Review of Research and Evaluation Related to Reading Recovery

 

Table 4: Review of Reading Recovery Studies
STUDY/SOURCE	*Reading Recovery. Pinnell, (1989) Elementary School Journal, 90 (2), pp. 159-181. Report of a study by Pinnell, Lyons, & DeFord
PURPOSE	(1) To explore whether RR could succeed with low-achieving children and (2) to determine whether those children maintained their gains (Summary of pilot year data and first full year data in Columbus, Ohio)
SAMPLE	First year study: Lowest achieving first grade children were randomly assigned either to RR or to a control group served daily in individual lessons taught by a trained paraprofessional. Both groups were compared with a random sample of average and high progress first graders (n=102) as an indication of average progress.
MEASURES	Diagnostic Survey (Clay, 1985) (all 6 tasks) Writing Sample Comprehensive Test of Basic Skills (2 subtests)
BASIC FINDINGS/ CONCLUSIONS	RR children performed better than control children (p<.05) on 7 of the 9 measures at the end of first grade. They compared well with the random sample group. In subsequent years, RR children continued to perform well on text reading. Effect sizes were reduced over the years.
COMMENTS	This study provided early evidence in the US of RR's immediate and long-term positive effects.
STUDY/SOURCE	*Comparing Instructional Models for the Literacy Education of High-Risk First Graders. Pinnell, Lyons, DeFord, Bryk, & Seltzer (1995) Reading Research Quarterly, 29 (1), pp. 8-39.
PURPOSE	To examine the effectiveness of RR as compared to three other instructional models for early intervention
SAMPLE	Lowest achieving first grade readers (N=324) were randomly assigned, within schools, to RR, a RR-like intervention with partially trained teachers, another skills-based individual intervention, group instruction by a RR teacher, or a control group.
MEASURES	Gates MacGinitie Woodcock Reading Mastery Dictation (Clay, 1993) Text Reading Level (Clay, 1993)
BASIC FINDINGS/ CONCLUSIONS	RR subjects performed significantly better than any other treatment and comparison groups on all measures. Essential differences were related to: • individual instruction • the lesson framework (combination of techniques) • teacher training
COMMENTS	RR emerged as most powerful of the tested interventions at the conclusion of the experiment and at the beginning of grade 2.
STUDY/SOURCE	Factors Affecting Progress in Reading: Key Findings from a Longitudinal Study, Rowe (1995) Literacy, Teaching and Learning, 1 (2), pp. 57-110.
PURPOSE	To provide information over a four-year period about factors affecting students' literacy development, with a particular focus on reading achievement, and to identify key factors affecting that development
SAMPLE	The sample included 5,092 students and 256 classes in 92 schools. The longitudinal design involved repeated measures nested within classes/schools and repeated measures on schools. The second design involved cross sections of students nested within schools that were changing over time.
MEASURES	Reading Achievement: Primary Reading Survey Test Test of Reading Comprehension English Profile Reading Bands
BASIC FINDINGS/ CONCLUSIONS	RR children benefited notably from participation. Some RR students were achieving beyond the 80th percentile level of their Non-RR peers. Lower limits of the distribution for achievement measures were higher for RR children. Gains of RR children seemed to have been sustained in Grades 5 & 6.
COMMENTS	RR appeared to be meeting its intended purpose for those students involved.
STUDY/SOURCE	An Evaluation of Reading Recovery. Center, Wheldall, Freeman, Outhred, & McNaught (1995) Reading Research Quarterly, 30, (2), pp. 240-263.
PURPOSE	To evaluate the effectiveness of RR in primary schools in New South Wales
SAMPLE	Low achieving children were randomly assigned to two groups: (1) RR (n=31) and (2) control (n=39), low progress students who had not entered RR by Nov. (3) A third group (n=39) consisted of students from 5 matched schools. By the end of the study sample sizes were (1) 23; (2) 16; (3) 32.
