toward modeling reading comprehension and reading fluency ... · between grade 1 phonological...

25
Toward modeling reading comprehension and reading fluency in English language learners Zohreh Yaghoub Zadeh Fataneh Farnia Esther Geva Published online: 21 August 2010 Ó Springer Science+Business Media B.V. 2010 Abstract This study investigated the adequacy of an expanded simple view of reading (SVR) framework for English language learners (ELLs), using mediation modeling approach. The proposed expanded SVR included reading fluency as an outcome and phonological awareness and naming speed as predictors. To test the fit of the proposed mediation model, longitudinal data from 308 ELLs from different linguistic backgrounds were analyzed using structural equation modeling. We examined the mediating role of Grade 2 word-level reading skills in the association between Grade 1 phonological awareness, naming speed, and listening compre- hension and Grade 3 reading comprehension and reading fluency. The results indicated that word-level reading skills fully mediated the association between phonological awareness, reading comprehension and reading fluency. Word-level reading skills partially mediated the association between naming speed and reading fluency. Listening comprehension contributed directly to reading comprehension and reading fluency. It appears that reading development in ELLs is better under- stood when reading fluency is added to the SVR framework as an outcome and naming speed as a building block of SVR. Theoretical aspects of the mediation model in relation to ELL reading development are also addressed. Z. Yaghoub Zadeh (&) Directions Evidence and Policy Research Group, 1055 Dunsmuir, Suite 1254, Four Bentall Centre, P.O. Box 48448, Vancouver, BC V7X 1A2, Canada e-mail: [email protected] F. Farnia Adolescent Biliteracy Development, Department of Human Development and Applied Psychology, The Ontario Institute for Studies in Education, Hincks-Dellcrest Centre/Institute, Department of Psychiatry, University of Toronto, 252 Bloor St, West Toronto, ON M5S 1V6, Canada E. Geva Department of Human Development and Applied Psychology, The Ontario Institute for Studies in Education, University of Toronto, 252 Bloor St, West Toronto, ON M5S 1V6, Canada 123 Read Writ (2012) 25:163–187 DOI 10.1007/s11145-010-9252-0

Upload: doankhue

Post on 28-Mar-2019

221 views

Category:

Documents


0 download

TRANSCRIPT

Toward modeling reading comprehension and readingfluency in English language learners

Zohreh Yaghoub Zadeh • Fataneh Farnia •

Esther Geva

Published online: 21 August 2010

� Springer Science+Business Media B.V. 2010

Abstract This study investigated the adequacy of an expanded simple view of

reading (SVR) framework for English language learners (ELLs), using mediation

modeling approach. The proposed expanded SVR included reading fluency as an

outcome and phonological awareness and naming speed as predictors. To test the fit

of the proposed mediation model, longitudinal data from 308 ELLs from different

linguistic backgrounds were analyzed using structural equation modeling. We

examined the mediating role of Grade 2 word-level reading skills in the association

between Grade 1 phonological awareness, naming speed, and listening compre-

hension and Grade 3 reading comprehension and reading fluency. The results

indicated that word-level reading skills fully mediated the association between

phonological awareness, reading comprehension and reading fluency. Word-level

reading skills partially mediated the association between naming speed and reading

fluency. Listening comprehension contributed directly to reading comprehension

and reading fluency. It appears that reading development in ELLs is better under-

stood when reading fluency is added to the SVR framework as an outcome and

naming speed as a building block of SVR. Theoretical aspects of the mediation

model in relation to ELL reading development are also addressed.

Z. Yaghoub Zadeh (&)

Directions Evidence and Policy Research Group, 1055 Dunsmuir, Suite 1254,

Four Bentall Centre, P.O. Box 48448, Vancouver, BC V7X 1A2, Canada

e-mail: [email protected]

F. Farnia

Adolescent Biliteracy Development, Department of Human Development and Applied Psychology,

The Ontario Institute for Studies in Education, Hincks-Dellcrest Centre/Institute,

Department of Psychiatry, University of Toronto, 252 Bloor St,

West Toronto, ON M5S 1V6, Canada

E. Geva

Department of Human Development and Applied Psychology, The Ontario Institute for Studies

in Education, University of Toronto, 252 Bloor St, West Toronto, ON M5S 1V6, Canada

123

Read Writ (2012) 25:163–187

DOI 10.1007/s11145-010-9252-0

Keywords English language learners � Reading comprehension �Reading fluency � Simple view of reading � Mediation modeling � Primary level

Introduction

According to the Simple View of Reading (SVR), reading comprehension is a

product of the joint effect of word-level reading skills (decoding) and linguistic

comprehension (Hoover & Gough, 1990; Gough & Tunmer, 1986). The SVR

framework has been the focus of numerous studies that examined its adequacy in

addressing the complexities of reading comprehension. For example Kirby and

Savage (2008) maintained that in spite of the broad appeal for SVR framework, it is

not sufficiently specified. This framework does not address the relationship between

reading comprehension and reading fluency, nor does it explicitly address the role of

underlying cognitive processes in reading comprehension. The adequacy of SVR

framework is not well understood in the context of English Language Learners

(ELLs), that is, students whose home language is different from English, the societal

and school language. The present study targeted ELLs, and examined longitudinally

the adequacy of an expanded mediation SVR framework that includes reading

fluency as an outcome, word-level reading as a mediator, and cognitive processes as

predictors of reading fluency and reading comprehension.

Considering a longitudinal expanded mediation SVR framework

Very few published studies (e.g., Gottardo & Mueller, 2009; Proctor, Carlo, August,

& Snow, 2005) have examined the reading comprehension of ELLs within the SVR

framework, though parts of the model have been examined in various second

language (L2) contexts. In particular, there is evidence that word-level reading and

reading comprehension skills are highly correlated in L2 learners, just as they are in

monolinguals (Chiappe, Siegel, & Wade-Woolley, 2002; Lesaux, Lipka, & Siegel,

2006; van Gelderen et al., 2004; Verhoeven, 2000), and that word reading fluency

(conceptualized in terms of accuracy and speed) correlates with reading compre-

hension (e.g., van Gelderen et al., 2004).

It is also well-documented in the L2 literature that oral language is strongly

related to literacy outcomes such as reading comprehension and reading fluency

(e.g., Droop & Verhoeven, 2003; Geva & Yaghoub Zadeh, 2006; Lesaux, Rupp, &

Siegel, 2007; Miller et al., 2006; Nakamoto, Lindsey, & Manis, 2008; Proctor et al.,

2005), but weaker in relation to accurate word-level reading skills (for a systematic

review, see Geva, 2006). Unlike children learning to read in their first language

(L1), ELLs have, by definition, less developed oral language skills to draw on when

they read for fluency and comprehension in their L2. Because reading for fluency or

comprehension may be a more challenging task for ELLs than for their monolingual

counterparts, they may need to rely more heavily on basic cognitive skills such as

phonological awareness and naming speed that are less dependent on language

proficiency to support the decoding of the written text. For example, in a study of

Grade 2 ELLs and monolingual English speaking (EL1) students, Geva and

164 Z. Yaghoub Zadeh et al.

123

Yaghoub Zadeh (2006) found that phonological awareness, rapid naming, accurate

word recognition, and oral language proficiency, concurrently predicted reading

fluency in ELLs, but for EL1s only rapid naming and word recognition predicted

reading fluency, and the contribution of language proficiency was negligible in this

group. This study, however, did not examine whether phonological awareness and

naming speed would make additional longitudinal contributions to reading fluency,

over and above their role in word-level reading skills. In another longitudinal study,

Lesaux et al. (2007) showed that there were associations between phonological

awareness, word recognition, and oral language assessed in kindergarten and Grade

4 reading comprehension.

Additional nuances concerning the direct or mediated nature of the relations

between underlying cognitive skills, word reading and reading comprehension, and

the validity of the SVR framework for L2 learners were reported in a recent study of

Spanish-speaking ELLs (Gottardo & Mueller, 2009). In this two-year, longitudinal

study, the relations between phonological awareness and language proficiency

assessed in Grade 1 in children’s L1 (Spanish) and their L2 (English) were used to

predict word reading accuracy and reading comprehension in Grade 2. The

researchers tested the SVR using structural equation modeling (SEM) and

concluded that the SVR framework is indeed a valid framework for understanding

the English reading comprehension skills of these children. In particular, the results

showed that oral language skills assessed in Grade 1 and word reading skills

assessed in Grade 2 contributed to Grade 2 reading comprehension. However, unlike

Lesaux et al.’s (2007) findings, phonological awareness measured in Grade 1 did not

contribute to reading comprehension directly but rather through accurate word

recognition in Grade 2.

