Does Extensive Reading Actually Work? The Evidence

If a colleague asks whether extensive reading really works, you need more than "studies show". This article is written for that sceptical reader. It sets out the main meta-analyses and landmark studies: what each measured, with whom, and how big the effects were. It then lists the weaknesses a careful reader should know about. What the findings mean for pupils and parents is covered in a separate article. This one is about the research itself.

Does extensive reading work?

Yes, on the measures researchers have used. Extensive reading improves reading comprehension, reading rate and vocabulary, with small to medium average effects. The evidence is strongest for adults learning English at college or university. For pupils aged 8 to 16 it is positive but thinner. For writing and speaking it is thin.

Most of this research reports results as an effect size, written d. It expresses the gap between two groups, or between before and after, in standard deviations. A d of 0.5 means the average reader in one group scored half a standard deviation above the average reader in the other.

For scale, Jeon and Day (2016) compare their overall result with other meta-analyses in language learning. Strategy instruction averaged d = 0.49 and corrective feedback d = 1.16. Extensive reading, at d = 0.57, sits among them. For what these gains mean in a classroom, see What Are the Benefits of Extensive Reading?

The main studies at a glance

The five studies below are the ones this article relies on most. Three are single studies and two are meta-analyses.

What the two meta-analyses found

Nakanishi (2015) included 34 studies: 32 research articles and two PhD dissertations. Together they gave 43 effect sizes and 3,942 participants. He found a medium effect against comparison groups (d = 0.46) and a larger one from before to after (d = 0.71). He concluded that extensive reading improves reading proficiency and should be part of language curricula.

Nakanishi also asked whether age and program length mattered. Jeon and Day (2016) summarise his answer: the effect was larger for older participants and for longer programs.

Jeon and Day (2016) set stricter rules. They kept only studies of English as a second or foreign language in which learners chose their own books. Their 49 studies gave 71 samples. The average effect was d = 0.57 against control groups and d = 0.79 from before to after.

By outcome, reading rate showed d = 0.83, comprehension d = 0.54 and vocabulary d = 0.47. The authors note that these differences were not statistically significant. Reading rate rested on only ten samples.

What made a difference

• Setting. Effects were larger where English is a foreign language (d = 0.65) than where it is a second language (d = 0.38).
• Program form. Extensive reading built into the curriculum showed d = 0.91. A stand-alone reading course showed d = 0.24, but that figure comes from a single study.
• Length. Short, medium and long programs gave similar effects (d = 0.51, 0.59 and 0.60). Only four samples came from long programs.

The landmark single studies

Meta-analyses average many studies together. Three single studies are worth knowing on their own.

Elley and Mangubhai (1983) is the classic school study. Year 4 and Year 5 pupils at eight rural Fijian schools read high-interest story books for about 30 minutes a day. The books replaced the drills of an audio-lingual program, which four control schools kept.

After eight months, the book groups had progressed in reading and listening comprehension at twice the rate of the control group. In the second year, the gap spread to every language test, including writing and grammar. Effect sizes on those tests ranged from 0.76 to 1.18.

Beglar, Hunt and Kite (2012) is a careful, year-long study of reading rate. It compared one intensive reading group with three pleasure reading groups of Japanese university students over a year. All pleasure reading groups gained more speed, and the heaviest readers gained most. Comprehension stayed high, and simplified texts gave larger gains than unsimplified ones.

Waring and Takaki (2003) is a sobering vocabulary study. Fifteen learners read one 400-headword graded reader. Most target words were not learned. The meanings of words met fewer than eight times were not remembered three months later. The authors concluded that building vocabulary needs a massive amount of graded reading.

Who was studied, and why it matters for schools

This is the biggest gap for a school reader. Jeon and Day (2016) grouped their samples by age. Thirty were adults at college level or above. Fifteen were adolescents in secondary school. Only six were children in primary school.

