3

English Teaching, Vo l. 65, No. 3 Autumn 2010

A Relationship Between Reading Comprehension and Reading Strategy Use: Meta-Analysis

뼈 삐·

p

s

o

μ

v‘ p

”쉬 “ n

‘

O』〕

h

“m 。。

n ”u

’n 「L

/，、

Park, Yong-Hyo. (2010). A relationship betwccn rcading comprehension and

reading strategy use: Meta-analysis. English teaching, 65(3), 3-22.

Substantial studics on the relationships bctween reading comprehension and reading

st:rategy use of ESUEFL leamers have bcen reportcd. However, the results of thc

studies were inconsistcnt and hardly conclusivc. This study was purposed to show stablc

estimates of the rclationships and also to investigate whcther ccrtain variables, such as

language context, L 1 or native culture, grade level , measure for reading comprchension, and research design, make differences in the relationships. In order to fulfill thc purpose, the present study mcta-analyzcd 18 unique studics prcviously conductcd on thc

relationships. This mcta-analysis showcd that there was a significant positive cOITclation

between reading comprchcnsion and reading strategy usc of the ESLlEFL leamers, and

the correlation was mcdium. Of the moderatin !! variablcs cxamincd. thc four variablcs 。

exccpt for thc research dcsi밍1 made a statistically significant difference in the

relationships. The findings are discussed and implications for rcsearch and pedagogy are

suggested. (146)

1. INTRODUCTION

Reading is probably one ofthe most important skills for langllage leamers including L2

leamers. Particularly, reading comprehension is a crucial component for the languagc

leamers ' academic success across edllcational contexts. Among diverse approachcs to

develop the language leamers' reading comprehension , reading sπategy lIse has been

widely discusscd as onc of the approaches. Within a largc research framework in reading

comprehension and strategy use, relationships betwcen reading comprehension and

reading strategy use of lang uage leamers have been intercsts of scholars. Carrell, G에dusek，

and Wise (1998) explain the reason for the scholars' interests as follows: “ Reading

strategies are of interest not only for what they reveal about the ways readers manage

4 Yong-Hyo Park

interactions with written text but also for how the use of strategies is related to effective

reading comprehension" (p. 97). Underlying assumption in investigating the relationships

is that reading sπategy use can improve reading comprehension and help leamers become

proficient readers.

Anderson (1991) found that Spanish speaking adult ESL students who used more

strategies comprehended better. On the other hand, Brantmeier (2000) conducted research

on reading strategy use of students leaming Spanish as a L2 and concluded that there was

no significant relationship between reading strategy use and reading comprehension. One

possible explanation for this difference in the results of these two studies is that other

variables, such as target su이ects， types of a comprehension test, and a language context,

might influence the results. Regard to this issue, Brantmeier (2002), who comprehensively

reviewed the studies on reading strategies, stated as follows: “ Because of the wide variety

of participants, tasks, and reading matetials employed in studies that examine L2 reading

strategies, it is difficult to compare results across studies" (p. 2). ln light of such diffic비ttes ，

synthesizing and analyzing the results with a different perspective, particular1y in terms of

a methodological approach, would be a meaningful study. As to the different

methodological approach, a quantitative meta-analysis, which is a powerful statistical tool

to synthesize and analyze the results of the previous studies, can be one good option. The

present study is an exploratory attempt to synthesize the available studies and re-analyze

them and to shed more light on the relationships between reading comprehension and

strategy use in ESLIEFL contexts.

Several definitions of ‘ reading strategy ’ are available in the literature on reading, but

there is no clear-cut definition. Gamer (1987) defines reading strategies as “ generally

deliberate, planful activities undertaken by active leamers, many times to remedy

perceived cognitive failure" (p. 50). Brantmeier (2002) characterizes reading strategies as

“ the comprehension processes that readers use in order to make sense of what they read"

(p. 1). F or the purpose of this study, reading strategies are defined as techniques, actions,

and procedures that readers deliberately employ to enhance their comprehension in reading

a tex t. With regard to an operational definition of read

A Relationship Between Reading Comprehension and Reading Stratcgy Usc: Mcta-Analysis 5

11. LlTERATURE REVIEW

1. Relationships Between Reading Comprehension and Strategy Use

Since Hosenfeld ’s (1976) early work, substantial studies on the relationships between

reading strategy use and reading comprehension have been performed in ESLIEFL

contexts (e.g., AI-N내aidi， 2003; Brantmeier, 2000; Darabie, 2000; Lee, 2007; Schueller, 1999; Sheorey & Mokhtari, 2001; Song, 1998, 1999; Wu, 2005; Zhang, 2001; Zhicheng, 1992). For example, Song (1999) investigated the relationship between reading strategy

use and reading comprehension ability of Korean EFL co l1ege students, and results of her

study showed a very strong positive correlation (r = .727). Darabie (2000) explored the

relationship with EFL leamers in Jordan and reported a very strong positive correlation (r

= .790) too. AI-Nujaidi (2003) conducted research on the relationship with EFL leamers in

Saudi Arabia and reported a weak positive correlation (r = .19). Lee (2007) investigated

cffects of strategy instruction on rcading proficiency of Korean EFL co l1 ege students, and

resu1ts of her study showed both weak and strong and both positive and negative

correlations (r = .11 for pre-test; r = -.11 for mid-tcst; r = .60 for final-test}-these

correlation coefficicnts wcre calculated from means, standard deviations, and t values

reported origina l1y in Lee ’s (2007) stl.ldy

As shown above through the brief rcvicw of some empirical stl.ldies, therc sccm

correlational relationships to some degree between reading comprehension and reading

strategy use of ESLlEFL leamers. However, the resl.llts were various, not consistent across

the studies; some studies showed a strong positive correlation, but some studies showed a

weak correlation and even a study showed a negative correlation. As for the results, it

seems hard to draw a simple and unified conc\usion about the relationships between

reading comprehension and reading strategy use of ESLlEFL leamers. Given the

apparently inconsistent results across the studies, synthesizing and re-analyzing the resl.llts

with a different methodological approach might provide more qualified results than any

narrative reviews or any single empirical study cOl.lld provide.