Proctor et al. (2005) examined the reading comprehension of Grade 4 Spanish-

speaking ELLs within the SVR framework. Using path analysis, these researchers

examined concurrently the contribution of two language proficiency measures

(vocabulary and listening comprehension), word reading fluency, and reading

comprehension. They reported that Grade 4 vocabulary contributed to reading

comprehension directly and indirectly through listening comprehension, but that

Grade 4 word reading fluency had a lesser effect on Grade 4 reading comprehension.

Evidence from studies involving monolinguals suggests that text reading fluency has

a stronger relationship with reading comprehension than does word reading fluency.

It has been argued that text reading fluency plays a more prominent role in reading

comprehension than word reading fluency because it is a more complex task that

draws not only on word-level accuracy and speed, but also on the understanding of

connected discourse (cf. Cutting, Materek, Cole, Levine, & Mahone, 2009; Jenkins,

Fuchs, van den Broek, Espin, & Deno, 2003). In light of this evidence coming from

the L1 literature, it may not be surprising that Proctor et al. (2005) did not find a

correlation between word reading fluency and oral language skills of ELLs.

The inconsistent findings concerning the role of reading fluency in L2 reading

comprehension may be due to different analytical and modelling approaches,

diversity in sample characteristics, the nature of the reading fluency tasks used,

different time frames (concurrent or longitudinal), and different research objec-

tives. Given that the nature of reading changes with schooling and development,

Mediation model of ELL reading 165

123

it is necessary to carry out research that delineates the longitudinal relations

between reading-related skills that develop early, reading competence that builds on

these early skills, and subsequent reading comprehension in ELLs. To the best of

our knowledge, to date, no longitudinal study of ELL reading has attempted to

expand the SVR framework by examining the role of word-level reading as

mediating between prerequisite skills that develop early, and the subsequent

emergence of higher order text processing outcomes, such as reading comprehen-

sion and reading fluency. Mediation modeling (Maxwell & Cole, 2007) is a useful

methodological tool for unpacking the complexity of longitudinal associations

between these reading components.

Reading fluency as an outcome in an expanded SVR framework

Reading fluency is often conceptualized as involving accuracy and speed of reading

words in isolation and in text (Crosson & Lesaux, 2010; Meyer & Felton, 1999;

Torgesen, Rashotte, & Alexander, 2001). This definition stems from automaticity

theories which posit that effortless reading results in less involvement of cognitive

resources in lexical retrieval, and leads to allocation of cognitive resources to higher

level reading comprehension (Perfetti, 2007).

Slocum, Street, and Gilberts (1995) reviewed correlational and experimental

research on the association between reading comprehension and reading fluency in

monolingual students. They concluded that although correlational studies point to

an association between reading fluency and reading comprehension, experimental

studies failed to show that enhancing students’ reading fluency (speed) improved

their reading comprehension. They also concluded that the extent of this association

may vary as a function of the type of reading comprehension measures used in

different studies. Relatedly, in a recent review of the research, Collins and Levy

(2008) discussed the nature of the relationship between reading comprehension and

reading fluency. They concluded that reading comprehension and reading fluency

develop side by side and share similar underlying factors such as text representation.

Studies that examined reading fluency as a predictor of reading comprehension

provide further evidence on the lack of association between reading fluency and

reading comprehension. For example, Adlof, Catts, and Little (2006) examined the

association between reading fluency and reading comprehension in monolingual

students in Grades 2, 4, and 8. Their findings indicate that reading fluency did not

add any unique variance to the SVR framework.

Research focusing on monolingual students has shown that, similar to reading

comprehension, text reading fluency is associated with oral language (e.g., Cohen-

Mimran, 2009; Cutting et al., 2009; Puranik, Petscher, Al Otaiba, Catts, & Lonigan,

2008) and with word-level reading skills (Biemiller, 1999; Carver & David, 2001;

Wolf & Katzir-Cohen, 2001). Such findings suggest that reading comprehension and

reading fluency draw on similar prerequisite processing skills such as phonological

awareness and naming speed that are related to word-level reading skills. In other

words, reading fluency and reading comprehension of ELLs may be considered as

two separate, complex aspects of reading that draw, to some extent, on similar

166 Z. Yaghoub Zadeh et al.

123

underlying predictors. Over time, however, these two components may become

mutually facilitating as is the case in EL1s (e.g., Jenkins et al., 2003).

Riedel (2007) examined the association between oral reading fluency and reading

comprehension in a large sample of children in first and second grade, the majority

of whom were EL1 students. Riedel found that students with adequate levels of oral

reading fluency but poor reading comprehension had lower vocabulary scores than

those with adequate levels of oral reading fluency and reading comprehension.

Similar results were reported for ELLs by Buly and Valencia (2002). They

conducted a cluster analysis to determine whether word identification, reading

fluency and reading comprehension were similar across the majority of students or

represented various patterns for different groups of students. Buly and Valencia

reported that in two clusters students had relatively stronger word recognition and

fluency skills than they did in reading comprehension, and that more than 60%

(n = 12) of the students in these two clusters were ELLs. These findings suggest

that the association between reading fluency and reading comprehension in ELLs is

not as strong as it is in EL1s. Buly and Valencia (2002) attributed the weak

association between reading fluency and reading comprehension in ELLs to the lack

of English language proficiency. In a similar vein, Wiley and Deno (2005) studied

the association between oral reading fluency and reading comprehension in Grade 3

and Grade 5 ELLs and EL1s. They found a stronger association between oral

reading fluency and reading comprehension in EL1s than in ELLs. They also

reported that the association between oral reading fluency and reading comprehen-

sion was stronger in the older ELLs than in younger ELLs.

A recent study by Crosson and Lesaux (2010) involving fifth grade Spanish-

speaking ELLs provides additional support for the notion that the relationship

between reading fluency and reading comprehension may not be identical in ELLs

and EL1s. They focused on the role of English language proficiency in the

concurrent association between reading fluency and reading comprehension.

Crosson and Lesaux reported that text reading fluency was associated with reading

comprehension in the case of ELLs with high levels of oral language proficiency,

but not for ELLs with low levels of oral language proficiency.

Taken together, these studies suggest that the relationship between reading

comprehension and reading fluency is not identical in EL1s and ELLs, and it

probably varies as a function of the age of the learners and their language

proficiency. In the early stages of learning to read, when oral language skills are not

well developed, the association between oral reading fluency and reading

comprehension may be low in ELLs. This body of research suggests that it may

be of theoretical value to consider an expanded SVR framework, in which reading

fluency and reading comprehension are treated as distinct, yet related, parallel

outcome behaviors.

Cognitive processing skills as predictors in an expanded SVR framework

Ample research involving the SVR framework supports the view that oral language

and word-level reading skills play an important role in understanding reading

comprehension. However, the SVR framework ignores cognitive processes that have

Mediation model of ELL reading 167

123

been shown to play a significant role in predicting reading comprehension in

monolingual (Cain, Oakhill, & Bryant, 2004) and bilingual learners (e.g., van

Gelderen, Schoonen, de Glopper, & Hulstijn, 2007). Previous research has shown that

both phonological awareness and naming speed are predictors of word-level reading

(e.g., Bowers, 1995; McBride-Chang, Wagner, & Chang, 1997; Wagner et al., 1997;

see also Vukovic & Siegel, 2006 for a review). Phonological awareness has been

shown to contribute to reading comprehension in monolingual (e.g., Cain, Oakhill, &

Bryant, 2000) and in second language learners (Carlisle, Beeman, Davis, & Spharim,

1999; Manis, Seidenberg, & Doi, 1999; Proctor et al., 2005; Verhoeven, 2000).

Furthermore, research on monolingual students has shown that processing skills

such as naming speed are related to reading comprehension concurrently and

longitudinally (Johnston & Kirby, 2006; Joshi & Aaron, 2000; Parrila, Kirby, &

McQuarrie, 2004). These studies provide support for the unique role that

phonological awareness and naming speed may play in reading comprehension in

monolingual children, over and above the known SVR components. What is not yet

clear is the extent to which these findings are applicable to models of ELL reading

comprehension, and whether phonological awareness and naming speed exert their

role on reading comprehension directly, or their influence is mediated through word-

level reading skills.

Mediation modeling: rationale and procedures

Mediation modeling is one of the best available statistical procedures to model

simultaneously the nature of the interrelationship between hypothesized precursors,

mediator(s), and outcomes (Shrout & Bolger, 2002). Mediation can be best

modelled when using longitudinal databases, because the sequence of data points

allows the direction of effect to be modeled. Importantly, modeling mediation

follows a specific procedure that does not require the inclusion of autoregressors in

the model (e.g., Baron & Kenny, 1986). It is notable that although mediation

procedures allow one to model the direction of the effects among various constructs,

it is important to be mindful of the fact that causal conclusions can only be made

with caution in the absence of an experimental design (Shrout & Bolger, 2002).

When testing integrative models using procedures such as SEM, intercorrelations

among the predictor variables are taken into account. Therefore, unlike regression

and path analysis approaches, SEM is considered to be an appropriate analytical

technique for multivariate data analyses that enables testing mediation models that

highlight longitudinal, developmental relationships among the components.

