Extensive Reading in English as Foreign Language by Adolescents and Young Adults: A Meta-Analysis

Stephen Krashen (2007)

International Journal of Foreign Language Teaching 3(2): 23–29. (2007) Also in: Free Voluntary Reading, S. Krashen, 2011. Libraries Unlimited

A review of studies of extensive reading adolescents and young adults studying English as a foreign language revealed a strong and consistent positive effect for both tests of reading comprehension and cloze tests. Students provided with more access to reading (titles per student) did significantly better on tests of reading comprehension, but there was no relationship between access and performance on cloze tests.

Despite the consistently positive results of extensive reading programs, there still seem to be doubts as to its effectiveness: Study after study says it works, but very few language programs have adopted it. This paper takes another look at the research, focusing on studies of extensive reading using adolescent and adult students of English as a foreign language. Older students of English as a foreign language (EFL) students are an appropriate group to study for both practical and theoretical reasons. Mastery of English is, of course, crucial for nearly all activities that involve any kind of international communication. Also, focusing on foreign language removes one potential confound, the easy availability of English outside the classroom.

   There are two goals of this meta-analysis. The first is to determine whether free reading has an overall positive effect. The second goal is to determine the factors that contribute to the overall effect. We will be able to accomplish the first goal and take some steps toward reaching the second goal.

   The studies examined here were all published in professional jour- nals or conference proceedings. Studies were done in Taiwan (Yuan and Nash 1992; Sims 1996; Sheu 2005; Hsu and Lee 2005, 2007; Lee 2005a, 2006; K. Smith 2006, 2007; Liu 2007), the Philippines (Lituanas, Jacobs, and Ranayda 1999), Japan (Mason and Krashen 1997), and Yemen (Bell 2001).

   In all studies, time was set aside for self-selected reading. Studies in which a significant percentage of reading was assigned are not included here (e.g., Lee 2005b; Lao and Krashen 2000), and only studies that included reading tests (see below) are included.

Two kinds of reading tests were considered, cloze tests and tests of reading comprehension. The impact of extensive reading was determined by computing effect sizes. The usual formula for the effect size is the mean of the experimental group minus the mean of the comparison group, all divided by the pooled standard deviation, based on posttest scores. This formula was used here, and when possible the effect of the pretest was taken into consideration by subtracting the effect size of the pretest from that of the post-test. The mean effect size for all studies is a measure of the overall impact of extensive reading.

   Data on two factors that could influence the impact of extensive reading was included in the analysis: Access to reading material and duration of treatment.

Access

For first language development, access to reading material has been consistently shown to be a predictor of how much students read and how well they read (Krashen 2004). In this analysis, book access was represented by total titles available to students, and the number of book titles per student.

Duration

Previous reviews have shown that longer SSR programs tend to be more effective than shorter programs (Krashen 2001). Duration is included here in terms of the number of weeks, months, or academic years the program lasted. This is a crude measure because it does not consider the amount of time set aside for reading each day or week.

   A number of other factors are undoubtedly relevant to predicting the impact of extensive reading programs, but their inclusion will await additional studies, when methods are developed for representing their contributions quantitatively.

Results

Table 1 presents data on access, duration of the program, and the results of cloze tests and reading comprehension tests.

Table 1 
Mean titles per student and effect size

	mean	sd	n
titles/S	17.5	32.1	18
Cloze	0.79 (.73)	0.46	13
RC	1.06 (.88)	0.87	9

n = number of studies
( ) = adjusted for sample size

Overall, extensive reading programs clearly produce positive effect sizes. All 13 effect sizes for cloze tests, and all nine effect sizes for reading comprehension were positive.

   For unknown reasons, sample size was negatively correlated with measures, with studies with fewer subjects producing larger effect sizes (for cloze tests, r = –.45, p = .11; for reading comprehension (r = –.81, p < .011). For this reason, weighted means were calculated, resulting in adjusted means of .73 for cloze tests and .88 for reading comprehension.

   Details of the studies, as well as notes on effect size calculations, are presented in table 2. In some cases, effect sizes were calculated for each experimental group in a given publication. This was not possible in other cases. Thus, the average values calculated here should be considered approximate. For studies with no results listed, it was either not possible to calculate effect sizes from the data provided or cloze tests or reading comprehension were not used as measures.

   In Liu (2007), sample size and titles per student were calculated on the basis of students and titles per class. More than one class was involved in these studies. The Liu (2007) effect size is an average calculated from five experimental classes and 12 comparison classes over four years.

   For Yuan and Nash (1992), the effect size in table 2 is the average of three methods of calculating the score (from the t-score of gains, comparison of mean gain scores, pre- and post-tests).

   The reading comprehension test used in K. Smith (2006) also included usage and listening, and was given five months after the course ended.

   In Sims (1996), two different experimental classes were used. The number of titles was estimated from the total number of books: 700 books were provided, and “most” were separate titles (Sims, personal communication).