2. Moderating Variables in Reading Strategy Use

Diverse variables intluencing ESLIEFL leamers' reading strategy use, such as age, grade level, gender, language context, nationality, persona1ity, and proficiency, have been

examined. Among those variables, this meta-ana1ysis se1ected grade level, language

context, and nationality as variables moderating the relationships between reading strategy

I.Ise and reading comprehension. These moderating variables were derived from

observation made in the primary studies that were inc\uded in this meta-analysis.

6 Yong-Hyo Park

The language context of the study may play a significant role in L2 leamers' language

behavior and leaming. As to the language context, the distinction between ESL and EFL

contexts has been traditionally used in both practical and research fields. Even though the

distinction is not clear as much as it was before and it is an on-going debate (Brown, 2000), it is still a valid and widely accepted distinction. In research on general language leaming

strategy use, the influence ofESL versus EFL differences has been reported. For example, Ri ley and Harsch (1999) reported that Japanese college students in an ESL contεxt showεd

a different pattem and frequency of strategy use from Japanese EFL college students.

Considering factors such as input and motivation, ESL readers might also have different

characteristics from EFL readers in strategy use and reading comprehension.

The grade level may play a part in strategy use and reading comprehension. Leamers in

a lower grade level could be cognitively less developed and still in the process of

developing their Ll reading skills, and also they may be less proficient in their L2 reading

comprehension than leamers in a higher grade level. These characteristics may influence

reading strategy use and the relationships between reading strategy use and reading

comprehension. Some studies showed that higher grade EFL students have a different

pattem in reading strategy use from lower grade EFL students (e.g. Kung, 2007; Y. -H.

Park, 20 1O)

According to studies on general language leaming strategy use, nationality/ethnicity

seems to influence the use of L2 leamer strategies (Oxford, 1996). However, researchers

have recently begun to show interests in cultural anψor national influences on L2 reading

strategies (Erler & Finkbeiner, 2007). Tailefer (2005) reported that national origin

in f1uences both reading comprehension and strategy use. Under a situation that very

limited studies on L2 reading strategy use from a cross-cultural perspective have been

reported, it might be interesting to determine in f1uences of Ll factor indirectly as a

moderating variable in this meta-analysis.

In addition to the moderating variables that are related to a language context and a

leamer, two additional variables about a research setting were included in the current meta­

analysis. One was related to a measure for reading comprehension, and the other one was

related to a research desi

A Relationship Between Reading Comprehension and Reading Sπategy Use: Mεta-Analysis 7

1) How large is a correlation between reading sσategy use and reading comprehension

ability of ESLlEFL leamers?

2) Ooes a language context (i.e. , ESL, EFL) influence the correlation?

3) Ooes a LI or native culture ofsubjects influence the correlation?

4) Ooes a grade level of subjects influence the correlation?

5) 00 types of a measure for reading comprehension (i .e., standardized vs. non­

standardized) make a difference in the correlation?

6) 00 types of a research design (e.g., experimental, correlational) make a difference

in the coπelation?

The first research question is the core question of this meta-analysis. The subsequent

questions are for deterrnining potential moderating variables that may influence the results

ofthe studies included in this meta-analysis.

III.METHOD

1. Sampling of Studies

Studies relevant and included in the present meta-analysis had to meet the following

criteria: 1) a target language of the studies had to be an English, not other languages, 2)

target subjects ofthe studies had to be ESL or EFL students, not native English speakers, 3)

the studies had to be written in English, 4) the studies should provide relevant statistics

(i .e., means, standard deviations, sample size) eligible for computing an Effect Size (ES)

unless the studies presented effect sizes (i.e., correlation coefficient, standardized mean

difference) specifically. 1n order to find relevant studies for this meta-analysis, articles were initially located by

searching multiple databases, such as ERIC, Linguistic and Language Behavior Abstracts

(LLBA), Proquest Oissertations and Theses (POT), with combination of keywords

“ reading", “ reading abiIity", “ reading comprehension", “ strategy", “ reading strategy", “ leaming strategy", and “ strategy instruction". Once studies were found, their

bibliographies were also examined to identify additional relevant studies for inc1usion.

Meanwhile, the searching was Iimited to the studies published or reported since 1990.

There was no theoretical reason for this Iimitation, but studies prior to 1990 were

frequently judged inappropriate because of uncharacteristic research design and analyses

and missing relevant statistics. Approximately 250 abstracts were read and analyzed, and

21 studies were identified to have the required information.