In the context of reading development in ELLs, a mediation approach facilitates

unpacking associations between precursors of reading (e.g., phonological aware-

ness, naming speed, language comprehension), the hypothesized mediator, namely,

word-level reading, and outcome variables, namely, reading comprehension and

reading fluency. Furthermore, the mediation approach allows for the possibility to

be examined that some early predictors exert their influence on reading outcomes,

whether directly and/or indirectly, through the mediator. This elaboration is

necessary in order to examine the adequacy of an expanded SVR framework for

understanding reading comprehension and reading fluency in ELLs.

168 Z. Yaghoub Zadeh et al.

123

To the best of our knowledge, no previous studies have used mediation modelling

as an analytical approach for understanding reading development in young ELLs.

From a theoretical perspective, a mediation model is juxtaposed with a direct model.

It is possible to think of the direct model as a benchmark in which all precursors

have a direct effect on all reading constructs, namely, word-level reading, reading

comprehension, and reading fluency. According to the direct model, the contribu-

tions of all hypothesized prerequisite cognitive and language skills to reading

measures are independent and direct.

When the mediation model provides the best fit, it may support partial or full

mediation. In the present context, partial mediation might show, for example, that

phonological awareness or naming speed not only contribute to the reading

outcomes through the mediator (in this case, word-level reading), but also that

contribute directly to the outcome measures. Alternatively, full mediation would

indicate that the only contribution of the prerequisites to the outcome measures is

through the mediator. Note that, regardless of what model is supported, it is

presumed that listening comprehension, an aspect of language proficiency, will be

directly related to the outcome measures (i.e., reading comprehension and reading

fluency). The extent to which the results of the mediation model support the SVR

framework depends on whether full or partial mediation is supported. Full mediation

of word-level reading between earlier phonological awareness and naming speed

and subsequent reading outcomes would confirm the adequacy of the SVR

framework. Support for partial mediation might suggest that the SVR is not

sufficient to understand the attainment of reading comprehension and reading

fluency in primary level ELLs. In this study we fitted two models to compare the

direct-effect and the mediation models. No direction of effect was proposed for the

concurrent hypothesized prerequisite constructs assessed in Grade 1 or for the

outcomes measured in Grade 3, though correlations between precursor measures

and outcome measures were assumed.

To examine the direct-effect model (see Fig. 1), we modeled all possible

longitudinal direct paths from Grade 1 predictors (i.e., phonological awareness,

naming speed, and listening comprehension) to the mediator (i.e., word-level

reading in Grade 2), and to the outcome variables (reading fluency and reading

comprehension in Grade 3). We expected that listening comprehension would

contribute directly to reading comprehension and reading fluency. However, given

previous research findings (August & Shanahan, 2006), a significant path was not

expected between listening comprehension and word-level reading. Similarly, based

on previous findings (Manis et al., 1999; Pennington, Cardoso-Martins, Green, &

Lefly, 2001; Torgesen, Wagner, Rashotte, Burgess, & Hecht, 1997; van Gelderen

et al., 2004), no significant path was expected from naming speed to Grade 3

reading comprehension.

In the mediation model (Fig. 1), we added two paths: one from the word-level

construct to reading fluency (path A), and one from the word-level construct to

reading comprehension (path B). We hypothesized that these two paths would be

significant. For a full mediation model to be supported, it was expected that the

direct paths from phonological awareness to reading comprehension and reading

fluency, and the path from naming speed to reading fluency, would not be

Mediation model of ELL reading 169

123

significant. For a partial mediation model to be supported, the magnitude of the

relationships between phonological awareness, naming speed, and the outcome

measures were expected to decrease significantly (in comparison to the direct

model).

In sum, we hypothesized that the mediation model would fit the data better than

the direct-effect model. More specifically, we expected that as ELLs gradually

develop their reading skills, (a) Grade 1 rapid naming would contribute both directly

and indirectly (through Grade 2 word-level reading skills) to reading fluency; (b)

Grade 1 phonological awareness would contribute to Grade 3 reading comprehen-

sion and reading fluency either indirectly through Grade 2 word-level reading (full

mediation), or both directly and indirectly through word-level reading (partial

mediation); and (c) Grade 1 listening comprehension would contribute directly to

Grade 3 reading comprehension and reading fluency.

Method

Participants

Longitudinal data from 308 ELLs from diverse linguistic backgrounds were

collected in three sequential cohorts. The students came from 12 schools spread

across four boards of education in a large metropolitan area in Canada. Thirty-five

classes were involved. Fifty-two percent of the participants were male. The

participants came from a variety of home language backgrounds comprising 33%

Punjabi, 23% Portuguese, 14% Tamil, 14% Cantonese, 11% from three language

groups (Urdu, Hindi, and Gujarati), and 5% from other language backgrounds.

All three cohorts were first assessed in Grade 1; they were drawn from the same

schools and there were no changes in schools’ curricula or policy during the study.

The data from the three cohorts were combined and all analyses were done on the

A

B

Listening Comprehension

Grade 1

Phonological Awareness

Grade 1

Naming Speed Grade 1

Word-levelReadingGrade 2

Reading Comprehension

Grade 3

Reading Fluency Grade 3

Fig. 1 Direct-effect model (benchmark) and mediation model (dotted lines are added for mediationmodel)

170 Z. Yaghoub Zadeh et al.

123

total sample. In order to determine ELL status, the information gathered from a

number of sources was triangulated. The identification began with school

nominations as ELLs. This information was gathered in order to distribute the

appropriate translated consent form to parents. We then checked official school files

for all the nominated students to confirm the information. We also asked teachers to

identify students in their classrooms who spoke a language other than English at

home. This information was verified through parental consent forms and child

questionnaires. In order to make sure that ELL students had sufficient knowledge of

the English language to understand the English instructions, the testers were

instructed to chat with students before administering the tasks while accompanying

them from their class to the testing room. Before administering any of the tasks, the

testers also monitored whether the participants followed the instructions and did as

they were asked in order to develop an index of the adequacy of the students’

English oral language.

Typically, in Canada, recent immigrants from non-English speaking countries or

with limited English proficiency are placed in regular English classrooms. In the

province where the study took place, ELL students with minimal command of

English are withdrawn from their classrooms daily for 30–40 min of English

language instruction, provided by teachers with English as a second language (ESL)

specialist training. The ESL classes comprise students of various ages and home

language backgrounds, and they are grouped by level of English language

proficiency. ELLs receive instruction in ESL classes for up to 2 years. For the

remainder of the day, the students are integrated into the regular classroom. Regular

classroom teachers are expected to make appropriate adaptations to the program-

ming and curriculum for their ELLs.

Demographic background

We were not able to obtain demographic information directly from parents.

However, we were able to access Canadian census data to obtain demographic

information in the neighborhoods where the schools were situated, by using relevant

postal codes. This information provided useful information that helped to

contextualize the study. According to the 2001 Canadian Census, about 58% of

the families living in the neighborhoods where the participating schools were

located reported a language other than English or French (the two Canadian official

languages) as the home language. About 91% of the families were first-generation

immigrants, and 68% of the adults immigrated when they were 20 years of age or

older. The average poverty rate in these neighborhoods was 23%, ranging from 0 to

50%. The median income of these families was considerably lower than the median

income for the metropolis in which they lived. There was also substantial variation

in the level of education of the adults living in these neighborhoods: 36% of the

individuals living in the relevant postal code blocks had not obtained a high school

diploma or had not finished high school, 13% had a high school diploma, 27% had

either a trade certificate or college education, and 20% had obtained at least a

bachelor’s degree.

Mediation model of ELL reading 171

123

Grade 1 measures

Phonological awareness

Two measures of phonological awareness skills were used: the Auditory Analysis

Task and the Oddity Task.

Auditory analysis task An adapted version of the Auditory Analysis Task (AAT)

developed by Rosner and Simon (1971) was used to measure students’ phonological

awareness. To minimize the effect of lexical knowledge, only high frequency words

were used for the initial stimuli and target responses (e.g., sunshine, picnic, leg).

The 20-item task consists of 3 subtests of progressive difficulty. In the first subtest,

students were asked to delete one syllable morpheme in either initial or final

position (e.g., ‘‘Say sunshine’’; ‘‘Say it again but don’t say ‘shine’’’). The second

subtest aimed at the isolation and deletion of initial or final single phonemes in one-

syllable words (e.g., ‘‘Say hand’’; ‘‘Say it again but don’t say the/h/’’). The third

subtest involved deletion of single phonemes in initial or final consonant blends

(e.g., ‘‘Say ‘left’’’; ‘‘Say it again without the/f/’’). The test was discontinued after

five consecutive errors. Each correct answer scored one point. The Cronbach acoefficient was 0.92 for the sample used in this study.1

Oddity task This is an experimental task in which children listened to a series of

three, single-syllable CVC pseudowords played on a tape-recorder (e.g., wom, wob,

vog) and were asked ‘‘Which one starts with a different sound? ‘wom, wob, vog’?’’.