Table 2
 Access, duration, and effect sizes

study	N	titles	titles/S	Duration	ES Cloze	ES RC
Yuan & Nash 1992	37	200	5.4	one year	0.38
Sims 1996	30	550	18.3	one year		0.81
Sims,1996	30	550	18.3	one year		0.65
Mason retakers	30	100	3.3	one sem	0.702
Mason Jr college	31	200	6.4	one year	1.47
Mason university	40	200	5	one year	1.11
Mason: response L1	40	550	13.75	one year	0.24	0.61
Mason: response L2	36	550	15.28	one year	0.63	0.48
Lituanas et al. 2001	30			6 months		1.7
Bell 2001	14	2000	142.9	one year	1.31	3.14
Sheu 2003	31	57	1.84			0.71
Sheu 2003	34	55	1.62			1.04
Lee 2005a	65	215	3.3	12 weeks	0.24
Hsu & Lee 2005	47	354	7.5	one year	0.58
K. Smith 2006	51	500	9.8	one year	0.47	0.39
Lee 2006	41	1200	29.3	one year	1.02
Hsu & Lee 2007	47	500	10.6	3 years
K. Smith 2007	41	500	12.2	one year	0.56
Liu 2007	46	450	9.8	one year	1.59

Effect size = Cohen’s d.

All effect sizes take pretests into account (ES post-test – ES pretest), except for Mason (Mason and Krashen, 1997) for reading comprehension which was based only on the post-test.

n = number of students in extensive reading group;  titles/S = number of separate book titles per student

Mason studies from Mason and Krashen (1997).

Mason: response L1 -  students wrote summaries in Japanese; response L2 - students wrote summaries in English

Number of titles in Mason, response in L1, response in L2 from Mason (personal communication)

Inspection of table 1 shows that there was little variability in duration in the studies in this sample. Most studies lasted for one academic year. Thus, duration was not examined as a predictor of effect sizes.

The relationship between total titles and titles per student was strong (r = .91). Thus, only titles per student was used in the analysis.

Titles per student was modestly correlated with cloze test effect sizes, and the correlation was close to statistical significance (r = .35, n= 13, p = .12, one-tail). Because of the influence of sample size, a multiple regression was done with sample size and titles/students as predictors (table 3). The relationship between access and reading comprehension was not significant in this analysis.

Table 3
 Predictors of effect sizes for cloze tests (13 studies): Multiple regression

predictor	b	se	t	p
n	0.016	0.014	1.13	0.142
titles/S	0.0009	0.0045	0.208	0.42

Adjusted r2 = .067
   n = sample size


title/S = titles per student

Eight studies provided data for both reading comprehension and titles per student. In contrast to the cloze test results, the two were nearly perfectly correlated (r = .95). Because of the influence of sample size on scores of reading comprehension, the impact of titles per student on reading comprehension scores was investigated using multiple regression, controlling for sample size. As presented in table 4, number of titles per student was a highly significant predictor.

Table 4
 Predictors of effect sizes for reading comprehension (8 studies)

predictor	b	se	t	p
n	0.021	0.015	1.42	0.11
titles/S	0.015	0.0033	4.51	0.003

Adjusted r2 = .91
n = sample size
titles/S = titles per student

According to these results, there is a strong impact of access: Setting the predicted number of students to the mean of the eight studies in table 2 (n = 33.3), doubling the number of titles per student (from 27 to 88) would increase the effect size for reading more than a third of a standard deviation (from .98 to 1.36).

Summary and Discussion

The clearest result of this study is that extensive reading is consistently effective. The average effect size for both measures was over .70. There were no negative effect sizes; the smallest effect size was .24.

The attempt to study factors contributing to successful extensive reading results has clearly only just begun. Data from eight studies using tests of reading comprehension showed that providing more titles per student had a substantial effect on the outcome of the study, but this relationship was not present for tests of reading comprehension.

It is surprising that even this much of a relationship between access and the effect of extensive reading was found, because so many other factors are probably at work.

First, access defined as titles per student, as noted above, is a crude measure. A modest number of books, if they are the right ones, can have a strong impact, and supplying large quantities of books will not help if the books are not interesting and comprehensible.

Also, as noted earlier, other factors undoubtedly play a role. These include

1. The duration of the program (Krashen 2001).

2. The length of time and frequency of each reading session, i.e., massed versus distributed sessions.

3. The extent of comprehension checking: Krashen (2007) has hypothesized that more frequent and more detailed comprehension checking will result in less interest in reading and less progress in literacy development.

4. Whether reading is encouraged by the use of read-alouds, conferencing, and discussion, all of which have empirical support (Krashen 2004).

5. Whether students are under pressure because of heavy academic loads and exams. Those in SSR programs do more pleasure reading on their own outside of class (Sims 1996), and it is likely that this contributes to the success of the program. Pressure from exams and other courses can reduce the amount of time students devote to reading. According to student reports, this was the case in Hsu and Lee (2007).

