within the classroom of each tutored student. The students were first graders in 20 Indianapolis schools. Most of the schools served low income populations, but the students were selected to be representative of their schools and did not necessarily have reading problems. The Directed Tutoring program did not use the programmed materials or highly structured procedures used in Programmed Tutorial Reading, but used remedial and supplementary materials more like those typically used in classrooms or in remedial reading programs.

     The results (see Table 9)indicate strong effects of the 30-minute Programmed Tutorial Reading Program on tests provided along with students' Ginn basals (mean ES - +.52), but effects on the standardized Stanford Achievement Test were near zero, as were overall effects of the 15-minute per day program. Small positive effects were found for the 15-minute per day Directed Tutoring program, but (oddly) effects of the 30-minute Directed Tutoring treatment were slightly negative. Another study, by Ellson, Barber, Engle, & Kampwerth (1965), compared 15 minutes per day of Programmed Tutorial Reading for a semester to an untreated control group. in this case, moderate positive effects were found on the three measures used.

     The largest methodologically adequate study of Programmed Tutorial Reading was done by McCleary (1971) in Lenoir County, North Carolina, In this study, low achieving fire graders were matched and assigned to experimental or control groups. The experimental students were tutored for the entire school year for 15 minutes per day. Positive effects on the Ginn reading test were found for the sample as a whole (ES - +.40) and for the poorest readers (ES - +.37). In addition, retention's in first grade were 55% lower in the tutored group than in the non tutored group. Taken together, the evaluations of Programmed Tutorial Reading suggest that the program has positive effects on student reading achievement, but the effects are smaller and less consistent than those for the programs that use certified teachers.

     Two issues need to be addressed regarding the Program Tutorial Reading project. The study done by Ellson and his colleagues found different results than the McCleary study. These differences could be due to the fact that McCleary had a better experimental design than Ellson et al. However, another explanation which should be considered is that the differences found were the result of differences in implementation of the program at different sites. This is only speculative since implementation data are not available on either study.

     In Programmed Tutorial Reading, the assessment used to measure outcomes matches the model of reading in the program. The Ginn Total Word Analysis and Vocabulary tests assess word identification and decoding skills, two skills that are emphasized in this program. The Ginn Total Comprehension consists primarily of reading short sentences or passages and answering questions. This section of the Ginn primarily assesses lower level comprehension skills that are taught specifically in this program. Unlike the comparison group, students in the Programmed Tutorial Reading group were taught tasks in tutoring which were similar to those used in evaluating the program.

     The five tutoring programs discussed here vary enormously in models of reading, curriculum, tutoring methods, duration, integration with regular classroom instruction, and other characteristics. The studies are equally diverse in populations, measures, and procedures. However, some patterns can be perceived.

     First, programs with the most comprehensive models of reading, and therefore the most complete instructional interventions, appear to have larger impacts than programs that address only a few components of the reading process. Reading Recovery and Success for All include in reading instruction several components of reading such 2S perceptual analysis, conventions of print, error correction strategies, decoding, comprehension, error detection, and reading strategies. Moreover, they have comprehensive approaches to teaching the complex process of reading. In contrast, the Prevention of Learning Disabilities program which focuses only on budding specific skills related to the reading process produced less consistent comprehension outcomes.

     Second, it is not enough that programs simply use tutors. The content of the reading program in addition to the form of instructional delivery may be important variables. Ellson et al. (1968), for example, found the Programmed Tutorial Reading model to be significantly more effective than a standard 'directed tutoring intervention, and Arnold et 21. (1977) found the Prevention of Learning Disabilities (TEACH) program to be considerably more effective than 'regular tutoring." Mantzicopoulos et al. (1990) faded to replicate the findings of the earlier studies of Prevention of Learning Disabilities, but similarly found few effects of a "standard" phonics-based tutoring approach. An Ohio statewide study of Reading Recovery failed to find any positive effects of two alternative models of one-to-one tutoring (Pinnell et al., 1991).These findings, plus the apparent advantage of tutoring by certified teachers over tutoring by paraprofessionals, provides support for the proposition that for tutoring to be y effective it must improve the quality of instruction, not only increase the amount of time, incentive value, and appropriateness to students' needs (see Wasik & Slavin, 1990).