The same vowel was used within all items in a set. As each item was presented, the

experimenter pointed to a corresponding wooden counter (e.g., a square, a star, or a

triangle). A tone separated each set of items and alerted children to the next set. To

ensure that children remembered the set of items, the entire sequence was presented

twice in a row. Three practice items and 19 test items were presented in a fixed

sequence. The raw score was used in the analyses. The Cronbach a coefficient was

0.70 for the sample used in this study.

Naming speed

Two subtests of rapid automatized naming (RAN) developed by Denckla and Rudel

(1976) were used to measure naming speed: letters and objects. Such tasks tap basic

lower level cognitive processes by estimating the speed with which participants

access the names of highly automatized printed symbols (Bowers, Golden,

Kennedy, & Young, 1994; Wolf, Pfeil, Lotz, & Biddle, 1994).

Letter naming This task consists of the presentation of a series of five highly

frequent letters of the English alphabet (O, A, S, D, P). Each letter appears 10 times

in random order. Participants are instructed to name the items as quickly and

1 Note that when data collection commenced, commercial, standardized measures of phonological

processing, such as the CTOPP, were not yet available.

172 Z. Yaghoub Zadeh et al.

123

accurately as possible. Accuracy and time (in seconds) in naming all 50 items were

recorded.

Object naming This task consists of the presentation of a series of five highly

frequent objects (i.e., table, door, box, ball, hat). Each of the items appears 10 times

in random order. Participants are instructed to name the objects as quickly and

accurately as possible. Accuracy and time (in seconds) in naming all 50 items were

recorded.

The standardized scores of the two naming speed measures were calculated by

converting the speed in seconds and the number of errors to respective Z scores.

Listening comprehension

Listening comprehension (LC), as an indicator of linguistic comprehension, is an

experimental measure adapted from the Durrell Analysis of Reading Difficulty

(Durrell, 1970). This measure comprises two short stories (about a paragraph in

length) that represent different difficulty levels (Merbaum & Geva, 1998). Each

story is read to the child, and the child is instructed to pay attention because he/she

will be asked to retell the story and answer some questions about it. LC was

evaluated in two complementary manners. There were eight idea units in each story.

After listening to each story, the child was asked to retell it, and answer one

inferential and four factual questions which were presented orally to the child. Both

Story 1 and Story 2 had a maximum score of 13.

Children’s story retelling and answers were tape-recorded. The recordings were

later transcribed and scored by two native English-speaking raters. For the free

recall component, children were given one point for each idea unit recalled. One

point was also given for each correctly answered oral comprehension question.

Children were not penalized for making grammatical errors in the free recall or the

question–answer components of this task. There was an 85% agreement rate

between the two raters. However, following discussion of answers that were not

initially agreed upon, the raters were able to reach a full consensus on all protocols,

and the resulting scores were used in the analyses. The Cronbach a coefficient was

0.76 for the sample used in this study.

Grade 2 measures

Word-level reading skills

Two measures were used to assess children’s word-level reading skills: a word

identification test, and a pseudoword decoding test.

Word identification The word identification subtest of the Wide Range Achieve-

ment Test-Revised (WRAT-R; Wilkinson, 1993) was used to assess children’s

ability to read isolated words in English. WRAT is a standardized test with an

internal consistence of 0.88 at Grade 2. This test consists of 42 monosyllabic and

polysyllabic words. The word items involve nouns, verbs, adjectives, and

Mediation model of ELL reading 173

123

prepositions. The test was discontinued after 10 consecutive errors. The total

number of correctly read words was considered as each child’s score on the test.

Pseudoword decoding The Word Attack subtest of the Woodcock Reading

Mastery Test-Revised (Woodcock, 1987) was administered to assess children’s

ability to employ grapheme-phoneme correspondence rules in decoding pseudo-

words. The test consists of 45 items that conform to the rules of English

orthography, but are not real words in English (e.g., ‘‘bufty’’, ‘‘mancingful’’). The

total of correctly read items was considered each child’s total score. The split half

reliability reported for Grades 1–3 ranged from 0.91 to 0.94.

Grade 3 measures

Reading comprehension

An experimental measure of reading comprehension was adapted from the Durrell

Analysis of Reading Difficulty (Durrell, 1970). Children were asked to read aloud

three short stories. They were instructed to pay close attention to the stories. These

were not the same stories used for the LC condition. Children were asked to retell

each story and then answer five open-ended questions, four of which were of a

factual nature (e.g., ‘‘What did the men look like?’’) and one which was inferential

(e.g., ‘‘Where was the money returned to?’’). The children’s story retelling and

responses to the questions were tape-recorded. As in the LC condition, they were

given one point for each idea unit recalled and one point for each correct answer.

There was an 87% agreement rate between the raters. However, following

discussion of answers that were not initially agreed upon, the raters were able to

reach a full consensus on all protocols, and the resulting ratings were used in the

analyses. The Cronbach a coefficient was 0.83 for the sample used in this study.

Reading fluency

Two subtests of the Biemiller Test of Reading Processes (Biemiller, 1981) were used

to measure reading fluency, oral text reading fluency, and oral word reading fluency.

Each subtest yields a measure of accuracy and a measure of speed of reading.

Oral text reading fluency Children were asked to read a short narrative text as

quickly as possible. The text consists of 100 primary level words.

Oral word reading fluency The children were asked to read, as quickly as possible,

a corresponding word list containing 50 randomly ordered words taken from the

narrative text described above.

The number of correctly read words within the word and text reading fluency

conditions yielded measures of word- and text-reading accuracy respectively, and

the number of seconds it took children to read the text and the words provided

corresponding measures of word and text reading speed. Errors and speed scores

were standardized to Z scores. The word fluency scores were based on the average

174 Z. Yaghoub Zadeh et al.

123

of the errors and speed Z scores (Stanovich & West, 1989). The same procedure was

used to calculate standard scores for text fluency. The lower the scores, the more

fluent the children are in reading words and texts.

Procedure

Consent forms in English and in the students’ home languages were distributed in

each of the participating classrooms. Only children with parental consent

participated. Students were tested on a large battery of tests, administered across

four testing sessions; each session lasted approximately 30 minutes. Students were

assessed in the winter/spring of each successive year. Testing was done on an

individual basis by fully trained graduate students and research assistants.

Results

Missing data points are unavoidable in longitudinal research. The sample size in Grade

1 was 308. The rate of attrition for data gathered in Grade 2 was about 27% (n = 225)

(i.e., word identification and word attack), and in Grade 3 about 42% (n = 179) (i.e.,

reading comprehension and reading efficiency measures). To examine whether

participants with partial data were different from participants with full data, we

compared the two groups on Grade 1 data. The two groups did not differ on measures

of listening comprehension, phonological awareness, naming speed or nonverbal

ability (see ‘‘Appendix’’). However, due to the bias that emerges from analyzing only

the data from participants with complete data, multiple imputation procedures were

used to estimate the missing data points. Multiple imputation is one of the best

procedures to deal with missing data (Allison, 2003; Collins, Schafer, & Kam, 2001;

Schafer & Graham, 2002). The LISREL 8.72 (Joreskog & Serbom, 2001) program was

used to impute the missing points using an expected maximization (EM) algorithm.

This procedure resulted in complete data for 308 ELLs.

All measures had normal distributions and nonsignificant skewness and kurtosis.

Table 1 presents means, standard deviations, and correlation coefficients for all

variables. On the whole, there were significant correlations among all variables of

interest. As can be seen in Table 1, there were significant bivariate correlations

among (a) early (Grade 1) cognitive and phonological processing predictors (i.e.,

phonological awareness, naming speed), and linguistic comprehension, and Grade 2

word-level reading (i.e., word identification and pseudoword decoding); (b) early

cognitive and phonological processing predictors and outcome variables (i.e.,

reading comprehension, reading fluency); and (c) measures of word-level reading

(i.e., word identification and pseudoword decoding), and outcome variables. In

addition, there were significant, albeit moderate, correlations between the two

outcome measures, reading comprehension and reading fluency.

Measurement model

In this study, we considered six latent variables: phonological awareness, rapid

naming, listening comprehension (assessed in Grade1), word-level reading skills

Mediation model of ELL reading 175

123

Tab

le1

Mea

ns,

stan

dar

dd

evia

tio

ns,

and

corr

elat

ion

coef

fici

ents

for

all

mea

sure

s

Mea

sure

s1

23

45

67

89

10

11

12

13

Gra

de

1

1.

LC

-S1

2.

LC

-S2

0.6

4–

3.

AA

T0

.33

0.3

5–

4.

Od

dit

y0

.14

0.2

00

.45

5.