Some studies (e.g., AI-Nujaidi, 2001; Darabie, 2000; Y. -H. Park, 2010; Song, 1999; Tsai,

8 Yong-HyoPark

2000; Torut, 1994) were specifically designed to measure correlation coefficients between

reading comprehension and reading strategy use, and they provided effect sizes of

correlation. Other studies (e.g., Abdelhafez, 2006; Bimmel, Van den Berg, & Oostdam, 2001; Brown, 1991; Chang, 2005; Dhieb-Henia, 2003; Kusiak, 2001; Lee, 2007; Ou, 2006;

Pappa, Zafiropoulou, & Metallidou, 2003 ; Y.-Y. Park, 1996; Prinz, 1998; Zhicheng, 1992)

were designed to measure mean differences of reading comprehension scores between a

treatment group receiving reading strategy instruction and a control group not receiving

reading strategy instruction. The forrn of effect sizes that can be obtained from these

studies is “ mean difference effect size". Since the mean difference effect size can be

converted into the coπelation effect size, studies reporting either one of the two types of

effect size were included in this current meta-analysis. On the other hand, some studies

(e.g., Salataci & Akyel, 2002; Singhal, 1999; Song, 1998) investigated effects of strategy

instruction on reading comprehension with a pre- and post-test research design without a

contro\ group. The forrn of effect sizes that can be obtained from these studies is “ mean

gain effect size". The mean difference and the mean gain effect size statistics cannot be

comparable, and they shou\d not be mixed in the same meta-ana\ysis (Lipsey & Wi\son, 2001). Thus, those studies reporting the mean gain effect sizes were excluded in this meta­

ana\ysis. Accordingly, 18 studies in tota\ were included in the current meta-analysis.

A majority of studies, unfortunately, investigating the relationships between reading

strategy use and reading comprehension, did not report effect sizes or appropriate statistics

(i .e., means, standard deviations, sample size) which are needed to compute the effect sizes.

Accordingly, those studies were not included in this meta-analysis, which \ed to smal1

sample studies for this meta-analysis even though a large number of studies were initially

searched

2. Calculation of Effect Sizes

This study used a correlation coefficient effect size for the meta-analysis. Pearson

product-moment coπelation coefficient is the most comrnon measure computed for the

relationship between two continuous variables (e.g. , scores of reading strategy use and

reading comprehension test scores). Point-biserial correlation coefficient is another

product-moment correlation coefficient computed for the relationship between a

dichotomous independent variable and a continuous dependent variable (Lipsey & Wilson, 2001). Some studies reported the Pearson product-moment correlation coefficient, and this

study used the reported correlation coefficient effect size. Some studies reported

standardized mean differences between groups, and thus the standardized mean difference

effect sizes were converted to the point-biserial correlation coefficient effect size. The

point-biseria\ correlation coefficient can be ca1culated with the following forrnula:

A Relationship Bctwccn Reading Comprehension and Reading Strategy Use: Meta-Analysis 9

파긁길)+ESm2

whcrc ESm is any standardized mean ditference etfect size, p is thc proportion of

subjects in Group 1, and 1- p is the proportion of subjects in Group 2. Some experimental

or qusi-experimcntal studies did not report the standardized mean ditference eff농ct size

specifically, but thc standardized mean ditference etfect size can be calculated from the

basic statistical information (i.e. , sample size, means, standard deviations) reported in

those studies according to a following formula:

?‘ -

G -

-4

-”얘

1·ι -

G -

-x -m 잠

where X Cl is the mcan for Group 1, X C2 is the mean for Group 2, and Sp is thc

pooled standard deviation. The pooled standard deviation is represented as

( nCl -l)S~l + (nC2 -1)S~2

(nCl -1) + (nC2 -1) Sn = p

where SCI is the standard deviation for Group 1, SC2 is the standard deviation for Group

2, nCI is the number of subjects in Group 1, and nC2 is the number of subjects in Group 2.

Etfect Size Determination Program (Wi lson, 2001) was used for the conversion and

computation of the etfect sizes. MS-Excel 2007 was used for additional mathematical

calculations.

10 y ong-Hyo Park

TABLE 1 Basic Descriptive Statistics for Meta-analysis

Studies Ne Nc Reading score

ESm ESr Pre Xe PreXc Post Xe Post Xc

Abdelhafez (2006) 40 40 29.60 24.00 1.471 592

AI-Nujaidi (2003) 226 6.16 190 8immel et al. (2001) 12 119 83.88 82.02 89.07 86.06 113 .033

8rown (1991) 15 15 28 27 32 30 .840 .387

Chang (2005) 39 38 14.08 12.82 .332 164

Darabie (2000) 300 14.7 .370 Dhieb-Henia (2003) 35 27 39.42 38.16 70.80 40 .42 2.456 .773 Kusiak (2001) 78 80 11.56 12.78 14.49 13.81 476 .232

Lee (2007) 38 34 57.1 1 56.62 74.96(mid) 76.12(mid) -.204 -.101

Lee (2007) 38 34 57. 11 56.62 81.20(final) 74.28(final) 613 .293

Ou (2006) 62 62 18.34 17.95 22 .82 20.56 .509 .247

Pappa et al. (2003) 30 29 54.9 6 1.5 60.9 62.0 489 .237

Park, Y.-H. (2010) 115 7.37 .182 Park, y.-y. (1996) 27 21 20.33 23. 10 F = 5.42 1.290 .610

Park, Y. -Y. 2 (1 996) 24 21 97.13 123.00 F = 6.28 1.458 .685

Prinz (1998) 12 14 14.00 19.85 12.33 15.62 124 062

Song3 (1999) 66 17.63 727

Torut (1 994) 468 13.23 083

Tsai (2000) 746 53.30 450

Zhicheng4 (1992) 15 14 F-values F-values .932 410

I.2Mean scores ofthe post-tests were not reported. ESm was calculated from the repoπed Fvalues

3 The score is the weighted grand mean of means of 3 levels. 4 Three reading passages were used for the study. For each reading passage, an obtained ESm was

converted into ESr and then the mean ESr was calculated.

