Experimental Evaluation

Experimental Evaluation 2017-06-26T13:57:54+00:00

Experimental Evaluation

An Experimental Evaluation of Reading Recovery

Center, Y., K. Wheldall, K., Freeman, L., Outhred, L., & McNaught, M. (1995). Reading Research Quarterly, 30, 240-263.

Center, Wheldall, Freeman, Outhred, and McNaught evaluated the effectiveness of Reading Recovery schools in New South Wales. Low-achieving children were randomly assigned to either Reading Recovery (n=31) or a control group (n=39) of low-progress students who had not entered Reading Recovery by November. A third group (n=39) consisted of students from five matched schools. By the end of the study, sample sizes were 23, 16, and 32 respectively. Measures used were Clay’s Diagnostic Survey, Burt Word Reading Test, Neale Analysis of Reading Ability, Waddington Diagnostic Spelling Test, Phonemic Awareness Test, Cloze Test, Word Attack Skills Test, and Woodcock Reading Mastery.

At post-test evaluation (15 weeks after the pre-test) an independent assessment showed that Reading Recovery students scored significantly higher on all tests measuring reading in context and in isolation. Of the eight measures reported, the only ones that did not differ significantly were a cloze test and a phonemic awareness measure. At short-term maintenance (15 weeks after the post-test) the Reading Recovery control group still scored significantly higher than the control group on six of the eight measures, including Clay’s test reading measure and several standardized measures of text and word reading. At this point the Reading Recovery group also scored significantly higher than the control group on phonemic awareness.

The study’s published results for medium-term maintenance (12 months after the post-test) appear to have errors. The authors report “no overall significant group effect, F(8,30) = 0.262, p = .0268)” (p. 253). There appear to be several typos and errors in this statistical statement beyond the inclusion of an additional closing parenthesis. An F value of 2.62 would match the probability level of .0268. Since the authors state that “significant multivariate results (alpha = 0.05) were followed up by univariate pairwise multiple comparisons (alpha = 0.01)” (p. 250), the conclusion should be that the MANOVA revealed an overall significant group effect in favor of Reading Recovery. Still, the only univariate result was for text reading. The authors point out that the reduced difference between the Reading Recovery and control groups found in the 12-month follow-up could be due to the fact that 15 of the 31 control group students (probably those with the lowest scores) had been eliminated from the control group to receive Reading Recovery instruction.

The study provides strong, independent replication of the pattern of results found in other research and in the U.S. national evaluation data for all participating students. The authors state that their “results clearly indicate that low-progress students, exposed to 15 weeks of Reading Recovery, outperformed control students on Clay book-level and Burt Word Reading tests and on all Set 2 tests which measure reading and writing words in context and isolation” (p. 256). Despite a number of qualifications related to metalinguistic measures, the article reports independently measured and extremely large effect size for text reading, 3.05 and 1.55 for post-test and short-term maintenance respectively (p. 253).

For more information:

Download Six Reading Recovery Studies: Meeting the Criteria for Scientifically Based Research (PDF)

Excerpt from What Evidence Says About Reading Recovery (2002). Columbus, Ohio: Reading Recovery Council of North America.