     Third, programs using certified teachers as tutors appeared to obtain substantially Larger impacts than those using paraprofessionals. Effect sizes for Programmed Tutorial Reading and the Wallach Tutorial Program generally fell in the range of +.20 to +.75, while those for the programs using certified teachers produced average effects from +.55 to +2.37 by the end of first grade. The teacher-delivered and paraprofessional-delivered models also differed in curriculum. Both the Wallach model and Programmed Tutorial Reading use highly structured, clearly described instructional materials, which in the latter program were explicitly patterned on programmed instructional methods usually designed for self-instruction. in contrast, the three teacher administered models rely on teachers' judgment, flexibility, and knowledge of how children learn.

     Only one program, Success for All, is designed to integrate completely with regular classroom instruction, and this program also produced some of the largest effect sizes. Although coordinating the tutoring sessions with classroom instruction is sensible in theory, empirical data need to be collected to determine its importance. The type of classroom instruction with which the tutoring was coordinated would also be an important factor. In addition, lack of consistency between how reading is presented in the classroom and how it is presented in tutoring may present a mismatch in the way reading is taught and result in confusion for the children. However, if Reading Recovery and Programmed Tutorial Reading were used both in the classroom and in tutoring, Reading Recovery might still have greater effects because its model of reading and delivery of instruction may be more effective. All of this remains to be determined in additional studies.

     Several studies evaluated the cumulative and lasting effects of one-to-one tutoring in the early grades. Studies of two Success for All schools (Slavin et al., 1992) found that as students continued into second and third grades, initial positive effects continued to grow. Similar cumulative effects were found for Prevention of Learning Disabilities in two studies (Silver & Hagin, 1979; Arnold et. al., 1977) but not in a third (Mantzicopoulos et al., 1990). Silver & Hagin (1979) also found that students who experienced Prevention of Learning Disabilities for a full year learned more than those who had it for a semester, and Ellson et al. (1968) found that gains were greater when students received 30 minutes per day of Programmed Tutorial Reading than when they received only 15 minutes.

     Because one-to-one tutoring (especially by a certified teacher) is expensive, the lasting effects of a approach are of great importance. Reading Recovery has been evaluated for lasting effects, and the results are positive but complex. On one hand, the raw score gains that students made on Text Reading Level in first grade have maintained through the end of third grade in two different cohorts (DeFord et al., 1988; Pinned, 1988). On the other hand, because standard notations of this measure increase each year, effect size estimates have diminished each year for both cohorts. A 1-year follow-up of Prevention of Learning Disabilities showed consistently positive effects for the third-graders for most measures with the exception of performance on the Gates-MacGinitie Comprehension test. The effects for reading comprehension decreased I year after the intervention.

     Two of the tutorial programs, Success for All (Slavin et al., 1992) and Programmed Tutorial Reading (McCleary, 1971) documented substantial reductions in retentions as a result of first-grade tutoring, and Success for All (Slavin et al., 1992) also showed reductions in special education referrals.

Is tutoring cost effective?
     It should not come as a surprise that one-to-one tutoring of primary grade students is effective. A more important question is whether it is effective enough to justify its considerable cost. One way to address this question is to compare tutoring to other expensive interventions. For example, experiments in Tennessee, New York City, Toronto, and Indiana have reduced class size by almost half. This is the same as hiring an additional teacher for each class, who could instead be used to provide one-to-one tutoring for 20 minutes per day to about 15 students. The best and most successful of these class-size experiments, a Tennessee statewide study, found a cumulative effect of substantially reducing class size from kindergarten to first grade of about +.25 (Word et al., 1990), less than that found in any of the tutoring models. A follow-up study I year later found lasting effects of 4 years of small classes to be only +.13 (Word et al., 1990). Other studies of halving class size have found even smaller effects (Slavin, 1999). The effects of having aides work in the classroom have been found to be minimal in many studies (see Scheutz, 1980; Slavin, in press); the same aides could be used as tutors using models designed for that purpose, or could be replaced by teachers for a greater impact.