Let

ter

nam

ing

-0

.05

-0

.02

-0

.10

-0

.13

6.

Ob

ject

nam

ing

-0

.08

-0

.13

-0

.15

-0

.12

0.3

9

Gra

de

2

7.

Dec

od

ing

0.2

40

.24

0.6

80

.47

-0

.19

-0

.20

8.

Wo

rdID

0.2

70

.27

0.6

70

.45

-0

.25

-0

.26

0.8

8–

Gra

de

3

9.

RC

-S1

0.3

40

.22

0.4

20

.25

-0

.01

0.0

40

.39

0.4

6–

10

.R

C-S

20

.43

0.4

00

.43

0.3

10

.15

-0

.07

0.4

70

.50

0.5

1–

11

.R

C-S

30

.46

0.3

70

.45

0.3

7-

0.0

8-

0.1

10

.48

0.5

30

.44

0.6

8–

12

.F

luen

cy-W

-0

.24

-0

.17

-0

.23

-0

.16

-0

.18

0.1

3-

0.2

7-

0.2

7-

0.2

7-

0.3

1-

0.3

5–

13

.F

luen

cy-T

-0

.28

-0

.24

-0

.11

-0

.09

0.2

60

.15

-0

.25

-0

.22

-0

.10

-0

.21

-0

.28

0.7

5–

Mea

ns

7.1

34

.30

7.0

21

0.6

60

.00

20

.01

11

4.0

32

4.6

96

.70

9.6

59

.65

-0

.03

-0

.03

SD

s3

.16

2.7

94

.21

3.6

50

.50

0.5

91

0.4

65

.11

1.4

42

.29

2.0

30

.33

0.2

8

All

corr

elat

ion

coef

fici

ents

above

0.1

1ar

esi

gnifi

cant

atp

\0

.001

;L

C-S

1li

sten

ing

com

pre

hen

sio

n-s

tory

1,

LC

-S2

list

enin

gco

mp

reh

ensi

on

-sto

ry2

,A

AT

aud

ito

ryan

aly

sis

task

,O

dd

ity

pse

udo

wo

rdfi

rst

ph

onem

eid

enti

fica

tio

n,

RC

-S1

read

ing

com

pre

hen

sio

n-s

tory

1,

RC

-S2

read

ing

com

pre

hen

sio

n-s

tory

2,

RC

-S3

read

ing

com

pre

hen

sio

n-s

tory

3,

Flu

ency

-Ww

ord

flu

ency

,F

luen

cy-T

tex

tfl

uen

cy

176 Z. Yaghoub Zadeh et al.

123

(assessed in Grade 2), and reading comprehension and reading fluency (assessed in

Grade 3). Each latent variable comprised two measures, except for reading

comprehension which consisted of three measures. We used confirmatory factor

analysis to test the measurement model. All the measures loaded significantly on the

respective latent variables. As shown in Fig. 2, factor loadings for Grade 1 predictor

measures ranged from 0.55 to 0.88; factor loadings for the two Grade 2 measures,

comprising the mediator, were 0.93–0.95, and factor loadings for Grade 3 outcome

measures ranged from 0.61 to 1.00.

Since chi-square is sensitive to sample size, we used fit indices that are less

sensitive to sample size to assess goodness of fit for the models. The ratio between

chi-square and degrees of freedom is considered a good fit when it is less than 3

(Cole, 1987; Kline, 1998). In this study, this ratio was 1.97 for the measurement

model. The root-mean squared error of approximation (RMSEA) is also one of the

indices that is less dependent on sample size, and a value of 0.06 or less indicates a

Phonological AwarenessGrade 1

Auditory analysis task

Oddity

.83

.55

Naming Speed Grade 1

Letters naming

Objects naming .58

.69

Word ID

Decoding

.95

.93

Word-level reading Grade 2

Listening Comprehension Grade 1

List. Comp-Story 1

List. Comp-Story 2

.88

.73

Word Fluency

Text Fluency

.76

1.0

Reading Fluency Grade 3

Reading Comp-Story 1

Reading Comp-Story 2

Reading Comp-Story 3

.61

.79

.85

Reading Comprehension Grade 3

Fig. 2 Measurement model: factor loadings on the six latent constructs. Note. v2 (46) = 90.03;RMSEA = 0.06; GFI = 0.96; AGFI = 0.92; CFI = 0.98; NFI = 0.97; NNFI = 0.97

Mediation model of ELL reading 177

123

model with good fit (Hu & Bentler, 1999). The RMSEA for the measurement model

was 0.06, indicating a good fit. Other indices of fit, independent of sample size, are

the model goodness of fit (GFI), adjusted goodness of fit (AGFI), comparative fit

index (CFI), normed fit index (NFI), non-normed fit index (NNFI); values of 0.90 or

higher indicated a good fit of the model. For the measurement model, all these

indices were above 0.90, indicating a good fit. Fit indices of the measurement model

(v2 (46) = 90.03; RMSEA = 0.06) indicated that the model fit the data well and it

was feasible to test the full models.

First, we fitted a direct and a mediation model to the data. We then compared the

two models in terms of their fit indices, including chi-square values and degrees of

freedom. This was done to determine which of the alternative theoretical models

best fit the data. The model with significantly lower chi-square would be the one that

best fits the data.

Direct-effect model

Figure 3 depicts the direct-effect model. For simplicity, only the structural models

with the estimated standardized coefficients for the paths is presented. The loadings

of the measures on the latent variables remained similar to the loadings presented in

Fig. 2. As expected, listening comprehension did not contribute to word-level

reading, but phonological awareness and rapid naming did. Of the three Grade 1

latent constructs, listening comprehension and phonological awareness were

directly related to reading comprehension in Grade 3, but rapid naming was not.

For reading fluency, the direct model indicated that all Grade 1 constructs were

related to reading fluency in Grade 3. The fit indices, and the ratio between chi-

square and degrees of freedom (2.01) indicated that the direct model fit the data well

(v2 (48) = 96.53; RMSEA = 0.06). This model explained 62% of the variance in

reading comprehension and 23% of the variance in reading fluency.

Naming SpeedGrade 1

.34

.58-.19

.95

.47

-.23 -.20

-.19

Listening Comprehension

Grade 1

Phonological Awareness

Grade 1

Word-levelReadingGrade 2

Reading Comprehension

Grade 3

Reading Fluency Grade 3

.30

Fig. 3 Direct-effect model: structural equation model indicating coefficients for all the significant paths.Dotted arrows indicate the non-significant path coefficients. v2 (48) = 96.53; RMSEA = 0.06,GFI = 0.95, AGFI = 0.91; CFI = 0.98; NFI = 0.97; NNFI = 0.97

178 Z. Yaghoub Zadeh et al.

123

Mediation model

To examine the fit for the mediation model (see Fig. 4), we added the paths from the

Grade 2 word-level reading construct to Grade 3 reading comprehension and

reading fluency. While the path from word-level reading to reading comprehension

was significant, the direct paths from Grade 1 phonological awareness and naming

speed to reading comprehension were nonsignificant. Instead, the word-level

reading construct fully mediated the association between phonological awareness,

naming speed, and subsequent reading comprehension. In other words, the effect of

phonological awareness and naming speed on reading comprehension was solely

through their effect on word-level reading. As hypothesized, the relationship

between listening comprehension and reading comprehension was direct, and not

mediated through word-level reading.

As for reading fluency, while the path from word-level reading to reading fluency

was significant, the direct path from Grade 1 phonological awareness to reading

fluency was nonsignificant. That is, the effect of phonological awareness on reading

fluency was solely through its effect on word-level reading. The standardized path

coefficient from naming speed in Grade 1 to reading fluency in Grade 3 decreased

from 0.30 to 0.20, once the path from word-level reading to reading fluency

construct was included, indicating that word-level reading partially mediated the

effect of naming speed on reading fluency. In other words, unlike phonological

awareness, naming speed made an additional contribution to reading fluency aside

from its contribution through word-level reading.

As hypothesized, the relationship between listening comprehension and reading

fluency was direct, and not mediated through word-level reading. The nonsignificant

path between the reading fluency and reading comprehension constructs in the

mediation and direct-effect models should be considered in conjunction with the

.42Listening Comprehension

Grade 1

Phonological Awareness

Grade 1

Naming Speed Grade 1

Word-levelReadingGrade 2

Reading Comprehension

Grade 3

Reading Fluency Grade 3

.46 .43

.87

-.23 -.20

.20

-.28

.51

Fig. 4 Mediation model: structural equation model indicating coefficients for all the significant paths.Note. Dotted arrows indicate the non-significant path coefficients; bolded arrows indicate mediationpaths. v2 (46) = 90.03; RMSEA = 0.06, GFI = 0.96; AGFI = 0.92; CFI = 0.98; NFI = 0.97;NNFI = 0.97

Mediation model of ELL reading 179

123

correlation tables. Table 1 indicates that there was a small but significant

association between measures of reading fluency and reading comprehension, prior

to fitting the structural model. The correlations hovered between 0.21 and 0.35 with

one exception (0.10). Therefore, the nonsignificant bidirectional path between the

reading fluency and reading comprehension constructs might be the result of

modeling the role of early predictors of these reading skills.