MEASURES	Clay's Diagnostic Survey (1985) Burt Word Reading Test Neale Analysis of Reading Ability Passage Reading Test Waddington Diagnostic Spelling Test Phonemic Awareness Test Cloze Test Word Attack Skills Test Woodcock Reading Mastery
BASIC FINDINGS/ CONCLUSIONS	At short-term evaluation (15 weeks), the RR scores were superior to control students on all tests measuring reading achievement but not on 2 of 3 tests of metalinguistic skills. At medium term (30 weeks) there were no longer significant differences between RR and control children on 7 of 8 measures. However, the book level test was significantly higher for the RR group.
COMMENTS	The authors suggested considerable caution when examining medium-term results because so few children from the original control group cohort were remaining (N = 16). Authors claimed that Clay's studies had excluded about 30% of children who were either removed or not discontinued from the program. Clay's 1979 data negate this claim. No children were dropped from her analyses. Clay responds to this claim in a letter to RRQ, Vol. 32 (1), 1997, p. 114.
STUDY/SOURCE	Phonological Processing Skills and the RR Program. Iversen & Tunmer (1992) Journal of Educational Psychology, 85 (1), pp. 112-126.
PURPOSE	To determine whether the RR program would be more effective if systematic instruction in phonological recoding skills were incorporated into the program. Three groups were compared: (1) children taught by teachers who received RR training; (2) children taught by teacher who received RR training that included phonological recoding skills as part of lesson; (3) children who received standard intervention
SAMPLE	Three matched groups of 32 at-risk readers each were formed: standard RR, modified RR, and standard intervention.
MEASURES	Diagnostic Survey (Clay, 1985) All 6 tasks Dolch Word Recognition Test (Dolch, 1939) Yopp-Singer Phoneme Segmentation Test (Yopp, 1988) Phoneme Deletion Test (Bruce, 1964) Pseudoword Decoding Task
BASIC FINDINGS/ CONCLUSIONS	The 2 RR treatment groups performed at very similar levels at discontinuing point. Both groups performed much better on all measures than children in the standard intervention group. Both RR groups often performed significantly better than classroom controls (especially on phonological segmentation and phoneme deletion). Results revealed that the modified RR group reached levels of performance required for discontinuing faster than the standard RR group.
COMMENTS	Authors acknowledged that both the standard and modified RR programs included explicit instruction in phonological awareness.
STUDY/SOURCE	Early Intervention in Children with Reading Difficulties: An Evaluation of Reading Recovery and a Phonological Training. Sylva & Hurry (1995) Literacy, Teaching and Learning, 2 (2), pp. 49-68.
PURPOSE	To evaluate the effectiveness of 2 different interventions (RR and Phonological Training)
SAMPLE	Almost 400 children from 7 English level authorities; diverse sample with inner-city over-represented nationally; 22 RR schools, 23 Phonological Intervention schools and 18 control schools
MEASURES	British Ability Scale Word Reading Neale Analysis of Reading Clay's Diagnostic Survey (5 tasks) Assessment of Phonological Awareness British Ability Scale Spelling Background information on each child
BASIC FINDINGS/ CONCLUSIONS	Intervention Year: Phonological Intervention effect was more specific than RR and not as secure. Only area where P.I. children significantly improved compared to control group was on test of phonological awareness. RR children made significantly more progress than control group on every measure of reading. Second Year: Phonological Intervention was less effective than RR and the effects narrowed.
COMMENTS	RR was the more powerful intervention and the more expensive. However, RR was particularly effective for socially disadvantaged children who are over-represented in special needs program. While cost of RR was more than other groups, the cost gap was narrowing and predicted to narrow further.