There are, at the moment, not enough studies to warrant a meta-analytic review of extensive reading studies done with children acquiring a second language. All studies using children that we have seen, however, have produced impressive evidence for “the power of reading” (Aranha 1985; Elley 1991; Elley and Mangubhai 1983; Cho and H. Y. Kim 2004; Cho and H. Kim 2005), with the exception of Williams (2007). 

Conclusion

This review provides more evidence that in-school self-selected reading works. It must be emphasized that effect sizes were uniformly positive and typically quite impressive. In-school self-selected reading is effective and its effects are robust


NOTE: After this paper appeared, an improved a method of calculating effect sizes that takes pretest scores into consideration was developed by Morris (2008):  Estimating Effect Sizes From Pretest-Posttest-Control Group Designs. Organizational Research Methods 11(2): 364-386. Morris concluded that his analysis "favored an effect size based on the mean pre-post change in the treatment group minus the mean pre-post change in the control group, divided by the pooled pretest standard deviation" (p. 364).

      References

Aranha, M. 1985. Sustained silent reading goes East. Reading Teacher 39(2): 14–217.

Bell, T. 2001. Extensive reading: Speed and comprehension. The Reading Matrix 1(1). http://www.readingmatrix.com/articles/bell/index.html

Cho, K. S., and H. Kim. 2004. Recreational reading in English as a foreign language in Korea: Positive effects of a 16-week program. Knowledge Quest 32(4): 36–38.

Cho, K. S., and H. J. Kim. 2005. Using the newspaper in an English as a foreign language class. Knowledge Quest 34(4): 47–49.

Elley, W. 1991. Acquiring literacy in a second language: The effects of book-based programs. Language Learning 41: 375–411.

Elley, W., and E. Mangubhai. 1983. The impact of reading on second lan- guage learning. Reading Research Quarterly 19, 53–67.

Hsu, Y. Y., and S. Y. Lee. 2005. Does extensive reading also benefit junior college students in vocabulary acquisition and reading ability? The Proceedings of the 22nd International Conference in English Teaching and Learning. Taipei: Crane Publishing Company, pp. 116–27.

Hsu, Y. Y., and S. Y. Lee. 2007. A three-year longitudinal study of in-class sustained silent reading with Taiwanese vocational college students. In Selected Papers from the Sixteenth International Symposium on Eng- lish Teaching, English Teachers’ Association—Republic of China. Taipei: Crane Publishing Company.

Krashen, S. 2001. More smoke and mirrors: A critique of the National Reading Panel report on fluency. Phi Delta Kappan 83: 119–23.

Krashen, S. 2004. The power of reading. Portsmouth, NH: Heinemann and Westport, CT: Libraries Unlimited.

Krashen, S. 2007. Free reading in school: Three hypotheses. Paper presented at CELC conference, Singapore, May 2007.

Lao, C. Y., and S. Krashen. 2000. The impact of popular literature study on literacy development in EFL: More evidence for the power of reading. System 28: 261–70.

Lee, S. Y. 2005a. The robustness of extensive reading: Evidence from two studies. International Journal of Foreign Language Teaching 1(3): 13–19.

Lee, S. Y. 2005b. Sustained silent reading using assigned reading: Is comprehensible input enough? International Journal of Foreign Language Teaching 1(4): 10–12.

Lee, S. Y. 2006. A one-year study of SSR: University level EFL students in Taiwan. International Journal of Foreign Language Teaching 2(1): 6–8.

Lituanas, P. M., G. M. Jacobs, and W. A. Renandya. 1999. A study of exten- sive reading with remedial reading students. In Language instructional issues in Asian classrooms, ed. Y. M. Cheah and S. M. Ng. Newark, DE: International Development in Asia Committee, International Reading Association, pp. 89–104.

Liu, C. K. 2007. A reading program that keeps winning. In Selected Papers from the Sixteenth International Symposium on English Teaching, English Teachers’ Association - Republic of China. Taipei: Crane Publishing Company.

Mason, B., and S. Krashen. 1997. Extensive reading in English as a foreign language. System 25: 91–102.

Sheu, S. P-H. 2004. Extensive reading with EFL learners at beginning level. TESL Reporter 36(2): 8–26.

Sims, J. 1996. A new perspective: Extensive reading for pleasure. The Proceedings of the Fifth International Symposium on English Teaching. Taipei: Crane Publishing Company, pp. 137–44.

Smith, K. 2006. A comparison of “pure” extensive reading with intensive reading and extensive reading with supplementary activities. Inter- national Journal of Foreign Language Teaching (IJFLT) 2(2): 12–15.

Smith, K. 2007. The effect of adding SSR to regular instruction. In Selected Papers from the Sixteenth International Symposium on English Teaching, English Teachers’ Association—Republic of China. Taipei: Crane Publishing Company.

Williams, E. 2007. Extensive reading in Malawi: Inadequate implementa- tion or inappropriate innovation? Journal of Research in Reading 30(1): 59–79.

Yuan, Y. P., and T. Nash. 1992. Reading subskills and quantity reading. Selected Papers from The Eighth Conference on English Teaching and Learning in the Republic of China. Taipei: Crane, pp. 291–304.