The fit indices of the mediation model indicated that the model fit the data well

(v2 (46) = 90.03; RMSEA = 0.06). The model explained 61% of variance in

reading comprehension and 25% of variance in reading fluency. Comparisons

between the fit indices for the direct-effect and mediation models indicated that the

mediation model had a significantly better fit than the direct-effect model

(Dv2 = 6.50, df = 2, p \ 0.05).

Finally, it is important to note that in both the direct model and the mediation

model, the correlation between reading comprehension and reading fluency was not

significant. As for the predictors, all correlations except the correlation between

naming speed and listening comprehension were significant.

Discussion

Findings of this study add to an emerging body of L2-based literature by suggesting

an expanded SVR framework. The study expands the SVR framework in three

interrelated perspectives. First, it affords a long-range perspective that delineates the

longitudinal relationships among component reading skills through a mediation

model. Second, it suggests that the expanded SVR framework needs to include

reading fluency and reading comprehension as outcomes, at least in the case of

young ELLs. Third, it draws attention to additional cognitive processes that underlie

reading comprehension and reading fluency in young ELLs. In what followswe

discuss the findings in turn, from each of these perspectives.

Component reading skills and subsequent reading outcomes: a mediation model

Research has shown that regardless of the L1s spoken by ELLs, cognitive-linguistic

processes, such as phonological awareness and naming speed, are more consistent

and potent predictors of subsequent word-level reading skills than are L2 oral

language skills (Geva, 2006). At the same time, well-developed language skills (in

conjunction with well developed word-level reading skills) are essential for deriving

meaning from texts. In this study we delineated the longitudinal relationships

among component reading skills that build on each other and develop gradually in

ELLs. The mediation model enabled us to examine the influence that early cognitive

and linguistic proficiency predictors (i.e., phonological awareness, naming speed,

listening comprehension) exert on reading outcomes, reading comprehension and

reading fluency (whether directly or indirectly), through word-level reading. By

design, and based on previous theoretical frameworks of reading development (e.g.,

Catts, Fey, Zhang, & Tomblin, 1999; Chall, 1983; Cutting & Scarborough, 2006;

Francis et al., 2005), in this study, certain skills were assessed at developmentally

180 Z. Yaghoub Zadeh et al.

123

appropriate times. The longitudinal, developmental perspective is an asset to

mediation modelling, but it raises interesting questions pertaining to causal

longitudinal relations between constructs. That is how findings might have differed

if for example reading fluency data were used at multiple time points and any

potential causal relations that could be explored may be the subject of future

research.

The expanded SVR framework demonstrates within a longitudinal framework

that the impact of oral language proficiency on subsequent reading comprehension

and reading fluency is direct, and independent of word-level reading skills. These

longitudinal relationships are present in ELLs whose language skills in English are

far from being at an optimal level. Importantly, even as ELLs continue to develop

their language skills, individual differences in their language proficiency is directly

related to subsequent reading comprehension and reading fluency. At the same time,

individual differences in language proficiency of ELLs are not related to the more

modularized word-level reading skills. Futhermore, the results point to the fact that

early predictors, such as phonological awareness and naming speed, are related to

subsequent reading outcomes in a more complex manner. Phonological awareness

exerts its influence solely through the mediator (word-level reading); naming speed

exerts its influence both directly and indirectly. That is, once ELLs have had

sufficient opportunities to develop their word-level reading skills, individual

differences in phonological awareness no longer contribute directly to the higher

level reading components (i.e., reading fluency and reading comprehension), though

they continue to do so, as shown, through the mediator. At the same time, individual

differences in naming speed continue to exert an influence on reading fluency

beyond their contribution to effortless word reading.

More generally, these findings should be considered in light of language exposure

and early literacy instruction. At the onset of systematic exposure of ELLs to

language and literacy skills in Grade 1, individual differences in underlying

processing skills, such as phonological awareness and naming speed, play a key role

in developing word-level reading skills. Gradually, with schooling, literacy

development, and systematic exposure to English, the word-level reading skills of

ELLs become more automatized, and their command of the societal and school

language improves. Improvement in word-level reading and language skills enables

ELLs to read texts with more fluency and ease, and with more comprehension.

Along with word reading skills, individual differences in language comprehension

continue to play a sustained role in reading fluency and reading comprehension.

An expanded SVR: reading fluency and reading comprehension as outcomes

As noted in the introduction, there is no agreement in the literature on the

relationship between reading fluency and reading comprehension. Some L1-based

researchers argue that reading fluency is a bridge from word identification to reading

comprehension (e.g., Bashir & Hook, 2009). Others maintain that reading fluency is

not merely a component of reading comprehension, but that it is an aspect of higher

level reading that is distinct from reading comprehension (Adlof et al., 2006;

Collins & Levy, 2008). Our ELL-based findings are in line with the latter position.

Mediation model of ELL reading 181

123

The univariate associations between reading comprehension and word and text

reading fluency in Grade 3 is significant but rather low. Once entered into the

mediation or direct-effect models, there is no significant association between these

two outcome constructs. In other words, once the prerequisite reading skills

(phonological awareness, naming speed, word reading, and language proficiency)

that underlie these two higher level reading components are modeled, the

association between them becomes nonsignificant.

These findings suggest that, to a large extent, the positive association between

reading fluency and reading comprehension depends on the factors that drive this

association. That is, at least in the case of young ELL students, the oft-cited

correlations between reading fluency and reading comprehension can be understood

in terms of common underlying factors. The findings support an argument for an

expanded SVR framework that takes a developmental stance, that includes reading

fluency and reading comprehension as outcomes, and that allows for direct and

indirect contribution of cognitive processes and language proficiency related skills to

the outcomes. Such a developmental framework provides a more complex, yet

parsimonious, model of the factors that contribute to subsequent reading achieve-

ment in ELLs. While reading comprehension and reading fluency draw on similar

processes, they are distinct constructs in the primary grades. As suggested elsewhere,

a closer alignment or amalgamation between reading comprehension and reading

fluency in ELLs is likely to emerge in later years (Wiley & Deno, 2005). This

distinction has important theoretical implications and implications for instruction.

While compelling, it is important to acknowledge that these conclusions might be

an artifact of the methodology used. For example, in this study, reading

comprehension was an untimed measure and the reading fluency measure focused

on accuracy and speed and not on meaning. The degree of association between these

two reading measures might have been stronger had we used a timed measure for

reading comprehension or a measure of reading fluency that included meaning. In

this regard it is also important to note that when testing for reading fluency, the

nature of the instructions might affect the results. Instructions of the kind given in

this study to read ‘‘as fast as you can’’ have been shown to affect participants’

performance as they are less likely to focus on accuracy or meaning (Colon &

Kranzler, 2006). In addition, reading development in ELL populations can be the

result of a complex interaction of linguistic and cultural factors which may impede

second language development. Lack of information on cultural factors is one of the

limitations of this study.

Cognitive processes that underlie reading comprehension and reading fluency

Notwithstanding contextual factors, such as instructional approaches, background

knowledge, and home literacy, that affect reading achievement and reading fluency

(beyond the scope of this paper), individual differences in language competence

underlie these longitudinal relationships. Even under optimal instructional and

contextual conditions, individual differences in L2 language competence exist. Even

under similar instructional conditions, some children will have the competence to

develop their English language skills faster and with more ease than others.

182 Z. Yaghoub Zadeh et al.

123

Naturally, these children are likely to attain subsequently better developed

comprehension of written language and more fluent reading skills. Before being

exposed systematically to the L2 in the school context, some ELLs are better

language learners than others. Good language learners acquire vocabulary faster, are

more sensitive to phonemic contrasts, are better at parsing morphemes and at

processing complex sentences, and have better developed metalinguistic skills. In

turn, in a cascading fashion, these early differences in language skills also underlie

the potential for acquiring the L2, and in the long run, enhance better reading

comprehension, more fluent reading, and further language development (for a

similar argument, see Sparks & Ganschow, 2001). In conjunction with language

skills, good word-level reading skills are essential for reading fluency and reading

comprehension. However, a model that includes only these building blocks is not

sufficient, in the case of reading fluency. Instead, reading fluency is better

understood when naming speed, an important underlying cognitive skill is added to

the SVR building blocks.

It is noteworthy that the expanded mediation SVR framework explains more than

twice as much variance in reading comprehension compared with reading fluency.