STUDY/SOURCE	The Development of Phonological Awareness and Orthographic Processing in RR Stahl, Stahl, & McKenna (under review)
PURPOSE	To determine whether techniques used in RR lessons promote progress in the metalinguistic areas of phonemic awareness and phonological recoding
SAMPLE	A total of 30 at-risk first-grade students were rank ordered. The lowest achieving children (n = 11) were entered into RR; a control group of 19 subjects eligible for RR was formed.
MEASURES	Observation Survey (Clay, 1993) Letter identification Dictation Task Pseudoword test (Stahl & Stahl, 1996) Yopp-Singer Test of Phoneme Segmentation (1995)
BASIC FINDINGS/ CONCLUSIONS	RR students made significantly greater improvement than control group on measures of phonological processing. Discontinued RR students demonstrated strategies similar to children in the alphabetic stage by the sixteenth week of grade 1.
COMMENTS	This study suggested that RR children acquire phonological awareness and phonological recoding within RR lessons. The inclusion of all RR participants and the utilization of measures other than Clay's responded to methodological concerns stated in other reports.
STUDY/SOURCE	Scaffolding in Reading Recovery Hobsbaum, Peters, & Sylva (1996) Oxford Review of Education, 22 (1) pp. 17-35.
PURPOSE	To explore the writing episode in the Reading Recovery lesson for aspects of scaffolding. Specific explorations: the structure of interaction whether interactions can be conceptualized as scaffolding procedures whether patterns of interactions change over time
SAMPLE	Data were drawn from a longitudinal study of 17 RR children and 7 teachers in different schools in London and the South of England.
MEASURES	Sources of data: Detailed field notes Transcribed audiotapes of lessons Writing books for each child
BASIC FINDINGS/ CONCLUSIONS	Researchers found that the scaffolding process does underpin teaching in the Reading Recovery intervention. They identified 3 phases within the program showing change across time: (1) teachers monitor and structure the learning within the task, (2) children independently identify their needs and teachers prompt to retrieve and make connections, and (3) children exercise increasing control over cognitive processes through the use of regulatory language. The phases indicate that the interactive framework within the writing portion of a RR lesson is a process of scaffolding learning.
COMMENTS	This study is an example of theoretical and pedagogical investigations within the RR context. Researchers in this study contributed to the theory of why RR succeeds and how writing fits into the success story. Distinctions were drawn between research on scaffolding within short-term experimental tasks where the goal is to solve a unique problem, and long-term, instructional contexts where the curricular goals are ever-increasing. (p.17)
STUDY/SOURCE	Descubriendo La Lectura: An Early Intervention Literacy Program in Spanish Escamilla, (1994) Literacy, Teaching and Learning, 1 (1), pp. 57-70.
PURPOSE	To examine whether the Descubriendo La Lectura (DLL) program achieved results with Spanish-speaking first graders equivalent to Reading Recovery programs in English
SAMPLE	All Spanish-speaking first graders who were receiving literacy instruction in Spanish in 6 elementary schools in an urban Arizona district were subjects for the study (N = 180). Four Schools had DLL and two did not. Subjects fell into 3 groups: children served by DLL (N = 23); a control group of children needing the service and not receiving it (N = 23): and a comparison group of all remaining 134 children in the sample.
MEASURES	Spanish Observation Survey (fall and spring) Aprenda Reading Achievement Test (fall and spring)
BASIC FINDINGS/ CONCLUSIONS	At the end of grade 1, DLL children had not only caught up to the comparison group on the Spanish Observation Survey, but surpassed them. Differences were statistically significant on all tasks except text reading. DLL students also significantly outperformed the control group (p <05) on all measures. On Aprenda, when standard scores were connected to percentiles, only the DLL and control groups made gains. In May, the DLL group was at the 41st percentile, the comparison group at the 31st percentile and the control group at the 28th percentile. When using average bands to determine how the 3 groups compared to the average progress of all first graders, progress of the comparison and control groups lagged statistically behind that of the DLL group. While findings were encouraging for DLL students, the study raised some concerns regarding quality of Spanish reading instruction within regular bilingual classrooms.