Table 1 provides basic infonnation ofthe studies that were inc1uded in the present meta­

analysis. Most experime따al studies were qusi-experimental, not real experimental, meaning that intact class were used, along with a pre and post-test, instead of random

assignment of subjects to experimental or control groups. Accordingly, in order to assure

that the degree of equality of groups for the qusi-experimental studies, the effect size

between the control and experimental group on the pre-test was subtracted from the effect

size between the respective groups on the post-test (i.e., mean d미iffe하r댄e히nce ESposf-t,

ιm’51)' This procedure was used for all qusi-experiments in the present meta-analysis (e.g., 8immel et al., 2001; 8rown, 1991; Dhieb-Henia, 2003; Kusiak, 2001; Lee, 2007; Ou,

2006; Pappa et al., 2003; Y.-Y. Park, 1996; Prinz, 1998). Randomized, πue experimental

studies without pre-test could simply be calculated with only the post-test results (e.g., Abdelhafez, 2006; Zhicheng, 1992). A qusi-experimental study with pre and post-test,

which did not report statistics of pre-test, was ca1culated with only the post-test results too

A Relationship Between Reading Comprehension 때d Reading Stratcgy Use: Mcta-Analysis 11

(e.g., C hang, 2005). Two effcct s izes from Lee ’s (2007) and y.-y. Park’s study (1996) were

rcspectively obta ined. In Lee ’s study (2007), one effect size was ca \culated betwccn thc

prc-test and mid-tcst rcsults, and the other one was calωlated between the prc-tcst and

final-test results. In y.-y. Park ’s study (1 996), she used two different measures for reading

comprehension , and thus cach effect size was calculated based on results from each

measure. According ly, 20 effcct sizes out of 18 individual studies in total were analyzed

for the current meta-analys is.

In order to adjust samples s ize bias, weights bascd on the standard error (SE) of each

cffect s ize are needcd for meta-analysis--called invcrse vari ance weight (Lipsey & Wilson, 200 1). For convenicnce in conducting this procedurc, the Fisher ’s Z - transformation was

app lied, and then the results were converted back latcr into regular correlation coeffic ients

for interpretation (see Table 2). Z, - transformed correlation is defined as

타'Zr= .5때

where r is the corre lation coe에cient and 10& is thc natural logari thm.

TABLE 2 Sample Size Bias Adjustment

Fisher’s Z-transformation

Studies ESr N ESz, SEZr WZr I*ESZr

Abdelhafez (2006) .592 80 0.681 114 52 .464

AI-Nujaidi (2003) .190 226 0.192 067 42.891 Bimmel et al. (200 1) .033 131 0.033 .088 4.226 Brown (199 1) .387 30 0 .408 .192 11.023 Chang(2005) .164 38 0.590 .169 20.655

Darabie(2000) .370 300 0.388 .058 115.362 Dhieb-Henia (2003) .773 62 1.028 130 60.637 KlIsiak(200 1 ) .232 158 0.236 .080 36.627 Lee (2007) -.101 72 -0.101 .120 -6.993 Lee (2007) .293 72 0.302 120 20.827 Oll (2006) .247 124 0.252 091 30.5 18 Pappa et al. (2003) .237 59 0 ‘ 242 134 13.529 Park, Y.-H. (20 10) .182 115 0. 184 094 20.6 14 Park, Y.-Y. ( 1996) .6 10 48 0.709 149 3 1.90 1 Park, Y.-Y. (1996) 685 45 0.838 .154 35.2 16 Prinz (1998) .062 26 0.062 209 1.428

Song (1999) ‘ 727 66 0.922 126 58.107 Torut (1994) .083 468 0.083 046 38.684 Tsai (2000) .450 746 0 .485 .037 360.132 Zhicheng (1992) .4 10 29 0.436 .196 11.326

I 1nverse variance weight (= I/SEz/ )

12 Yong-Hyo Park

3. Homogeneity Te5t

In order to examine homogeneity of distribution of the observed effects sizes (i.e., the

variabi1ity among the effect sizes is 깐om subject-1eve1 samp1ing error a1one), homogeneity

based on a Q statistic was tested. The formu1a for Q is

2 , (L Wi ESi)2 Q=CI. wi ES{) - ~τ←-

4 ‘ VVl

where ES; is the individua1 effect size for i = 1 to k (the number of effect sizes) and Wj is

the individua1 weight for ES;. A test for homogeneity yie1ded a Q statistic of 162.377, which exceeded the critica1 va1ue for a chi-square, X( 19) = 30.14 at a = .05, and thus the

nu11 hypothesis of homogeneity was rejected, which indicates a heterogeneous distribution.

l va1ue was 94 .457, which means that 94 % of variance cannot be explained by subjects­

level sampling error. This information is summarized in Tab1e 3

M

SE 95% CI

Q

TABLE3 Statistics from Fixed Effects Model

338 019 .302

162.377 94.457

ESZr

.375

326 .019 .293

ESr

358

Accordingly, a fixed effects model cannot be he1d, which means that an effect size

observed in a study might have study-1eve1 sampling error as well as subjects-level

samp1ing error. Thus, this meta-ana1ysis fuπher ana1yzed effect size variation in terms of

characteristics of the observed studies whi1e assuming the fixed effects model. In other

words, this meta-ana1ysis exp10red systematic, not random, variability of the observed

effect sizes that can be exp1ained by moderator variables.

4. Coding of Studies

To account for possible differences across studies, the studies were categorized

according to following variables: language context of study, L 1 or native culture of

subjects, grade level of subjects, measure of reading comprehension, and research design.

The coding categories are listed in Table 4. The values assigned to each study for each

coding category are displayed in Table 5.