Other factors, not included in this study, such as short-term memory (Cohen-

Mimran, 2009), morphology (Cohen-Mimran, 2009), orthographic speed (Wood,

2009), orthographic representation (Berninger et al., 2010), and orthographic pattern

recognition (Katzir et al., 2006) may explain additional variance in reading fluency

and contribute further to this model. The results pertaining to reading fluency in

ELLs are in line with L1-based research pointing to a ‘‘complex view of reading

fluency’’ (Katzir et al., 2006, p. 77). Clearly, more research is needed to understand

what cognitive processes contribute to the reading comprehension and reading

fluency of ELLs, in addition to those associated with language comprehension and

word-level reading skills (Cain et al., 2004; Kirby and Savage, 2008; van Gelderen

et al., 2007).

This study expands the SVR framework for young ELLs coming from different

language backgrounds. However, because the sample size for students from

different language backgrounds was not large enough, it was not possible to

examine the mediation model for different language groups in this study. The extent

to which the predictability of this expanded mediation model might be upheld,

regardless of typological language differences and across different ages, is open for

further investigation.

These findings have practical implications for assessment of at-risk ELLs. Our

findings suggest that phonological awareness, naming speed, and oral language

measured in Grade 1 ELLs have predictive power for how well their reading

comprehension and reading fluency will develop subsequently. While mindful of

their ELL status, poor performance of young ELLs on phonological awareness,

naming speed and oral language can be a warning sign of potential difficulties in

their subsequent word reading, reading fluency, and reading comprehension. The

model suggests that early identification can take place even before ELLs

demonstrate reading problems. When failing to complete preliteracy tasks, such

as phonological awareness and speed of processing, ELLs should be supported to

develop these skills to avoid word reading problems. If this support is accompanied

Mediation model of ELL reading 183

123

with activities to enhance their linguistic comprehension, ELLs may be less likely to

develop difficulties in reading comprehension and reading fluency. By Grade 2,

additional information about risk status can be determined if students have

difficulties with word-level reading skills. These findings could be used as a starting

point for identification and validation of screening tools for ELLs with reading

difficulties.

Appendix

See Table 2.

References

Adlof, S. M., Catts, H. W., & Little, T. D. (2006). Should the simple view of reading include a fluency

component? Reading and Writing: An Interdisciplinary Journal, 19, 933–958.

Allison, P. D. (2003). Missing data techniques for structural equation modeling. Journal of AbnormalPsychology, 112, 545–557.

August, D., & Shanahan, T. (2006). Introduction and methodology. In D. August & T. Shanahan (Eds.),

Developing literacy in second language learners: Report of the National Literacy Panel onLanguage-Minority Children and Youth (pp. 1–42). Mahwah, NJ: Erlbaum.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of Personalityand Social Psychology, 51, 1173–1182.

Bashir, A. S., & Hook, P. E. (2009). Fluency: A key link between word identification and comprehension.

Language, Speech, and Hearing Services in Schools, 40, 196–200.

Berninger, V. W., Abbott, R. D., Trivedi, P., Olson, E., Gould, L., Hiramatsu, S., et al. (2010). Applying

the multiple dimensions of reading fluency to assessment and instruction. Journal of Psychoed-ucational Assessment, 28, 3–18.

Table 2 Means, standard deviations, F-value, and p-value for missing group and the group with com-

plete data

Variables Mean (SD) for

missing group

Mean (SD) for

complete group

F p

LC-S1 7.26 (3.43) 7.36 (3.31) 0.06 0.80

LC-S2 4.25 (2.92) 4.66 (3.04) 1.27 0.26

AAT 7.36 (4.28) 7.37 (4.59) 0.001 0.98

Oddity 10.81 (3.85) 10.91 (4.01) 0.05 0.82

Letter naming 0.03 (0.49) -0.09 (0.68) 2.94 0.09

Object naming 0.04 (0.61) -0.11 (0.71) 3.09 0.08

Decoding 98.82 (15.26) 95.56 (17.12) 2.46 0.12

WID 100.77 (15.21) 98.66 (16.95) 1.05 0.31

MAT (ss) 97.34 (11.92) 98.07 (11.73) 0.14 0.71

LC-S1 listening comprehension-story 1, LC-S2 listening comprehension-story 2, AAT auditory analysis

task, Oddity pseudoword first phoneme identification, WID word identification, MAT (ss) standardized

non-verbal IQ. The sample size for missing group was 83 for the Grade 1 measures and 129 for the Grade

2 measures. For complete group, the sample size was 225 for Grade 1 measures and 179 for Grade 2

measures

184 Z. Yaghoub Zadeh et al.

123

Biemiller, A. J. (1981). Biemiller test of reading processes. Toronto, ON, Canada: University of Toronto

Press.

Biemiller, A. (1999). Language and reading success. Cambridge, MA: Brookline.

Bowers, P. G. (1995). Tracing symbol naming speed’s unique contribution to reading disability over time.

Reading and Writing: An Interdisciplinary Journal, 7, 189–216.

Bowers, P. G., Golden, J., Kennedy, A., & Young, A. (1994). Limits upon orthographic knowledge due to

processes index by naming speed. In V. W. Berninger (Ed.), The varieties of orthographicknowledge: Vol. 1. Theoretical and developmental issues (pp. 173–218). Dordrecht, The

Netherlands: Kluwer.

Buly, M. R., & Valencia, S. W. (2002). Below the bar: Profiles of students who fail state reading

assessments. Educational Evaluation and Policy Analysis, 24, 219–239.

Cain, K., Oakhill, J. V., & Bryant, P. (2000). Phonological skills and comprehension failure: A test of the

phonological processing deficit hypothesis. Reading and Writing: An Interdisciplinary Journal, 13,

31–56.

Cain, K., Oakhill, J. V., & Bryant, P. (2004). Children’s reading comprehension ability: Concurrent

prediction by working memory, verbal ability, and component skills. Journal of EducationalPsychology, 96, 31–42.

Carlisle, J. F., Beeman, M., Davis, H. L., & Spharim, G. (1999). Relationship of metalinguistic

capabilities and reading achievement for children who are becoming bilingual. Applied Psycho-linguistics, 20, 459–478.

Carver, R. P., & David, A. H. (2001). Investigating reading achievement using a causal model. ScientificStudies of Reading, 5, 107–140.

Catts, H. W., Fey, M. E., Zhang, X., & Tomblin, B. (1999). Language basis of reading and reading

disabilities: Evidence from a longitudinal investigation. Scientific Studies of Reading, 3, 331–361.

Chall, J. S. (1983). Stages of reading development. New York: McGraw-Hill.

Chiappe, P., Siegel, L. S., & Wade-Woolley, L. (2002). Linguistic diversity and the development of

reading skills: A longitudinal study. Scientific Studies of Reading, 6, 369–400.

Cohen-Mimran, R. (2009). The contribution of language skills to reading fluency: A comparison of two

orthographies for Hebrew. Journal of Child Language, 36, 657–672.

Cole, D. A. (1987). Utility of confirmatory factor analysis in test validation research. Journal ofConsulting and Clinical Psychology, 55, 584–594.

Collins, W. M., & Levy, B. A. (2008). Developing fluent text procession with practice: Memorial

influences on fluency and comprehension. Canadian Psychology, 49, 133–139.

Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and restrictive strategies

in modern missing data procedures. Psychological Methods, 6, 330–351.

Colon, E. P., & Kranzler, J. H. (2006). Effect of instructions on curriculum-based measurement of

reading. Journal of Psychoeducational Assessment, 24, 318–328.

Crosson, A. C., & Lesaux, N. K. (2010). Revisiting assumptions about the relationship of fluent reading to

comprehension: Spanish-speakers’ text-reading fluency in English. Reading and Writing, AnInterdisciplinary Journal, 23, 475–494.

Cutting, L. E., Materek, A., Cole, C. A. S., Levine, T. M., & Mahone, E. M. (2009). Effects of fluency,

oral language and executive function on reading comprehension performance. Annals of Dyslexia,59, 34–54.

Cutting, L. E., & Scarborough, H. S. (2006). Prediction of reading comprehension: Relative contributions

of word recognition, language proficiency, and other cognitive skills can depend on how

comprehension is measured. Scientific Studies of Reading, 10, 277–299.

Denckla, M. B., & Rudel, R. G. (1976). Rapid automatized naming (R.A.N.): Dyslexia differentiated

from other learning disabilities. Neuropsychologia, 14, 471–479.

Droop, M., & Verhoeven, L. (2003). Language proficiency and reading ability in first- and second-

language learners. Reading Research Quarterly, 38, 78–103.

Durrell, D. D. (1970). Durrell analysis of reading difficulty. New York: Psychological Corporation.

Francis, D. J., Fletcher, J. M., Stuebing, K. K., Lyon, G. R., Shaywitz, B. A., & Shaywitz, S. E. (2005).

Psychometric approaches to the identification of LD: IQ and achievement scores are not sufficient.

Journal of Learning Disabilities, 38, 98–108.

