A close look at the research on how to teach writing.
Writing is a key skill. The ability to put words down on paper is a critical skill for success in college and career, but few students graduate high school as proficient writers. Less than a third of high school seniors are proficient writers, according to the National Assessment of Educational Progress (NAEP). Low-income, and Black and Hispanic students fare far worse and less than 15 percent scoring proficient.
So how can students learn to write better? What tools do they need to become better at making an argument on paper? In this document, we’re largely focused on argumentative writing, which is central to, well, just about every aspect of modern life, from emails to workplace memos to empowered citizens.
Teaching argumentative writing involves teaching both argumentation and writing. This means there’s research on how to teach writing, there’s research on how to teach argumentation, and there’s research on the intersection of the two.
Earlier research in this topic often referred to it as “persuasive” writing. More recent research seems to prefer and use the term “argumentative” writing. One reason why is that “argumentative” is seen as a broader term. Persuasion is a particular goal that argumentative writing might have. But argumentative writing is seen to encompass many possible goals: persuasion, but also analysis, collaborative problem-solving, and conceptual clarity.
A series of research syntheses and meta-analyses make broadly overlapping recommendations on best practices for teaching writing of all types across grade levels.
There are important nuances further below, but the general picture of best practice based on experimental and quasi-experimental research looks something like the following:
Learning to Write
Writing is a complex skill involving the interaction of lots of different kinds of knowledge (e.g., “genre knowledge”, “subject matter knowledge”, “writing process knowledge”). This complexity makes it challenging to study.
There is also an increasing recognition of the discipline- and genre-specific nature of writing. It’s not about teaching a single, relatively discrete skill — “writing” — but about teaching ways writing in science, ways of writing in history, ways of writing in literature. Writing in any genre is not a fully transferable skill—it requires contextual knowledge (about the purpose and the audience of the work), content knowledge (about the subject of the work), and writing process knowledge.
Methods and Approaches
Before diving into the research, a quick primer on the research methods in this space.
Experimental and quasi-experimental work in this area has been the subject of many meta-analyses, with slightly differing emphases. These are fairly comprehensive quantitative summaries of the work in this area, but a substantial number of the underlying studies have methodological problems of one sort or another: e.g., no random assignment, no controlling for the Hawthorne effect, no measures of whether the intervention was implemented with fidelity, or simply no descriptions of the control condition.
More recent studies tend to be of higher quality, and also result in higher effect sizes than older studies (suggesting that earlier studies may have been understating the effect of some of the proposed interventions).
Meta-analyses look at broad swaths of different studies and group studies into categories to be able to calculate the effects of individual types of interventions. These categories are not always mutually exclusive. For instance, the distinction between “writing process” models (which focus on the process of writing — prewriting, organizing, drafting, editing, revising, etc.), “self-regulation” models (which focus on setting writing goals and teaching students to monitor their writing and editing to achieve those goals), and “peer writing” models (where student co-write and edit each others work) can overlap substantially. In other cases, single categories encompass quite a diverse set of practices. For instance, the “strategy instruction” model includes interventions involving pre-writing activities, peer writing, persuasive writing structures, etc.
Another complexity is how writing outcomes are measured. Rubric-based instruction may help students improve their writing when measured by the rubric, but not necessarily when evaluated through other means.
Research evaluating the popular 6+1 trait writing model, for example, shows little improvement in student outcomes when those outcomes are not measured with the 6+1 writing rubric.
A final qualification on the meta-analyses is that studies used as the basis for the meta-analysis often use differing control conditions. This makes estimating effect sizes particularly challenging because the baseline comparison is different across studies. And there’s usually not enough studies in each category to look at control condition as its own factor.
Main Takeaways About Learning To Write
Teachers don’t assign enough writing.
One recent survey of teachers and students on the current state of writing instruction in middle schools and high schools suggests that students simply don’t write often enough, that writing assignments tend to be too short, and that standardized tests (such as AP exams) incentivize teachers to dilute writing instruction to match test standards. These behaviors are important because students need lots of opportunities to write (and incorporate feedback) to become confident writers.
The survey’s results, however, suggest some improvements in teacher practice compared to 40 years ago. Compared to an assessment made in 1979-80, teachers are giving students more questions that ask for more original analysis, they’re incorporating more research-backed strategies like process writing, explicit strategy instruction, and collaborative writing in the curriculum, and they use technology more often.
These findings dovetail with another recent large-scale survey of 361 high school teachers across several disciplines. Again, the main finding is that students don’t compose texts that are long enough (most writing practice asks students to compose texts one-paragraph in length or less), and that few writing assignments as students for analysis and interpretation. Moreover, most teachers say they are not prepared to teach writing (71% say they had either no or minimal college-level training in the subject in their teacher preparation programs).