13 A Relationship Bctwccn Reading Comprehension 때d Reading Strategy Use: Meta-Analysis

TABLE4 Variables and Codes

Codes

1 - ESL

2 - EFL

Variablcs

A. Langllage context of Stlldy

1 - African

2 - Arabian

3 - East Asian

4 - European

5 - Mix

B. L 1 or native cllltllre of SlIbjects

1 - Secondary

2 - Post-secondary

C. Grade Level ofS lIbjccts

1 - Standardizcd Tcst

2 - Non-standardizcd Tcst

O. Measllre of Reading Comprehcnsion

1 - Experimental or QlIsi-Expelimental

2 - Correlational

E. Research Oesign

앙 -

-때

-

-많 K -

뼈 -@ -c -2

2

l 2

2

2

2

l 2

2

2

l

2

2

2

I 2

l 2

2

F3

-」u-

%

--따 -

쩔 얘 -

nμ 땐

” ·” =

A A -

1 R 」

u= 0 g

c -%

-뿔

V

E

2

2

2

2

2

2

D-2

2

2

2

l 2

2

2

2

2

l 2

l

l 2

l 2

2

l 2

B-l

2

4

3

3

2

l

4

3

3 3

4

3

5

5

5 3

3

3

5

A-2

2

2

2

2

2

2

2

2

2

2

2

2

l

l

1

2

2

2

l

Studies

Abdelhafez (2006)

A I-Nujaidi (2003)

Bimmel et a l. (2001)

Brown ( 1991 )

Chang (2005)

Oarabie (2000)

Ohieb-Henia (2003)

KlIsiak (2001)

Lee (2007)

Lee (2007)

Oll (2006)

Pappa et a l. (2003)

Park Y.-H. (2010)

Park Y.-Y. ( 1996)

Park Y.-Y . ( 1996)

Prinz ( 1998)

Song (1999)

Torut ( 1994)

Tsai (2000)

Zhicheng ( 1992)

14 Yong-Hyo Park

IV. RESULTS

Under the fixed effects model, the meta-analytic estimate of the population coπelation

between reading comprehension and reading strategy use was r = .326 (p < .01), and the

95% confidence interval was [.293 , .358] (see Table 3). These res비ts indicated that there

was a significant correlation between reading comprehension and reading strategy use of

ESLÆFL learners. According to a widely used interpretation key of the magnitude of

effect sizes that was established by Cohen (1988), the correlation effect sizes fell into the

following ranges: r < .10, small; r = .25, medium; r > .40, large. Thus, the mean

correlation effect size ofthe current meta-analysis, r = .326 was medium.

While assuming the fixed effect model, Hedges' (1982) analog to the analysis of

variance (ANOVA) was selected to model between-study variance. Accordingly, the Q

statistics was partitioned as within (Q w) and between (Qs) variance for a set of categories

based on the coding of the studies. This analysis was done separately for the five

moderating variables. To put it in another way, if the between (Qs) variance for groups on

the basis of each moderating variable is significant, the moderating variable accounts for

variability in the observed effect sizes across the groups

1. Language Context of Study

Table 6 shows that the mean of effect sizes for the studies which were conducted in ESL

contexts was r = .528 (p < .01), whereas the mean of effect sizes for the studies which

were conducted in EFL contexts was r = .314 (p < .01), and both the means of effect sizes

were statistically significant. The Q between groups, Qs (i .e., QTolol - Qw= 162.377 -

(10.258 + 143 .265)) was 8.853 , which exceeded the critical value for a chi-square, X(I) =

3.84 at a = .05, and thus the Qs was statistically significant at p < .05, indicating a

significant between-groups effect. Accordingly, whether the studies were conducted in the

ESL contexts or EFL contexts made statistically significant differences in the observed

correlations between reading strategy use and reading comprehension.

TABLE6 Statistics from Fixed Effects-Partitioned Model: Language Context

r

ESL (k = 4)

EFL (k = 16)

M-없 M

SE

086 019

95%CI .396 .638 .280 .348

iv-졌

M

[im

…에 Between 짤쁘~ Q'8 = 8.853, i (l) = 3.84, p < .05

A Relationship Between Reading Comprehension and Reading Strategy Use: Meta-Analysis 15

2. L 1 or Native Culture of Subjects

Table 7 shows that thc mcan of effect sizes for the studies was vary according to L I or

nati ve culture of su이ccts in those studies, and all the means were statistically significant (p

< .0 1). The Q between groups, Q8 was 53.164, which exceeded the critical value for a chi­

square, ;(4) = 9.49 at a = .05 , and thus the Q8 was statistically significant at p < .05, indicating a significant between-groups effect. Accordingly, different Ll or native culture

of subjects made statistica lly significant differences in the observed correlations between

reading strategy use and rcading comprehension.

TABLE 7 Statistics from Fixed Effects-Partitioned Model: L 1 or Native Culture

African (k = 2) Arabian (k = 2) East Asian (k = 9) European (k = 3) Mix(k = 4)

M

‘ 681 ‘ 295 .3 14

‘ 159 ‘ 528

r

SE 95% CI ili: 086 ‘ 581 ‘ 761 4008 044 .2 15 .372 4.897 024 ‘ 270 .356 86.7 11 054 054 .261 2.339 086 396 .638 10.258

Q8 = 53. 164, f(4) = 9.49, p < .05 Between groups

3. Grade Level of Subjects

Table 8 shows that the mean of effect sizes for the studies which were conducted with

secondary students was r = .114 (p = .00 1), whereas the mean of effect sizes for the studies

which were conducted with post-secondary students was r = .406 (p < .0 1), and both the

means of effect sizes were statistically significant. The Q between groups, Q8 was 58.623 , which exceeded the critical value for a chi-square, ;(1) = 3.84 at a = .05, and thus the Q8

was statistically significant at p < .05, indicating a significant between-groups effect.