The average effect was largest for adults (d = 0.70), then children (d = 0.52), then adolescents (d = 0.35). The authors suggest that adults bring more background knowledge and vocabulary to their reading. They also say more research with younger learners is needed to confirm the pattern.

Place matters too. Most of their studies came from Asia and the Pacific, with Japan, Taiwan and Korea the most common foreign-language settings. Results from Japanese university students may not transfer neatly to a class of 11-year-olds elsewhere.

The Fiji study is the best-known evidence with young pupils. It is also more than forty years old, and its comparison group followed an audio-lingual drill program. For how the method itself should run in a school, see the extensive reading guide.

Weaknesses a sceptic should know

None of these weaknesses means that extensive reading does not work. They mean you should be precise about what the evidence shows.

1. Short studies. Jeon and Day (2016) found only four samples from programs longer than an academic year, so long-term effects are largely unknown.
2. Adult samples. Most samples were adults at college or university, and few were children.
3. Mixed definitions. Jeon and Day (2016) point out that some studies in Nakanishi (2015) used assigned reading of long texts, not free choice, and some were not about English. Two studies with the same label may test different things.
4. Narrow outcomes. The meta-analyses measured reading. They tell you little about writing or speaking.
5. Small samples. Single studies are often small. Waring and Takaki (2003) tested 15 learners.
6. Different comparison groups. Control groups did intensive reading, drills or ordinary lessons. An effect against one of them is not an effect against all.

Then there is advocacy. Krashen (2004) argues that free voluntary reading is one of the most powerful tools in language education. He also argues that pupils in school free reading programs do as well as or better than pupils taught by traditional methods. His book is a strong case built from many studies, but it is an argument. The meta-analyses support a more modest claim.

What the evidence lets you claim

When you present extensive reading to colleagues or leaders, match each claim to its evidence:

• Safe: extensive reading improves reading rate, comprehension and vocabulary, with small to medium average effects.
• Safe: the learners who read most tend to gain most.
• Careful: effects for children and teenagers are positive but rest on few studies.
• Avoid: claims that it improves speaking, or that it can replace the teaching of grammar and writing.

A trial like this is the honest answer to thin evidence. How to Run a Pilot of a Reading Platform covers how to set one up so the result means something. Why enjoyment keeps pupils reading long enough to gain is covered in the benefits of reading for pleasure.

How iRead handles this

No reading platform can settle the research questions above. What iRead offers is a way to test extensive reading with your own pupils. A school can run a free pilot with one class for 4 to 6 weeks, with dashboards for attendance and retention. Every story, quiz and game is levelled to the CEFR, from A1 to C2. A pilot that short will not show a year's gains, but it will show whether pupils read.

Key takeaway

Tell colleagues exactly what the research shows: steady gains in reading, mostly measured in adults over a year or less. Then gather your own evidence with your own pupils.

Frequently asked questions

Sources

1. Nakanishi, T. (2015). A meta-analysis of extensive reading research. TESOL Quarterly, 49(1), 6-37. doi.org/10.1002/tesq.157
2. Jeon, E.-Y., & Day, R. R. (2016). The effectiveness of ER on reading proficiency: A meta-analysis. Reading in a Foreign Language, 28(2), 246-265. doi.org/10.64152/10125/66901
3. Elley, W. B., & Mangubhai, F. (1983). The impact of reading on second language learning. Reading Research Quarterly, 19(1), 53-67. doi.org/10.2307/747337
4. Beglar, D., Hunt, A., & Kite, Y. (2012). The effect of pleasure reading on Japanese university EFL learners' reading rates. Language Learning, 62(3), 665-703. doi.org/10.1111/j.1467-9922.2011.00651.x
5. Waring, R., & Takaki, M. (2003). At what rate do learners learn and retain new vocabulary from reading a graded reader?. Reading in a Foreign Language, 15(2), 130-163. doi.org/10.64152/10125/66776
6. Krashen, S. D. (2004). The Power of Reading: Insights from the Research (2nd ed.). Libraries Unlimited. www.abc-clio.com/products/F2619P/