A meta-analysis of recent research in the teaching of writing: Workshops, computer applications, and inquiry
Dianne L. Atkinson
School: Purdue University (0183)
Degree: Ph.D. (Educational Psychology)
Date: 1993; pp: 167
Advisor: Asher, J. William; Lauer, Janice M.
Source: DAI-A 54/09, p. 3354, Mar 1994
Subjects: Education, Language And Literature (0279); Education, Technology (0710)
ProQuest Document Number: 304104055
UMI Number: AAT 9403647
Comparison-group research with a focus on workshops, computer applications, or the teaching of inquiry skills as instructional programs in writing are summarized, using a quantitative integration based on the calculation of mean effect sizes. The meta-analysis procedures are based on the work of Hedges and Olkin (1985). The discussion relates the results of other meta-analytic work, with a special emphasis on the first meta-analysis of research in the teaching of writing completed by Hillocks (1986).
The identification of comparison-group research was made through the study of over 700 abstracts. Any study using writing samples as the outcome measure and for which it was possible to calculate the ratio of differences between group means and their pooled variances (‘effect size’) was included in the meta-analysis.
Of the three instructional strategies meta-analyzed, the most effective were workshop approaches with a mean weighted effect size of.519, followed by the teaching of inquiry skills with a mean weighted effect size of.452, and lastly, by computer applications with a mean weighted effect size of.318. The cumulation is based on 55 individual effects sizes, 30 workshop effect sizes, 20 computer effect sizes, and 5 inquiry effect sizes.
Additionally a number of methodological differences were examined, in order to ascertain whether these differences were associated with differences in effect sizes. Almost without exception, the study-level characteristics were associated with differences in effect size, and so contribute to understanding the heterogeneity observed within primary research populations. Study level differences considered included: duration of instruction, grade level of participants, and other student population characteristics. Design characteristics were also examined, including the random assignment of students to treatment, the number of measures, the number of raters, and whether more than one teacher was included.