Accordingly, whether the studies were conducted with secondary or post-secondary

students made statistically significant differences in thc obscrved correlations between

rcading strategy use and reading comprehension.

TABLE8 Statistics from Fixed Effects-Partitioned Model: Grade Level

l

Secondary (k = 5) Post-secondary (k = 15)

M 114

.406

또-m …ι

95%C[ 와E 4.573 99.181

046 . [48 369 .442

Between 융띤브~ P8 = 58.623, i (l) = 3.84, p < .05

16 Yong-Hyo Park

4. Type of Measure of Reading Comprehension

Table 9 shows that the mean of effect sizes for the studies which were conducted with

standardized tests was r = .406 (p < .01), whereas the mean of effect sizes for the studies

which were conducted with non-standardized tests was r = .274 (p < .01), and both the

means of빠ect sizes were statistically significant. The Q between groups, QB was 15.018, which exceeded the critical value for a chi-square, i (1) = 3.84 at a = .05, and thus the QB

was statistically significant at p < .05, indicating a significant between-groups effect.

Accordingly, whether the studies were conducted with standardized or non-standardized

tests made statistically significant differences in the observed correlations between reading

strategy use and reading comprehension

TAßLE 9 Statistics from Fixed Effects-Partitioned Model: Measure of Reading Comprehension

r

Standardized (k = 6) Non-standardized (k = 14)

M-%

갱

SE

030

.023

95% Cl Qw

.355 .455 20.333 230 317 127.026

Between gr쁘~ QB = 15.018, 상(1) = 3.84, p < .05

5. Type of Research Design

As shown in Table 10, the mean of effect sizes for the studies which were experimental

research was r = .3 34 (p < .01), whereas the mean of effect sizes for the studies which

were correlational research was r = .322 (p < .01), and both the means of effect sizes were

statistically significant. The Q between groups, QB was .103 , which did not exceed the

criticaI value for a ch냥quare， 상(1) = 3.84 at a = .05, and thus the QB was not statistically

significant at p < .05, indicating no significant between-groups effect. Accordingly, whether the studies were experimental or correlational did not make statistically

significant differences in the observed correlations between reading strategy use and

reading comprehension.

TABLE 10 Statistics from Fixed Effects-Partitioned Model: Research Design

r

M SE 95%CI ili:: ExperimentaI (k = 14) .334 033 275 .389 86.469 Correlational (k = 6) .322 023 281 ‘ 362 75.805

Between groups QB = .103, i(1) = 3.84, p > .05

A Relationship Between Reading Comprehension and Reading Strategy Use ν!eta-Analysis 17

v. DISCUSSION

This study meta-analyzed 20 correlation effect sizes dcrived from 18 independent

studies. Several points can be discussed on thc basis of thc rcsults drawn from thc current

meta-analysis. Overall, there was a positive coπclational relationship between reading

strategy use and reading comprehension ability of ESLlEFL leamers, and thc con-elation

coefficient (r = .326) was moderate. This finding, previously f1 uctuant, now providcs more

stable estimates of how much the reading strategy use is coπelated to reading

comprehension ability of ESLÆFL leamers

There was a statistically significant difference between studies that were conducted in

ESL and EFL contexts. The correlation between reading strategy use and reading

comprehension ability of ESL leamers (r = .528) was much stronger than that of EFL

leamers (r = .314). The language context in which the respective studies were conducted

seemed to moderate the results of the studies. This might be because leamers in ESL

contexts could be more advanced than leamers in EFL contexts in a perspective of English

reading proficiency. As usual characteristics of advanccd language leamers, the ESL

leamers might rely much on using reading strategies in their reading comprehension

process rather than discrete or lower-level language skills that less advanced language

leamers usually rely much on. This finding seems to be noteworthy in temlS of

pedagogical implications for less advanced readers in EFL contexts, particularly in Korea.

Korean EFL teachers might introduce advanced readcrs ’ characteristics of relying much on

reading strategy use to their students, especially lower level students, and encourage them

to execute reading strategy use actively rather than to rely much on discrete language skills

1n terrns of an implication for research, no study, compating Korean ESL leamers' reading

strategy use to Korean EFL leamers' , has been reported. Thus, it is expected to see a study

covering this issue in the future, and the study would provide more insights on the

in f1uence of ESL versus EFL differences in reading strategy use and reading

comprehension.

This meta-analysis found a statistically significant difference among studies that were

conducted with subjects having different Ll or native cultures. The studies with su이ects

having an A한ican L 1 background showed the strongest correlation, followed by the

studies with subjects ha

18 Y ong-Hyo Park

between studies according to the grade levels of the subjects participated in the studies

The correlation between reading strategy use and reading comprehension ability of post­

secondary leamers (r = .406) was much stronger than that of secondary leamers (r = .114).

Probably, the post-secondary leamers, who are believed to be more cognitively developed

than the secondary leamers, might more actively execute reading strategy use, which is

cognitively demanding, in their reading comprehension process.

The instrument of the individual study to measure reading comprehension resulted in a

statistically significant difference between studies. The studies using standardized reading

comprehension tests (r = .406) showed the stronger correlation between reading strategy

use and reading comprehension ability than the studies using non-standardized reading

comprehension tests (r = .274). lt is difficult to figure out why this difference was made.