Although experts argue that argumentative writing should happen across all disciplines, English teachers have often shouldered the main responsibility for teaching writing skills of all sorts (at least as judged by the surveys mentioned earlier). This emphasis, however, clashes with a tradition of emphasizing the literary tradition in English language arts classes (e.g., fiction and narrative), rather than argumentation. One way of providing students with more opportunities to write is to have teachers across many disciplines assign writing tasks. This has been a common aim of several reforms.
Time and expertise seem to be the main obstacles to teachers assigning more writing. Teacher workloads are high, while providing meaningful feedback on long pieces of writing takes considerable effort. Teachers report simply not having enough time to assign long pieces of writing, provide feedback on them, and providing lots of opportunities for revision.
Many teachers also report not being very confident in their ability to teach writing, not having the training or experience to teach writing well, or both. This lack of expertise (or at least lack of confidence) also likely discourages teachers from assigning more challenging pieces of writing.
Research has also revealed a number of best practices. The following best practices all improve student outcomes, and are listed in approximate order of effect size (largest to smallest) based on several recent meta-analyses of existing studies
Explicit strategy instruction helps students.
One of the most robust findings in the literature is that students benefit from explicit strategy instruction. These writing strategies can range from ideation strategies (e.g., brainstorm and select which ideas are main ideas and which ideas are supporting ideas) to drafting strategies (e.g., imitate an author’s writing style for a paragraph, write several sentences and pick out the best one) to monitoring strategies (e.g., asking yourself if there’s a clear beginning, middle, and end?).
That said, authors of the survey view the overall amount of class time spent on research-backed strategies to be insufficient: for instance, about 3 minutes out of every 50 minute English class period would be devoted to one of the most support strategies—explicit writing strategy instruction.
Teachers should illustrate each strategy and facilitate students’ independent use of each strategy over time. For instance, a pre-writing activity like brainstorming might be initially presented to the students: the teacher can explain what the strategy is, why writers do it, and present a model of the practice to students. Then, the teacher would have students practice brainstorming in small groups, with occasional teacher guidance. After that, students would brainstorm on their own, while the teacher would provide occasional support. Finally, the teacher would encourage students to use brainstorming completely independently (revisiting aspects of the strategy if needed).
Self-regulation strategies can improve outcomes.
Setting writing goals and helping to monitor their progress towards these goals is also a well-established strategy with empirical support. Goals might be something like “add three new ideas to your paper when revising it” or “address both sides of an argument, providing three reasons supporting your position and two reasons supporting the opposing position.”
The most popular self-regulation strategy is the SRSD (self-regulation and strategy development) model. This integrates both explicit strategy instruction and self-regulation strategies into a classroom context. Teachers don’t just introduce a strategy—they track student progress towards mastery in using the strategy over time, aiming to motivate students and build self-confidence. The SRSD approach uses mnemonics, graphic organizers, and other scaffolds, in addition to developing prior knowledge about specific writing genres as students are taught them. Evidence suggests that the program is one of the most effective classroom programs, but outcome measures stop short of long trajectories (stopping at six months).
Process Writing improves student outcomes. Process writing involves getting students involved in pre-writing, organizing, drafting, revising, re-drafting, and editing their papers. It’s about engaging students in a relatively long-term writing process involving multiple drafts, rounds of peer and/or teacher feedback, and a finalized product. It’s also helpful to have students explain why they made their edits.
Process writing seems to be a key part of many successful writing interventions (such as SRSD), but the evidence on the efficacy of process writing alone is somewhat contradictory.
Most meta-analyses that evaluate process writing as a separate category of intervention find that process writing is a best practice and has substantial effect sizes. This effect, however, seems modulated by professional development — when teachers are not taught how to use process writing approaches effectively, the benefit disappears. A meta-analysis that focused solely on process writing found modest effects for students of average ability, but no significant effects on writing outcomes for struggling writers, which contrasts with qualitative research suggesting the benefits of process writing for struggling writers in particular. Many of the studies in the process-writing-specific meta-analysis also had weak study designs making the claim that process writing doesn’t help struggling writers at best preliminary.
Feedback Drives Achievement.
Feedback is an essential part of nearly any process writing or self-regulated approach. Feedback is best when it comes regularly during the process of writing longer written work, and when it’s linked to established learning objectives.
For feedback to be effective, students must be motivated to use feedback to improve and they must know how to use the feedback effectively. The first is a challenge of classroom culture. The second is a challenge of student knowledge.
Surveys suggest that many students feel like they don’t have enough guidance on what to do with the feedback they receive. One study at the undergraduate level suggests that the main factor influencing whether students implemented changes based on feedback was whether the problem described by the feedback was understood. Problem understanding increased when teachers identified where the problem was, proposed some solutions, and gave a summary of the students’ work.