     On the other hand, it is not yet established that a heavy investment in first grade will pay off in permanent gains for at-risk students. The Reading Recovery and Prevention of Learning Disabilities results hold out some hope for lasting gains, and the cumulative effects of Success for All also show promise for maintaining initial gains. Reductions in retentions and special education referrals, seen in two of the tutoring models, have both immediate and long-term impacts on the costs of education for low achievers. Substantial savings due to reduced retentions and special education placements have been shown for Reading Recovery (Dyer, 1992) and for Success for All (Slavin et al., 1992). However, if first-grade tutoring models prove to have long-term effects either without additional intervention (as in Reading Recovery) or with low-cost continuing intervention (as in Success for All), cost effectiveness win not be the only criterion for deciding to use these models. For if we know that large numbers of students can be successful in reading the first time they are taught, and that the success not only lasts but also builds a basis for later success, then educators and legislators may perceive an obligation to do whatever it takes to see that all students do in fact receive that which is necessary for them to succeed.

Future research
     In many ways, research on preventive tutoring models is in its infancy. Although the studies reviewed here clearly indicate a strong positive effect of well designed tutoring models, there are many important issues to be understood.

     On the programmatic side, one important set of questions concerns how much reading failure can be prevented using resources short of one-to-one instruction by certified teachers. Could one-to-two or one-to-three instruction be nearly as effective? Could forms of tutoring using paraprofessionals be devised that would be nearly as effective as forms requiring certified teachers? Must tutoring be done daily, or could it be done less frequently. How much time must be allotted to tutoring each day? These issues need to be empirically tested.

     More work is clearly needed on long-term effects of tutoring, not only on achievement but also on special education referrals and need for long-term remediation, critical elements in any consideration of cost effectiveness. Also, studies of alternative approaches to tutoring are needed. Successful models range from the phonemic rigidly prescribed Programmed Tutorial Reading, to the "learning to read by reading' emphases of Reading Recovery and Success for All, to the focus on specific perceptual deficits of Prevention of Learning Disabilities. In the studies, it may be that each of these types of approaches would be successful with different children, and that someday we may know which type of program will work best with children of a given profile.

     The issue of selection of assessment measures based on what is being caught in the programs has been - Recently, researchers have been calling for more authentic measures for assessing what children learn. One possible way of trying to establish some understanding across programs would be to assess children in each program on the same measures, both standardized tests and perhaps more importantly, ongoing literacy performance measures (see Taylor, 1990). This information would allow some cross-program comparisons and also help in determining generalizability of what is taught to other forms of assessment.

     A great deal of work is needed to understand why tutoring is effective. The rudimentary explanation offered in this article must be replaced by a far more sophisticated understanding of cognitive and motivational processes activated in tutoring that are not activated to the same degree (at least for at-risk children) in the regular classroom. Understanding how at-risk children learn to read in tutoring would contribute to an understanding of how at-risk children team in general; the tutoring setting provides an ideal laboratory in which the process of learning to read can be observed as it unfolds over time. Microanalysis of tutor/child discourse could contribute to our understanding of how children team to read (Green & Weade, 1985; Handerhan, 1990).

     This qualitative understanding of tutoring would also help address the important issue of implementation across tutoring sessions. Only Reading Recovery has attempted to assess implementation and the effect this has on outcome data. Understanding how instruction is delivered will also help in tutor training. Every tutor is different and brings to the tutoring session his or her own unique understanding of that child and reading. However, each program has specific prescribed theories of reading and how these theories translate into practice. The goal is to ensure that instruction is in concert with the model of reading and is consistent from one tutoring session to the next.

