EEF Blog: Improving reading comprehension through strategy instruction

In this blog, Robbie Coleman – a secondary school English teacher and a Senior Associate at the EEF – looks at the debates and recent developments about explicitly teaching thinking strategies in lessons.

In my previous blog-post, I declared myself a big fan of Barak Rosenshine. Rosenshine’s writing is sharp and illuminates some interesting ideas around the murky educational debates between teaching skills versus knowledge, and the relative merits of teacher- and student-led activities.

Rosenshine’s support for explicit teacher-led skills instruction shows why a binary split between “traditional” and “progressive” teaching is unhelpful. Rosenshine proposed that a central aim of education is to help novices become experts – and a fundamental characteristic of an expert is the way they use skills (or ‘cognitive strategies’) to deal with challenging or unstructured tasks.

In addition to boosting students’ knowledge, Rosenshine believed that effective teachers should make these strategies explicit to students and practise them in their lessons. 

For example, Rosenshine argued that in addition to teaching knowledge, such as new vocabulary, English teachers should spend time teaching comprehension skills by modelling the habit of regularly pausing to summarise the key ideas in a text.

I want to explore some of the developments in the field of cognitive strategy instruction since 1997, when Rosenshine’s paper,The Case for Explicit, Teacher-led, Cognitive Strategy Instruction, was published. 

During my research, I also discovered that Barak Rosenshine died in May this year so I was keen to understand how his ideas had influenced new research.

The headlines

Lots of new research into strategy instruction has been completed in the last two decades. The overall conclusion from this has stayed the same – that explicitly teaching strategies improves learning. 

In the 1990s, Rosenshine and colleagues found that the average effect size for explicitly teaching strategies was 0.32, based on the average from 56 studies (summarised in Rosenshine, 1997). This is equivalent to around +4 additional months’ progress over the course of a year, using the EEF’s metrics. A 2010 meta-analysis, which focused on strategies for supporting reading comprehension, found a very similar overall effect, 0.36 (Davis, 2010).

This consistency is very encouraging. Although it might sound dull that the same effect was repeated over several decades, such a pattern is rare. Few areas of education have been studied in as much depth as strategy instruction, and where they have, it is usually the case that as the quality of evaluations improves, the average impact decreases – often dramatically.

What’s new?

Beyond this overall strengthening of the research base, however, there have been two specific developments worth highlighting.

First, there is much more support available to teach strategies in lessons. In 1997, Rosenshine concluded that teaching students to generate their own questions about a text improved reading comprehension. But he also highlighted the lack of more detailed information about applying that idea, for example, about the types of questions that should be taught.

Andy Tharby’s excellent Making Every English Lesson Count (2017) is an example of the type of support teachers now have. For example, it includes an extensive section on question prompts. 

Likewise, in its guidance report on primary literacy published earlier this year, the EEF gave advice on teaching strategies for reading (including prediction, clarifying and summarising) and writing (planning, drafting, revising). They also gave additional guidance on how teachers can model and scaffold new strategies.

Second, researchers now commonly use the term ‘meta-cognition’ when discussing some of the skills that were previously classified under the broader heading of ‘cognitive strategies’. Meta-cognition is thought about thinking, including planning how to complete a task, monitoring learning, and evaluating whether a particular approach to a task was effective.

For example, students who are taught to summarise a passage have been provided with a cognitive strategy that can be used to improve comprehension. 

However, students who are taught both to summarise and to generate questions may also develop the meta-cognitive ability to plan how best to complete the task by selecting between the two strategies. At the end of the task, they may also be able to evaluate whether the strategy they chose was effective.

Studies where students receive instruction to develop meta-cognition alongside cognitive strategies have consistently been found to have a strong positive impact on learning. 

A 2014 meta-analysis^ of 58 studies published between 2000 and 2012 concluded that outcomes were almost always improved when students were taught more than one strategy and the inclusion of an explicit focus on meta-cognition boosted academic achievement.

For example, when reading an unseen poem, in addition to explaining a “5Ws” questioning strategy and the value of re-reading, a teacher might model the process of trying to ‘crack’ the poem, including narrating the process of selecting and evaluating the strategies used to do so.

An outstanding question

While there appears to be broad consensus about the value of explicitly teaching thinking strategies, one area of contemporary debate regards how much time should be spent on their instruction.

Writing  in 2014 and again this year, Daniel Willingham argues that strategies are a “one-time improvement” – once students are familiar with some basic strategies, further instruction is unlikely to provide additional benefit. He recommends that strategy instruction should be explicit, but brief – specifically suggesting that two weeks should be enough.

This prescription has been challenged by some, including the academic Timothy Shanahan, who  argued that the evidence is not yet strong enough to enable a robust comparison between brief and more extended studies of strategy instruction. 

In a discussion of this evidence  in 2015, Willingham appeared to concede that two weeks was a “lowball” estimate, though the recommendation was restated in his 2017 book. 

Further study of duration, and of the integration between teaching knowledge and skills, would be valuable.

* A version of this blog was first published by the Chartered College of Teaching here.

^ Donker, A., De Boer, H., Kostons, D., Dignath van Ewijk, C. and Van der Werf, M. (2014). Effectiveness of learning strategy instruction on academic performance: A meta-analysis in Educational Research Review 11 (1-26).