Meanwhile, this result might have important implications as for selecting a measure for

reading comprehension. Researchers planning to do research on reading strategy use and

reading comprehension might have to consider this result when they choose a measure for

reading comprehension in their research.

There was no statistically significant difference between studies that employed an

experimental and correlational research design. Both the studies employing experimental

(r = .334) and correlational (r = .322) design showed medium correlational effect sizes

between reading strategy use and reading comprehension ability. This result is noteworthy

in terms of that consistent correlational effect sizes can be anticipated regardless of a

research design, either experimental or coπelational.

Among the five moderating variables, the variables of Ll or native culture and grade

level relatively seem to make clearer differences in the observed effect sizes across groups

than the other variables. This result might be because these two variables can be directly

related to L2 proficiency which is believed to be a critical factor for L2 leamers' reading

strategy use.

VI. CONCLU510N

This meta-analysis synthesized and reanalyzed an individual result of each study about

relationships between reading strategy use and reading comprehension ability of EFLÆFL

leamers and provided more stable empirical evidence that reading strategy use is

moderately correlated to reading comprehension ability. 1n addition, the current meta­

analysis showed that there could be diverse variables moderating the relationships.

The current meta-analysis has its own limitations as follows: The nurnber of the effect

sizes included in this meta-analysis was not big enough. If more effect sizes had been

included, more potential moderator variables and more sophisticated categorization of the

A Relationship Betwccn Reading Comprehension and Reading Strategy Use: Meta-Analysis 19

moderator variables would have been possible, and thus more surc conclusions would

have been made. In tcmlS of potential l110derator variables, reading l11atcrials, tasks, and

L2 proficiency have bccn considered as critical factors for L2 leamers' reading stratcgy

use. Thus, it is suggestcd that those factors , particularly L2 proficiency, be includcd as

moderator variables in a further meta-analysis stlldy in the future. 1n spite of these

Iimitations, this study is significant, particlllarly in terrns of a research methodology. There

have been no efforts to analyze and synthesize variolls studies on the relationships between

reading comprehension and reading strategy llse. As Brantl11eier (2002) presented those

studies descriptively bllt not analyzed or synthesized, she indicated that it is neither easy

nor interesting to dircctly compare results obtained from the studies based on different

conditions sllch as variolls target su셰ects and langllagc contexts. This study, as sllch an

effort, applied a different methodological approach called a quantitative Meta-Analysis.

With applying the mcta-analysis, this study could ovcrcome to some degree the problem of

direct comparison.

Lastly, one of advantagcs of the qllantitative meta-analysis is easiness of replicate. AII

the stlldies inclllded in this meta-analysis are c1early citcd and can be readily accessed.

Furtherrnore, the statistics, samples sizes, methods and SOllrces for the calculation of the

effect sizes, and weighting methods have been already presented. I f some readers disagree

with the way of coding that this meta-analysis llsed, they can easily llse those data

presented here and apply their own coding to test their own hypotheses. Other readers who

can access more studies that this l11eta-analysis could have missed may add effect sizes

derived from those additional studies into the effect sizes in the current meta-analysis and

re-analyze all of them with the coding of this meta-analysis or their own coding. It is

expected to see another l11eta-analytic study on the relationships between reading strategy

use and reading comprehension of ESLIEFL leamers in the future.

REFERENCES References marked with an asterisk indicate stlldies inclllded in the meta-analysis.

*Abdelhafez, A. M. M. (2006). The 행ect 01 a sugóested lraining program in some

metacognitive language learning slralegies on developing listening and reading

comprehension ojuniversity EFL sludenls (ERIC Document Reproduction Service

No. ED 498262)

*Al-N내aidi ， A. H. (2003). The relationship between vocabulary size, reading strategies, and reading comprehension 01 EFL learners in Saudi Arabia. Unpublished doctoral

dissertation, Oklahoma State University.

Anderson, N. 1. (1991). lndividual differences in strategy use in second language reading

20 Yong-Hyo Park

and testing. Modern Language Journal, 75, 460-472.

*Bimmel, P. E., Van den Berg, H., & Oostdam, R. J. (2001). Effects ofstrategy training on

reading comprehension in first and foreign language. European Journal 01 PsycholO양 of Education, 16, 509-529.

Brantrneier, C. (2000). The relationship between reader 's gender, passage content, comprehension, and stralegy use in reading Spanish as a second language.

Unpublished doctoral dissertation, Indiana University, Bloomington, ll'ι Brantmeier, C. (2002). Second language reading sπategy research at the secondary and

uni versity levels: Variations, disparities and generalizability. The Reading Matrix, 2(3), 1-14.

*Brown, R. L. (1 991). The 얘ects of teaching a multicomponent reading strategy on

university ESL students ' reading comprehension and reposts 01 reading strategies.

Unpublishεd doctoral di ssertation. University of Lava l.

Carrell, P. L., Gajdusek, L. , & Wise, T. (1998). Metacognition and EFLIESL reading.

Jnstmctional Science, 26, 97-112 .

*Chang , D. M. (2005). Comparing the 행ects of traditional vs. non-traditional reading

instruction on level 01 reading comprehension, and use of metacognitive strategies

in EFL learning in Taiwan. Unpublished doctoral dissertation, La Sierra University.

Cohen, 1. (1988). Statistical power analysis for the behavioral sciences (2nd ed.) Hillsdale,

NJ: Erlbaum.

*Darabie, M. Y. (2000). The relationshψ between college-level Jordanian students ’

metacognitive awareness strategies and their reading comprehension achievement

in Englβh as a foreign language. Unpublished doctoral di sseπ.ation， Ohio

University, OH.