Geva, E. (2006). Second-language oral proficiency and second-language literacy. In D. August &

T. Shanahan (Eds.), Developing literacy in second-language learners: Report of the NationalLiteracy Panel on Language-Minority Children and Youth (pp. 123–139). Mahwah, NJ: Erlbaum.

Mediation model of ELL reading 185

123

Geva, E., & Yaghoub Zadeh, Z. (2006). Reading efficiency in native English-speaking and English-as-a-

second-language children: The role of oral proficiency and underlying cognitive-linguistic

processes. Scientific Studies of Reading, 10, 31–57.

Gottardo, A., & Mueller, J. (2009). Are first and second language factors related in predicting school

language reading comprehension? A study of Spanish-speaking children acquiring English as a

second language from first to second grade. Journal of Educational Psychology, 101, 330–344.

Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and reading disability. RASE: Remedial andSpecial Education, 7, 6–10.

Hoover, W. A., & Gough, P. B. (1990). The simple view of reading. Reading and Writing: AnInterdisciplinary Journal, 2, 127–160.

Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:

Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.

Jenkins, J. R., Fuchs, L. S., van den Broek, P., Espin, C., & Deno, S. L. (2003). Sources of individual

differences in reading comprehension and reading fluency. Journal of Educational Psychology, 95,

719–729.

Johnston, T. C., & Kirby, J. R. (2006). The contribution of naming speed to the simple view of reading.

Reading and Writing: An Interdisciplinary Journal, 19, 339–361.

Joreskog, K. G., & Serbom, D. (2001). LISREL 8: User’s reference guide. Chicago: Scientific Software.

Joshi, R. M., & Aaron, P. G. (2000). The component model of reading: Simple view of reading made a

little more complex. Reading Psychology, 21, 85–97.

Katzir, T., Kim, Y., Wolf, M., O’Brien, B., Kennedy, B., Lovett, M., et al. (2006). Reading fluency: The

whole is more than the parts. Annals of Dyslexia, 56, 51–82.

Kirby, J. R., & Savage, R. S. (2008). Can the simple view deal with the complexities of reading? Literacy,42, 75–82.

Kline, R. B. (1998). Principles and practice of structural equation modeling. New York: Guilford Press.

Lesaux, N. K., Lipka, O., & Siegel, L. S. (2006). Investigating cognitive and linguistic abilities that

influence the reading comprehension skills of children from diverse linguistic backgrounds. Readingand Writing: An Interdisciplinary Journal, 19, 99–131.

Lesaux, N. K., Rupp, A. A., & Siegel, L. S. (2007). Growth in reading skills of children from diverse

linguistic backgrounds: Findings from a five-year longitudinal study. Journal of EducationalPsychology, 99, 821–834.

Manis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). See Dick RAN: Rapid naming and the longitudinal

prediction of reading subskills in first and second graders. Scientific Studies of Reading, 3, 129–157.

Maxwell, S. E., & Cole, D. A. (2007). Bias in cross-sectional analyses of longitudinal mediation.

Psychological Methods, 12, 23–44.

McBride-Chang, C., Wagner, R. K., & Chang, L. (1997). Growth modeling of phonological awareness.

Journal of Educational Psychology, 89, 621–630.

Merbaum, C., & Geva, E. (1998, December). The relationship between listening and readingcomprehension in L1 and L2 Grade one children. Paper presented at the annual meeting of the

National Reading Conference (‘‘The role of oral language proficiency in the development of English

as a second language reading skills of young children’’), Austin, TX.

Meyer, M. S., & Felton, R. H. (1999). Repeated reading to enhance fluency: Old approaches and new

directions. Annals of Dyslexia, 49, 283–306.

Miller, J. F., Heilmann, J., Nockerts, A., Iglesias, A., Fabiano, L., & Francis, D. J. (2006). Oral language

and reading in bilingual children. Learning Disabilities Research and Practice, 21, 30–43.

Nakamoto, J., Lindsey, K. A., & Manis, F. R. (2008). A cross-linguistic investigation of English language

learners’ reading comprehension in English and Spanish. Scientific Studies of Reading, 12, 351–371.

Parrila, R. K., Kirby, J. R., & McQuarrie, L. (2004). Articulation rate, naming speed, verbal short-term

memory, and phonological awareness: Longitudinal predictors of early reading development.

Scientific Studies of Reading, 8, 3–26.

Pennington, B. F., Cardoso-Martins, C., Green, P. A., & Lefly, D. L. (2001). Comparing the phonological

and double deficit hypotheses for developmental dyslexia. Reading and Writing: An Interdisciplin-ary Journal, 14, 707–755.

Perfetti, C. (2007). Reading ability: Lexical quality to comprehension. ‘‘What should the scientific study

of reading be now and in the near future?’’ [special issue]. Scientific Studies of Reading, 11(4),

357–383.

Proctor, C. P., Carlo, M., August, D., & Snow, C. (2005). Native Spanish-speaking children reading in

English: Toward a model of comprehension. Journal of Educational Psychology, 97, 247–256.

186 Z. Yaghoub Zadeh et al.

123

Puranik, C. S., Petscher, Y., Al Otaiba, S., Catts, H. W., & Lonigan, C. J. (2008). Development of oral

reading fluency in children with speech or language impairments. A growth curve analysis. Journalof Learning Disabilities, 41, 545–560.

Riedel, B. W. (2007). The relation between DIBELS, reading comprehension, and vocabulary in urban

first-grade students. Reading Research Quarterly, 42, 546–567.

Rosner, J., & Simon, D. P. (1971). The auditory analysis test: An initial report. Journal of LearningDisabilities, 4, 383–392.

Schafer, J., & Graham, J. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7,

147–177.

Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New

procedures and recommendations. Psychological Methods, 7, 422–445.

Slocum, T. A., Street, E. M., & Gilberts, G. (1995). A review of research and theory on the relation

between oral reading rate and reading comprehension. Journal of Behavioral Education, 5, 377–398.

Sparks, R., & Ganschow, L. (2001). Aptitude for learning a foreign language. Annual Review of AppliedLinguistics, 21, 90–111.

Stanovich, K. E., & West, R. F. (1989). Exposure to print and orthographic processing. Reading ResearchQuarterly, 24, 402–433.

Torgesen, J., Rashotte, C., & Alexander, A. (2001). The prevention and remediation of reading fluency

problems. In M. Wolf (Ed.), Dyslexia, fluency, and the brain (pp. 333–355). Cambridge, MA: York

Press.

Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Burgess, S., & Hecht, S. (1997). Contributions of

phonological awareness and rapid automatic naming ability to the growth of word-reading skills in

second- to fifth-grade children. Scientific Studies of Reading, 1, 161–185.

van Gelderen, A., Schoonen, R., de Glopper, K., & Hulstijn, J. (2007). Development of adolescent

reading comprehension in language 1 and language 2: A longitudinal analysis of constituent

components. Journal of Educational Psychology, 99, 477–491.

van Gelderen, A., Schoonen, R., de Glopper, K., Hulstijn, J., Simis, A., Snellings, P., et al. (2004).

Linguistic knowledge, processing speed and metacognitive knowledge in first and second language

reading comprehension: A componential analysis. Journal of Educational Psychology, 96, 19–30.

Verhoeven, L. (2000). Components in early second language reading and spelling. Scientific Studies ofReading, 4, 313–330.

Vukovic, R. K., & Siegel, L. S. (2006). The double deficit hypothesis: A comprehensive review of the

evidence. Journal of Learning Disabilities, 39, 25–47.

Wagner, R. K., Torgesen, J. K., Rashotte, C. A., Hecht, S. A., Barker, T. A., Burgess, S. R., et al. (1997).

Changing relations between phonological abilities and word-level reading as children develop from

beginning to skilled readers: A five-year longitudinal study. Developmental Psychology, 33,

468–479.

Wiley, H. I., & Deno, S. L. (2005). Oral reading and maze measures as predictors of success for English

learners on a state standards assessment. Remedial and Special Education, 26, 207–214.

Wilkinson, G. S. (1993). Wide range achievement test–Revised (WRAT 3-R) (3rd ed.). Wilmington, DE:

Wide Range.

Wolf, M., & Katzir-Cohen, T. (2001). Reading fluency and its intervention. Scientific Studies of Reading,5, 211–239.

Wolf, M., Pfeil, C., Lotz, R., & Biddle, K. (1994). Toward a more universal understanding of the

developmental dyslexias: The contribution of orthographic factors. In V. W. Berninger (Ed.), Thevarieties of orthographic knowledge: Vol. 1. Theoretical and developmental issues (pp. 137–171).

Dordrecht, The Netherlands: Kluwer.

Wood, D. E. (2009). Modeling the relationships between cognitive and reading measures in third and

fourth grade children. Journal of Psychoeducational Assessment, 27, 96–112.

Woodcock, R. W. (1987). Woodcock reading mastery test. Circle Pines, MN: American Guidance

Service.

Mediation model of ELL reading 187

123