     As discussed in the introduction of this article, it would have been helpful to discuss the differences in each program's theory of reading. However, this was not possible because only Reading Recovery his made attempts at outlining a clear, coherent theory of reading. Instead, these programs take a pragmatic approach; that is, the evaluations focus on producing data to indicate that the programs work, not why they work. However, articulating a theory of reading based on empirical evidence is a valuable contribution to the field of reading. This area is ripe for theory development. It would be important to begin to understand how the interaction between the tutor and the student results in learning to read. Clarifying a theory of reading would add to a fundamental understanding of why the components included in a particular program make the program effective.

     Finally, several limitations to this type of research synthesis need to be addressed. First, when only tutoring programs are reviewed, research on other effective interventions for preventing reading failure is not addressed. Other class wide reading programs also have shown some effectiveness. Also, in a best evidence synthesis, programs are grouped together and examined in terms of effectiveness. However, each program has very distinct characteristics and has been tested on different populations. Although we have attempted to look at some specific similarities and difference among programs, a review of this kind does not examine the nuances of each program nor does it address the qualitative differences that exist in the tutor-child dyad. Also, to test the relative effectiveness of these programs, studies need to be conducted in which children are randomly assigned to alternative programs and a control group, and in which the children's success on a variety of measures is assessed.

     Although we want to know much more about how tutoring works and how to maximize its effectiveness (and minimize its cost), it appears from the research reviewed in this article that one-to-one tutoring is a potentially effective means of preventing student reading failure. As such, preventive tutoring deserves an important place in discussions of reform in compensatory, remedial, and special education. If we know how to ensure that students will learn to read in the early grades, we have an ethical and perhaps legal responsibility to see that they do so. Preventive tutoring can be an alternative for providing a reliable means of abolishing illiteracy among young children who are at risk for school failure.

REFERENCES

ARNOLD, L.E., BARNEBEY, N., MCMANUS, J., SMELTZER, DJ., CONRAD, A., & DESGRANGES, L. (1977). Prevention of specific perceptual remediation for vulnerable first-graders. Archives of General Psychiatry, 34,1279-1294.

BAUSELL, R., MOODY, W., & WALZL, F. (1972). A factorial study of tutoring versus classroom instruction. American Educational Research Journal. 9,591-597.

BLOOM, B.S. (1981).All our children learning. New York; McGraw-Hill.

CARROLL, J.B. (1963). A model for school learning. Teachers College Record, 64,723-733.

CARTER, L.F. (1904). The sustaining effects study of compensatory and elementary education. Educational Researcher, 13, 4-13,

CLAY, M.M. (1979). Reading: The patterning of reading Difficulties. Exeter, NH: Heinemann.

CLAY, M.M., & CAZDEN, C. B. (1990). A Vygotskian interpretation of Reading Recovery, In L. Moll (Ed.), Vygotsky and education. New York: Cambridge University Press.

CLOWARD, R.D. (1967) Studies in tutoring. Journal of Experimental Education, 36,14-25.

COUNCIL OF CHIEF STATE SCHOOL OFFICERS. (1987). Elements of a model state statute to provide educational entitlements for at-risk students. Washington, DC; Author.

COUNCIL OF CHIEF STATE SCHOOL OFFICERS. (1989). Success for All in a new century. Washington, DC: Author.

DEFORD, D., PINNELL, G.S., LYONS, C., & YOUNG, P. (1988). Reading Recovery.- Volume IX, Report of the follow-up studies. Columbus, OH: Ohio State University.

DEVIN-SHEEHAN, L., FELDMAN, R. S., & ALLEN, V L. (1976), Research on children tutoring children: A critical review. Review Education Research, 46, 355-385.

DORVAL, B., WALLACH, L., & WALLACH, M.A. (1978). Field Evaluation of a tutorial reading program emphasizing phoneme identification skills. The Reading Teacher, 31, 748-790.