Even generic automated feedback to scientific arguments can improve students subsequent arguments (e.g., feedback that reads something like “You haven’t provided enough evidence for your argument. You should add more evidence.” The effects, however, were most pronounced for students that had high scores to begin with, suggesting that receptiveness to feedback may also be playing a role.
One of the values of outside feedback (from either peers or teachers) is that students are more likely to make deeper, meaning-related changes in response to outside feedback than they are when making individual revisions.
Peer Collaboration Can Make A Difference.
Most process writing approaches involve peer collaboration, either during the pre-writing, drafting, editing, and/or revising process. Peer collaboration has a consistently positive effect on writing outcomes. One meta-analysis found seven studies exploring peer writing specifically (as distinct from process writing and other strategies); all seven had positive effects.
Peer feedback offers several proposed benefits. Students often perceive peer feedback as more understandable (and therefore, more actionable) than teacher feedback. Students often feel that teacher feedback is insufficient or unhelpful, and, in higher education at least, students report that peers provide more (and more helpful) feedback. Peer feedback is also a way of providing more overall feedback to students (on early or intermediate drafts of a piece of writing, for instance), as it requires less teacher time. Some reports even suggest that peer feedback encourages students to redouble their efforts because of the social pressure to do well on the assignment.
There are, however, also some challenges to peer feedback. Peers can offer valuable feedback, but it’s not quite up to the standards of expert feedback. For instance, in a study where peers used a predefined assessment guideline, peer and expert feedback was structurally comparable, although peers suggested fewer changes, gave more positive judgments and gave less evidential support for their negative judgments.
Another challenge is to develop a classroom culture where feedback is perceived to be helpful and non-threatening. In one study, English Language Learners report liking peer feedback, but also being concerned about hurting their partners feelings or lacking enough knowledge to be helpful. In that study, students preferred online forms of feedback over face-to-face feedback.
Teaching students how to give valuable feedback is a critical step. Not only does this make the peer feedback more valuable for the writer, but it also helps the peer reviewer develop their own revision skills. For instance, a teacher might give peer reviewers some questions to guide their revision process. Questions like: Is the piece suitable for the intended audience—does it use the right tone, the right kinds of vocabulary words and text complexity? Is the overall structure of the argument clear? Etc.
In some cases, learning to provide peer feedback seems to have more benefits to students than receiving feedback. In a study from 2009, one group of students gave feedback on written work throughout the semester while the other group received feedback. Both groups improved over the course of the semester, but the givers improved more than the receivers when they had no prior experience in giving peer feedback.
Foundational skills and knowledge enable students to engage deeply with the material.
At the earliest ages, foundational skills like spelling, handwriting, and typing help develop overall writing skill. Even creative visualization exercises for young children seem to improve writing outcomes. Development of these skills should be linked to writing practice (for instance, having students review whether they spelled the words in their essay correctly). Similarly, teaching students to combine simple sentences into more complex ones seems to improve overall writing quality.
Students also need to know genre-specific writing structures and topic-specific vocabulary. For instance, young children have a better understanding of the narrative genre compared to other genres of writing (argumentative writing, scientific reports, poems). When students don’t know enough about the genre expectations, they can’t produce work that’s recognizably within the genre (e.g., when asked for a science report or an argument, they might produce something that’s more like a story).
Genre learning is something that continues throughout careers and later schooling (e.g., imagine learning about the “business case” genre, the “legal memo” genre or the “economic theory paper” genre); it’s part of the overall development of meta-cognitive knowledge. Explicitness about genre expectations also helps marginalized groups (who may not be as familiar with the genres they’re asked to produce) participate more fully.
Word processing lets students edit and revise more easily.
All of the evidence suggests that using word processors (instead of paper and pencil) to write papers improves writing outcomes, although this effect seems small in the earliest grades [replace link with slavin or wwc doc]. In those grades, students insufficient familiarity with word processing software and low typing skills may get in the way. Word processing seems to help students regardless of ability (although dictation can also help learning disabled students).
The commonly accepted explanation is that word processing makes editing and organizing a paper easier, and so encourages revision.
This finding (and the presumed explanation) contrasts sharply with evidence suggesting that teachers frequently ask students only to type up their final product using word processors, but not to draft and edit their paper using word processors.
Although seventy-five percent of teachers report having students turn in their final draft through a computer, less than half of students use word processing for drafting, editing, and collaborating on written work. This is particularly notable given the presumed mechanism behind the advantage of using word processors over paper-and-pencil to draft and edit: word processors make editing and re-organizing far easier.
More Frequent Practice. Perhaps the simplest intervention that seems to improve writing outcomes is more practice (especially given the evidence on the little amount of practice that students typically get at long-form writing). Establishing writing routines (for example, having students write for fifteen minutes more a day than they currently do) and ensuring that the writing tasks are interesting to students — that the writing has some larger relevance or purpose — seem like important keys to having students write more.