COMMENTS	This study provided positive evidence for the potential of the DLL program, or Reading Recovery in Spanish. Results demonstrated that the program has a great deal of promise in assisting children who are struggling to become literate in Spanish. The author cautioned that the study was limited by sample size and encouraged additional studies. She also called for studies to explore the sustaining of initial gains across grade levels and as children transition from Spanish to English instruction. Note: Annual national data on DLL outcomes have supported Escamilla's early findings. Longitudinal studies are in progress. Published accounts of these studies are forthcoming.
STUDY/SOURCE	*Ohio Fourth Grade Proficiency Results for Two Cohorts of Students The Ohio State University
PURPOSE	To determine the performance of former RR students on tests of proficiency at fourth grade
SAMPLE	Subjects were children served by RR in 1991 (Reading Test N = 2714; Writing Test N = 2813) and in 1992 (Reading Test N = 2994; Writing Test N = 3002). Of all districts eligible for the study, 69% reported data.
MEASURES	Ohio Test of Fourth Grade Proficiency
BASIC FINDINGS/ CONCLUSIONS	For the 1991-92 cohort, 71% were at or above proficiency in reading and 75% in writing. For the 1992-93 cohort, 76% were at or above proficiency in reading and 69% in writing.
COMMENTS	This study included all children served by RR, not just discontinued children.
STUDY/SOURCE	*Massachusetts Follow-up Studies: Grades 2 & 3 Lesley College
PURPOSE	To compare the subsequent achievement of former RR children with achievement of randomly selected non-RR peers
SAMPLE	Sample included randomly selected former RR discontinued students and randomly selected non-RR children in grades 2 & 3 (3 years of data). Number of subjects in groups ranged from 74 to 220.
MEASURES	Test of Oral Text Reading Story Retelling Slosson Test of Word Recognition Dictation Task Gates MacGinitie Classroom teacher ratings
BASIC FINDINGS/ CONCLUSIONS	Across all 3 years, RR children compared well with random sample group on oral reading and retelling measures. RR children attained mean Gates-MacGinitie scores within average bands of achievement (non-RR group). On the Slosson and Dictation Task, RR children were not within average bands of random sample achievement in grade 2 but they were within an average band in grade 3. Class teachers perceived most children to be average on literacy behaviors.
COMMENTS	This study supported the notion that somewhat tentative literacy performance immediately after the intervention seems to become stronger in subsequent years.
STUDY/SOURCE	*A Four-Year Follow-Up Study of Reading Recovery Children in New York State: Preliminary Report. Jaggar & Simic (1996) New York University, School of Education
PURPOSE	To determine the long-term impact of RR by determining whether children who successfully complete the program in first grade sustain their gains in reading at the end of grades 2 and 3 when compared to a grade-level random sample of their peers
SAMPLE	Four cohorts of RR children who successfully discontinued across 4 academic years were tested at end of grade 2 and three cohorts were tested again at the end of grade 3. Data were collected on a total of 1,596 RR second graders (74% of the total who had successfully discontinued across the 4 years at participating sites) and a total of 604 third graders (58% of those discontinued in the 3 cohort years in participating sites). Randomly selected groups of grade level peers (1,236 second graders and 402 third graders) were chosen to compare progress with the RR groups.
MEASURES	Text Reading Slosson Oral Reading Test (Sort-R)
BASIC FINDINGS/ CONCLUSIONS	The mean text reading levels of the 4 cohorts and total group of second graders were, in all but 1 case, slightly higher than the means of the random sample group. Mean text levels of the 3 cohorts indicate that RR students continued to progress and perform as well as their random sample classmates. RR children's mean achievement levels on the word recognition test (SORT-R) reflected 'average' or expected performance for students at the end of grade 2 and slightly higher than average performance at the end of grade 3. Almost all of the RR children scored at or above grade level on Text Reading. A large majority of RR students (69% of second graders and 72% of third graders) scored at or above level on SORT-R, practically the same as the random sample group.