*Dhieb-Henia, N. (2003). Evaluating the effectiveness of metacognitive strategy training

for reading research artic les in an ESP context. Englishfòr Specific Purposes, 22, 387-4 17.

Erler, L., & Fi따cbeiner， C. (2007). A review of reading strategies: Focus on the impact of

first language. In A. D. Cohen & E. Macaro (Eds.), Language Learner Strategies

(pp. 187-206). New York: Oxford University Press.

Gamer, R. (1987). M강tacognition and reading comprehension. Norwood, NJ: Ablex.

Hosenfeld, C. (1 976). Leaming about leaming - discovering our students ' strategies.

Foreign Lan밍IGge Annals 9(2), 117-129.

Kung, S. S. (2007). A stu껴y of the relationship between reading comprehension and the

use of reading strategies among EFL college students. Unpubl때ed doctoral

dissertation, Texas A&M U

A Relationship Between Reading Comprehension and Rcading Stratcgy Use: Meta-Analysis 21

in the EFL contex!. Unpublished doctoral dissertation , Univers ity of Maryland, College Park

Lipsey, M. w., & Wilson, D. 8 . (2001). Practical meta-an이ysis (Vo l. 49): SAGE

Publication, lnc.

*Ou, F. C. (2006). The 갱ec!s 0/ summari::alion stra!egy on reading comprehension 0/ non-pr，ψcienl Taiwanese university EFL learners. Unpub1ished doctoral

dissertation , Uni vcrs ity of Kansas.

Oxford, R. L. (Ed.). (1996). Language learning stralegies around the world: cross­

cullural perspeclives. Manoa: Uni versity of Hawai’i Press.

*Pappa, E., Zafiropouloll , M., & Mctallidou, P. (2003). Intervention on strategy use and on

motivation of Grcek pllpils' reading comprehcnsion in Engli sh classes. Perceplual

and Moto Skills. 96. 773-786.

*Park, Y.-H. (2010). Korcan EFL colleges stlldents’ rcad ing strategy use to comprchcnd

authentic expository/technical texts in Engli sh. Unpllblishcd doctoral disSCItation.

University of Kansas.

*Park, Y.-y. (1996). Sclεregulatcd strategy training in sccond langllage reading: Its cffccts

on reading comprehension, strategy lIse, rcading attitlldcs, and leaming sty lcs of

university ESL students. Foreign Language Educalion 2 (1), 59-80.

*Prinz, P. (1998). The 띠fluence 0/ slrategic leaching on reading in a second language

Unpllblished doctoral disscltation. 80ston Univcrsity

Riley, L., & Harsch, K. (1999). Enhα1cing Ihe learning experience with slralegy journals

Supporling the diverse learning styles 0/ ESL/EFL sludenls. Paper presentcd at

HERDSA Annuallntemational Conference, Melboume.

Sa[ataci, R., & Akyel, A. (2002). Possible effects of strategy instruction on L 1 and L2

reading. Reading in a Foreign Language, 14(1), 1-17.

Schuel\er, 1. (1999). The 행ects 0/ two types 0/ slrategic training on /oreign language

reading comprehension: An analysis by gender and pn야ciency. Unpublished

doctora1 di ssertation, The University ofWisconsin-Madison, WI

Sheorey, R., & Mokhtari, K. (2001). Differences in the metacognitive awareness of

reading strategies among native and non-native speakers. System, 29(4),43 1-449.

Singhal, M. (1999). The ψects 0/ reading stralegy instruction on Ihe reading

comprehension, reading process and slralegy use 0/ adu /t ESL readers

Unpublished doctoral disscrtation, Uni vcrsity of Arizona

Song, M. 1. (1998). Tcaching rcading strategies in an ongoing EFL university reading

classroom. Asian Journal 0/ English Language Teaching, 8, 4 [-54.

*Song, M. 1. ( 1999). Readin

22 Yong-Hyo Park

*Torut, B. (1 994). Metacognitive strategy awareness and EFL reading comprehension: A

study of Thai secondary students. Unpublished doctoral dissertation, Southem

Illinois University, at Carbondale.

*Tsai, C. -H. (2000). Comprehension strategies used by technical college EFL students in

Taiwan in comprehending English expositolγ texts. Unpublished doctoral

dissertation, University of Mississippi.

Wilson, B.D. (2001). 댐ect size determination program, from the World Wide Web:

http ://mason.gmu.edω'-dwilsonb/downloadsIES Ca1culator.xls

WlI, c.-P. (2005). An investigation of metacognitive reading strategies used by EFL

Taiwanese college students to comprehend familiar versus u까miliar Chinese and

English texts. Unpublished doctoral dissertation, University ofIdaho

Zhang, L. 1. (2001). Awareness in reading: EFL students’ metacognitive knowledge of

reading strategies in an acqllisition-poor environment. Language Awareness. 10(4), 268-288.

*Zhicheng, Z. (1992). The 행ects of teaching reading strategies on improviη'g reading

comprehensionfor ESL learners. Paper presented at the annual meeting ofthe Mid­

South Educational Research Association. Knoxvi l1e, TN. November 11-13. (ERIC

Document Reproduction Service No. ED 356 643).

Applicable levels: secondary and post-secondary education Key words: reading comprehension, reading sπategy， meta-analysis

Yong-Hyo Park 684-1 Jurye-Dong, Sasang-Gu, Busan 617-839 Te l.: 010-5322-5690 Email: lamesa5812@yahoo.co.kr

Received in June, 2010 Reviewed in Ju1y, 2010 Revised version received in August, 2010