Multiple drafts improve essay quality.
More practice alone, however, does not guarantee improvement. One striking example comes from a two-year program to improve middle-schoolers argumentation practice. In the intervention condition, students explored a series of issues through individual work, collaborative peer work, and debate. In the control condition, students explored the same issues, but through teacher-led whole-group discussions. Students in the control group wrote a total of 14 essays over the course of the year. Students in the intervention group wrote a total of 4 essays. Nevertheless, essays from students in the intervention group displayed more nuanced and advanced argumentative structures.
Writing can also improve reading.
When students write about what they have read, their comprehension for the read material improves. When students improve their writing skills, their reading skills improve, too. And simply increasing the amount of writing that students do has a beneficial effect on overall reading skill as well. These interventions aimed at improving reading skill through writing are all set against reading instruction conditions. In other words, in some cases writing seems to improve reading outcomes over and above the impact of direct reading instruction.
Researchers have proposed several explanations for the deep relationship between reading and writing. They both involve common processes. Spelling, for example, helps students read and write. Sentence-combining helps students write complex sentences and comprehend more complex sentences. The writer also has to re-organize what she’s read, leading to deeper reflection of the material. And finally, writing provides insight into what a writer might be thinking, which helps students interpret written work.
Evaluations of programs that integrate reading and writing programs show positive writing outcomes. These programs typically involve professional development and peer collaboration on reading and writing assignments. Part of the explanation may be that reading helps students anticipate what readers may believe as they write. Reading texts that students will write about also increases their prior knowledge of the issues.
Grammar instruction is important but not that helpful for more advanced writers.
Although there are several fundamental skills improve overall writing skills, such as spelling, typing, and sentence-combining, generally speaking grammar instruction does not have a large impact. Most meta-analyses find that grammar instruction actually has a negative effect size. That’s because in many studies grammar instruction is the control condition, and the intervention conditions tend to do better than the control conditions. Not surprisingly, grammar instruction does help students on grammar-related measures. But these measures have little to do with good writing (at least, according to most researchers).
Research on common practices among exceptional writing teachers largely complement or confirm the best practices listed above. Exceptional teachers:
Still, Some Open Questions About Learning to Write
As we;’ve outlined above, there’s a lot that we do know about how to teach writing. But there’s also a lot that we do now know
For instance, at what age should students begin formulating structured arguments (either through discussion, as preparation for later writing assignments, or in writing itself)? Under certain contexts, even young children can reason well, which suggests that early, more comprehensive practice could substantially improve student reasoning in the long term. But there are also developmental considerations. There’s relatively little evidence on this question.
What’s the proper role of writing rubrics? Rubrics can help students set goals and perceive differences between their work and an ideal piece of writing. But such rubrics can become formulaic substitutes for deep engagement with the issues.
How should argument-structuring scaffolds be used? There’s little doubt that argument diagrams can help students formulate counter-arguments and rebuttals, and that they help lower-ability students in particular. Whether the benefits of such scaffolds extend beyond the assignments that use the scaffolding, however, is an open question: in some transfer tasks, students who have been using argument diagrams don’t write any better arguments than those who haven’t. They can, however, also limit reformulations and of the argument by making the issue seem “black and white” as opposed to “gray” and promote linear, less open approaches, particularly for higher ability students. How to fade these scaffolds over the long trajectory of student development is largely unexplored.
How should we be evaluating written argumentation skills? Researchers have used a variety of methods to distinguish “good” essays from “bad” essays, but the diversity of measurement approaches makes it challenging to compare studies. Articulating counter-arguments has been a major focus, but so have measures of elaboration and organizational structure. In some cases, it’s difficult to tell what aspect of the essays have improved because the descriptions are simply too vague.
What are interesting cases and authentic contexts for students in different classroom settings? A big hurdle for teaching written argumentation seems to be engagement. Students who feel that the writing has some larger purpose (and have some relevant experience) produce stronger written arguments, and presumably become better arguers because of it. But, as far as I can tell, there is no collection of vetted cases across the variety of classroom situations (and levels of prior knowledge) that teachers encounter.
Successful writing intervention programs involve extensive professional development, but what kinds of professional development is best, and how should successful professional development programs be scaled?
Are there other complexities to teaching students with learning disabilities that haven’t surfaced because of lack of research? The research on teaching writing to students with learning disabilities is particularly sparse. A relatively recent review of the research (2009) found only 9 studies that met their inclusion criteria for a total of 31 students.
- Ulrich Boser
Interview with Pooja Agarwal
Ulrich Boser on TEDx Talk
Interview with Ken Koedinger
The Learning Curve publishes articles about how people learn. Please reach out with any ideas for articles on the research on learning