DYER, P (1992). Reading Recovery: A cost-effectiveness and educational outcomes analysis. ERS Spectrum., 10(1), 10-19

EDMONDS, R.R. (1981). Making public schools effective. Social policy, 12, 56-60,

ELLSON, D G., BARBER, L., ENCLE. T L., & KAMPWERTH, L. (1965), Programmed tutoring A teaching aid and a research tool. Reading Research Quarterly 1, 77-127

ELLSON, D G., HARRIS, P, & BARBER, L. (1968). A field test of programmed and directed tutoring Reading Research Quarterly, 3, 307-367.

FARR, B., QUILLING, M, BESSEL, R., & JOHNSON, W (1987). Evaluation of PRIMETIME 1986-1987 Final Report Indianapolis, IN: Advanced Technologies, Inc.

FRESKO, B., & EISENBERG, T. (1983). The effect of two years of tutoring on mathematics & reading achievement. Journal of Experimental Education, 17,89-100.

GLASS, G., MCGAW, B., & SMITH, ML. (1981). Meta-analysis in social research. Beverly Hills, CA: Sage.

GOODMAN, K., WATSON, B., & BURKE, C. (1987). Reading Miscue Inventory, 2nd Ed. New York: Richard C. Owen. GREEN, J., & WEADE, R. (1985). Reading between the words: Social . Theory into Practice, 24, 14-21

GREENFIELD, P.M (1994). A theory of the teacher in the learning activities of everyday life. In B. Rogoff & J. Lave (Eds ), Everyday cognition: Its development in social context. Cambridge, MA: Harvard University Press,

GREENWOOD, C.R., DELQUARDI, J.C., & HALL, R.V. (1989). Longitudinal effects of classwide peer tutoring. Journal of Educational Psychology, 81, 371-383.

HAGIN, R.A., SILVER, A.A., & BEECHER, R. (1978). Scanning, diagnosis, and intervention in the prevention of learning disabilities: II. Teach: Learning tasks for the prevention of learning disabilities. Journal of Learning Disabilities 11(7), 54-57.

HANDERHAN, E. (1990). Reading instruction as defined 'successful' teachers and their first-grade students within an early intervention program. Education Abstracts International, No. AAC910512.

HARRIS. P. (1974). Programmed tutorial reading project. (Submission to joint Dissemination Review Panel) Washington, DC: U.S. Department of Education.

HUCK, C.S., & PINNELL, G.S. (1986). The Reading Recovery Project in Columbus, Ohio. Pilot Year, 1984-85. Columbus, OH: Ohio State University.

KENNEDY, M.M., BIRMAN, B.F., & DEMAUNE, R.E. (1906). The effectiveness of Chapter 1 services. Washington, DC: Office of Educational Research and Improvement, U.S. Department of Education.

LOYD, D.N. (1978). Prediction of school failure from third-grade data. Educational and Psychological Measurement, 38, 1193-1200.

LYONS. C.A. (1991). A comparative study of the teaching effectiveness of teachers participating in a year-long or 2-week inservice program. In the 40th Yearbook of the National Reading Conference, Learner factors l Teacher factors: Issues in Literacy research and instruction. Chicago: National Reading Conference.

LYONS, C., PINNELL, G S., DEFORD, D., MCCARRIER, A., & SCHNUG, J. (1909).The Reading Recovery Project in Columbus, Ohio. Year 3: 1988-1989. Columbus, OH: Ohio State University.

MADDEN, N.A., & SLAVIN, R.E. (1909). Effective pullout programs for students at risk. In R.E. Slavin, N.L. Karweit, & N.A. Madden (Eds.), Effective programs for students at risk. Needham Heights, MA: Allyn & Bacon.

MADDEN, N.A., SLAVIN, R.E., KARWEIT, N.L., & LIVERMON, B. (1987). Success for All: Teacher's manual for reading. Baltimore, MD: Johns Hopkins University, Center for Research on Elementary and Middle Schools,

MADDEN, N.A., SLAVIN, R,E, KARWEIT, N L., DOLAN, L.J. & WASIK, B A. (1992, April) Success for All Longitudinal effects of a restructuring program for inner-city elementary schools. Paper presented at the annual meeting of the American Educational Research Association, San Francisco.

