January 2014: A Meta-Analysis of Writing Instruction for Elementary Students

Overview

What are the most effective ways to teach writing to students in grades 1–5? Although there are many opinions and testaments about best practices, a scientific approach that uses research for decision making would rely on empirical studies of outcomes associated with different writing interventions. Researchers would then try to synthesize the cumulative evidence across multiple studies. Such empirical approaches to synthesis are commonly called “meta-analysis.” These methods have been discussed previously in Education Research Matters—for example, our January 2013 and February 2013 editions. In a meta-analysis, a search uses defined parameters for including and excluding studies. The characteristics of the sample, methods, and interventions are coded, using a system the researchers develop. Then, averages of the size of the effects of different interventions can be computed. Meta-analysis also examines whether the effects are homogeneous; if not, sources of heterogeneity are evaluated, such as differences in the sample, research designs, and interventions.

In the meta-analysis of writing interventions discussed here, the authors asked, “What writing treatments improve the quality of writing produced by students in the elementary grades?” (p. 880). They focused on studies with random assignment to different interventions (“true experiments”) as well as quasi-experiments that used other methods for creating a comparison of treatments. Only about a third of the studies employed a randomized control design. The key was that the study had a “control” condition against which the intervention could be compared and a formal measure of writing quality, thus allowing the researchers to calculate an effect size for the intervention or instructional practice. They focused on interventions involving students in grades 1–6 in typical school settings, excluding students served primarily in special schools (e.g., deaf education).

Results

Using these criteria, the researchers identified 115 qualifying studies. To further understand the interventions, the researchers categorized the interventions into 15 instructional practices and programs, based on their predominant characteristics, which are listed here with the mean effect size. A common rubric for understanding effect sizes is that effects of .25 are considered small but practically important. Effect sizes of .25 to .80 are considered medium, and effects higher than .80 are considered large.

The authors list the categories of interventions, the number of studies in each, their effect sizes (weighted by sample size), and confidence intervals for each in Table 2 (p. 885), which is partially reproduced here:

Implications

This meta-analysis shows the importance of teaching written expression as an explicit set of strategies that support prewriting, composing, and editing, and that include a component for goal setting, monitoring, and other aspects of self-regulation. This approach had the largest effect sizes in the meta-analysis. Components of this type of instruction also had moderate to large effect sizes when compared to nontreatment groups.

One example of such an approach is the Self-Regulated Strategy Development program, developed by Steven Graham, Karen Harris, and their colleagues. This program has four main components, which are taught explicitly: (1) pick a topic, (2) organize a plan, (3) modify the plan when writing, and (4) self-regulate, or set goals and monitor progress. This program is an overarching “metacognitive” approach to text generation. A practitioners’ guide to writing by the developers of Self-Regulated Strategy Development (Harris, Graham, Mason, & Friedlander, 2008) has specific lesson plans not only for composition and text generation, but also for transcription, which is essentially handwriting and spelling. These lesson plans are organized into short 20- to 30-minute lessons to integrate into language arts instruction. For example, a practitioner can download writing lesson plans and then organize the writing activities around the novels and related materials introduced in reading instruction.

Other writing activities were found to be effective, including stimulating creativity and developing mental imagery. The key here is the explicit nature of the instruction, usually with teacher modeling. Assessment of written products by adults was much more effective than assessments by peers, although having peers work in a group on writing activities was effective. Forming goals, adding a self-regulation component, focusing on background and other prewriting activities, and teaching about text structure and genre were all effective instructional activities. Many of these activities are incorporated into Self-Regulated Strategy Development.

Teaching handwriting and spelling can be controversial, but this meta-analysis clearly supported their effectiveness, especially considering that to be included, the studies had to measure the quality of written expression. This finding is not surprising because efficient transcription skills allow the writer to allocate more resources to the process of composing. Forming letters and spelling become automatic. In children who struggle with written expression, handwriting and spelling can be bottlenecks. Interestingly, in the longitudinal study of Berninger et al. (2006), transcription skills in grade 4 were strongly related to the quality of compositions in grade 8. Explicit instruction in transcription in the early grades should not be discounted.

Clearly ineffective for written composition was explicit teaching of grammar. The most common approach to teaching writing in school, the less-explicit process approach, was found to be effective but modest in relation to other activities and even in relation to more explicit programs. Extending time for writing was not strongly effective, nor was the teaching of word processing skills.

There are three important caveats to this meta-analysis. First, although many components had large numbers of studies yielding effects, some activities had relatively few studies. Effect sizes are less reliable with smaller numbers of effects and may change as the research base grows. Second, there was a paucity of experimental studies with random assignment to conditions. These studies tended to give lower effect sizes, a common finding when true experiments and quasi-experiments are compared. Finally, the overall pattern of effect sizes is larger than commonly observed in educational interventions in reading and math. A key issue is the nature of the comparison group, which in many of these studies was “business as usual,” or the school’s typical instruction. If writing is not effectively taught, a common situation in many schools, effect sizes will be larger because there is no adequate instructional condition against which the results of the intervention can be compared.

The most important implication is that written expression can be taught effectively. Instructional practices that teach strategies explicitly in small groups, with teacher modeling and demonstration, and with attention to transcription are most promising. Such approaches are easily integrated into language arts instruction and parallel strategies often used to teach reading.

References

Berninger, V. W., Abbott, R. D., Jones, J., Wolf, B. J., Gould, L., Anderson-Youngstrom, M., . . . Apel, K. (2006). Early development of language by hand: Composing, reading, listening, and speaking connections; three letter-writing modes; and fast mapping in spelling. Developmental Neuropsychology, 29, 61–92.

Harris, K. R., Graham, S., Mason, L. H., & Friedlander, B. (2008). Powerful writing strategies for all students. Baltimore, MD: Paul Brookes.