COMMENTS	Authors concluded that RR children, after becoming average or better readers in first grade, continue to learn along with their classmates and make significant progress in reading after the specialized tutoring is discontinued.
STUDY/SOURCE	*Texas Follow-Up Study: Grades 2, 3, 4 Texas Woman's University
PURPOSE	To examine subsequent literacy performance of RR children and to compare their performance to a random sample of grade level peers
SAMPLE	Sample included randomly selected discontinued RR children in 48 schools in grades 2, 3, and 4 and randomly selected children never served by RR in same schools (N ranged from 88-103 across the groups).
MEASURES	Gates MacGinitie Reading Tests (1989) Texas Assessment of Academic Skills (TAAS) in grades 3 and 4 only Test of Oral Text Reading Written Retelling Classroom teacher questionnaires
BASIC FINDINGS/ CONCLUSIONS	Scores on standard measures (Gates and TAAS) increased across grade levels. In grade 4, approximately 70% of former discontinued RR children had scores considered average or meeting passing criteria. There were no significant differences between the 2 groups on tests of text reading at grades 3 and 4 or on retelling measures at all levels. Classroom teachers perceived most former RR children as performing within average range on literacy tasks.
COMMENTS	This study supported the notion that somewhat tentative literacy performance immediately after the intervention seems to become stronger in subsequent years (cross-sectional study).
STUDY/SOURCE	*Texas Longitudinal Studies A and B (Preliminary Report) Texas Woman's University
PURPOSE	To examine subsequent literacy performance of 2 cohorts of children who participated in RR in grade 1 (through grade 4) To compare the literacy performance of these RR children with a cohort of their classroom peers
SAMPLE	The 2 studies represent 2 cohorts of children. In Study A, 150 schools were randomly selected from 698 RR schools in Texas in the spring of 1995. Within each school, 2 children were randomly selected in 2 categories: discontinued RR children and grade-level peers not served by RR. In Study B, 50 schools were randomly selected from the more than 800 RR schools in Texas in the fall of 1995. Within each school, 12 children were selected as possible RR subjects (those placed in RR at beginning of year and the children in 'next group up' if needed to get sample to 12). Six children were randomly selected from across the first-grade population in these schools.
MEASURES	Observation Survey (Clay) for fall and spring data on both groups in grade 1 Gates-MacGinitie Reading Test - for fall and spring data on both groups in grade 1 and then at end of each year of study Test of Oral Text Reading Texas Assessment of Academic Skills (TAAS) in grades 3 and 4 Classroom teacher questionnaires
BASIC FINDINGS/ CONCLUSIONS	Preliminary findings across both studies indicate that oral text reading performance of RR children improves across years, matching that of random sample peers in grade 3. In both studies, RR children were generally perceived to be average by their classroom teachers. Gains on the Gates-MacGinitie were similar for both groups each year, indicating that RR children continued to progress with their peers. In grade 3, 69% of the former RR children passed the reading subtest of the Texas Assessment of Academic Skills, a robust measure of reading performance. In grade 4, 78% of the former RR children passed this measure. Responding to Shanahan and Barr's recommendation, Study B included entry data (beginning of grade 1) for both groups on the Survey and on Gates, providing evidence of dramatic differences between the 2 groups. The mean entry stanine for RR children on Gates was stanine 1; the mean for the random group was stanine 4. Gains of RR children during the intervention year far exceeded those of the random group.
COMMENTS	Study A will be completed in 1998 and Study B in 1999 after subjects at the end of grade 4 again take the Gates, a test of oral text reading, and TAAS reading and writing tests. Classroom teachers will again complete questionnaires. Early results appear promising for the children who began first grade as the lowest-achieving in their grade cohort. Study B includes implementation data to be reported in final analyses.

*Indicates that author is directly involved with Reading Recovery.