MADDEN, NA.. SLAVIN, R,E, KARWEIT. N L, DOLAN L.J., WASIK, B A., SHAW A, LEIGHTON, M, &.MAINZER, K L (1991. April). Success for All Third-year results. Paper presented at the annual convention of the American Educational Research Association, Chicago.

MANTZICOPOULOS, P., MORRISON, D, STONE, E., & SETRAKIAN. W (1990). Academic effects of perceptually and phonetically based intervention for vulnerable readers. Paper presented at the American Educational Research Association, Boston.

MCCLEARY, E. (1971), Report of results of Tutorial Reading Project. The Reading Teacher, 24, 556-559

PINNELL. G S. (1985). Helping teachers help children at risk: Insights from the Reading Recovery program. Peabody Journal of Education, 62(3), 70- 85.

PINNELL. G S, (1988, April) Sustained effects of a strategy-centered early intervention program in reading. Paper presented at the annual convention of the American Educational Research Association, New Orleans.

PINNELL, G.S. (1989). Reading Recovery: Helping at-risk children learn to read. Elementary School Journal, 90, 161-182.

PINNELL, G.S., DEFORD. D.E., & LYONS, C.A. (1988). Reading Recovery and Early intervention for at-risk first-graders. Arlington, VA: Educational Research Service.

PINNELL, G S., LYONS, C.A., DEFORD, D.E., BRYK, A.S., & SELTZER, M. (1991). Studying the effectiveness of early intervention approached for first grade children having difficulty in reading. Columbus: Ohio State University, Martha L. King Language and Literacy Center.

PINNELL, G.S., SHORT, A.G., LYONS, C.A., & YOLNG, P. (1986). The Reading Recovery Project in Columbus, Ohio. Year 1: 1985-1986 Columbus, OH: Ohio State University.

POLLOWAY, E.A., CRONIN, M.E., & PATTON, J,R. (1986) The efficiency of group versus one-to-one instruction. Remedial and Special Education. 7(l), 22-30.

SCHEUTZ, P. (1980). The instructional effectiveness of classroom aids. Pittsburgh, PA: University of Pittsburgh, Learning Research and Development Center.

SCRUGGS, T.E., & RICHTER, L. (1985). Tutoring learning disabled students: A critical review. Learning Disability Quarterly, 8, 274-286.

SHAVER, J.P., & NUHN, D. (1971). The effectiveness of tutoring underachievers in reading and writing. Journal of Educational Research, 65, 107-112.

SILVER, A.A., & HAGIN, R.A. (1979). Prevention of learning disabilities. (submission to Joint Dissemination Review Panel) Washington, DC: U.S. Department of Education.

SILVER, A.A., & HAGIN, R.A. (1990). Disorders of learning in childhood . New York: John Wiley & Sons.

SILVER, A.A., HAGIN, R.A., & BEECHER, R. (1981). A program for secondary prevention of learning disabilities: Research in academic achievement and in emotional adjustment. Journal of Preventative Psychiatry, 1, 77-87,

SLAVIN, R. E. (1986). Best-evidence synthesis: An alternative to meta-analytic and traditional reviews. Educational Researcher, 15(9), 5-11.

SLAVIN, R.E. (1989). Class size and student achievement Small effects of small classes. Educational Psychologist, 24, 99-1 10.

SLAVIN, R. E. (in press). School and classroom organization in beginning reading: Class size, aides, and instructional grouping. In R. E. Slavin, N. L. Karweit, B. A. Wasik, & N. A. Madden (Eds.), Preventing early school failure: Research on effective strategies. Boston: Allyn & Bacon

SLAVIN, R.E., MADDEN, N.A., KARWEIT, N.L., DOLAN, L., & WASIK, B.A. (1990, April). Success for All, Effects of variations in duration and resources of a schoolwide elementary restructuring program. Paper presented at the annual convention of the American Educational Research Association. Boston.